AIM-AP. Accurate Income Measurement for the Assessment of Public Policies. Citizens and Governance in a Knowledge-based Society

Transcription

1 Project no: AIM-AP Accurate Income Measurement for the Assessment of Public Policies Specific Targeted Research or Innovation Project Citizens and Governance in a Knowledge-based Society Deliverable 1.2g Public education in the Netherlands Due date of deliverable: January 2007 Actual submission date: February 2007 Start date of project: 1 February 2006 Duration: 3 years Lead partner: CentERdata Revision [draft]

2 Accurate Income Measurement for the Assessment of Public Policies (AIM-AP): Project 1. Non-cash incomes WP1.2 Public education Deliverable D1.2g Education report on the Netherlands Klaas de Vos CentERdata Tilburg University PO Box LE Tilburg The Netherlands October 2006 Abstract This report analyses the distributional impact of public education expenditures in the Netherlands. On the basis of the Euromod baseline simulation for 2001 we examine the effect on the income distribution of adding public education expenditures per student to household income. In particular, the effects on relative poverty and income inequality are discussed, decomposed according to household type, employment status, age group and education of the head of the household. 1

3 0. Introduction One principal objective of the AIM-AP project is to enrich the definition of household income with non-cash components such as education, housing and health care. In this first AIMAP report on the Netherlands we concentrate on public education expenditures. On the basis of OECD figures we calculate the education expenditures per student in primary, secondary and tertiary education. Using micro-data on household income from the Euromod baseline 2001 we compare the baseline income distribution with the income distribution where these expenditures are counted as additional income. 1. The Dutch education system Table A provides a very brief description of the Dutch education system. For most children, formal education starts at age 4. Primary school lasts 8 years. In secondary education, there are various levels for children with varying talents. Compulsory full-time education ends at the end of the school year in which the child turned 16. In addition, one year of part-time education is compulsory. Table A. Education in the Netherlands Primary education 1. Basic education (po): Primary school Duration of studies: 8 years Secondary education 1. Secondary education (vo): a. preparatory intermediate level vocational education (VMBO) Duration of studies: 4 years b. higher general education (HAVO) Duration of studies: 5 years c. preparatory scientific education (VWO) Duration of studies: 6 years 2. Intermediate level vocational education (MBO) Duration of studies: 2-4 years Ages 4-11 Ages Ages Compulsory from age 5, 99% attends at age 4 Compulsory until age 16, + 2 days per week age 16 Tertiary education 1. Scientific education : Universities Duration of studies : 4-5 years 2. Higher vocational education (HBO) Duration of studies : 4 years Ages 18+ Ages 17+ 2

4 There are state schools ( openbaar onderwijs ) and non-state schools ( bijzonder onderwijs ), but by law almost all education is paid for by the state. From age 18 (until 2005: from age 16), tuition fees are charged but they are by no means sufficient to cover the costs of education. Next to tuition fees, schools may charge for e.g., extracurricular activities. In addition to the schools described in table A there is special education meant for children with serious handicaps and/or learning difficulties. In this report we will not be able to distinguish children in special education. Table B summarizes the numbers of students and the amounts of public spending on education in Table B Number of students and cost structure of the Dutch education system (2001) Students (10 6 ) Current public spending (10 9 Euro) Average spending per student (1000 Euro) Primary education Secondary education Tertiary education (excl. R & D) Source: Statistics Netherlands, Statline, Notably, the figures in the third column cannot be calculated by dividing the first and the second column. In fact, the comparable figures for 2003 are consistent with the OECDfigures 1 which include tuition fees and some other non-public expenditures and refer to fulltime students. In our calculations of education expenditures to be added to be added to household disposable income, we will use the figures in the third column, subtracting tuition fees if relevant and taking account of part-time students as follows. For students in secondary education, 8 hours of paid work are assumed to be possible in full-time education, but for students working more than 8 hours the amounts are reduced proportionately: For students working 16 hours, we take 75% of the amounts mentioned in the table, and for students working 24 hours, we take 50%. For students in tertiary education, we assume below 16 hours of paid work all are in full-time education, between 16 and 24 hours of paid work, a weighted average of the amounts for full-time and half-time education is taken, whilst above 1 OECD (2006), Education at a glance, OECD Indicators, Paris. The 2003 figures in this OECD report form the basis for the results in the comparative report. 3

5 24 hours of paid work the amounts are weighted averages of half-time and no education. Persons working 40 hours per week are assumed not to profit from public education expenditures. In a sensitivity analysis we will check to what extent this assumption affects the results. Notably, the available data do not allow differentiation within education levels. In particular, we assume that all children aged 4-11 participate in (full-time) primary education, and all children aged participate in (full-time) secondary education. For persons aged 16 or over who are currently in education we assume that they are in public education unless they follow a correspondence course or a course paid for and or organised by their employer. For persons in secondary and tertiary education we cannot distinguish between levels, subjects, trades or professions, nor can we distinguish post-secondary non-tertiary education from secondary education. In this report we add total amounts per student to household income. In fact, one could question whether these gross amounts should be added to net income. In theory, one might prefer to include the (average) effect of a year of additional education on the net present value of (net) lifetime income. This would not solve the problem that it is not clear for whom education is successful, and who is, e.g., participating longer than the normal number of years. It should also be noted that we cannot identify students in private education. We assume that adults following correspondence courses are in private education but neglect the (very small) number of children enrolled in private schools. The most important earlier studies on the distributional effects of public expenditures in the Netherlands have been carried out by the publicly funded Social and Cultural Planning office (SCP): SCP (2003) 2 and SCP (2006) 3 include the distributional effects of public education expenditures. Unfortunately, the results are very hard to compare with ours because the household income distribution used by SCP does not use equivalence scales to correct for differences between households of different sizes and composition. As a result, according to the SCP study, a relatively large part of the public education expenditures goes to the higher income strata. 2 E. Pommer and J. Jonker, Profijt van de Overheid, SCP, Den Haag, B. Kuhry and E. Pommer, Publieke productie en persoonlijk profijt, SCP, Den Haag,

6 2. Data The data we use as our baseline is the Euromod simulation for 2001, updated from the Socio- Economic Panel (SEP) data collected by Statistics Netherlands in The SEP is a representative panel survey in which information on income, wealth, housing and work was collected on a yearly basis ( ) from all household members aged 16 or over. For these persons, we also know educational achievement and the level of current education, if any. For persons younger than 16 we derive their participation in education from their age, for older persons we assume that the information on educational participation collected in 2000 is also representative for It should be noted that households with one or more members who did not fill out the income questionnaire (unit non-response) have been dropped out of the Euromod baseline sample without adjusting the original weights produced by Statistics Netherlands. In particular, this causes some underrepresentation of households with members aged As a result, the number of households with members in primary education is relatively high, whilst the number of students in tertiary education in the sample is somewhat lower than the population figures. Compared to the enrolment figures in table B, children in primary education are slightly overrepresented (1.8 vs 1.7 million pupils) whilst students in tertiary education are underrepresented. (0.3 vs. 0.4 million students). Notably, only very few tertiary education students would be counted as belonging to the institutionalized population, and, as such, would fall outside the sample frame. 5

7 3. Results 3.1. Beneficiaries by quintile Table C1 shows which parts of the income distribution will be most affected by including public education expenditures, by providing the population shares of the beneficiaries by quintile. It also shows the percentages in the age groups 4-11, 12-17, and 30+ that participate in education by quintile. Table C1 Quintile Populations shares of beneficiaries % potential beneficiaries age 4-11 age age age 30+ primary secondary tertiary all (primary) (secondary) (sec+tert) (sec+tert) Total It turns out that almost half of the children in primary education are in the second and third quintile of the income distribution, whilst the highest quintile contains relatively few children in primary education. Persons in secondary education are overrepresented in the first and second quintile, with respectively 28 and 25% of all persons in secondary education. Tertiary education is associated with the first quintile: 39% of students in tertiary education belongs to the lowest 20% of the income distribution. Obviously, it is questionable to what extent the measured income of students is a correct representation of their level of well-being. Moreover, for many of them, the low income will be a transitory phenomenon. All-in-all, almost half of the persons in education is in the first two quintiles of the income distribution, whilst only 11% is in the fifth (and highest) quintile. Hence it may be concluded that adding education expenditures to income will have the largest effects in the lowest strata of the income distribution. By construction, 100% of the persons aged 4-11 in all income quintiles are in primary education, whilst all persons aged are in secondary education. All-in-all, more than 96% of the age group is in secondary education in all quintiles. The shares in (secondary or tertiary) education in the age group drop from almost half in the lowest quintile to less than 15% in the highest. Above age 30, the percentage of persons in education is less than 4% in all quintiles. 6

8 3.2 Income effects by quintile Table C2 confirms that the effect of adding education expenditures to income has the largest effects in the lower strata of the income distribution. All in all the share of total income increases in quintiles 1 to 3 whilst it decreases in particular in quintile 5. Disposable income increases by more than 20% in the lowest quintile and by less than 3% in the highest. Obviously, the relative increase would also have been lower in the highest quintiles if the participants would have been distributed evenly across the income distribution, given that we add a fixed amount per participant. Columns G thru J show that the income increase is also lowest in the highest quintiles in absolute terms, as a result of the lower shares of persons in education. Table C2 Quintile Income share % increase in disp inc Mean transfer per capita (1000 Euro) A B C D E F G H I J Total A. Baseline distribution (equivalised disposable income per capita) B. Distribution of equivalised (disposable income + public education expenditures) per capita C. % Increase in disposable income : Primary Education D. % Increase in disposable income : Secondary Education E. % Increase in disposable income : Tertiary Education F. % Increase in disposable income : All G. Mean transfer per capita: Primary Education H. Mean transfer per capita: Secondary Education I. Mean transfer per capita: Tertiary Education J. Mean transfer per capita: All 3.3 Overall effects on inequality and poverty Table D shows the effects of adding education expenditures to income on a number of indicators of inequality and poverty. Columns A and B show the value of the indices before and after adding primary, secondary as well as tertiary education expenditures. In columns C, D and E the proportional changes of adding these expenditures separately are given, whilst column F gives the total change of adding all three components. Overall income inequality, as measured by the Gini coefficient shows a decrease of 11% when all three components are added, whilst the poverty rate decreases by almost 6%, when the poverty line is drawn at 60% of the median of equivalised disposable income. In fact, adding primary education expenditures only, the poverty rate slightly increases, and, albeit to a lesser extent, the same is true for the normalised poverty gap. The Atkinson indices and the Foster-Greer- Thorbecke(2) index are most sensitive to shifts in the lowest parts of the income distribution. 7

9 They show considerable decreases in inequality and poverty in all cases, up to more than 20% for the Atkinson indices and more than 30% for the FGT2 index.. Table D. Inequality and poverty indices Value of the index Proportional change (%) A B C D E F Gini Atkinson Atkinson Poverty rate Normalised poverty gap FGT A. Baseline distribution (equivalised disposable income per capita) B. Distribution of equivalised (disposable income + public education expenditures) per capita C, D and E: Proportional change in index when ONLY primary/secondary/tertiary education transfers are added to income F. Proportional change in index when all public education transfers are added to income 3.4. Effects differentiated by socio-economic characteristics Table E1 shows the effects of adding education expenditures to income differentiated according to four important socio-economic characteristics: household type, socio-economic category of head, education level of head and age of household member. Differentiating by household type, we would expect that income level and income distribution of elderly will be hardly affected, given that participation in education is negligible in this group. Indeed we find that income decreases relative to the national mean and that income inequality is hardly affected. A fairly similar result is found for younger singles and couples. This group contains a number of tertiary students but is largely made up by single persons and couples of working age (up to 65). As a result, the increase in disposable income and the decrease in inequality is modest. In fact, the share of the two groups in overall inequality increases from less than 49% before adding education expenditures to more than 56% when education expenditures are added. Obviously, this is mainly due to the considerable decrease in inequality in the remaining household type groups, namely couples with children aged up to 18, monoparental households and other household types. Since the latter group consists mainly of households with children aged above and below 18, participation in education is high in all these household type groups. As a result, disposable income increases considerably (up to 47% for monoparental households), whilst income inequality shows a substantial decrease. All in all, by adding education expenditures we find a notable increase in the share of within group income inequality, whilst income inequality between groups has decreased. 8

10 Table E1. Effects on inequality decomposed by socio-economic characteristics A B C D E F G H I Household type Older single/couple Younger single/couple Couple, ch Monoparental hh Other hh types % within % between Socioeconomic Group of HH Head Paid empl Self-employed Unemployed Pensioner Other % within % between Educational level of HH Head Tertiary Higher second Lower second Primary or less % within % between Age of HH Member Below Over % within % between All Distribution 1: Baseline distribution Distribution 2: Distribution of equivalised (disposable income + publication expenditures) per capita A: Population share B, C. Mean equivalent income relative to the national mean, distributions 1 and 2 D % increase in equivalent income E, F: Inequality index (2 nd Theil index) distributions 1 and 2 G: % change in inequality H, I: Contributions to aggregate inequality, distributions 1 and 2 9

11 Notably, a reverse result is found if we differentiate according to the socio-economic category of the head of the household: here, the share of within group inequality decreases, mainly as a result of a decrease in inequality among persons in households with employed heads, whilst the share of between group inequality increases. In this case, the largest effect on disposable income occurs in the small group of unemployed, whilst the largest decrease in inequality is found in others which a.o. includes students. Consistent with the previous result, and as expected, a very small effect on both income and inequality is measured for pensioners. When subdivided according to education of the head, the share of between group inequality is the largest of the four classifications in this table, and it increases by adding education expenditures to income. Remarkably, the percentage increase in disposable income is quite close to the national average of almost 11% in all four education groups, but inequality decreases fastest within the group of persons in households with a head with tertiary education, whilst it shows a less than average decrease in the group with primary education only. Next to tertiary education, higher secondary education also shows a more than average decrease in inequality, and as a result, the share of inequality within these two groups in overall inequality decreases by about 4% points in total. The subdivision according to age of the head results in the highest share of inequality within groups, and this is hardly affected by adding education expenditures. As might be expected because most of this group is affected, income rises fastest, and inequality declines most in the age group below 25. Income shows a slightly less than average increase among persons aged This group mainly consists of persons with children in education on the one hand, and persons in households without persons in education on the other. The decrease in inequality is also slightly below the national average. In the oldest age group, income and inequality again show very little effect of adding education expenditures. Shifting the focus from inequality to poverty, table E2 shows the effects of adding education expenditures to income on the poverty rate (FGT0), the normalized poverty gap (FGT1) and the poverty index FGT2, subdivided according to the same characteristics as table E1. 10

12 Table E2. Effects on poverty decomposed by socio-economic characteristics Household type A B C D E F G H I J K L M N O P Older single/couple Younger single/couple Couple, ch Monoparental hh Other hh types Socioeconomic Group of HH Head Paid empl Self-employed Unemployed Pensioner Other Educational level of HH Head Tertiary Higher second Lower second Primary or less Age of HH Member Below Over All Distribution 1: Baseline distribution Distribution 2: Distribution of equivalised (disposable income + publication expenditures) per capita A: Population share B, C: Poverty index (FGT0 poverty rate, distributions 1 and 2), D: % change in poverty rate E, F: contribution to aggregate poverty, distributions 1 and 2, G, H: Poverty index (FGT1 normalized poverty gap, distributions 1 and 2), I: % change in poverty gap J, K: % contribution to aggregate poverty, distributions 1 and 2 L, M: Poverty index (FGT2, distributions 1 and 2), N: % change in poverty O and P: % contribution to aggregate poverty (FGT2, distributions 1 and 2) 11

13 As we saw in table D, the overall poverty rate shows a marginal decrease when we redraw the poverty line after adding education expenditures. As is obvious from table E2, this small overall effect hides considerable shifts in the poor population. In the first panel, we see that among older single persons and couples, poverty almost doubles, whilst poverty is cut by almost 80% in monoparental households and by almost 50% in couples with children up to 18. Since adding education expenditures in most cases does not affect the elderly, it can be concluded that many of them have an income slightly above the old poverty line (excluding education expenditures) but slightly below the new line. In fact, even though the poverty gap increases by a factor 2.5, the resulting poverty gap of 0.04 shows that on average, the incomes of the poor are still quite close to the poverty line. Similarly, with the increase of the FGT2 index by a factor 4.6, it should be taken into account that the absolute value of this index remains small. Relative to the total poor population, the poverty gap and the FGT2 index of the elderly also remain small, as can be seen from the fact that their share in the total index are below their share in the poor population. By contrast, young singles and couples face a moderate increase in the poverty rate, a moderate rise in the poverty gap ratio and a small fall in the FGT2 index but their share in the poverty gap and the FGT2 index increases, and remains above their share in the poor population. Likewise, whilst all three poverty indices of couples with children are almost halved, their share in the poverty gap and the FGT2 index remains above their share in the poor population. For monoparental households the shares in all three poverty indices after adding education expenditures is smaller than their share in the total population, whilst the reverse held before education expenditures were added. In the category Other household types, largely consisting of households with children older than 18, the poverty gap ratio and the FGT-2 index decrease even faster than the poverty rate. Differentiated by socio-economic category of the head, the moderate decrease in the overall poverty rate is again accompanied by considerable changes in the various groups. Consistent with the previous panel, the poverty rate among pensioners increases by more than 80%. Poverty gap ratio and FGT2 index increase even faster but their share remains below their share in the poor population. For the members of households with heads in paid employment, all three poverty indices decrease by about 40%. The headcount ratio of self-employed decreases by about 30% to a figure close to the average for the total population but the other two indices decrease less fast, and the share of this subgroup in these indices remains clearly above their population share. For the (small group of) unemployed the poverty rate after adding education expenditures is about a third lower than before but still almost half of them are counted as poor. Their poverty gap ratio decrease almost equally fast, and their FGT2 index even faster, but both remain the highest of all the groups in question. Others face a remarkably moderate drop in the poverty rate, and a much sharper decline in the poverty gap ratio and the FGT2 index. Apparently a considerable number of poor in this group see their income increase by adding education expenditures, but the increase is not often sufficient to end up higher than the (increased) poverty line. Differentiated by education level, the changes are generally less dramatic than in the previous two panels, but still the changes in the group-specific poverty rates are much bigger than the change in the overall poverty rate. The largest decrease in the poverty rate is found for members of households with 12

14 heads with higher secondary education, whilst the largest increase is found for primary education or less. The latter result is not surprising since the elderly are overrepresented in this group. More of a surprise is the fact that the poverty gap ratio and the FGT2 index do not show the large jumps that they show among older single persons and couples and among pensioners. However, it can also be seen that the baseline values for these indicators are much higher for primary education than for pensioners and older persons and couples, which can probably mainly be attributed to the presence of a number of poor non-elderly in this group. The share of the total poverty gap and the FGT2 index of this group in the baseline is much higher, but after adding education expenditures it remains below the share in the poor population. By contrast, the poverty rate of members of households with heads with tertiary education decreases, but their share in the total poverty gap and the FGT2 index remains higher than their share in the poor population. Differentiated by age of household member, the results are again consistent with the figures differentiated according to household type. For persons over 65, the results are quite close to those for older singles and couples. For persons below 25, who are most likely to be in households where the addition of education expenditures results in an increase in income, the poverty rate decreases by more than 50%, and the same holds for the poverty gap ratio and the FGT2 index. In the age group 25-64, poverty remains relatively stable. On the basis of the first panel it can be safely assumed that this is the result of a strong decrease in poverty for persons in households with children and an increase in poverty for persons in households without children. 4. Sensitivity analysis 4.1 Tertiary students living at home of parents only One possible objection against the analysis of the previous chapter is that the low income of many students in tertiary education does not necessarily indicate that they are at risk of social exclusion. Indeed, for many of these students living away from home, the low income will be temporary, perfectly adequate to participate in society, and/or supplemented by income in kind from their parents. Whilst for students living at the home of their parents, their parents income is supposed to reflect their position in the distribution of well-being, it is likely that the income of the parents also affects the level of well-being of students living away from home. However, the data in use do not include the income of the parents of students living away from home. To obtain an idea to what extent the results are driven by students in tertiary education living away from home we have performed an analysis without these students. From table A4 we conclude that more than two-third of students in tertiary education do not live with their parent(s). Notably, in the Netherlands the number of students living in institutions is quite limited. Many students live in rented rooms in student houses but would qualify as independent households for the purpose of the panel survey from which the data are derived. In particular almost all the students in the lowest quintile live away from the parental home. Table A4. Students in tertiary education 13

15 Quintile living at home of parents living away from home Total Dropping students living away from the parental home from the analysis, we obtain a new quintile distribution. We see (table C1a) that students in tertiary education living at the home of their parents are not overrepresented in the first, but in the second and third quintile. Because the total number of students in tertiary education is small relative to the number of students in primary and secondary education, the overall distribution of beneficiaries across quintiles is hardly affected. The decreased share of beneficiaries in tertiary education in the first quintile is also reflected in decreased percentages of potential beneficiaries in this quintile, especially in the age group Table C1a Quintile Populations shares of beneficiaries % potential beneficiaries age 4-11 age age age 30+ primary secondary tertiary all (primary) (secondary) (sec+tert) (sec+tert) Total The income shares of the various quintiles before and after adding education expenditures are hardly affected by dropping tertiary students living away from home (table C2a). As expected, the total increase in disposable income drops fastest in the lowest quintile (from 22.9 to 20.0%). The mean transfer per capita is still highest in the lowest quintile but drops by about 200 Euros to 1300 Euros per year, about as high as in the second quintile. 14

16 Table C2a Quintile Income share % increase in disp inc Mean transfer per capita (1000 Euro) A B C D E F G H I J Total A. Baseline distribution (equivalised disposable income per capita) B. Distribution of equivalised (disposable income + public education expenditures) per capita C. % Increase in disposable income : Primary Education D. % Increase in disposable income : Secondary Education E. % Increase in disposable income : Tertiary Education F. % Increase in disposable income : All G. Mean transfer per capita: Primary Education H. Mean transfer per capita: Secondary Education I. Mean transfer per capita: Tertiary Education J. Mean transfer per capita: All Dropping students living away from the parental home, the baseline indicators of inequality show a slight decrease compared to table D. After adding education expenditures, the indicators are hardly different from table D. Most of them still move in the direction of a lowering of poverty and inequality, albeit to a somewhat lesser extent than in the case where students living away from the parental home were included. In particular, the decrease in the poverty rate, the normalized poverty gap and the FGT-2 index is considerably lower, but this should largely be ascribed to the fact that the baseline figures are different. Table Da Inequality and poverty indices Value of the index Proportional change (%) A B C D E F Gini Atkinson Atkinson Poverty rate Normalised poverty gap FGT A. Baseline distribution (equivalised disposable income per capita) B. Distribution of equivalised (disposable income + public education expenditures) per capita C, D and E: Proportional change in index when ONLY primary/secondary/tertiary education transfers are added to income F. Proportional change in index when all public education transfers are added to income 4.2 Non-compulsory education only In a second sensitivity analysis we examine the distributional effects of adding education expenditures on non-compulsory education only. We approximate this by only looking at the education expenditures 15

17 for persons aged 4 and persons age 16 or over. We do not take into account that for persons aged 16 two days of education per week may be compulsory. Table C1b Quintile Populations shares of beneficiaries % potential beneficiaries age 4 age age age 30+ primary secondary tertiary all (secondary) (sec+tert) (sec+tert) Total Comparing table C1b with table C1, we see that the share of beneficiaries of non-compulsory secondary education in the second quintile is smaller and the share in the highest quintile is higher than the respective shares of beneficiaries of all secondary education. As a result, the share of beneficiaries of all non-compulsory education is also a few percentage points higher in the highest quintile. On the other hand, given that students in tertiary education are concentrated in the lowest quintile, the share of beneficiaries of non-compulsory education is highest in the lowest quintile. Table C1b also shows that at least 88% of year olds are in secondary education in all quintiles. From table C2b we infer that limiting ourselves to non-compulsory education only, the income shares of the various quintiles after adding education expenditures hardly deviate from the original distribution. The largest decrease, 0.7% is in the highest quintile, compared to a decrease of more than 2% in the share of the highest quintile when we add all education expenditures (cf. table C2). The largest percentage increase is 9% in the lowest quintile (compared to 23% in table C2). Likewise, the mean transfer per capita decreases from about 1500 Euro to 600 Euro in the lowest decile, and the relative decline is even faster in the second to fourth quintile. Table C2b Quintile Income share % increase in disp inc Mean transfer per capita (1000 Euro) A B C D E F G H I J Total A. Baseline distribution (equivalised disposable income per capita) B. Distribution of equivalised (disposable income + public education expenditures) per capita C. % Increase in disposable income : Primary Education 16

18 D. % Increase in disposable income : Secondary Education E. % Increase in disposable income : Tertiary Education F. % Increase in disposable income : All G. Mean transfer per capita: Primary Education H. Mean transfer per capita: Secondary Education I. Mean transfer per capita: Tertiary Education J. Mean transfer per capita: All Table Db shows that by adding non-compulsory education only, the decrease in the Gini index and the Atkinson indices is much smaller than when all education expenditures are added. The difference is relatively small for the poverty rate, whilst the decrease in the normalized poverty gap is larger in comparison to the case where all education expenditures were added. The FGT-2 index again shows a smaller decrease. A more in depth analysis could reveal to what extent these differences can be explained by differential changes in the poor population on the one hand and differential changes in the incomes of the poor on the other. Table Db Inequality and poverty indices Value of the index Proportional change (%) A B C D E F Gini Atkinson Atkinson Poverty rate Normalised poverty gap FGT A. Baseline distribution (equivalised disposable income per capita) B. Distribution of equivalised (disposable income + public education expenditures) per capita C, D and E: Proportional change in index when ONLY primary/secondary/tertiary education transfers are added to income F. Proportional change in index when all public education transfers are added to income % of education expenditures A third sensitivity analysis concerns adding 50% of the total education expenditures to income instead of 100%. One interpretation could be that only half of the education expenditures is effective, another that the net price a household would have to pay for the participation in education would be lower than 100% since the education expenditures would be tax deductible. Because the beneficiaries are the same as in the original exercise we do not reproduce table C1. From comparing table C2c with table C2 we see that the income shares of the first quintile after adding education expenditures is almost as high as when the full education expenditures are added. In addition, the decrease in the highest quintile is more than half of the decrease in the original exercise. Probably this is largely caused by the fact that some households end up in a different quintile when adding 50% instead of 100% of the education expenditures. As expected, the percentage 17

19 increase in disposable income is exactly half of that in table C2, whilst the mean transfer per capita is also halved. Table C2c Quintile Income share % increase in disp inc Mean transfer per capita (1000 Euro) A B C D E F G H I J Total A. Baseline distribution (equivalised disposable income per capita) B. Distribution of equivalised (disposable income + public education expenditures) per capita C. % Increase in disposable income : Primary Education D. % Increase in disposable income : Secondary Education E. % Increase in disposable income : Tertiary Education F. % Increase in disposable income : All G. Mean transfer per capita: Primary Education H. Mean transfer per capita: Secondary Education I. Mean transfer per capita: Tertiary Education J. Mean transfer per capita: All Interestingly, as appears from comparing table Dc with table D, the effect of adding only 50% of education expenditures on the inequality and poverty indices is by no means equal to 50% of the original effect. The effect on the Gini and Atkinson indicators is about 70 to 80% of the original effect, whilst the effect on the poverty rate is larger than the original effect: instead of a decrease in the poverty rate by 0.7% we see a decrease of 1%. Likewise, the poverty gap decreases almost twice as fast as in the original exercise, whilst the FGT-2 index shows a decrease of about 90% of the original decrease. Table Dc Inequality and poverty indices Value of the index Proportional change (%) A B C D E F Gini Atkinson Atkinson Poverty rate Normalised poverty gap FGT A. Baseline distribution (equivalised disposable income per capita) B. Distribution of equivalised (disposable income + public education expenditures) per capita C, D and E: Proportional change in index when ONLY primary/secondary/tertiary education transfers are added to income F. Proportional change in index when all public education transfers are added to income 18

20 4.4 Part-time education with full-time job A final sensitivity analysis is meant to check the sensitivity of the results to the assumption on the income effects of education of persons with a full-time job. In the original exercise, persons with a fulltime job are assumed not to benefit from public education expenditures, even if they indicate that they are currently in education as well. Implicitly, it is assumed that these persons are either in private education, in a job related education or following a course as a hobby or leisure activity but not in regular secondary or tertiary education. In this sensitivity analysis we assume that persons with a fulltime job who indicate that they are in education receive 20% of the public education expenditures of those who are in full-time education. Table C1d Quintile Populations shares of beneficiaries % potential beneficiaries age 4-11 age age age 30+ primary secondary tertiary all (primary) (secondary) (sec+tert) (sec+tert) Total From table C1d we see that the distribution of beneficiaries of secondary education across quintiles is only marginally affected by this exercise, whilst the distribution of beneficiaries of tertiary education shows a clear shift towards the higher quintiles. The total effect is a marginal decrease in the shares of the lowest quintiles and a slight increase (of 0.5%) in the share of the highest quintile. As expected, the percentage of potential beneficiaries in the age group is hardly affected, the percentage of potential beneficiaries in the oldest age group increases by 0.3% overall (by 0.8% in the highest quintile), whilst the percentage of potential beneficiaries in the age group increases by 2%. Here the largest increase (2.7%) is in the second quintile. Table C2d Quintile Income share % increase in disp inc Mean transfer per capita (1000 Euro) A B C D E F G H I J Total A. Baseline distribution (equivalised disposable income per capita) B. Distribution of equivalised (disposable income + public education expenditures) per capita C. % Increase in disposable income : Primary Education D. % Increase in disposable income : Secondary Education 19

21 E. % Increase in disposable income : Tertiary Education F. % Increase in disposable income : All G. Mean transfer per capita: Primary Education H. Mean transfer per capita: Secondary Education I. Mean transfer per capita: Tertiary Education J. Mean transfer per capita: All Table C2d shows the income effects in the various quintiles. Compared to the original exercise (table C2) the differences are minimal. This also holds for the effects on the inequality and poverty indices (table Dd). Hence, we may conclude that the assumption that full-time workers do not benefit from public education expenditures is not crucial for the results of adding public education expenditures to income. Table Dd Inequality and poverty indices Value of the index Proportional change (%) A B C D E F Gini Atkinson Atkinson Poverty rate Normalised poverty gap FGT A. Baseline distribution (equivalised disposable income per capita) B. Distribution of equivalised (disposable income + public education expenditures) per capita C, D and E: Proportional change in index when ONLY primary/secondary/tertiary education transfers are added to income F. Proportional change in index when all public education transfers are added to income 5. Concluding section Adding public education expenditures to the disposable income of households in the Netherlands has non-negligible effects on the income distribution. The effects are strongest in the lowest quintile, both in relative as in absolute terms, since students in secondary and especially tertiary education are overrepresented there. In general, indicators of poverty and inequality decrease when public education expenditures are added to disposable income, but the relative size of the decrease shows considerable variations. As could be expected, the increase in disposable income is largest in the socio-demographic groups which contain many household members in education. As a result, poverty shows a decrease in these group whilst it shows a considerable increase in groups with few or no members in education, such as pensioners. Some sensitivity analyses show the varying degree to which adding education expenditures affects the distribution of income, depending on some crucial assumptions about the level of the addition and 20

22 the beneficiaries to be included. Although the overall picture is fairly stable, there are also notable differences when we drop tertiary students living away from the parental home, only take into account non-compulsory education or halve the education expenditures to be added to household income. 21