See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/254423474

The demand for private schooling in England: The impact of price and quality

Article · January 2010

CITATIONS READS

12 2,094

3 authors:

Richard Blundell Lorraine Dearden

University College London University College London
532 PUBLICATIONS 71,876 CITATIONS 134 PUBLICATIONS 6,471 CITATIONS

SEE PROFILE SEE PROFILE

Luke Sibieta
Institute for Fiscal Studies
37 PUBLICATIONS 648 CITATIONS

SEE PROFILE

All content following this page was uploaded by Richard Blundell on 24 June 2014.

The user has requested enhancement of the downloaded file.

 The Demand for Private Schooling in
England: the impact of price and
quality1
Richard Blundell
University College London and Institute for Fiscal Studies

Lorraine Dearden
Institute of Education, University of London and Institute for Fiscal Studies

Luke Sibieta
Institute for Fiscal Studies

Abstract
In this paper we use English school level data from 1993 to 2008 aggregated up to small
neighbourhood areas to look at the determinants of the demand for private education in England
from the ages of 7 until 15 (the last year of compulsory schooling). We focus on the relative
importance of price and quality of schooling. However, there are likely to be unobservable factors
that are correlated with private school prices and/or the quality of state schools that also impact
on the demand for private schooling which could bias our estimates. Our long regional and local
authority panel data allows us to employ a number of strategies to deal with this potential
endogeneity. Because of the likely presence of incidental trends in our unobservables, we employ
a double difference system GMM approach to remove both fixed effects and incidental trends. We
find that the demand for private schooling is inversely related to private school fees as well as the
quality of state schooling in the local area at the time families were making key schooling choice
decisions at the ages of 7, 11 and 13. We estimate that a one standard deviation increase in the
private school day fee when parents/students are making these key decisions reduces the
proportion attending private schools by around 0.33 percentage points which equates to an
elasticity of around -0.26. This estimate is only significant for choices at age 7 (but the point
estimates are very similar at the ages of 11 and 13). At age 11 and age 13, an increase in the
quality of local state secondary reduces the probability of attending private schools. At age 11, a
one standard deviation increase in state school quality reduces participation in private schools by
0.31 percentage points which equates to an elasticity of -0.21. The effect at age 13 is slightly
smaller, but still significant. Demand for private schooling at the ages of 8, 9, 10 and 12, 14 and 15
are almost entirely determined by private school demand in the previous year for the same
cohort, and price and quality do not impact significantly on this decision other than through their
initial influence on the key participation decisions at the ages of 7, 11 and 13.

1
The authors would like to thank the Australian Research Council (ARC) and Economic and Economic and
Social Research Council (ESRC) for funding this research under their ARC-ESRC collaborative bid framework
(Ref: RES-000-22-2524). Blundell would also like to thank ESRC-funded Centre for the Microeconomic Analysis
of Public Policy at IFS (grant number M535255111) for support. The authors would also like to thank Susan
Dynarski and conference participants who attended the IFS event on “Determinants of Private Schooling” on
th th
June 17 2010 and the EALE/SOLE conference presentation at University College London on June 19 2010
for constructive comments and suggestions.

1
1. Introduction

This paper looks at the impact of private school fees and school quality on the demand for private

secondary schooling in the UK. This topic has not been examined before in the UK and almost all

research on this issue has used US data.

There are some papers looking at the demand for private schooling in the UK. A recent paper by

Blow, Blundell and Machin (2010) using UK Family Expenditure Data shows that household demand

for private schooling is positively related to income, but also positively to regional inequality. Recent

work by Dearden and Sibieta (2010) using the British Household Panel Survey shows that that the

probability of attending a private schooling is related to household income and parental education

but is also more likely if one of the child’s parents went to a private school when they were young.

Neither of these papers, however, explicitly consider the impact of private school fees or school

quality on the demand for education. A number of papers have looked at how parental preferences

for state schools are related to school quality (recent examples include Hansen and Machin (2010)

and Burgess and Vignoles (2009)) but they do not consider choices between the state and private

sector, only choices within the state sector. In a series of papers, Gibbons and Machin (2003, 2006,

2008) examine the relationship between local state school quality (at primary level) and local house

prices, and attempt to recover the implicit price of attending high-performing state schools.

However, they also are unable to consider the private sector in such analysis.

The key problem with trying to estimate the causal impact of price and quality is that there are likely

to be unobservable factors that are correlated with private school prices and/or the quality of

private and state schooling as well as the demand for private schooling which would bias any

estimates. Moreover, it is highly likely that these unobservable factors are changing over time.

Coming up with a credible way of controlling for this potential endogeneity is therefore a key issue.

One paper which does this in a convincing way is Dynarski et. al. (2009). They only have cross-

sectional data but use variation in private school tuition that arises through sibling discounts and use

2
this within-neighbourhood variation in tuition price to identify the price elasticity of demand (as this

within neighbourhood variation means that unobserved determinants of demand can be controlled

with a neighbourhood fixed effect). They estimate the price elasticity of demand for private

schooling in the US and find that a standard deviation decrease in tuition prices increases the

probability of a family sending their child to private schools by 0.5 of a percentage point which

translates into an elasticity of -0.19.

In this paper we use rich English schools data which records the number of children in every state

and private school by age and gender from 1996 to 2008 from the age of 7 until 15. We use this data

to calculate the proportion of children in private schooling at a fine neighbourhood level over a 13

year period from 1996 to 2008. In the paper, we consider private school participation from the age

of 7 through to 15. At age 7, most children have the option of attending a private primary school

whereas before that time private schooling is not always an option. Age 15 is the last year of

compulsory schooling in the UK. As might be expected, we find that private school attendance is a

dynamic process, and understanding what drives the initial decision to enter private schooling is a

crucial part of understanding what determines the demand for private schooling at subsequent ages.

It turns out that private school decisions at age 7, 11 and 13 are the key points at which fees and

state school quality can impact on demand for private schooling. For other years, attendance in the

previous year is the key determinant of the demand for private schooling.

Our neighbourhood measure is at the local authority. There are 150 local authorities in England,

each of which is responsible for state school in their area. Unlike in the US, schools within each local

authority are not funded through local taxation. Instead, they are largely funded via grants from

central government, raised through centrally collected taxes. Around 90 per cent of children attend

state secondary schools in their own local authority and local authorities decide on admissions

policies for most state schools in their authority. This allows us to identify how variation in prices

and school quality over time impact on the demand for private schooling at each age.

3
In the UK, the overall demand for private schooling has remained very flat at around just over 7 per

cent for the period under examination (and as Blundell et. al. (2010) show this is also true for the

last 30 years). However over this period there has been changes in the regional patterns of private

school attendance with some areas seeing large increases, other large decreases and yet others not

much change at all.

This means we need to come up with a methodological approach which can account for these

differing trends in the observed and unobserved changes in regional demand for private education.

In Dynarski et. al. (2009) the unobserved determinants of demand can be controlled with a local

authority fixed effect. However they are dealing with just one cross-section. In our reasonably long

panel this may not be a credible strategy. In this paper we develop a system GMM panel data

estimation method that allows for incidental trends in the unobserved determinants of demand for

private schooling and this turns out to be important for the question we are looking at.

We find that the demand for private schooling at age of 7 (the first year when the majority of

students enter private schools for primary education) is inversely related to private school fees. We

estimate that a one standard deviation increase in fees reduces the demand for private schooling at

age 7 by 0.33 percentage points which equates to an elasticity of -0.26. Between the ages of 8 and

10, private school attendance in the previous year is the key determinant of private schooling and

controlling for incidental trends is crucial in deriving consistent estimates.

At age 11, the demand for private schooling is related to private school attendance in the previous

year, as well as being inversely related to state school quality and private school fees (though the

fee effect is not significant at conventional levels). An increase in the quality of local state secondary

schools of one standard deviation when the child is 10 reduces the probability of 11 year olds

attending private schools by 0.31 percentage points, which equates to an elasticity of -0.21. The

price elasticity at this age is similar to that found at age 7 but is not statistically significant. Demand

for private schooling at the age of 13 is also inversely related to price and state school quality,

4
 however the price effect is not significant and the impact of quality is significant but lower than at

age 11. Demand for private schooling at the ages of 12, 14 and 15 is almost entirely determined by

demand in the previous year for the same cohort and price and quality to not impact on this decision

other than through their initial influence on the participation decision at ages 7 and/or 11 and/or

13.

Our paper proceeds as follows. In section 2 we describe the data we use in the paper and our

empirical approach. In section 3 we discuss the results of our modelling. In section 4 we conclude.

2. Data and Estimation Strategy

2.1 Introduction

In this paper we make use of detailed English schools administrative data from 1993 to 2008

to look at the determinants of the demand for private education.

We use two sources of school administrative data. The first is the school census data (LEASIS

data) which from 1996 records the number of pupils in every school in England for every age

group in the private and state sector. It also records information on the number of children

receiving and eligible for free school meals in all of the state schools (a measure of socio-

economic status) as well as information on the proportion of children in the school with

special educational needs and items like authorised and unauthorised absences. It also has

information on pupil teacher ratios at the school (for schools in the state and independent

sector).

From 1993 to 2008 we know the results of GCSE exams taken by 15/16 year olds in England

for every secondary school in the country (the last exam before children can leave school).

5
In this paper we use as our quality measure the proportion of children achieving the

expected level at age 15/16 (which is 5 or more GCSEs with a mark of A*,A, B or C).

In this paper we concentrate on looking at the demand for private schooling between the

ages of 7 and 15 in England. We start at age 7 (Year 3 of primary school) as a significant

proportion of private schooling begins at this age. In most local authorities secondary school

starts in year 7 when all children are aged 11 at the beginning of the school year. However,

in some local authorities they have middle schools, and children do not start at secondary

school until Year 9 (age 13 at the beginning of the school year). Also, a significant proportion

of boys’ private secondary schools, have large intakes at Year 9 or indeed only start at Year 9

(Eton perhaps being the most famous example). However, students also attend private

schools for primary education, so demand for secondary private schooling will also depend

on earlier primary school choices. Hence our modelling takes a dynamic approach and looks

at the demand at each age, beginning at age 7. It turns out that these dynamics are very

important in explaining private school demand.

The second data source we use is the annual census of the Independent Schools Council.

This goes back to 1983 and contains the average private school fee for boarding and day

schools across broad regions in the UK. We focus on the fee level of day schools within

these broad regions, as boarding schools are much more likely to be attended by pupils

from all over the country rather than just those in the region or local authority. We have fee

information dating back to the mid 1980s.2

2.2 Estimation Strategy

2
At this stage we only have historical private school information at 7 broad regional levels. For 2008 we have
individual school data and we are in the process of obtaining this information back to the early 1990s.

6
We begin by modelling the demand for private schooling at age 73 (measured as the

proportion of children in private schools) as a function of the known fees and school quality

when the child was aged 5 or 6 (when parents were making the decision). This is generally

the first time parents will consider private schooling (although some children will attend

private preparatory schools before this age).

The next major decision point is when the child is aged 10 (generally in October when the

child has just commenced in Year 6) when parents must decide whether they want their

child to be educated up to age 15 in a state or private secondary school. However, a

significant proportion of children choosing private education at 11 will already be in the

private sector so we need to account for this within a dynamic framework. In some areas

and for some private schools, secondary schooling starts at age 13, so this is an important

age for some parents if they are going to switch sectors.

Our modelling set-up is very general and models the demand for private schooling from all

ages from 7 to 15. However, as mentioned earlier, we need to account for the fact that

school quality and fees are potentially endogenous or pre-determined. It is also highly likely

that unobserved determinants of demand have changed over this long time period and if we

do not take this into account our estimates could be biased. So we begin with a general

model of the form:

(1)

where is a vector of strictly exogenous variables (such as time dummies and other

exogenous determinants of demand), is a vector of pre-determined covariates as well as

3
At age 7 we mean children who are 7 at the beginning of the beginning of the school year and who will turn 8
sometime during the year. These children are in Year 3 of primary school.

7
potentially endogenous covariates such as school quality measures in the state and private

sector and private school fees (with some lag). This is the model developed by Blundell and

Bond (1998) and others, the exception is that we allow the unobserved group level effect

to vary incidentally with time t. First differencing removes the unobserved fixed effect

however it does not remove the incidental trend (if it is present). If, however, we double

difference the incidental trend is removed.

Double differencing equation (1) gives:

( 2)

Clearly in this model and are correlated with , violating one of the moment
conditions and meaning OLS estimation of (2) is severely biased (see Han and Phillips (2010) ). But,
and are not correlated with so can be used to instrument and
. Clearly in this the error term is likely to be subject to second order autocorrelation but if
our model is correctly specified it should not have third order autocorrelation. We test for this in our
estimation procedure.

Of course, if there is no incidental trend, we need only first difference and use standard system
GMM estimation. However if there is an incidental trend, then the first difference moment
conditions are violated and we would expect our traditional system GMM estimates to be upward
biased if the unobserved incidental trend is positively related to our variable of interest (which is
what we would expect for lagged participation and fees as it is highly likely that unobserved trend
increases in private school demand are positively correlated with increases lagged participation, and
increases in private school fees) and downward bias coefficients if negatively correlated to this
incidental trend (which is what we might expect with state school quality as it is highly likely that
unobserved trend increases in private school demand are inversely related to changes in state
school quality).

We estimate (2) using a modified version of the Blundell and Bond (1998) Generalised Method of
Moments (GMM) system estimator where the levels equation is now a first difference equation and
the difference equation is a double difference equation, and where we now add the first difference

8
equation to the system together with our double difference equation. This is related to the double
difference estimators suggested by Han and Phillips (2010) where there is an incidental trend.

2.2 Data Description

For our identification strategy to work we need variation within region in private school fees and

school quality over time. This is because we need to control for differences across regions in factors

such as average income, parental education and taste for private education using regional fixed

effects with an incidental trend. We have just under 150 local authorities in our data but we

aggregate it up to 9 broader regions to demonstrate the variability we have (we demonstrate it

empirically in the next section).

Broadly speaking, private school attendance in England has remained largely unchanged at just over

7 per cent for the last 30 years. Over the period we consider, private school attendance at secondary

schools reduced slightly from 7.3% in 1996 to 6.9% in 2001 and 2002, before increasing to 7.4% in

2008. However the patterns differ by age group as can be seen in the Table 2.1 below (where we

show the proportion attending at age 7, age 11 (first year of secondary school), age 13 and age 15

(last year of compulsory school). Over the same time period real private schools fees have increased

fairly rapidly with average real growth of 3.8% per year. However this increase has been far from

smooth. In 2004 real day fees went up by 6.8 per cent whereas in 1998 the comparable figure was

1.8 per cent. The proportion of pupils achieving 5 GCSEs at A*-C (the expected level at age 15) has

risen steadily in the state sector and much faster than in the independent sector, but from a much

lower base. We do not use independent school quality in our analysis as a number of independent

schools have started taking exams at 16 which are not included in the standard measures of school

quality and hence from 2006 this data is not reliable. However over this period the proportion

getting GCSEs at A*-C rose in independent schools rose from around 81 per cent to 91 per cent.

9
TABLE 2.1: Summary statistics (mean and [standard deviations]) for key variables

Year Proportion Proportion Proportion Day fee Propn Propn Propn

Private Private Private Private Private A*-C Eligible
at age 7 at age 11 at age 13 at age 15 (£ pa) State FSM
1996 0.048 0.069 0.073 0.078 6399 0.427 0.184
[ 0.042] [ 0.053] [ 0.054] [ 0.059] [ 662] [ 0.078] [ 0.107]
1997 0.049 0.07 0.071 0.077 6565 0.435 0.182
[ 0.043] [ 0.052] [ 0.053] [ 0.058] [ 705] [ 0.077] [ 0.105]
1998 0.05 0.068 0.07 0.076 6680 0.449 0.176
[ 0.044] [ 0.052] [ 0.054] [ 0.058] [ 770] [ 0.082] [ 0.105]
1999 0.051 0.066 0.07 0.073 6978 0.468 0.169
[ 0.045] [ 0.051] [ 0.054] [ 0.056] [ 714] [ 0.080] [ 0.100]
2000 0.051 0.066 0.069 0.073 7227 0.479 0.166
[ 0.045] [ 0.052] [ 0.053] [ 0.057] [ 812] [ 0.080] [ 0.102]
2001 0.052 0.067 0.068 0.072 7589 0.49 0.16
[ 0.046] [ 0.054] [ 0.053] [ 0.056] [ 875] [ 0.077] [ 0.099]
2002 0.054 0.069 0.069 0.072 8025 0.506 0.151
[ 0.047] [ 0.054] [ 0.053] [ 0.056] [ 924] [ 0.075] [ 0.094]
2003 0.054 0.069 0.07 0.071 8378 0.522 0.148
[ 0.046] [ 0.054] [ 0.054] [ 0.055] [ 1016] [ 0.069] [ 0.093]
2004 0.054 0.07 0.072 0.072 8943 0.531 0.147
[ 0.047] [ 0.054] [ 0.055] [ 0.056] [ 1103] [ 0.066] [ 0.093]
2005 0.054 0.07 0.071 0.073 9261 0.561 0.144
[ 0.046] [ 0.054] [ 0.054] [ 0.056] [ 1114] [ 0.060] [ 0.093]
2006 0.054 0.072 0.071 0.074 9470 0.585 0.139
[ 0.047] [ 0.056] [ 0.054] [ 0.057] [ 1177] [ 0.055] [ 0.089]
2007 0.055 0.073 0.071 0.073 9667 0.614 0.137
[ 0.048] [ 0.055] [ 0.055] [ 0.056] [ 1212] [ 0.051] [ 0.088]
2008 0.055 0.075 0.074 0.075 9926 0.656 0.136
[ 0.049] [ 0.059] [ 0.056] [ 0.057] [ 1293] [ 0.048] [ 0.085]
All years 0.052 0.07 0.071 0.074 8134 0.52 0.156
[ 0.046] [ 0.054] [ 0.054] [ 0.056] [ 1566] [ 0.097] [ 0.097]
Note: fees are measured in 2007-08 prices , and are measured yearly as three time the termly fee.

This summary masks significant variation in the level of private school attendance, fee levels and

average state school quality across regions. In Appendix 1 we show the variation in private school

attendance (at age 15) for each of our 149 local authorities but in Figure 1 below we show

differences in private school participation (at age 15), private school real day fees and state school

quality across nine English regions between 1996 and 2008. As can be seen, even at these broad

regional levels there is significant variation across region. This is even more true at the local

authority level (as seen in Appendix 1) and it is this variation across local authority that we exploit in

this paper.

10
 Figure 1: Proportion of 15 year olds in private schools 1996-2008
.1
.08
Proportion

.06
.04

1990 1995 2000 2005 2010

Year

North Yorkshire
East Midlands East Anglia
South East South West
West Midlands North West
London

Figure 2: Private school day fees, 1990-2008 (2007/08 prices)

Fees
12
10
8
6
4

1990 1995 2000 2005 2010

Year

North, NW, Yorks East Midlands

East Anglia South East
South West West Midlands
London

11
 Figure 3: Proportion of 15 year olds reaching expected level in state schools, 1993-2008
.7
.6
Proportion

.5
.4
.3

1990 1995 2000 2005 2010

Year

North Yorkshire
East Midlands East Anglia
South East South West
West Midlands North West
London

3. Results

The results of our preferred double difference GMM system estimation procedure is shown in Table

3.1. It turns out that using a double difference model is important in this context as the system

GMM estimates of the lagged dependent variable for a number of ages is close to a unit root and in

some cases unstable. This is probably due to there being an incidental trend in our unobserved fixed

effect which renders the traditional system GMM invalid (because of the violation of the required

moment conditions). The results of doing the traditional system GMM estimation procedure is

shown in Table A2 in Appendix 2.

12
TABLE 3.1: System GMM Double Difference Estimates of the Demand for Private Schooling for 7, 11, 13 and 15 year olds.

Variable Estimate Elasticity SD Estimate Elasticity SD Estimate Elasticity SD Estimate Elasticity SD

Increase Increase Increase Increase
0.591 0.747 0.945
[ 0.067] [ 0.082] [ 0.035]
0.004 0.075 0.093 -0.006 -0.081 -0.132 0.009 0.118 0.198 0.004 0.048 0.083
[ 0.012] [ 0.210] [ 0.262] [ 0.014] [ 0.180] [ 0.294] [ 0.012] [ 0.155] [ 0.260] [ 0.010] [ 0.124] [ 0.214]
-0.015 -0.263 -0.328 -0.017 -0.224 -0.366 -0.014 -0.181 -0.303
[ 0.007] [ 0.123] [ 0.153] [ 0.015] [ 0.204] [ 0.333] [ 0.013] [ 0.164] [ 0.274]
0.002 0.014 0.016 -0.032 -0.211 -0.304 0.003 0.019 0.029 -0.009 -0.054 -0.082
[ 0.008] [ 0.069] [ 0.076] [ 0.012] [ 0.081] [ 0.117] [ 0.007] [ 0.046] [ 0.067] [ 0.010] [ 0.064] [ 0.098]
0.013 0.117 0.129 -0.015 -0.098 -0.141 -0.019 -0.125 -0.185
[ 0.012] [ 0.105] [ 0.116] [ 0.018] [ 0.121] [ 0.173] [ 0.010] [ 0.065] [ 0.095]
0.005 0.015 0.052 -0.067 -0.169 -0.751 0.007 0.018 0.083 -0.045 -0.106 -0.5
[ 0.016] [ 0.052] [ 0.177] [ 0.027] [ 0.068] [ 0.302] [ 0.019] [ 0.046] [ 0.211] [ 0.035] [ 0.083] [ 0.390]
-0.007 -0.022 -0.076 -0.11 -0.276 -1.228 -0.015 -0.036 -0.163
[ 0.033] [ 0.109] [ 0.369] [ 0.044] [ 0.110] [ 0.490] [ 0.028] [ 0.069] [ 0.314]

No. Of observations 1788 1639 1639 1639

AR(1) p-value 0.000 0.000 0.000 0.000
AR(2) p-value 0.001 0.015 0.028 0.019
AR(3) p-value 0.290 0.596 0.838 0.983
GMM Lags 3 to 5 3 to 6 4 to 7 4 to 7
Hansen test p-value 0.142 0.170 0.054 0.114

Note: All regressions include time dummies and use robust two stage system GMM estimation (see Windmeijer (2005)). There are 149 groups in our panel. Frt refers to day fees in region r at
time t, Qrt to proportion of children obtaining the expected level in state secondary schools at time t in region r and FSMrt refers to the proportion of children eligible for free school meals in
state secondary schools in region r at time t.

13
If we start by focusing on the results at age 7 we see that fees at age 5 (when parents were making
their primary school choice) have a negative impact on private school attendance at age 7. A one
standard deviation increase in fees reduces the demand for private schooling at age 7 by 0.33
percentage points which equates to an elasticity of -0.26. This is very close to the elasticity found by
Dynarski et. al. (2009). Secondary state school quality has no effect on the primary school decision
but impacts on demand at ages 11 and 13. We have not included primary state school quality in our
regressions as this information is not available for the time period under consideration. No other
factors impact on demand (after we have done our double differencing).

At ages 8, 9 and 10, the demand for private schooling is determined by participation in the previous
year. No other factors influence demand4.

If we move to participation at age 11, we see that fees in the previous two years have a negative
impact on participation, but this effect is not significant at conventional levels. In particular, fees at
age 9, when parents are making their secondary school decision, have a similar elasticity to that
found at age 7 (-0.22), but this estimate is not significant at conventional levels. However, now the
quality of state secondary schools impacts on demand for private schooling with an increase of one
standard deviation in state school quality at age 10 reducing private school demand at age 11 by
0.31 percentage points which equates to an elasticity of -0.21.

If we move on to look at demand at age 13, another key moving point in the English education
system, we similar negative but slightly smaller effects of price and quality on demand, although
once again the fees effect is not significant at conventional levels. A one standard deviation increase
in state school quality at age 11 decreases participation in private schooling at age 13 by 0.19
percentage points which equates to an elasticity of -0.13.

At age 15, the final year of secondary school, it is only participation in the previous year that explains
demand for private schooling. Similar findings are found for participation at age 12 and age 14.

Hence price and quality impact on the demand for private education in the expected way. However,
private school demand in England is a dynamic process and prices and quality only have direct
impacts at key moving points in the education cycle at the ages of 7, 11 and 13. For other ages, fees
and quality only impact on the demand for private schooling via their effect on lagged demand.

4
These results are available from the authors.

14
4. Conclusions

In this paper we use English school level data from 1993 to 2008 aggregated up to small
neighbourhood areas to look at the determinants of the demand for private education in England
from the ages of 7 until 15 (the last year of compulsory schooling). We focus on the relative
importance of price and quality of schooling. However, there are likely to be unobservable factors
that are correlated with private school prices and/or the quality of state schools that also impact on
the demand for private schooling which could bias our estimates. Our long regional and local
authority panel data allows us to employ a number of strategies to deal with this potential
endogeneity. Because of the likely presence of incidental trends in our unobservables, we employ a
double difference system GMM approach to remove both fixed effects and incidental trends. We
find that the demand for private schooling is inversely related to private school fees as well as the
quality of state schooling in the local area at the time families were making key schooling choice
decisions at the ages of 7, 11 and 13. We estimate that a one standard deviation increase in the
private school day fee when parents/students are making these key decisions reduces the
proportion attending private schools by around 0.33 percentage points which equates to an
elasticity of around -0.26. This estimate is only significant for choices at age 7 (but the point
estimates are very similar at the ages of 11 and 13). At age 11 and age 13, an increase in the quality
of local state secondary reduces the probability of attending private schools. At age 11, a one
standard deviation increase in state school quality reduces participation in private schools by 0.31
percentage points which equates to an elasticity of -0.21. The effect at age 13 is slightly smaller, but
still significant. Demand for private schooling at the ages of 8, 9, 10 and 12, 14 and 15 are almost
entirely determined by private school demand in the previous year for the same cohort, and price
and quality to not impact significantly on this decision other than through their initial influence on
the key participation decisions at the ages of 7, 11 and 13.

15
5. References

Arellano, Manuel & Bond, Stephen, (1991), ‘Some Tests of Specification for Panel Data: Monte Carlo
Evidence and an Application to Employment Equations,’ Review of Economic Studies, Blackwell
Publishing, vol. 58(2), 277-97.

Blundell, R. & S. Bond (1998), ‘Initial conditions and moment restrictions in dynamic panel data
models’, Journal of Econometrics 87, 115–143.

Burgess, S., Greaves, E., Vignoles A., and Wilson, D. (2009), ‘Parental choice of primary school in
England: what `type’ of school do parents choose?’ CMPO Working Paper No. 09/224, November.

Blow, L., Blundell, R. and Machin, S. (2010), ‘The demand for private schooling in the UK: the role of
income and inequality’, IFS, mimeo.

Dearden , L. and Sibieta, L. (2010), ‘What determines private school choice? Evidence from the
British Household Panel Survey’, mimeo, IFS.

Dynarski, S., Gruber, J. and Li, D. (2009), ‘Cheaper by the dozen: using sibling discounts at Catholic
schools to estimate the price elasticity of private school attendance’, NBER Working Paper no.15461,
October.

Gibbons and Machin (2003), ‘Valuing English primary schools,’ Journal of Urban Economics, 53(2),
197-219.

Gibbons, S. and Machin, S. (2006), ‘Paying for primary schools: Supply constraints, school popularity
or congestion’, Economic Journal, vol 116, C77-C93.

Han, C. and Phillips, P.C.B. (2010), ‘GMM Estimation for Dynamic Panels with Fixed Effects and
Strong Instruments at Unity’, Econometric Theory, 26, 119-151.

Hansen and Machin (2010), ‘Demand for School Quality in the Early Years’, forthcoming CEP Working
Paper, October 2009.

Windmeijer, Frank, 2005. ‘A finite sample correction for the variance of linear efficient two-step
GMM estimators,’ Journal of Econometrics, vol. 126(1), pages 25-51.

16
 Appendix 1 – Private school attendance by local authority 1996-2008

201 202
Proportion

Proportion

.1 .2
.05.15
.1 .2
.05.15
0

0
1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

203 204
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

0
1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

205 206

.1 .2
Proportion

Proportion

.05 .15
.1 .2
.05.15
0

1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

7 yr olds in private schools 11 yr olds in private schools

12 yr olds in private schools 13 yr olds in private schools
14 yr olds in private schools 15 yr olds in private schools

207 208
.1 .2
Proportion

Proportion

.05 .15
.1 .2
.05.15
0

1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

209 210
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

211 212
.1 .2
Proportion

Proportion
.05 .15

.1 .2
.05 .15
0

1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

7 yr olds in private schools 11 yr olds in private schools

12 yr olds in private schools 13 yr olds in private schools
14 yr olds in private schools 15 yr olds in private schools

17
 213 301

.1 .2
Proportion

Proportion

.05 .15
.1 .2
.05.15
0

0
1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

302 303
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

0
1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

304 305
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

0
1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

7 yr olds in private schools 11 yr olds in private schools

12 yr olds in private schools 13 yr olds in private schools
14 yr olds in private schools 15 yr olds in private schools

306 307
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

308 309
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

310 311
.1 .2
.1 .2
Proportion

Proportion

.05 .15
.05 .15
0

1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

7 yr olds in private schools 11 yr olds in private schools

12 yr olds in private schools 13 yr olds in private schools
14 yr olds in private schools 15 yr olds in private schools

18
 312 313
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

0
1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

314 315

.1 .2
.1 .2
Proportion

Proportion

.05 .15
.05 .15
0

0
1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

316 317
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

0
1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

7 yr olds in private schools 11 yr olds in private schools

12 yr olds in private schools 13 yr olds in private schools
14 yr olds in private schools 15 yr olds in private schools

318 319
.1 .2
Proportion

Proportion

.05 .15
.1 .2
.05.15
0

1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

320 330
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

331 332
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

7 yr olds in private schools 11 yr olds in private schools

12 yr olds in private schools 13 yr olds in private schools
14 yr olds in private schools 15 yr olds in private schools

19
 333 334
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

0
1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

335 336
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

0
1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

340 341
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

0
1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

7 yr olds in private schools 11 yr olds in private schools

12 yr olds in private schools 13 yr olds in private schools
14 yr olds in private schools 15 yr olds in private schools

342 343
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

344 350
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

351 352
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

7 yr olds in private schools 11 yr olds in private schools

12 yr olds in private schools 13 yr olds in private schools
14 yr olds in private schools 15 yr olds in private schools

20
 353 354
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

0
1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

355 356
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

0
1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

357 358
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

0
1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

7 yr olds in private schools 11 yr olds in private schools

12 yr olds in private schools 13 yr olds in private schools
14 yr olds in private schools 15 yr olds in private schools

359 370
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

371 372
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

373 380
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

7 yr olds in private schools 11 yr olds in private schools

12 yr olds in private schools 13 yr olds in private schools
14 yr olds in private schools 15 yr olds in private schools

21
 381 382
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

0
1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

383 384
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

0
1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

390 391
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

0
1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

7 yr olds in private schools 11 yr olds in private schools

12 yr olds in private schools 13 yr olds in private schools
14 yr olds in private schools 15 yr olds in private schools

392 393
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

394 800
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

801 802
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

7 yr olds in private schools 11 yr olds in private schools

12 yr olds in private schools 13 yr olds in private schools
14 yr olds in private schools 15 yr olds in private schools

22
 803 805
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

0
1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

806 807
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

0
1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

808 810
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

0
1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

7 yr olds in private schools 11 yr olds in private schools

12 yr olds in private schools 13 yr olds in private schools
14 yr olds in private schools 15 yr olds in private schools

811 812
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

813 815
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

816 820
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

7 yr olds in private schools 11 yr olds in private schools

12 yr olds in private schools 13 yr olds in private schools
14 yr olds in private schools 15 yr olds in private schools

23
 821 825
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

0
1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

826 830
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

0
1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

831 835
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

0
1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

7 yr olds in private schools 11 yr olds in private schools

12 yr olds in private schools 13 yr olds in private schools
14 yr olds in private schools 15 yr olds in private schools

836 837
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

840 841
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

845 846
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

7 yr olds in private schools 11 yr olds in private schools

12 yr olds in private schools 13 yr olds in private schools
14 yr olds in private schools 15 yr olds in private schools

24
 850 851
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

0
1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

852 855
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

0
1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

856 857
.1 .2
Proportion

Proportion
.05 .15

.1 .2
.05.15
0

0
1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

7 yr olds in private schools 11 yr olds in private schools

12 yr olds in private schools 13 yr olds in private schools
14 yr olds in private schools 15 yr olds in private schools

860 861
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

865 866
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

867 868
.1 .2
.1 .2
Proportion

Proportion

.05 .15
.05 .15
0

1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

7 yr olds in private schools 11 yr olds in private schools

12 yr olds in private schools 13 yr olds in private schools
14 yr olds in private schools 15 yr olds in private schools

25
 869 870
.1 .2
Proportion

Proportion
.05 .15

.1 .2
.05 .15
0

0
1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

871 872
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

0
1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

873 874
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

0
1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

7 yr olds in private schools 11 yr olds in private schools

12 yr olds in private schools 13 yr olds in private schools
14 yr olds in private schools 15 yr olds in private schools

875 876
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

877 878
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

879 880
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

7 yr olds in private schools 11 yr olds in private schools

12 yr olds in private schools 13 yr olds in private schools
14 yr olds in private schools 15 yr olds in private schools

26
 881 882
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

0
1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

883 884
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

0
1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

885 886
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

0
1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

7 yr olds in private schools 11 yr olds in private schools

12 yr olds in private schools 13 yr olds in private schools
14 yr olds in private schools 15 yr olds in private schools

887 888
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

889 890
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

891 892
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

7 yr olds in private schools 11 yr olds in private schools

12 yr olds in private schools 13 yr olds in private schools
14 yr olds in private schools 15 yr olds in private schools

27
 893 894
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

0
1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

908 909
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

0
1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

916 919
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

0
1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

7 yr olds in private schools 11 yr olds in private schools

12 yr olds in private schools 13 yr olds in private schools
14 yr olds in private schools 15 yr olds in private schools

921 925
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

926 928
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

929 931
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

7 yr olds in private schools 11 yr olds in private schools

12 yr olds in private schools 13 yr olds in private schools
14 yr olds in private schools 15 yr olds in private schools

28
 933 935
.1 .2

.1 .2
Proportion

Proportion
.05 .15

.05 .15
0

0
1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

936 937

.1 .2
Proportion

Proportion
.1 .2

.05 .15
.05 .15
0

0
1990 1995 2000 2005 2010 1990 1995 2000 2005 2010
Year Year

938
.1 .2
Proportion

.05 .15
0

1990 1995 2000 2005 2010

Year

7 yr olds in private schools 11 yr olds in private schools

12 yr olds in private schools 13 yr olds in private schools
14 yr olds in private schools 15 yr olds in private schools

29
Appendix 2 – Traditional System GMM estimates

TABLE A.2: Traditional System GMM Estimates of the Demand for Private Schooling for 7, 11, 13 and 15 year olds.

Variable Estimate Elasticity SD Estimate Elasticity SD Estimate Elasticity SD Estimate Elasticity SD

Increase Increase Increase Increase
0.918 0.942 1.009
[ 0.093] [ 0.035] [ 0.011]
0.105 1.823 2.273 -0.004 -0.055 -0.09 0.031 0.395 0.661 0.006 0.074 0.128
[ 0.031] [ 0.548] [ 0.683] [ 0.031] [ 0.412] [ 0.672] [ 0.013] [ 0.168] [ 0.281] [ 0.002] [ 0.028] [ 0.048]
-0.004 -0.069 -0.086 0.011 0.143 0.234 -0.02 -0.254 -0.426
[ 0.027] [ 0.475] [ 0.592] [ 0.031] [ 0.415] [ 0.678] [ 0.011] [ 0.145] [ 0.243]
0.078 0.681 0.748 -0.038 -0.255 -0.367 -0.004 -0.028 -0.041 0 0.003 0.004
[ 0.024] [ 0.209] [ 0.230] [ 0.020] [ 0.131] [ 0.188] [ 0.007] [ 0.045] [ 0.066] [ 0.003] [ 0.022] [ 0.033]
0.109 0.951 1.045 0.013 0.086 0.123 -0.008 -0.052 -0.076
[ 0.027] [ 0.087] [ 0.297] [ 0.020] [ 0.050] [ 0.225] [ 0.011] [ 0.026] [ 0.121]
0.004 0.013 0.045 -0.007 -0.019 -0.084 -0.013 -0.033 -0.15 -0.001 -0.002 -0.007
[ 0.039] [ 0.129] [ 0.440] [ 0.025] [ 0.062] [ 0.278] [ 0.010] [ 0.024] [ 0.108] [ 0.003] [ 0.008] [ 0.037]
0.053 0.173 0.588 0.037 0.094 0.418 -0.007 -0.018 -0.081
[ 0.038] [ 0.126] [ 0.429] [ 0.024] [ 0.061] [ 0.272] [ 0.008] [ 0.020] [ 0.093]

No. Of observations 1937 1788 1788 1788

AR(1) p-value 0 0 0 0
AR(2) p-value 0.074 0.205 0.933 0.505
AR(3) p-value 0.443 0.533 0.532 0.896
GMM Lags 3 to 5 3 to 6 3 to 6 2 to 5
Hansen test p-value 0.01 0.107 0.272 0.126

Note: see notes to Table 3

30
31
 32

View publication stats