# A realistic look at educational expenditure

Hugo Elbers and Remco Kaashoek

In the period 1997–2007 real expenditure by subsidised educational institutes increased by 9 billion euro from 20 to 29 billion euro. Some 3.2 billion euro can be attributed to changes in demography and participation in education, and 5.8 billion euro to increased spending per participant. A breakdown of the changes in spending within the education sector shows that the extent to which each factor contributed differed greatly per sector.

## 1. Introduction

Every year Statistics Netherlands provides statistics about spending on education, showing which aspects of spending changed. All sorts of things play a role in these changes. In order to explain the changes in spending on education, Statistics Netherlands developed a decomposition model to study which factors caused the greatest changes. The table below shows the factors with the greatest effect on spending per education sector.

### Table 1: Spending factors per education sector

 1997-2007 Primary Secondary Tertiary Total million euro Participation effects 254 1219 1685 3158 Demography 217 785 -297 706 Absolute    participation 1 322 1569 1892 Relative participation 35 112 413 560 Expenditure per participant 2684 3052 59 5795 Material expenditure 625 1520 163 2307 Labour costs 907 1108 804 2819 Ratio employee / participant 1153 424 -908 669 Total 2938 4271 1744 8953

Spending on subsidised education not corrected for inflation increased from almost 16 billion in 1997 to 29 billion euro in 2007. Adjusted for inflation and expressed in 2007 prices, spending on subsidised education came to 20 billion euro in 1997. So in eleven years time, spending on education increased by 9 billion euro or 45 percent in real terms. During the same period GDP increased by 29 percent in real terms to 569 billion euro. So spending on education also increased as a percentage of GDP.

We explain the changes in certain variables through a decomposition analysis. This is done with several predefined contributors, of which the variable is either the product or the sum. To these factors we attribute the effects that have contributed to the total change. In this way each effect shows to what extent the variable was influenced by the change of the factor in question. As with all explanatory analyses, the answers depend on the selection of factors in the analysis. The effect of inflation has been filtered out of the decomposition analysis. The results in this paper therefore purely show the effects of the factors defined.

The term spending on education includes all spending including reserves and investments by government subsidised institutions. This includes spending on research and development, as it is mostly the universities and to a lesser extent the colleges (hogescholen) that are involved in R&D spending. Spending by private education and spending by households on materials used in education are not included, nor is spending by companies training students as apprentices.

## 2. Changes broken down into factors

Breaking down the changes in spending on education starts with a simple breakdown into two explanatory factors. The first is the number of participants in education, the second is spending per participant. These two are further broken down into several components. The figure below shows which factors were analysed and to what amount they contributed to the total change in spending on education in the period 1997–2007.

### Table 2: Breakdown of effects in billion euro, 1997–2007

Demographic developments describe the changes in population numbers per age category during the period discussed. This shows if the potential number of participants in education is increasing or decreasing. It is the most important factor that cannot be influenced by education policy.
The demographic effect shows which influence on education spending can be attributed to increase or decrease of the population. Suppose that 25 percent of a certain age category participates in education. Ten years later the number of people in that age category is up by 10 thousand, while the percentage in education is still 25 percent. This means an increase of 2.5 thousand students is attributed to the demographic effect, and price tagged.
This effect should not be mistaken for participation in education. There may be very different causes for changes in the participation rate in education, for instant extending compulsory education by one or two years.
Participation in education shows how the potential number of participants in education is translated into the actual number of participants. In the analysis this factor is broken down into:
1. The absolute participation rate. This is the ratio between the total number of people per age category and the number of participants in education in that age category. The factor shows how much spending on education changes because the demand for education in general per age category increases or decreases percentage-wise.
2. The relative participation rate. This is the ratio between the total number of participants in education per age category in the Netherlands and the number of participants per education type in that age category. The factor shows the financial effect of a relatively higher or lower participation rate in the different types of education, such as special or higher education.

Spending per participant is broken down into two types of spending: material spending per participant and personnel spending per participant.
Material spending is all spending by the institutions not spent on personnel. Material spending refers to investments and costs incurred by the institution. Investments consist of spending on buildings, land, inventory, equipment and educational tools. Institutional costs refer to spending on rent, energy, cleaning, taxes, administration, management, presentation costs etc.
The effect of spending on personnel per participant in the decomposition model is dependent on the number of employees in fulltime equivalents (FTE) required per participant for the education, and on the costs per employee (FTE). These costs in turn are broken down by the effects brought about by the wage developments and changes in the other employer costs. The wages also include the effect of the changes in the collective (CLA) wages and the other wage effects. The trend in the collective wages is known and the other wage effects are derived from this.

### Graph 1: Breakdown of spending on education

Further on we will discuss the various effects per factor in more detail, but a preliminary remarkable find is that all factors studied in the period 1997–2007 exerted real upward pressure on spending. In other words, the increase in the spending on education can be explained by the fact that each underlying development helped increase the total costs.
Graph 1 shows that the effect of most factors was not constant throughout the entire period but fluctuates, sometimes wildly. In the graph each year is related to the start, 1997. This means that each annual change is added to that of the previous year, which results in a cumulative change.

## 3. Demography and participation in education

In the period 1997–2007 almost 3.2 billion euro of the real increase in total spending of the education institutions could be attributed to the demographic effect and participation in education. This means that the changing number of participants in education explains over a third of the entire increase in spending. The number of participants in subsidised education increased by over 300 thousand during this period.

3.1 Demography
In the period 1997–2007 the number of potential participants in education increased. This involved the age group 4-65, which consisted of about 12.8 million people in 1997 and about 13.4 million people in 2007. This is an increase of 4.1 percent. The changes are not the same for each age group though. For instance, the number of 4 to 13 year-olds increased by 4.4 percent whereas the number of 13-17 year-olds increased by 10.5 percent. On the other hand, age group 21-25 decreased by 8.4 percent and 25-30 by 23.2 percent.

### Graph 2: Demographic changes and the demographic effect, 1997–2007

The effect of the demographic component on the real increase of spending by the institutions was over 700 million euro according to the decomposition analysis. The demographic effect for the group 4-13 year-olds was over 250 million euro. The largest effect was for the 13- 21 year-olds: over 890 million euro. In the 21-30 age group the effect was negative: –415 million euro.
There were of course also demographic changes among the over 30s, but due to their limited participation in education the effects of this on spending was marginal.

The increase in the population of the 4-13 year-olds and the effect of this on education spending is manifested almost one to one in the spending of primary education. The increase of the potential number of participants in education aged 13-21 is of course mainly seen in secondary education, where spending increased by 785 million euro for this reason. The reduction of the 21-30 year-old population mainly had an effect on spending in tertiary education, where spending fell by over 295 million euro due to the demographic trend. Tertiary education may expect a spending increase soon due to the ‘demographic mini-boom’ that went through secondary education recently.

3.2 Participation in education
The size of the population relevant for education increased, which raised spending on education, but relatively speaking the participation in education increased far more. In the period 1997–2007 the effect of the participation in education on the real increase in total spending of subsidised education was close to 2.5 billion euro. This amount consists of two components: an effect related to absolute participation and an effect related to relative participation.

3.2.1 Absolute participation
In the period 1997–2007 the effect of absolute participation on spending amounted to nearly 1.9 billion euro. The absolute participation effect is about how much spending on education changes because of the increasing or decreasing demand for education in general per age category, in percentages. In primary education the absolute participation effect is almost nil, because there is compulsory education for all children of primary school ages 5-12. So changes in spending in primary education are brought about not by the participation effect but by the demographic effect.

This is slightly different in secondary education. Since 2003 the absolute participation effect has been positive and between 1997 and 2007 it has contributed 300 million euro to the increase in spending on education. The amount can be attributed almost fully to the 17-21 year-olds in vocational and adult education. In adult education the increase in the number of students is concentrated in this age bracket.

However, by far the largest effect is in tertiary education: nearly 1.6 billion euro. Divided over the two types of education tertiary education the effect on colleges is 480 million euro and on university 1.1 billion euro. Instead of taking a job when they finish secondary education, more and more students continue in tertiary education. The favourable position of the highly educated on the job market must have played a major role in their decision.

### Graph 3: Demographic and participation effects, 1997–2007

3.2.2 Relative participation
In the period 1997–2007 the effect of the relative participation on spending was nearly 0.6 billion euro. The relative participation effect refers to the question what the financial effect is of a relatively higher or lower participation rate in the various types of education. In primary education there is a modest effect of 35 million euro caused by the fact that relatively more pupils, 11 thousand, went to the expertisecentra rather than special primary schools, thereby causing a reduction in special primary school participation by the same number. Education in the expertisecentra has a higher price tag than education in special primary schools.

In secondary education the total effect is over 100 million euro. The main contributor is leerwegondersteunend and praktijkonderwijs, in which relatively many students participate; the effect of this for the entire period is nearly 470 million euro. However, this is compensated by ‘regular’ secondary education, in which the relative participation effect is close to –400 million euro.  Adult education hardly changed in relation to the other types of education, which results in a modest relative participation effect of 20 million euro for this type of education.

As is the case with absolute participation most the relative participation effect, over 400 million euro, is from tertiary education. In the previous paragraph we indicated that the number of participants in tertiary education due to demographic trends and, all things being equal, so did spending. However, this trend was more than compensated by the great increase in the participation in education. Not only did the absolute participation in education increase in the age group relevant for tertiary education, but the share of higher education also increased the relation relative with the other education sectors.
All in all tertiary education in 2007 had nearly 140 thousand more students than in 1997, of whom 93 thousand in college and 46 thousand at university. This resulted in a total extra spending pressure of 1.7 billion euro.

## 4 Spending  per participant

The effect of total spending per participant on the total increase in spending on education is 5.8 billion euro. Broken down by type of education sector the increase in primary education is 2.7 billion euro, 3.1 billion in secondary education and a mere 60 million euro in tertiary level education.

4.1 Material spending  per participant
Some 2.3 billion euro of the total spending increase can be attributed to the effect of changes in the material spending per participant. About 40 percent of material spending consisted of investments, including spending on buildings, equipment and educational tools, and about 60 percent went to current spending such as rent, energy, cleaning etc.

Compared to the total spending increase there are major differences per education sector on what was spent extra per participant on investments and current spending. In primary education the effect of material spending was 625 million euro. Here spending per pupil on rent, energy etc increased by over 50 percent. In secondary education the effect was considerably larger: 1.5 billion euro. Not only did spending per participant for buildings, equipment and educational tools doubled in the period 1997–2007 but so did spending on rent, energy etc. In tertiary education there was a modest effect of 165 million euro. Here spending per student for rent, energy etc increased slightly. Spending in the investments sphere fell slightly.

4.2 Spending  on personnel per participant
In the analysis of spending on personnel per participant we distinguish two questions: 1. how many employees per participant are needed to provide the education, and 2. what are the average cost per employee? The number of employees is expressed in full-time equivalents (FTE). The group consists of teaching and other personnel, such as management and support. This brings us to the question how costs per employee developed and how it contributed to the changes in spending on education.

4.2.1 Labour cost effect
Of the total spending increase in subsidised education in the period 1997–2007 over 2.8 billion euro can be attributed to the average labour cost effect. Labour costs include social premiums payable by the employer as well as the salaries paid to employees.
The social premiums contributed most with 1.7 billion euro. They include the premiums payable by the employer for unemployment insurance (WW) and (WIA), as well as the contributions made by employers to the pensions of their employees. The pension premiums increased in particular.
The increase in the real salaries accounted for 1.1 billion euro. A distinction can be made in the salary component between the effect of the collective wage developments (CAO) and the other wage effects. The increased collective wages in education contributed 300 million euro to the total change in spending on education according to the decomposition results, and the other wage effects 800 million euro.

### Graph 4: Labour cost effect, 1997–2007

The concept ‘other wage effects’ actually just includes the development of the average wage costs that cannot be derived from the collective wage developments. In the first place these are incidental wage developments such as special payments and individual bonuses. A second phenomenon is the change in the wage structure driven by the personnel structure. For instance there are the consequences of the division between full-time and part-time work, the aging workforce, the ratio between managers and teaching staff and other changes that influence the division of employees over the different salary scales.
In the current setup of the decomposition model these elements are not detailed as separate effects. However it is important to keep in mind that they play a key role in the average wage developments of the people employed in education. The increased spending is certainly not just an improvement of the income position of the individual employee.

There are differences between the education sectors in terms of the effect of the average wages and the previously mentioned social premiums. This has to do with the differentiated collective agreements for the different types of education in the CAO. There is also a difference in the personnel structure in terms of age and positions. On the whole the effect of these differences is limited, and all sectors follow the trends described above.

4.2.2 Employee/participant ratio
Of the total spending increase in subsidised education in the period 1997–2007 nearly 700 million euro can be attributed to the fact that in most types of education employment in FTEs rose faster, in relative terms, than the number of participants. The ratio used in this decomposition analysis is not the same as the better known teacher/student ratio which is usually used to indicate the average size of classes. The former concept is wider because it also includes non-teaching staff. The concept ‘class size’ is not very helpful in sectors outside primary education and part of secondary education because the concept of ‘class’ hardly applies.
In primary education the effect contributed 1.2 billion euro to the increase in total spending of this sector. The greatest effect was observed in the period 2000–2003, when the size of classes was reduced in primary schools.
In secondary education the effect of the employee/participant ratio contributed over 420 million euro to the increase in this education sector. One explanation is the growth of the leerwegondersteunend education. More money was spent in secondary vocational education on reducing the number of participants dropping out without a diploma.
In tertiary education the ratio in the period 1997–2007 fell almost continuously and the calculated effect of this is more than –900 million euro. This shows that the increase in spending in tertiary education by the increase in the number of participants was compensated largely by the fact that the growth in employment did not keep up.

## 5 Summary per education sector

In the previous paragraphs the changes in the spending on education in the period 1997–2007 were explained through developments in the number of participants in education and spending per participant, as well as the derived factors and their effects. This paragraph summarises the different trends in primary, secondary and tertiary education.
In Table 3 each year is related to 1997, which means that each annual change is added to that of the year before. So the right-hand column shows the total change for all components in the period 1997–2007.

### Table 3: Developments in spending on education 1997–2007 by sector Developments in spending on education by sector, 1997�2007   1997-1998 1997-1999 1997-2000 1997-2001 1997-2002 1997-2003 1997-2004 1997-2005 1997-2006 1997-2007 billion euro Total change Primary 655 915 1,360 1,977 2,564 3,166 3,085 3,221 3,041 2,938   o.w. participation 71 126 163 203 245 260 261 265 257 254   o.w. expenditure per participant 584 789 1,197 1,774 2,320 2,906 2,825 2,956 2,784 2,684 Total change Secondary 641 1,241 1,526 2,054 2,490 2,803 2,978 3,243 3,658 4,271   o.w. participation 28 74 173 288 408 553 738 898 1,078 1,219   o.w. expenditure per participant 612 1,168 1,353 1,766 2,083 2,250 2,240 2,345 2,580 3,052 Total change Tertiary 265 345 421 672 803 957 1,154 1,314 1,341 1,744   o.w. participation -1 89 243 425 638 875 1,191 1,440 1,578 1,685   o.w. expenditure per participant 265 256 178 247 166 83 -36 -127 -237 59 Total education sectors 1,561 2,501 3,307 4,703 5,858 6,926 7,218 7,778 8,041 8,954

As indicated above, each of the factors we studied contributed in the period 1997–2007 to the increase in spending on education as a whole. Graph 5 shows how spending per participant by education sector changed between 1997 and 2007.

### Graph 5: Spending per participant by sector, 1997–2007

5.1 Primary education: smaller classes
Primary education contributed 2.9 billion euro to the increase in spending on education. This can be attributed almost entirely to increased spending per participant. Substantial investments were made in personnel in primary education during 1997–2007. The relative increase in the number of teachers, assistants etc, that resulted from the reduction of the class size alone was responsible for 1.2 billion euro.
Table 3 clearly shows that the increase mainly took place in the period 2000–2003. The same goes for material spending. Part of this is undoubtedly due to the smaller classes or groups, but there were also major investments in teaching tools such as computers etc. Almost the entire increase in material spending took place in the period until 2004.
There was a small increase in the number of pupils due to the demographic effect. The participation effect is hardly relevant because all children have to be in primary education from age four.

5.2 Secondary education: buildings and inventory
Secondary education contributed 4.3 billion euro to the increase in spending on education. Here too most of the increase, about three quarters, can be attributed to increased spending per participant. In contrast to primary education, increased spending mainly went to the material spending per participant and to a lesser extent to personnel. The major investments were in buildings and inventory, such as furniture, teaching materials and computers.
The other quarter of the spending increase is divided as follows: two thirds for the demographic effect and one third for the increased participation in education. The demographic effect manifests itself primarily in general secondary education. Increased participation in education mainly plays in vocational secondary education.

5.3 Tertiary education : growing number of  participants
Tertiary education contributed over 1.7 billion euro to the increase in spending on education. The driving force behind the spending increase is the rising number of students because of the higher participation rate in education. If this had been the only relevant factor in 1997–2007 spending in tertiary education would have increased by 2 billion euro. But spending also increased due to growing real labour cost, just like in the other sectors.
However several factors compensated for the growth in spending. Material spending per participant saw very modest growth, in contrast to the other education sectors and demographic trends had a negative effect. The main compensation was the fact that employment did not keep up with increasing student numbers.

## Technical explanation:

The formulas used to relate the defined factors and total education spending are detailed below. Each factor is part of one formula in the end that leads to total education spending when it is written out. This allows the following decomposition.

U = Σi Li*Ki  Where:
U= Total Education spending
Li = Number of participants per education type
Ki  = Spending per participant in each type of education

Li = Σj Dj*Pi,j  Where:
Li = Number of  participants per type of education
Dj  = Number of  people in the population per age category  (demographic effect)
Pi,j  = Percentage of people participating in education per type of education by age category

Pi,j = (Lj/Dj)*(Li,j/Lj)  Where:
Pi,j  = Percentage participants in education per type of education per age category
(Lj/Dj) = Percentage of participants in education per age category (absolute participation effect)
(Li,j/Lj)  = Number of  participants per type of education per age category as a percentage van number of  participants in education per age category (relative participation effect)

Ki = (Oi/Li)*Xi  +Mi Where:
Ki  = Spending  per participant in each type of education
Oi = Labour volume (in FTEs) of education employees by type of education
Li = Number of  participants per type of education
(Oi/Li) = Employee/participant ratio per type of education (ratio effect)
Xi  = Average cost per employee (FTE) by type of education
Mi  = Material costs and investments per participant by type of education (material spending effect)

Xi = Yi + Zi  Where:
Xi  = Average cost per employee (FTE) by type of education
Yi  = Average wage sum per employee (FTE) by type of education (salary cost effect)
Zi = Average social premiums paid per employee (FTE) by type of education (social premiums effect)

In this method total education spending is the result of the factors mentioned. The yearly changes in education spending are attributed through a breakdown to the yearly changes in each factor. There is no rest component, because the interaction effect has been attributed proportionally to the factors. This was done in a method that, according to the literature is not unusual. The principle of the attribution for four factors, for instance, can be described as follows: If it is true that  a=b.c.d.e, then the influence of the yearly changes in b (bt1 - bt0) on the yearly changes in a (at1 - at0) can be expressed as : Effect of Δb on Δa = Δb.ct0.dt0.et0 + ½.Δb.(Δc.dt0.et0 + ct0.Δd.et0 + ct0.dt0.Δe) + ⅓.Δb.(Δc.Δd.et0 + Δc.dt0.Δe + ct0.Δd.Δe) + ¼.Δb.Δc.Δd.Δe.
The effect of the wage sum changes (Yi) is related directly to the known relative changes in the CLA wages, so that an effect of the CLA wage changes and a ‘rest’ (‘wage structure’) effect can be derived.

In the analysis we used 16 types of education and 31 age categories. The types of primary education are: elementary education, special elementary education and expertisecentra at the elementary level.
In secondary education : regular secondary education (vmbo, havo, vwo),  leerwegondersteunend- en praktijk education, expertisecentra at the secondary level, adult education at vmbo and havo level, adult education at vwo level, beroepsbegeleidende leerweg in mbo, full-time beroepsopleidende leerweg in mbo and part-time beroepsopleidende leerweg in mbo.
In tertiary education: full-time higher vocational college, part-time higher  vocational college, full-time university education, part-time university education and post graduation education . The age categories include ages 4-29 by year as well as age categories 30-34, 35-39, 40-44, 45-49 and 50-65.

## Sources:

StatLine: Education statistics. From this we took the number of participants and spending  by subsidised education.
StatLine: Arbeidsmarkt en sociale zekerheid, Arbeidsmarkt, Cao wage index.
StatLine: Population by age and marital status.
StatLine: Macro-economic National Accounts yearly figures, Labour Accounts on wage figures. Labour Accounts also contain employee numbers. Total Labour Accounts spending  is not by definition a one on one match with spending  on personnel of the education statistics. When applicable data in the breakdown of the education statistics and the Labour Accounts wage data are leading and the trend in the number of employees derived.

## Definitions:

Decomposition: attributing changes of a variable in time to changes in a number of  defined underlying factors.

Deflate: recalculating amounts with the help of the GDP index figures from a time series so that it becomes possible to make a real comparison between the value changes.

Delta (Δ): the difference between two bits of data.

Expertisecentra: institutions providing special  education under the law on expertisecentra.

Institutions of subsidised education: education institutions paid by the government.

Education sector: primary, secondary and tertiary education.

Type of education: the different types of education in a sector, such as elementary education, general secondary education (havo) or colleges (hbo).

Educational expenditure : spending  by subsidised educational institutes on their education task and on R&D.

Primary education:  elementary education , special  education and special  education under the law on expertisecentra provided to children aged 4-12r

R&D: research and development, activities at universities and colleges.

Real increase/decrease: a change in which the effect of inflation has been eliminated.

Secondary education: general, vocational and adult education.

Tertiary education: higher vocational university and post graduate education

Spending  per participant: total spending  divided by the number of  participants to which the spending applies.

Employee/participant ratio: the ratio expressing how many participants there are to one employee (FTE). This is different from the student teacher ratio which expresses the average size of classes.