Causes of Corruption

So, what do we know about
what causes corruption?

One of the best pieces on
this topic is by Daniel Treisman.

Let's look at its results
a little more closely.

He rates countries on the basis
of how corrupt they are,

using three different indices of
corruption from a group called

Transparency International.

Most researchers regard these
indices as actually pretty accurate.

He then looks to see which features
of a country are positively correlated

with the country having
a high rating for corruption.

We'll look at the different
hypotheses in this paper

noting, of course, that hypothesis
doesn't mean it's true;

it's simply something
we're going to investigate.

For instance, how much does having
a common law system matter?

Actually, it turns out
this result is unclear.

If a country is a colony of British origin,

it turns out that having
a common law system

is correlated with lower corruption.

But if the country was not
previously a British colony,

having, a common law system is
correlated with higher corruption.

So, on that, the effect of having
a common law system is quite unclear.

One of the most significant variables
for explaining low levels of corruption,

is having had the heritage
of being a British colony.

This holds up even if we adjust
for greater openness to trade,

or democracy, or religious tradition.

It seems that the former British colonies,
somehow, had instilled in them

greater protections against official abuse.

There are two striking things
about this correlation.

First, in the former British colony,

people are especially likely to have
confidence in the quality of judges

more than they have confidence
in the quality of politicians.

Second is that, from
a poll of businessmen,

they're more likely to have confidence that
the legal system will be enforced fairly.

So, if we try to think about which features
of having been a British colony

are the ones that are important;

maybe it's not about electoral politics

but it's somehow the quality
of the judiciary and its fairness.

Countries with a protestant religious
tradition also are less corrupt on average.

Why might this be?
We really don't know.

One interpretation is that Protestantism
may allow for a greater tolerance

for challenges to authority.

The second possibility is the stress of
Protestantism on casting out the wicked.

Some say that Protestantism allows for
greater focus on the individual

rather than the family and
this leads to less corruption.

Finally, it's been suggested that
Protestant societies

tend to have greater separation
of church and state,

and that may be another factor.

But the data don't themselves
allow us to judge on these issues.

Democratic countries also
tend to be less corrupt,

but there's something very
interesting in the data here.

Simply having democracy now
doesn't make the country less corrupt.

The democracies which are less
corrupt are those which have had

uninterrupted democracy
for 40 years or more.

So, actually, it's the distant
past which matters,

it's the traditions which are fostered by
having long periods of democratic rule.

You cannot create all the benefits
of a democracy overnight.

The wealthier and more literate
countries are also less corrupt

and this is indeed a very strong effect.

For instance, Treisman,
in the paper, notes that

a tenfold increase in
the 1990 per-capita GDP;

to go from, say, that of
El Salvador to that of Canada;

would lead to a drop in the corruption
rating of between 4.16 and 4.76 points.

That would bring El Salvador
up to somewhere about

to the level of Hong Kong or Ireland in
terms of having a low level of corruption.

Wealth here really matters.

Is it the case that corruption is low
when public salaries are relatively high,

as has proved to be the case in Singapore?

Well, in this data set,
it turns out we simply can't tell.

We'll be returning to
this question later in another unit.

What about political instability?

Well, that actually doesn't
turn up as being significant

in any of these statistical regressions.

So, on that, in this paper,
Treisman remains agnostic.

What about government
intervention in the economy?

Does that cause more corruption?

Well, on this again,
the author isn't sure.

He found that in his 1996 data,

there was a connection between
government intervention and corruption,

but, when he looked at
the 1997 and 1998 data sets,

it turned out there wasn't a connection.

So, on this, as with political
instability, he remains agnostic.

What about openness to trade?

Well, when exposure to imports is high,
corruption is somewhat lower,

but this is actually quite
a small effect quantitatively.

How about lots of natural resources?
Do they tend to make a country more corrupt.

This is, again, a case
where it's hard to say.

When we look at data from the 1980s,
it appears there is a relationship.

But when we look at data from the 1990s,
it appears that relationship has gone away.

This is, yet, another case where
some agnosticism is an order.

How about ethnic division or
what is, sometimes, called

ethno-linguistic fragmentation?

Well, it turns out that this doesn't
seem to predict corruption.

Once we adjust for
the wealth of an economy,

it turns out that ethnic division doesn't
matter for explaining corruption.

Corruption is correlated with
having a federal structure

much as the United States does.

Why might this be?

We're not sure, but it could be
that in a very decentralized society

there are all different levels of
officials who can get in your way

or stop something,

and maybe that makes it more likely
that you'll need to bribe one of them.

So, overall, what are the most robust
variables for predicting corruption?

They turn out to be British heritage,
Protestant tradition, economic growth,

having a federal structure, uninterrupted
democracy for 40 years or more,

and openness to imports.

These variables, together, actually
can account for more than 89 %

of the variation in the Transparency
International indices of corruption.

That's actually a somewhat
impressive result.

You might also wonder, which of
the countries which are more corrupt

or less corrupt than the model predicts?

Well, if we look at the 1980s,

the countries which turn out
to be more corrupt

than their underlying
variables would predict;

those are Thailand, Mexico,
Egypt, Indonesia, Haiti, and Zaire.

In the 1990s, the list changes.

It's Italy, Belgium and, sorry to say,
my country, the United States.

Interestingly, over this period of time,

Africa, on the whole, is less corrupt
than the model predicts.

Given its relatively low level
of per capita income,

we would expect Africa to be
more corrupt than it really is.

What can we say about
what's cause and what is effect?

For instance, if wealthier
nations are less corrupt,

is it the case that because they are
wealthy they are less corrupt?

Or is it the case that because they are
less corrupt they are wealthy?

Arguably, both effects are operating.

This is a tricky problem in
many statistical investigations.

Treisman attempts to overcome it
by focusing a lot of his attention

on very long-term variables which are not
changing in the short run with corruption.

For instance, if you think about our list
of the variables which really mattered:

40 years or more of democracy,

a British colonial heritage,
and Protestant religious tradition;

there's really not a plausible way to say
that low corruption brought about

the protestant religious tradition
from the distant past.

It's far more likely to be the case

that the Protestant religious
tradition is the active variable,

in some manner, driving
the lower level of corruption.

This paper leaves a lot of open questions
and questions unanswered or hanging

or, maybe, areas where we're agnostic.

But, still, it's considered to be one
of the most important studies

of the causes of corruption.