Education Subsidies, Who Are We Helping?
An Empirical Investigation on the Redistributional Effects of Public Education
Subsidies in the OECD.
Master’s Thesis Economics Jacob Veenstra (st. nr. 1540580) Supervisor: Prof. Dr. C. G. M. Sterks Date: 14-05-2010
I Abstract
This paper investigates the redistributional effects of public education subsidies among OECD-countries, using a cross sectional analysis. In contrast to most empirical work in this field, this paper adopts a macro perspective to estimate whether government intervention in education is in essence pro- or regressive. The paper finds that on the whole, public subsidies reduce inequality, mainly because of the result that the direct subsidy receipts end up at poor households. Whether or not education subsidies reduce inequality as a result of changes in learning decisions is unclear.
Keywords: Education subsidies, Income inequality, Static effect, Dynamic effect
2 II Introduction
Economic inequality touches upon the core of social and ethical thinking. Whereas economics probably is the science most suitable for making testable and measurable
predictions on the causes and consequences of inequality, it is obvious that all social sciences have a place in the debate concerning the most suitable government actions of redistributing ‘the economic pie’.
In this paper, I will attempt to set out a framework that investigates the role of
education subsidies as an equalization device. If education subsidies increase (or more
generally, education becomes cheaper), investing in human capital (and thus labour quality) becomes more attractive, both for poor and rich people. However, which group will benefit the most from this is ambiguous as it may depend both on the specifications of the subsidy schemes and the ways in which individuals react to them.
Perhaps even more important however, is the way in which we are actually measuring the progressivity of the subsidy. That is, before one can measure the redistributive effects of education subsidies, it should be clarified from which perspective we are measuring
3 The static perspective examines whether the direct subsidy receipts mainly end up at poor or rich households. If for example students’ access to education depends positively on household income, it may be the case that rich households deliver a relatively large amount of children at higher education institutions. As a result then, rich households would receive a more that proportional amount of the subsidy receipts1. Whether or not subsidies are progressive, is measured by the change in household income inequality before, and directly
after the subsidies are imposed. Clearly then, this measurement is conducted at a point in time
and therefore can be classified as static.
The dynamic perspective focuses on the general equilibrium aspect of the public subsidy schemes. That is, it is acknowledged that subsidies are implemented to influence learning decisions of the young, thereby increasing their productivity and lifetime earnings potential. Now, to measure whether subsidies increase or decrease inequality, we would have to sit and wait for all the decisions and effects to run through the economy. We could then test whether or not they yield a more equal income distribution, compared to cases where no subsidy exists. Intuitively, it is unclear whether this general effect leads to a more equal income distribution. It depends on the interplay of two effects, which we will entitle from now on as the ‘ability effect’ and ‘wealth effect’ (see figure 1).
On the one hand, persons with a high innate ability might experience a high elasticity of education demand, because they have a high ‘return to learning’. Thus the persons, who’s earning potential is already highest, receive an extra incentive to study through the subsidy. On the other hand, this elasticity could be high for poor persons as well, to the extent that without a subsidy, they would be constrained in their decisions on education. If a subsidy is introduced, poor persons can get access to schooling and escape from the poverty trap. If the first (ability) effect dominates, education subsidies would increase inequality, while we would expect the opposite if the second (wealth) effect is of greater importance. Of course the
‘wealth effect’ stands in close connection with the static effect, as they both depend on the way in which persons have access to schooling. However, the actual measurement is quite different.
1
4 So, already in our first careful steps in this area of conduct, it becomes clear that there is an enormous lacuna between the perspectives and methods of research. In the literature this gap remains open wide up and until the day of writing. Authors departing from the theoretical view have presented multiple models revealing the general equilibrium effects of subsidies on inequality. Their intergenerational models explaining the dynamic issues are highly
sophisticated. In sharp contrast with the theoretical work, the empirical literature has not been able to reach beyond a measurement of the static effect.
The main attempt of this paper is to take a bird’s eye view and oversee both the gap in the literature and the two shores at the same time. Next, the paper investigates whether it is possible to throw out a bridge, connecting the two different worlds. Section III reviews the literature and thereby explores both worlds carefully. The main question then is whether it makes sense to combine the views into a coherent framework.
Subsequently, it may be useful to provide a simple model to better understand the area of conduct. Whereas the static view is a purely empirical issue, the dynamic view might need some theoretical support. Therefore, to lay the foundations, section IV presents a model to attempt to better understand the behavioural consequences of education subsidies. I will consider a theoretical model, in which persons are exogenously given their ability and initial wealth. They will divide their time between working and learning. This decision is influenced by the price of education (which is affected by a subsidy on education) and their exogenous characteristics. Whether the subsidy will lead to lower or higher inequality, will partially depend on the exogenous parameters.
The empirical sections then, will zoom out and investigate the progressivity of education subsidies from a macro - or cross-country - perspective. Section V presents the data, while section VI gives the most important empirical results and interpretations. This section presents approximations both from the static as well as the dynamic perspective. For (all) 30 OECD-countries we examine relationships between government outlays for education subsidies and relevant measures of income inequality.2 Finally, section VII concludes.
2
5 III Review of the literature
General Motives for Subsidizing Education
Most countries reduce the private cost of education substantially by large subsidy schemes. Although in reality, the existence and justification of subsidies is hardly ever debated upon, the theoretical and empirical underpinnings are not clear-cut. The field of government intervention in the educational sector is a cumbersome road map with vague borders and a countless number of crossroads. Both positive and normative issues play a role in the debate. To clarify where this paper navigates, it is useful to briefly summarize the main arguments for the government intervening in education.
First, education can been considered a merit good, upon which neither the people that are most suitable to consume (children), nor their parents can decide properly. On these grounds, government intervention often boils down to publicly supplying education and making it obligatory at least for some years. Note that this argument is in essence normative but still widely accepted, especially for primary and secondary schooling. For a more elaborate view, see Friedman (1955).
Secondly, the social benefits of education may exceed the private benefits so that private demand would be too low. As Friedman argues, schooling provides these so-called externalities through either (a) improved citizenship (for example, the crime rate lowers if more schooling is taken up) or (b) labour productivity (i.e. person I becomes more productive if person II learns more). To stimulate education demand, the government could publicly supply education or impose subsidies. Econometric research has not led to a clear cut justification of education subsidies on these grounds (see for example: Ciccone and Peri (2002)).
Finally, it is argued, education subsidies can contribute to a more equal income
distribution (being part of a redistribution scheme). The interesting notion is that - in essence - is it not clear whether education subsidies in itself are pro- or regressive. As noted above, there are two potential (counter)effects, of which it is not clear - if they exist at all - which one dominates. It is often claimed that in an economy without market failures, subsidizing
6 An often heard counterargument focuses on the imperfection of capital markets. If there is a restriction to borrow money for education, poor (but able) persons cannot invest in human capital up to the efficient amount. This makes initial wealth relative to the height of
direct education costs, a factor in deciding on schooling decisions. Subsidies now open access
to education for the poor, leading to a progressive result. As noted, this paper focuses on this final rationale attempting to reveal the workings of the two counter effects. Having finished navigating, it is time to zoom in to the details of these issues.
Education Subsidies as Equalizers?
A long list of literature has been developed to unravel the redistributive effects of government intervention in education. Many of them yield important conclusions, but at the same time they also leave a big hiatus behind. Admitted, this paper is not sanctifying in that respect. To clarify the main arguments, it is relevant to break down the literature into
theoretical and empirical work, as both have their very distinct view of analysis. Whereas
theoretical research adopts a dynamic or general perspective in most cases, the empirical research still prefers the static method of measurement, using so-called 'net transfer calculations'.
The theoretical part of literature focuses on the behavioural consequences of changes in the education subsidies. The possible coexistence of two counter effects, along with the intergenerational structure and difficulties of unraveling causes and effects3, makes this issue a perfect example of the complexity of economics. The part of literature focusing on the ‘ability effect’ fails to yield unambiguous implications for public policy.
For example, Wigger (2001) argues that under an optimal labour tax scheme,
education should actually be taxed rather than subsidized, if the government searches for the most efficient way to redistribute. While abstracting from intergenerational considerations, Wigger considers an economy consisting of low and high talented people, the latter only being capable of investing in human capital. In a world free of market failures, a tax on education is most efficient. Although this tax will reduce investment in human capital, the decrease in labour income tax that comes with it increases the incentives to invest in human capital.
7 As a result, the output from the educational sector is used in a more productive way and therefore the capacity of the high talented people is utilized efficiently. This indeed summarizes and confirms the ‘regressivity’ effect of education subsidies. The crucial assumption leading to this result is that innate ability and education are complementary. In a more complex extension to this view, Maldonado (2007) reworks the conditions for which education subsidies are regressive. In a model with endogenous labour supply, he shows that the elasticity of education with respect to changes in wages is an important
parameter for the government to consider. Thus taking into account that changes in education subsidies and income tax may change both labour supply and education, it may be the case that subsidies are more optimal.
Another strand of literature focuses on the interaction between subsidy and tax policies from a slightly different point of view. Dur and Teulings (2001, 2003) consider an economy free of failures but still find that - as long as there is some call for redistribution - a subsidy on education can be justified, since it lowers the human capital premium. To get the intuition, think of an economy where people are divided between two wage platforms, either low (unskilled) or high (skilled). If an increase in the education subsidy makes you jump from the low to the high platform, your wage will rise, but because there will be people making the same jump, the high platform will have to carry a larger weight and starts to go down.
As a result, the pre-tax wage gap shrinks. Less distortionary taxes are then needed to reach a certain distribution. In their paper, Dur and Teulings (allowing for a continuum of ability-types) state that individuals do not take into account this beneficial effect of decreased distortion, so education demand is too low4. Education subsidies are an efficient device to reach the desired outcome.
A similar argument is brought to the fore by Bovenberg and Jacobs (2001), who state that to minimize distortions of a distribution scheme, subsidies on education should be imposed to ensure an efficient amount of education taken up by individuals, after which an income tax is imposed to redistribute. Concluding, in a world free of market failures, at first glance education subsidies seem regressive, but this result is not robust to changes in model specifications.
4
8 The literature on the ‘wealth effect’ has a more empirical character, but still considers the intergenerational aspect of the field. Here, the main question is to investigate whether indeed initial wealth (approximated by parental income) is a possible constraint in schooling decisions. If the answer is affirmative, it means that capital markets are imperfect and indeed initial wealth is a relevant factor in the decision making process. As noted above, education subsidies - even if regressive in a world without market failures - can be justified on grounds of these capital market imperfections. It is argued that without subsidies, education is not within reach of poorer students, since they cannot obtain the funds necessary through the capital markets. But if even they could, it is argued often, poor people would not borrow because of high risk aversion. In the case of accessibility then, normative and positive arguments start meeting each other again. Proponents of subsidies would argue that young persons cannot properly decide on risk-return investments and access to higher education is a good that everybody has the right to obtain.
Most empirical literature on this issue concludes that initial wealth indeed is a factor in deciding on the amount of schooling. There are plenty of results available revealing that parental income is positively correlated to income of their children, suggesting that children of poor parents don’t have equal access to education compared to their rich counterparts. However, Shea (1997) notes that such a positive correlation is not enough to justify the view that low income families are trapped in poverty. It may be the case that parental income is related to their ability, and parents pass their ability to their children. On the whole then, children from poor households do not suffer from inequality of opportunities. Indeed, after controlling for ability, Shea finds that parental income does not matter, unless for the poorest income groups. This exception to the main result is crucial, since the ‘inequality of
opportunity problems’ are most likely to be relevant for exactly the poorest families. In a similar paper, Plug and Vijverberg (2003) adopt a natural experiment to test the causality between parent’s and children’s education. On the whole, there is a positive
correlation between the two, and the question is whether this linkage is caused by inheritance effects or because parents with high education provide a better environment for their children. Plug and Vijverberg compare the schooling success of biological children and adopted
9 Plug and Vijverberg (2001) explicitly state that children from poor families do not have the same access to education as their richer counterparts. Lower education for poor children will also result in lower earnings for themselves, so that the children’s children will potentially be restricted as well. To break this chain, Plug and Vijverberg propose that at least the poor should be supported by government intervention in education.
If one thing becomes clear from the above, it is that consensus on the most appropriate government actions in the education sector has not been reached. Thus, when focusing on the dynamic method, the effects of education subsidies on economic inequality are debatable. Although Wigger (2001) claims the regressivity of education subsidies to be ‘a well documented fact’, broad consensus among scholars has not been reached until now.
Shifting the focus to the empirical literature - thus concentrating on the static
10 In an influential paper, Barbaro (2002) summarizes the bulk of the empirical research conducted on this subject and concludes that most of the studies found a progressive
redistribution as a result of the government education financing. Caution should be kept when comparing the literature one to one, since research differs with respect to the regions
investigated, and different regions are most likely to have very different tax and subsidy schemes. Also, the research methodology is not the same and the data availability is far from complete in most cases. Barbaro himself conducted a net-transfer calculation for West-Germany in 1997. He split up the population of West-West-Germany into income deciles, and for every decile the net benefits (education subsidies minus tax incidence) are calculated. Barbaro then states that whether the policy scheme is pro- or regressive depends on the interplay of two effects.
On the one hand, the ‘structural effect’ states that children from richer deciles have a higher probability of entering University College, but the ‘level effect’ states that the highest deciles contain the lowest percentage of households with children. A priori then, it is not clear which effect dominates. It may well be the case that the level effect dominates, so that the student representation is highest for low income deciles. Consequently, the poor deciles would receive the largest net benefits. For the West-German case, the net-transfer calculations indeed confirm this, leading Barbaro to conclude that the subsidy and tax scheme is
progressive.
The most important shortcoming of the empirical field, also acknowledged by
11 In conclusion, most of the empirical work employs net-transfer (or static) calculations in search for the redistributional effects of subsidies. As noted above, most of them find a progressive impact. But in fact, this result leads us no further than to conclude that the static measurement method yields a progressive result. Whether the dynamic method can give conclusions in support of this view is unclear. This paper is a first attempt to include the general perspective in empirical work. Before we are able to do this, it might be useful to briefly explain the theoretical underpinnings of the dynamic perspective.
IV Theoretical Underpinnings of the Dynamic Perspective
Before diving into theory, let us one last time clarify the focus of this paper. I attempt to set out and test a framework that investigates the redistribution effects of the education subsidy schemes in the OECD. To have some guideline, I set up a simple model to show that the general effect of education subsidies on inequality is ambiguous.
The main intuitive idea runs as follows: people can differ exogenously among two dimensions: wealth and ability5. That is, from the literature review, we take the conclusion that initial wealth and innate ability are relevant parameters in schooling decisions, as given. Therefore, it is unnecessary to redevelop a model to show why this is the case. We can thus suffice with a very simple model, attempting to show that the effect of education subsidies on the Gini-coefficient (being our measure of income inequality) is ambiguous.
We will first consider the case in which only ability is relevant. The expectation is that a rise in education subsidies leads to more inequality (‘ability effect’). Secondly, we consider the case where persons only differ in terms of initial wealth. To the extent that initial wealth constrains persons in choosing the optimal amount of education, the expectation is that
subsidizing education provides equality of opportunity and therefore has a progressive impact.
5
12 We consider a 2-person economy. People decide on their time allocation between working and learning in order to maximize disposable income given by6:
) 2 , 1 ( ) ( ) , ( 0 Y H E N b E T i Yi i i i i i i i (1a) tax sum lump T and subsidy the for corrected education of price b time working N time education E ability innate ty productivi labour H wealth initial Y income disposable Y Where , , , , , , , : 0
In words, disposable income is the sum of initial wealth, real production and the net benefits received from the government sector (which may be negative). Note that education has both direct (b-σ) and indirect costs (forgone earnings). Furthermore, assume that:
) , ( i i i E
H =iEi (1b)
so that
H
E =1, and thus ability and education are complements in productivity. Now we distinguish between two simple cases. First, I will present a model where only ability is relevant and secondly, the focus is shifted to the case where only wealth is relevant.Case 1: Let us start with the simplest model possible, where we try to find the behavioural effects of a change in education subsidy. We consider an economy of two persons, who only differ in ability. We suppose initial wealth is no factor of importance, that is: people are not constrained in the amount of education spending by current wealth, so we can set Y0 0. For now, assume it is unknown how the tax burden is shared between the
persons at the time of their choice on E and N, so their decisions cannot be influenced by T. If we normalize total time available to 1, which is divided between N and E, we find that
optimizing behaviour leads to7
i i i i b E and b N 2 2 1 2 2 1 (2) Using (2) in (1b) we have i i i i
b
N
H
4
)
(
(.)
2 2
(3) 6Because we consider a one-period model, the distinction between disposable income and consumption is not needed.
7
13
Define the Gini-coefficient8 as
,
1 1 2 2 1 1 2 2
N
H
N
H
N
H
N
H
G
(4)which is our measure of inequality. Now compare persons 1 and 2, and assume 2 1. Now clearly, for γ>1 (<1) person 2 (1) has higher ability and income. Using this assumption in (3), and substituting the result in (4) we get a Gini-coefficient of:
2 2
1 2 2 1 ) ( ) 1 ( ) ( ) 1 ( b b GTaking the first-order derivative with respect to the education subsidy, σ yields:
2 2 2 1 2 1 ) ) ( )( 1 ( ) )( 1 ( 4 b b d dG for γ>1, and 2 2 2 1 2 1 ) ) ( )( 1 ( ) )( 1 ( 4 b b d dG for γ<1
which is always negative.Thus, the income inequality actually lowers if the subsidy rises.9 The intuition is that a low ability person reacts more to a change in the price of education, since even a slight change in the price means a lot if your absolute income is low. Thus, concentrating purely on a decrease in the price of schooling, the inequality in productivity lowers.
The main argument of opponents of public education subsidies however, is that inequality actually rises as a result of the subsidy, since poor persons in fact pay for the subsidy of the rich through taxes. That is, although in the previous model, poor persons face a larger price elasticity of education, still the rich persons consume a larger amount of education in absolute terms. If the tax system then is not progressive enough, the poor (rich) pay more (less) in taxes than their benefit from the subsidy receipts.
8 In a two-person economy, the Gini-coefficient would actually be
) ( ) ( 5 . 0 1 1 2 2 1 1 2 2 N H N H N H N H
We multiply the Gini-coefficient by a factor 2 to ensure again that it is between zero and one. 9
14 To examine whether it is possible to let education subsidies have a regressive effect, take a specific case of our model, in which we assume that each agent pays a(n) (equal) lump sum tax to finance the education subsidies. Agents maximize (1a) as before but take into account that their tax burden will satisfy:
) ( 2 1 2 2 1 T T E E T (GBC) (5)
Substituting (5) and (1b) into (1a), we take the usual FOC’s and eventually find10:
i i b E
4 2 2 1 and i i b N
4 2 2 1 i=1,2 (6)Substituting all this into (1a) we have
2 16 ) 2 ( 2 16 2 4 2 b b b Y j i i i
(7)Now define the Gini-coefficient as:
1 2 1 2
Y
Y
Y
Y
G
(8)And assume as before that2 1.Using this assumption in (7) and substituting the result in (8), the Gini coefficient is:
1 2 1 2 2 1 2 16 4 ) 2 ( 2 ) 2 ( ) 1 ( 4 ) 2 ( 2 ) 2 ( ) 1 ( b b b b b G To conclude a regressive effect of the education subsidy, we should have that the first-order derivative of the Gini-coefficient with respect to the education subsidy is positive. It can be verified that this holds if:
1 1 2 6 8 G b
for γ>1, and (9a)
10
Now, for an interior solution, we assume
2 ) 2
(
15 1 1 2 6 8 G b for γ<1 (9b)
Clearly, condition (9) is always fulfilled, since the RHS is negative, and the LHS is positive as long as b>σ which is reasonable. Thus, if the tax burden is divided equally, indeed the subsidization of education increases inequality. To conclude, whether the ‘ability effect’ will be pro- or regressive depends on the specifications of the tax and subsidy schemes.
Case 2: As noted, the main argument of proponents of education subsidies is that they provide equal access to education and subsequently lead to a more equal income distribution. To model this specific case, we need to incorporate a measure of initial wealth. Assume for simplicity that persons differ only in terms of their initial wealth. It is then straightforward to show by means of a simple example that, as long as the initial wealth indeed constrains one of the persons, an education subsidy would provide a more equal income distribution. Assume again, the tax schedule is unknown at the moment of decision making and normalize φ to 1. Furthermore assume that person 1 has a favourable position because of higher wealth. Maximization leads to (2) again, but note that there is an upper limit on education spending:
0
)
(b Ei Yi (10)
That is, persons cannot spend more on education than the height of their initial wealth.11 For simplicity, let (10) hold with equality for both persons, but assume that it is a real
constraint for person 2 only. That is, initial wealth of person 1 is just high enough to afford an optimal education amount, whereas person 2 would prefer to rise his amount of education, but is constrained. Optimization leads to total disposable income of:
2 0 0 ) ( ) ( b Y b Y
Yi i i , giving a Gini-coefficient of:
) ( ) ( ) ( ) ( 0 2 0 2 0 1 0 1 0 2 0 2 0 1 0 1 Y b Y Y b Y Y b Y Y b Y G (11a)
16 If the government now raises the subsidy so that (10) does not constrain either person, decisions on working and learning will be equalized among the individuals, so that their incomes will only differ to the extent that initial wealth is unequal. The Gini-Coefficient is:
0 2 0 1 0 2 0 1 Y Y Y Y G Where 2EiNi 2(b)Ei (11b) Comparing (11a) and (11b) we see that the Gini-coefficient under 11b - that is, with increased subsidy - has actually lowered. That is, if a rise in the subsidy frees a fraction of the population from the initial wealth constraint, inequality will be lowered. Note that in this example we measured the total effects of the subsidy, after all decisions have been made and the effects have ran through the economy.
Although the models presented above are rather specific in nature, they are powerful enough to provide our previously mentioned notion that education subsidies may be
progressive as well as regressive. Concerning the ‘ability effect’, we found that the distributional nature of the subsidies depends on the specifications of tax and subsidy schemes. The ‘wealth effect’ on the other hand, clearly points to a progressive impact of education subsidies. From a theoretical viewpoint then, if wealth and ability interact, it is unclear what the general effect of education subsidies on inequality will be.
V Data
17 VI Empirics
In order to identify the effects of education subsidies in the OECD, we zoom out and focus on the subject from a macro-perspective. As noted, because of a lack of adequate data, a complete picture of the effects of government intervention in education is hard to sketch. Therefore we should always keep in mind carefully what the results tell us and which fields will remain to lay fallow.
To get some intuition, the empirical section begins with a very brief summary of government spending figures in the OECD. A first question of interest is whether on average, subsidies are actually paid back by the recipients through higher taxes. If earnings are a function of investment in human capital, which may be subsidized, we can estimate whether on average tuition subsidies are paid back because of higher taxes. This does not mean
however that the tax system has to be progressive.12 In an extensive report, Education at a Glance (2009), the OECD provided estimates of
subsidy outlays to tertiary schooling. Together with the information on the increase in tax burden as a result of higher earnings, we compute the average net benefit for tertiary students. As table 1 (see appendix) reveals, for the bulk of the OECD-countries, the extra tax burden as a result of increased earnings outweighs the total subsidies received. Although not very informative, we can conclude from this figure that (on average) persons with tertiary schooling pay back their subsidies completely because they pay more tax. However, if we were to compare total spending on education per person, with increased tax burdens, we find that for most countries governments actually pay more per student than can be compensated by later tax revenues.
12
18 Now, to investigate the most important relationships concerning education spending among OECD-countries, a first useful step is to consider whether the height of subsidies positively influences the school enrollment rates. The paper investigates OECD and World Bank data and relates several measures of education subsidies to tertiary school enrollment rates in order to have a first guess what the best approximation for public education subsidies is. Table 2 presents four possible approximations: (1) 'public grants and scholarships', (2) 'public loans', (3) 'public (or total) financial aid to students' and (4) 'the level of public education spending to private entities'. Note that the first three measures are interconnected since 'public financial aid' is simply the sum of 'grants and scholarships' and 'public loans'.
As table 2 reveals, all explanatory variables used seem to be reasonable
approximations. That is, all 4 measures of subsidies have a positive and significant effect on the enrollment rate. This suggests that the level of initial (direct) costs of schooling - although often considered to be of minor importance compared to indirect costs - is an important factor in tertiary school enrollment.
Note however that when splitting up 'total financial aid' and simultaneously testing for 'grants and scholarships' on the one hand and 'public loans' on the other, we find from column 3 that only the former remains a relevant variable. In conclusion, 'total financial aid' from the government mainly supports students through the beneficial effects of 'grants and
scholarships'. This implies that the measure of 'grants and scholarships' is preferred over 'total financial aid'. Our final measure; the 'level of public education spending to private entities' is significant as well so the paper continues to use this measure in the empirical section.
19 Table 3 concentrates on the static effect. As noted, most empirical work suggests this effect to be progressive (i.e. most of the government transfers end up at poor(er) households). In essence, the method used in this paper tries to answer the same question as the bulk of the empirical literature in the past. It is investigated whether the direct subsidy receipts mainly end up at poor or rich households. However, instead of explicitly calculating the total subsidy receipts for different income groups, this paper simply questions whether or not the inclusion of education subsidies in our measure of income inequality, lowers this measure.
The OECD provides a measure of the Gini-coefficient, both before and after
government transfers and taxes. If indeed, the static effect is progressive, we would expect for
20 The dependent variable is defined as:
transfers and taxes before t coefficien Gini transfers and taxes after t coefficien Gini ratio Gini
From table 3, there seems to be a negative relationship between the Gini ratio and the level of public spending on grants and scholarships. Indeed, a significant negative relationship between the variables is observed, suggesting a progressive static effect as a result of public education subsidies. If we include total public education spending to private entities (instead of grants and scholarships only), we find an insignificant - though still negative -relationship. So the level of grants and scholarships - being subsidies directly benefiting the students - seems to be the most suitable explanatory variable.
We should keep in mind however, that apart from education spending, most
governments impose a number of other public spending facilities aiming at achieving a more equal income distribution. To control for this, we include a measure of social spending of the government as a percentage of GDP. Indeed as the table reveals, a large portion of change in the Gini ratio is explained by social spending. However, grants and scholarships still have substantial explanatory power. Finally, including a measure of tax progressivity13 does not change the main result, leaving our conclusion of the progressivity of the static effect untouched.
The dynamic effect is somewhat more cumbersome to test empirically. As noted above, we would like to estimate the behavioural consequences as a result of public education expenditure policies. That is, the core question of analysis is whether the amount of public education subsidies influences inequality measures in subsequent years. Theoretically, it is unclear how many years it takes for an education policy to work through in labour markets, and its influence on inequality hereafter.
13
The measure of tax progressivity is defined as: ‘marginal tax rate of a person earning 200% of average
21 Moreover, proper data on our subsidy measures is available only for the period after 1997, whereas intuitively, we would expect the process of public policies seeping through the economy to last for more than the 10 years of data we are dealing with.14 However, there is always a bright side to the story. As said, before 1997, no proper data is available. But if the policies are persistent enough among countries over time, the level of education subsidies may be good proxies for years before. Indeed, the data tell us that public spending on grants and scholarships is strongly correlated over time.15 Therefore, the relationship between public education policies and measures of inequality, even within short time intervals, is likely to have a valuable economic interpretation.
For completeness, we estimate simple regression estimates that link public grants and scholarships for the years 1997-2005 to our main inequality measure, the Gini-coefficient. At first glance from table 4, we see that in all cases the negative relationship between public spending and the Gini-coefficients appear. That is, from a cross country perspective, the results hint at a progressive effect of grant and scholarship spending. Examining more closely, two general trends immediately become apparent. First, when reading along the columns we see that the magnitude of the coefficients decreases over time, i.e. it seems that indeed public policies have their strongest influence with a certain time lag.
Secondly and most important, when reading along the rows, we observe a large relevance of our use of the inequality measure. The relationship between spending on grants and scholarships to the Gini-coefficient before transfers and taxes indeed is negative, but never significant. If this relationship were to approximate the ‘ability effect’, we thus fail to find a significant effect. However, if we were to replace our measure of the Gini-coefficient
with the incorporation of transfers and taxes, the relationship becomes significantly negative
for all periods. Thus, the inclusion of the direct static effect leads to a clearly progressive result. Tables 3 and 4 together then, let us conclude that in total, public spending on grants and scholarships lead to a more equal income distribution. The most significant channel through which this effect flows seems to be the direct benefits that the poor receive because they receive a more than proportional share of the financial student aid. Although totally different in methodology, the results at least support the notion found in the bulk of empirical work from the past.
14 We do have data on ‘total public expenditure on education’ before 1997, but as noted, ‘public spending on grants and scholarships’, or ‘financial aid to students’ seem to be the more proper measures. Indeed, concerning these two measures, we only have data availability from 1997 onwards.
22 Let us focus more thoroughly on the total effect. We again notice that the Gini-
coefficient after transfers and taxes may be influenced by other (social) transfers and taxes. Incorporating our measures of total social spending and tax progressivity mentioned above, we find that there is still a significant explanatory role for education policies. Table 5 shows that although the tax progressivity has no explanatory power, the incorporation of a social spending measure is relevant. Indeed, if we incorporate this, we see that the coefficient for public grants and scholarships only remains significant for the most recent years.16 This stands in contrast with the findings of table 4, where statistical significance appears for all years, no matter how far in the past. Thus, the inclusion of other measures of social spending is important, leaving only recent education subsidies as true equalizers. This supports the view that the ‘static effect’ is the main channel through which the progressivity runs. Because it has a short period of influence, statistical significance only appears for more recent years. Thus, when considering the effect of public grants and scholarships in 1997 on the Gini-coefficient of 2005, we are actually measuring the dynamic effect only. Indeed, we see that in this case the relation is negative but insignificant.
A final relevant test might be to replace our measure of spending on grants and scholarships by a measure of total public spending on education. Although this measure may be less appropriate as an approximation for financial aid to students, the benefit is that data is available for more years. Table 6 provides estimates for this measure, leaving our conclusion untouched. For recent years the effects of education expenditure show the highest
significance, suggesting again a leading role for the static effect. Indeed, if we were to replace our measure of the Gini-coefficient after transfers and taxes by the Gini-coefficient before
transfers and taxes, not a single significant relationship appears (results not presented).
Furthermore, if we were to control for other relevant measures such as the level of social spending, the significance of education spending also falls rapidly (again, results not
presented). This result supports the notion that the level of grants and scholarships is the most appropriate measure to approximate education subsidies.
23 In general, the empirical section suggests that education subsidies in the form of grants and scholarships are progressive in nature, mainly because of the direct impact. Although the data is still far for complete, it is improving rapidly so that in the near future empirical work in this area is likely to provide stronger conclusions. It would then for example be interesting to estimate the effects of public education spending in developing countries as opposed to OECD member states. Also, if more attention were to be paid to the specific workings of carry-over effects we might yield more powerful conclusions in future research.
VII Conclusion
The field of government intervention in the education sector suffers from a lack of coherent research. Whereas the authors of theoretical literature have developed a large bulk of intergenerational models, explaining the general equilibrium effects of changes in education subsidies, empirical work is restricted by lack of data. As Conlisk (1977) noted, data on matched parent-child lifetime earnings would be needed to reach to the core of the field from an empirical viewpoint. This paper attempted to bridge the gap between the standard
empirical work which focuses on the static effect and the theoretical literature, with a dynamic aspect. The methodology of zooming out and estimating the effects from a cross-country or macro-perspective is fragile, but might at least offer us a new point of view through which the general relations become clear.
Our results hint at a progressive effect of public education subsidies. The main channel through which the progressivity is achieved seems to be the static effect. That is, public education subsidies mainly benefit the poor because of the direct transfers they receive. Abstracting from this direct effect, we find that the subsidies have no significant effect on inequality; it is only after incorporating transfers and taxes to our definition of inequality that the true progressive nature of the subsidies comes to the fore.
24 Furthermore, we should keep in mind that all policy implications depend upon the preferences of the government. Most models have in common that the government searches for redistribution to a certain extent. But other aspects such as equality of opportunity and social mobility may play a role as well. So again, the mixture of positive and normative arguments appears. In that aspect, decisions on public education subsidies are no different than all other political processes; coloured by differences in ideology.
25 VIII References
Barbaro, S. (2002). The Distributional Impact of Subsidies to Higher Education – Empirical Evidence from Germany, Departmental Discussion Paper, University of Goettingen, No. 114
Bovenberg, A.L and Jacobs, B. (2001). Redistribution and Education Subsidies are Siamese Twins, CEPR discussion paper No. 3309.
Ciccone, A. and Peri, G. (2002). Identifying Human Capital Externalities: Theory with an Application to US Cities, CEPR Discussion Paper No. 3350
Conlisk, J. (1977). A Further Look at the Hansen-Weisbrod-Pechman Debate, The Journal of
Human Resources, Vol. 12, No. 2 (Spring 1977), pp. 147-163
Dur, R.A.J. and Teulings, C. N. (2001). Education and Efficient Redistribution, CESifo
working paper No. 592
Dur, R.A.J. and Teulings, C.N. (2003). Are education subsidies an efficient redistributive device? Tinbergen Institute, Erasmus University Rotterdam discussion paper (TI 2003-024/3)
Friedman, M. (1955). The Role of Government in Education. in: Solo, R. A. (Ed.) Economics
and the Public Interest (NJ, Rutgers University Press).
Hansen, W.L. and Weisbrod, B.A. (1969). The Distribution of Costs and Direct Benefits of Public Higher Education: The Case of California, The Journal of Human Resources Vol. 4,
No. 2 (Spring 1969), pp. 176-191
Jacobs, B. (2002a), Equity Participation in het Hoger Onderwijs, ESB, 484-487
Jacobs, B. (2002b), An investigation of education finance reform -Graduate Taxes and Income Contingent Loans in the Netherlands, CPB Discussion Paper, No. 9
Maldonado, D. (2007). Education policies and optimal taxation, Int Tax Public Finance
(2008) 15:131-143
OECD (2009), Education at a Glance: OECD Indicators 2009, OECD
Pechman, J.A. (1970). The Distributional Effects of Public Higher Education in California,
26 Plug, E. and Vijverberg, W (2001). Schooling, Family Background, and Adoption: Does Family Income matter? IZA discussion paper No. 246.
Plug, E. and Vijverberg, W. (2003). Schooling, Family Background, and Adoption: Is It Nature or Is It Nurture? Journal of Political Economy 2003, vol. 111, No. 3
Rosen, H. S. (2005). Public Finance (seventh edition). McGraw-Hill, International Edition,
New York
Shea, J. (1997). Does parents’ Money matter? NBER Working paper No. 6026.
Wigger, B. U. (2001). Higher Education Financing and Income Redistribution, CESifo
27 IX Appendix
28
Table 2. Effects of public education subsidies on enrollment rates (tertiary schooling). Enrollment Rate17
Level of grants and scholarships18 0.013 (3.176)***
0.010 (2.365)** Level of Public loans19 0.003
(2.334)**
0.001 (0.878) Level of total financial aid
(=grants and scholarships + loans)
0.004 (1.917)* Level of public education spending
to private entities20 0.004 (2.651)** Constant 35.984 (5.690)*** 44.191 (8.419)*** 39.138 (6.672)*** 38.409 (6.682)*** N 28 27 27 29 29 R2 0.143 0.039 0.127 0.071 0.101
- Numbers between brackets show absolute t-statistics
- * Significant at 10% level, ** significant at 5% level, *** significant at 1% level - Standard errors corrected for heteroskedasticity when necessary
17 People enrolled in public tertiary education as a percentage of the total population eligible for tertiary education.
18
Measured as total government outlays on grants and scholarships divided by the total population eligible for tertiary education. All monetary measures in nominal US dollars.
19 Measured as total government outlays on education loans divided by the total population eligible for tertiary education.
20
29
Table 3. Static effects of public education subsidies on income inequality.
Gini Ratio (Mid 2000)21 Level of grants and
scholarships (2005) (in millions of US Dollars)
-32.728 (4.700)*** -17.536 (1.910)* -31.247 (4.070)*** -18.112 (1.564) Level of public education
spending to private entities (2005)
(in millions of US Dollars)
-10.676 (1.585) Social spending of government as a percentage of GDP (2005) -0.011 (4.986)*** -0.013 (5.966)*** Tax progressivity (2005) 0.018 (1.284) 0.015 (1.673) Constant 0.713 35.184*** 0.695 (31.463)*** 0.930 (19.011)*** 0.670 (20.576)*** 0.929 (19.314)*** N 28 29 27 27 26 R2 0.203 0.071 0.624 0.257 0.691
- Numbers between brackets show absolute t-statistics
- * Significant at 10% level, ** significant at 5% level, *** significant at 1% level - Standard errors corrected for heteroskedasticity when necessary
21
30
Table 4. Dynamic and general effects of public education subsidies on income inequality.
Public spending on grants and scholarships in:
Gini-coefficient before transfers and taxes (mid 2000)
31
Table 4 (cont.) Dynamic and general effects of public education subsidies on income inequality.
2002 Coefficient -12.987 (0.970) 0.469 (30.971)*** -28.079 (4.133)** 0.331 (26.553)*** Constant N R2 28 0.035 28 0.224 2003 Coefficient -14.732 (1.250) 0.472 (31.431)*** -26.584 (3.870)*** 0.334 (26.557)*** Constant N R2 29 0.055 29 0.246 2004 Coefficient -12.103 (1.161) 0.472 (30.480)*** -24.001 (4.160)*** 0.334 (26.053)*** Constant N R2 28 0.049 28 0.266 2005 Coefficient -8.469 (0.897) 0.464 (30.917)*** -19.361 (4.222)*** 0.329 (27.407)*** Constant N R2 28 0.030 28 0.232 - Numbers between brackets show absolute t-statistics
32
Table 5. General effects of education policies on inequality.
Gini-coefficient after transfers and taxes (mid 2000) Social spending (2005) -0.004 (2.718)** -0.005 (3.331)*** -0.004 (3.160)*** -0.002 (1.662) Tax progressivity (2005) 0.004 (0.895) Grants and scholarships (1997) -18.751 (1.180) Grants and scholarships (2001) -17.989 (1.704) Grants and scholarships (2005) -12.359 (2.001)* -12.193 (4.437)** Constant 0.408 (12.409)*** 0.423 (14.059)*** 0.410 (14.339)*** 0.355 (10.697)*** N 23 28 27 26 R2 0.412 0.454 0.451 0.367
- Numbers between brackets show absolute t-statistics
33
Table 6. General effects of education policies (total public spending).
Public spending per student (share of per capita GDP) in:
Gini-coefficient after transfers and taxes (mid 2000)
1975 -0.019 (0.506) 1985 -0.090 (1.917)* 1995 -0.142 (3.492)*** 2005 -0.176 (2.804)*** constant 0.312 (15.230)*** 0.355 (13.303)*** 0.344 (20.547)*** 0.363 (18.942)*** N 20 22 25 26 R2 0.009 0.061 0.254 0.153
- Numbers between brackets show absolute t-statistics
