Testing for Cream Skimming in India

In the previous video

- the rise of private education in India

we showed the beginning in the 1990s

there was a massive increase in
demand for education,

which was met by private schools.

So much so that now in India

27% of the primary age children
are in private schools,

and in urban regions it is as
high as almost 50%.

We also showed that on educational tests

the kids in private schools do better

than those in public schools.

But the question is: is this because
the private schools are better

or is it because they draw
from better students?

This is a question of cream skimming.

It's part of what I looked at in
one of my papers,

and it's what we are going to talk
about in this video.

Lots of papers have tried to
control for cream skimming

using American data or data
from other countries,

but they haven't always
convinced the critics.

And the reason for this is that

when the private schools are
just 10% of the market,

the kids in the private schools are

in fact very likely to be different

than the kids in the public schools.

With only 10% of the market,

It's easy to imagine that
it's a different 10%,

it's the best 10%,

or a portion of it it's the best 10%

However, India offers a unique opportunity

to test for cream skimming,

because in many districts in India --

and these are just a few --

a majority of the kids are
in private schools.

Well, if the majority of the kids

are in the private schools,

these kids must be much closer to average,

because they're the majority.

If the majority of the kids

are in the private schools,

they can't all be the cream.

That's the intuition,

let's look in a little more detail at a model.

Here's a simple model of cream skimming.

We are going to make 3 important assumptions:

First, there's some distribution of child ability -

some kids have higher ability than

do other kids.

Second, we're going to assume that

private schools offer no
educational advantages.

Private schools do not increase ability or

testing level anymore
than do public schools.

They are equal.

Now, given this assumption,

the average ability
taken over all children,

whether they are in public schools

or in private schools,

will not vary with the private share.

Since private schools offer no advantages

over public schools,

as we increase the private share,

the mean over the entire population

will be absolutely fixed.

Third assumption, we're going to assume

that the private schools draw from

the top of the public school distribution

or draw to a greater extent from the top

of the public school distribution.

With these assumptions, when the

private share is low,

it's quite easy for the private mean

to be well above the public mean.

Simply because private schools draw

cream skim from the top

of the public distribution.

Notice however that as the private share

increases, the private mean must fall and

get closer and closer to the
population mean.

Similarly, since the private schools are

drawing from the top of the
public distribution,

the public mean must fall lower and lower

as the private share increases.

Okay, here's a second model of education:

The Higher Productivity Model.

In this model, we assume that the private

schools are better than
the public schools.

They actually raise educational quality

and achievements.

We're also going to assume
that the private

schools pull randomly from
the public schools

So, there's no cream skimming going on

in this model.

What this means is that when the

private share of education is low,

the population mean is close to

the the public mean.

As the private share increases
however, and

as we get more and more kids entering the

better schools, you get an increase in the

population mean until as you get almost

everybody in the private schools, the

population mean is equal to
the private mean.

Key here is that as you
increase the private

share, you get more and more kids in the

higher productivity schools so the

population mean increases.

Okay, what do the data say?

Remember, the cream skimming model says

that as the private share of schooling

increases, you're just changing how the

children are distributed,
so there should be

no change in the mean score.

The higher productivity model says that as

the private share increases,
you want to see

an increase as you get more and more

children into the private sector,
where the

educational quality is higher.

So what we see for reading scores is that

as the private share increases,

the mean reading score increases,

suggesting higher productivity.

Notice, we do not see any
big changes in the

scores in the publics schools.
We see a slight

decline in the scores in the
private schools,

suggesting that a little bit of
cream skimming

may be going on; nevertheless, we see

increases in the mean, which does suggest

higher productivity.

Now this is just a crude look at the data,

let's take a closer look to see if we can

pinpoint where cream skimming might be

occurring, if it is occurring.

In this slide, we're showing
the percentage

of children scoring at the highest reading

level in the private and in the
public schools.

And what is very market here, is that as

the private share of schooling increases,

from about 0 to say 15%, we see a market

decline in the number of children in the

private schools scoring at the
highest level.

So this suggested that cream skimming is

going on. When private share
is really low,

When private share is really low, they

really are getting the
best of the students.

As the private share increases, they are

getting lower and lower
quality of students.

However, after about 10-15%,
notice that this

level's off, and that we see no decline in

the private share and also
no decline in the

public share.

So what this suggests is
that there's a little

bit of cream skimming going
on, when private

schools are small, but when the private

schools reach about 20% or so, they are

drawing randomly from the population

and they have greater achievement.

The private schools
actually are increasing

educational quality.

I think this story makes a lot of sense.

When the private schools
are small, of course

they draw from the cream,

but once the private schools
are drawing from

over 20% of the population
they're basically

drawing from the same pool
as are the public

schools and what we
see is that in the limit

there is an improvement
in the reading quality

of the kids.

Same is also true from mathematics,

by the way.

So if you go back to the previous video:

The Rise of Private Education in India,

you'll see that the achievements
level of the

kids in private schools are much
higher than

those in the public schools.

What we've seen now is that some of this

difference is due to cream skimming, but

a modest, nevertheless, a significant amount

is also due to the higher productivity in the

private schools.

And indeed, that shouldn't be surprising

after all the private schools have lower

absentee rates.

If teachers who are actually

in the classroom teaching

more often, so it makes sense

that there would be some difference.

How much is this difference?

Well like I said, this difference is

modest but significant.

So switching children to private

schools, would increase

by 11.8 percentage points,

or 38.5 percent the number

of children who can read at the
very highest testing level.

Looking at the bottom side,

it would decrease, switching kids

into the private schools, would

decrease by about 5 percentage
points, or 55%,

the number of children performing

at the lowest tested levels.

The same thing is true

for arithmetic, look at the paper for

details, but basically switching the kids

to the private schools would increase

by 30% the number of children who

can do arithmetic at the highest tested

level. And it would decrease by 26%
the number of children performing

at the lowest arithmetic level.

Okay, let's put these results in context

The Indian population is very young

370 million people, or 32% of the
population below the age of 15

What that means is that there are about

100 million children being education in

private schools

In fact, this is the largest experiment

in private education, probably since 19th
century Great Britain.

So how's the experiment going?

Well, what my paper suggests, as
well as others,

is that private schools improve
educational performance

modestly, but significantly.

Even if they don't improve education

performance, even if they're just as good

as the government schools, they're
considerably cheaper, so that's

a big advantage.

Despite the fact, or perhaps because of
the fact that this experiment

in private education, which happened
spontaneously,

which happened from the ground up,

was NOT driven by government policy

was a response of the people,

was a response of entrepreneurs and

the people for a demand for education

despite the fact that this has gone well,

the Right to Education Act of 2009

may threaten this experiment.

The Right to Education Act imposes

conditions on the private schools

many of which do not apply to the

public schools.

We'll have to see what happens in
the future.

Here's some further reading.

My paper of course, I mentioned James
Tooley before and his co-authors

and I've also mentioned Nindan Nilekani's
book before.

I should also point out that not everybody
finds that private education

raises achievement levels. There's a
paper by

Chudgar and Quin, they find that

private and public schools are about
the same,

once you control selection factors

I actually find this somewhat difficult
to believe

since we do know that the private schools
do have much lower absentee rates

the public schools have much higher
absentee rates.

Nevertheless, they find once you

control for other factors,

achievement levels are about the same

However, the private schools are much
cheaper than the public schools.

So when a productivity advantage

that would still go to the private schools

Kingdon finds, as I do, that the private

schools in India have a higher

quality and efficiency.

Thanks very much.