Digitized by the Internet Archive
in
2011 with funding from
Boston Library Consortium
Member
Libraries
http://www.archive.org/details/twopovertiesOObane
Technology
Department of Economics
Working Paper Series
Massachusetts
THE
Institute of
TWO
POVERTIES
Abhijit Banerjee
Paul J. Gertler
Maitreesh Ghatak
Working Paper 01-1
March 9, 2000
Room
E52-251
50 Memorial Drive
Cambridge, MA 02142
This paper can be downloaded without charge from the
Social Science Research Network Paper Collection at
http:/ /papers. ssm. com/paper. taf?abstract_id=271 731
MASSACHUSETTS INSTITUTE
OF TECHNOLOGY
AUG
1 5 2001
LIBRARIES
Massachusetts Institute of Technology
Department of Economics
Working Paper Series
THE
TWO
POVERTIES
Abhijit Banerjee
Working Paper 01 -16
March 9, 2000
Room
E52-251
50 Memorial Drive
Cambridge, MA 02142
This paper can be downloaded without charge from the
Social Science Research Network Paper Collection at
http://papers.ssrn.com
The Two
Poverties
Abhijit Banerjee*
MIT, Department
March
9,
of
Economics
2000
This Version December
8,
2000
Abstract
There are
at least
two distinct and inconsistent view of poverty.
These views, which can be
called "poverty as desperation"
as vulnerability", have different implications
icy. It is
and "poverty
about anti-poverty pol-
important to confront the conflict between them before data
can be applied to
tell
us whether any of the views are right or even
interesting.
*I
am
members
grateful to Esther Dufio,
of
my
14.772/2390b class
Campe Goodman, Karla
for
many
1
helpful
HofF,
comments.
Omer Moav, and
to the
Introduction
1
Are the poor
you and
just like
Hemingway's famous
invert
me
except in that they have less money, to
Or
line?
is it
them
useful to think of
as being
subject to different pressures from the rest of the population and therefore
sometimes making choices that are very different?
For a long time the dominant view in economics was that the distinction
had only descriptive
anyone
else.
usefulness, that behaviorally the poor were
One of the most important developments
much
like
two decades
of the last
within formal economics, fueled by attempts to rigorously study the evolution
of the distribution of consumption
this position.
There
is
and wealth,
is
a movement away from
growing emphasis on poverty as a distinct analytical
concept rather than purely as a category of description. 2 That
are the poor" and
The
roots of this shift are complex but an important part
there
risk
A
made
why
that affect the poor more than everybody else.
The
of asset
market
failures
assumption that the poor were workers and owned no
easier to focus
on the
choices, albeit
under
A
This
fact that the poor, like
less
it
everybody
is
to
be found
not to say that descriptions are unimportant.
Sen, 1999) and others on
their automatic
assets, also
else,
make
made
it
life-time
favorable circumstances.
very influential statement of this view
is
came from
better understanding of preferences to-
waning of the neo-Marxist and neo-Ricardian models, with
2
"who
easier to see
and the sources
may be problems
1
less
more "what do they do and why"?
developments in microeconomics:
wards
is,
who we should
in Schultz (1964).
The work
of
Amartya Sen
(see
think of as the poor have had an enormous impact
on how we assess the effectiveness of anti-poverty
policies.
with the work of people
and Loury (1981).
goal of this paper
process,
it
will
is
(1979), Kihlstrom
usually implicit,
than
in the model, rather
Kanbur
like
explicit:
to
is
i.e.,
"the right
in the
way
least
that
we
form of an assumption
to
model poverty
explicit.
and "poverty as vulnerability".
I
is...".
In the
two distinct and, prima
inconsistent views of poverty in these models, which
different implications
was launched
and Laffont (1979)
make the conceptualization
emerge that there are at
as desperation"
literature
remains true, however, that the conceptualization of
It
poverty in this literature
The
new
has been more than twenty years since this
It
facie,
will call these "poverty
These views have rather
about anti-poverty policy and this makes
it
important
confront the conflict between them.
Poverty as Desperation
2
The poor
lose.
3
To
they have too
are different because they are desperate:
give formal content to this idea, imagine a world
where there
one good produced and a population of identical people who each
period and always have one child. Each person starts
which her parent gave
beginning of her
we will
life,
Her
she chooses
life is
This idea
is
it
among income
to her child or eat
implicit in the
Banerjee and
Newman
and Sjostrom (1998).
(1993), Galor
It is
it
models that draw the
and the persistence of poverty.
See,
live for
among
to
is
one
with an endowment
simple, verging on the drab.
At the
earning opportunities (which
describe later). At the end she decides on
income - she can leave
tions
her.
life
little
what to do with her
herself.
link
realized
For simplicity, assume
between credit market imperfec-
other papers, Aghion and Bolton (1997),
and Zeira (1993), Piketty (1997) and Ghatak, Morelli
explicit in Banerjee
and
Newman
(1994).
that she has Cobb-Douglas-like preferences over consumption
(c)
and bequest
(by.
U(c,b)
AlJ-Pb
= -±-
1
-'*
}
i
—a
1
,0</3<
1,1
>a>0,A>0
Since she allocates her end-of-period wealth between these two uses, this
immediately implies that
if
her end-of-period income (or wealth)
c=(l-/%,
It
b
=
(3y.
follows that her (indirect) utility from having
l
The
an income of y
—a
uj,
though to understand the exact nature of the depen-
we would need to say something about the nature
opportunities in this economy. Moreover, choices
lifetime,
is:
end-of-period income, y, should clearly depend on beginning-of-
period endowment,
dence,
is y:
made
C, and luck in the form of some random shock,
For the time being, we just write the income function
y
=
y(u,6,C),
While we do not pretend that
this
yu
is,
>
of income-earning
during the person's
8,
must play a
as:
0.
or ought to be,
framework does capture important aspects of
it.
role.
A
all
of
large part of
life,
this
what we
start out
and
is
life
with - health, education, land, money - comes from our parents,
mostly a result of a choice that they make (or rather, one we make
together).
4
While the label we attach to the bequest may vary (and, indeed,
there are usually multiple types of bequests), to assume that one starts
life
with a bequest from one's parent and then makes choices about the kind of
income one
With
will get fits the pattern of
this
most people's
somewhat elaborate preamble, we are now ready to return to
the problem of conceptualizing poverty.
idea that the poor are desperate
how low
V(-) can get.
That
is
In other words,
<
¥->
if
is
The
is
some
system. Or
it
may
V_,
effective utility function,
rather than V(y).
may
just
V_.
reflect private generosity (or
fall
could be the result
It
embody the guarantees
neighbors will simply not allow anyone to
it
given by the welfare
the lack of
below
it)
- friends and
Or, less obviously,
V_.
could reflect the failure of imagination: perhaps people cannot conceive
of being worse off than
V_.
In other words,
beyond some
4
To the above
list
of possible types of bequest,
we might add
gives up) or
"culture"
where culture
interpreted narrowly as attitudes toward work and corruption.
of culture either
from one's parents or from one's peer group, and
learn reflects choices
made by our
point, having less
body
to eat either really stops mattering (perhaps because the
is
some bound on
is
the end-of-period income ends up being such that
There are several possible interpretations of
V_
to capture the
= max\£,V(y)].
the person's utility will be
of social policy:
way
simplest
to assume that there
to say, there
V{y)
V(y)
lives.
One
"learns" this kind
in either case
what we
parents (which include our choice of peer group).
stops mattering in the
of
it
of
someone who
is
not yet there but
thinking
is
as a possibility.
The
c
mind
interesting case for us, obviously,
and some random
is
the one in which for some choice
realization 9:
V(y(0,9,c))<V.
That
close
To
someone who
for
is,
enough to
see
model
what
starts with wealth 0,
as well, the constraint that V(-)
this
can
tell
us about poverty,
let
and therefore
>V
is
for those
sometimes binding.
us use this assumption in a
of credit.
Consider the following specialization of our basic model: Assume that
once one gets one's bequest, one has the choice of putting
(where
it
it
in the
earns a gross interest rate r) and going to work for a wage,
bank
W,
or
starting a business. If one starts a business, the rate of return on each dollar
invested in the business
is
up to /
is
R>
r,
never any reason to invest more than
more work:
let
/.
u>
is
but
however
effort,
E,
disutility.
remains to say something about credit
wealth of
Starting a business
us assume that starting a business has disutility of
whereas working has no
It
the market interest rate, and there
may want
to invest an
itself.
amount I >
Our
u.
typical agent has a
The
constraint
comes
from the possibility of borrower misbehavior. Once a borrower has borrowed
and invested the money, she has no obvious reason to want to repay. What
stops her from defaulting
(legally or otherwise)
that a borrower
who
is
and
tries
the fact that the lender will
will try to extract
the money.
not to repay gets away with
6
it
come
after her
Let us assume
with probability
q.
With
probability
confiscated,
i.e.,
1
-
she gets caught and suffers having her entire income
q,
she ends up with an end-of-period income of
< V.
V(0)
A
borrower
who
which she
will
y*hanest
=
Therefore, the
w
is
not to default
y*( IR
_
_
(j
To avoid keeping
>
=
y*
+
(1
-
q)V,
yR _
V_.
(
J
_ w ) r ).
if:
^
> q V*(IR) +
V*(IR -
track of
many
{I
lot)
-
for
can invest starting with wealth of
w)r)
I
—
u» is
=
drawn
assume
(1
u> is
-
q)V.
large
enough
have:
(1
in figures la
what she borrows.
+
qV*(IR)
V* we now
of/.
u>,
- q)V = V* dlshonest
> qmax{V{IR),V} +
sides of this equation are
'Since she has wealth
(l
cases, let us
Using the expression
V(I(R -r) +
The two
q V* (I R)
maximum amount anyone
given by J(w), defining
that V(ur)
=
utility of:
compare with: 5
y*honest
will prefer
an expected
defaults therefore has
ydiskonest
She
Assume
0.
-
q)V.
(1)
and lb as functions
Insert Figures la
These are the two
cases:
raises the
Figure lb there
here.
In Figure la the two curves intersect, which
An
gives us a finite investment cap I(uj).
and therefore
and lb
increase in
u>
investment cap. Richer people get to invest more. In
no intersection and consequently no
is
pushes V* honest up
amount
limit to the
she can borrow.
The
relatively low values of u, the left-hand side
As a
V_.
comes from the
basic logic behind this credit limit
result,
may
while she
may
the borrower cannot lose very
indeed gain a
lot.
As
uj
and her investment cap moves up as a
This intuition
is
becomes
not be
much by
much
fact that for
greater than
trying to default,
larger, she
has more to lose
result.
confirmed by looking at
V_.
Raising V_ raises y* dtshonest
and therefore exerts downward pressure on the investment cap.
rower has
A
less to lose,
she gets to borrow
related point emerges
when we
If
the bor-
less.
look at the condition for getting our
Figure la rather than Figure lb. For our assumed preferences, for the case
where I
is
large enough, the condition turns out to be:
R-r
The
reader will recognize
Note that as a goes up,
6
a
6
< q^R.
to be the coefficient of relative risk aversion.
this condition
become harder and harder to
This follows from directly substituting the expression for the
inequality
large.
V(I(R -
r)
+
wr)
< qV(IR) +
(1
-
q)V_
and looking
satisfy,
utility function into the
at the effect of
making /
and
a
for
close to
1,
^
since g 1
is
close to 0,
it
cannot be
words, the more risk averse the investor, the less likely
Once
credit limit.
again, this ought to
from the fact that she
is
make
satisfied.
7
In other
that she faces a
it is
the credit limit comes
sense:
too willing to gamble. Making
costlier to
it
gamble
(by making her risk averse) acts as an obvious antidote.
We
summarize
Claim
1 People
this discussion in:
may face
investment caps in this model. Richer people will
face less stringent investment caps.
level
Making
the
minimum
higher makes the investment cap more stringent.
guaranteed
utility
more
Finally,
risk
averse people will face less stringent investment caps.
The
second and third part of this result would hold in most models
first,
of imperfect credit markets.
to a class of models,
increases risk,
What
-
though given that
V_
=
among the poor? To
0. It
will clearly
'
The
driven by the fact that defaulting
1
that I
linear in wealth.
=
specifically,
a
\u>
where A
We
=
=
A(-)
y
=
Someone who has a wealth
anyone who has a wealth of
we have only allowed a
to take all non-negative values.
utility is negative
us about
be able to invest a very small amount. Therefore, she
reader will notice that
usual to allow
tell
so that V(y)
not want to give up the wage income she could get
More
meaningless.
is
a =
fix ideas, let
follows from Equation
The investment cap
close to zero will only
invest.
it is
probably more specific
is
does this model of credit limits and investment caps
and assume
)r
fourth property
ought to be relatively robust.
it
investment behavior
r -i\-
The
and therefore the
assume a <
1
will get
to be less than 1
The reason
V_ constraint
u>
is
if
that for
she did not
a
utility of
whereas
a
it is
more
greater than
1,
always binds. This makes our exercise
to avoid this problem.
Xlu(R
- r) + wr - E
else. Clearly,
For those
if
she invests and
+
who
start with a wealth of
W+u
Those who
u> t r).
children will start
case
u) t+ \
=
if
she goes to work for someone
she will not invest unless her wealth was more than
period income will be
(3(W
W + ur
life
P[I(R
-
and
r
tu t+1
+u
in period
their children will start
u > u
start with a wealth
with
r)
t
< u
uj t
= 0[X(R - r) + r}ui
unless
}Y+
\(R—
r)
, .
their end-of-
t,
life
with
will invest
t
E
u=
—
Aw > J 8
uj t+1
and
,
in
=
their
which
t r].
Under the assumption that 0[X(R -
+
r)
r]
<
these dynamics can look
1
Figure 2a.
like
Insert Figure 2a here.
This
so
is
being
given by
trap.
is
At
the classical poverty trap diagram. Poverty
rich.
[0,a;]
u>*,
Both
and
ui*
and [u,a]
a;**
is
a steady state and
are attractors with basins of attraction
respectively,
an extreme version of the poverty
no one starts a business, we think of u* as poverty. Poverty
self-sustaining because the
poor are credit constrained and as a
is
result choose
not to invest. Consequently, they earn low rates of return on their investment
and do not accumulate wealth
Recall that /
9
The
is
the
fact that the
maximum
income
fast
it is
shifts
enough
jump
in
income
is
is
9
ever worth investing.
up discontinuously
poverty trap possible. Otherwise, since 0[X(R
45° line again. This case
to get out of the trap.
—
r)
therefore only possible
+
at
r]
when
to is
<
1,
E
a part of what makes the
the
is
w (+1
will
not cross the
sufficiently large since the
just the reward for the extra effort involved in running a business.
might wonder whether parents, faced with such a discontinuity
children would not always
want
in
to increase their bequest slightly,
such an outcome cannot ever be an equilibrium
10
if
One
the outcome for their
and therefore perhaps
the parents were forward-looking. This
This
is
Zeira (1993),
different
name
easy to see that this case
implies that raising
is
most
V_ and reducing
the investment cap, also
when A
is
&
(1994),
education in
Banerjee-Newman.
in
likely to obtain
a,
Newman
to this investment:
and capital
Galor-Zeira; health in Dasgupta-Ray;
It is easy,
papers, including Galor
Dasgupta and Ray (1986) and Banerjee and
though they each give a
is
many
the essence of the story told by
small,
It
which
both of which, as we have seen, lower
make the poverty trap more
likely.
of course, to point to the aspects of reality that are missing
from this narrative. For example, savings/bequests ought to be responsive to
the rates of return. If
we were
to
add
this ingredient to our model,
it
would
mostly reinforce the poverty trap since the poorest face the lowest return on
their savings in this
model and therefore should have the lowest
more convexities to the production function
On
(a set-up cost, for example) also
the one hand, an economy with lots of poor people will tend to
have low wages and,
get out of poverty.
all else
On
An
reward
for those
Of course we do not have
not, however, correct.
The
it
also
moves
u
it
harder to
down, which makes
endogenous interest rate also cuts both ways:
high interest rates reduce A and
also raise the
make
being the same, low wages
the other hand,
escape from poverty easier.
is
Adding
Making wages endogenous has a more nuanced
reinforces the poverty trap.
effect.
/?'s.
make
who put
it
less
their
rewarding to invest, but they
money
to have a poverty trap.
in the bank.
One
fact that the production technology
is
alternative config-
non-convex necessarily
makes the parents' preferences over bequests non-quasi-concave and therefore even
if
the
parents were forward-looking, their decision rule over bequests would be discontinuous and
therefore the wealth of the next generation could be discontinuous as function of parental
wealth.
11
uration, which
valid
is
under the condition fi[X(R
where everyone ends up
rich, is
- r) + r] >
1
but
fir
<
l,
10
depicted in Figure 2b.
Insert Figure 2b here.
Even
this
if
were the economy, we would care about how long
get out of poverty. Raising V_
down
ut+i curve
This
(by lowering A) makes
of course
is
and reducing
all
One can imagine many
he could borrow and
it
a,
by reducing
u_
and moving the
a longer process.
other ways in which a borrower could misbehave:
fail
to put effort into
fails,
he
will
it
on the assumption that
not have to repay. 11
gamble on bad, high-risk projects, again with the view that
with very
is left
takes to
based on a very specific model of the credit market.
as a result, the investment
he
it
little,
he
will
up the lender once the investment
if,
Or he could
if
as a result
simply default. 12 Or he could try to hold
is
sunk, arguing that the lender needs his
cooperation to get returns on his investment, thereby forcing the lender to
lower the interest rate after the fact. 13 In each of these cases, richer people
will find
it
easier to invest the
part of the
money
at stake
is
same absolute amount simply because a
their own,
which makes them more
have the right incentives. Moreover, a lower
poor by making
10
The
11
This
latter
is
Legros and
it
V_
and a higher a
easier for the lender to punish borrowers
assumption guarantees that no one accumulates an
larger
likely to
will help the
who misbehave.
infinite wealth.
the essential idea of the models of credit in Aghion and Bolton (1997) and
Newman
(1996).
12
As
in
Bernanke and Gertler (1990), Hoff and Lyon (1995).
13
As
in
Hart and Moore (1990).
12
The
would therefore be more or
rest of the story
less
the same as what
we
have here.
Poverty as Vulnerability
3
The poor
are vulnerable: they are afraid of any losses because losses cause
them too much
pain.
same framework
as
We can,
14
we had
it is
and as a
result everyone repays.
they
and
all
the
it is
way up
to
Finally,
that with probability
with probability
As
1
impossible to get away with not repaying,
Everyone can therefore invest as much as
convenient to assume that everyone
/.
—
add some
v *nsky =
q >y*(j
for
invests, invests
by assuming
risk to the investment
he earns
0.
before, the alternative to investing
and to go and work
who
the investor earned a return R' on his investment,
q'
q'
the
before. However, in order to avoid dealing with
the credit market, assume that
like
much
fortunately, capture this idea using
a wage
R +
>
W. Our
^_ J) r
(uj))
+
is
to put the
decision
-
(1
money
in the
bank
maker now compares
q')V*((u
-
I)r(u))
-
E
with
V* safe
1J
This conceptualization
(1979),
and
Newman
(1997), Jacoby
much more
is
more
(1995). It
is
=
V*(ur
+ W),
or less explicit in
Kanbur
also implicit in Banerjee
(1979), Kihlstrom
and
Newman
and Laffont
(1991), Banerji
and Skoufias (1997), Morduch (1990, 1994), Ravallion (1988) and,
applied context, in Walker and
Ryan
13
(1990).
in
a
where r{u)
since they
with
u>
<
is
do not borrow,
/,
we have
investment
fails will
that q'r(u)
=r
Assume
to invest
v *risky =
who have
the (gross) effective interest rate. For those
W
= r.
For those
who have
to borrow,
i.e.,
to adjust the interest rate for the fact that those
not be able to repay at
=
or r{uj)
that
r(uj)
is
all.
Therefore
>
lu
I,
those
whose
must be true
it
r/q'.
high enough that
V(W) >
V. Our agent
will
choose
if:
g V(7(i2'
-
r)
+ lot) +
(1
-
q')V*{ujr)
-E> V{W + ur) = V*
safe
.
(1)
The two
of
u
sides of this equation are
under the assumption that
V
is
drawn
in Figures
3a and b as functions
—
low enough that q'V(I(R'
r))
+
—
(1
q)V-E<V{W).
Insert Figures 3a
The
and 3b
here.
fact that the curves eventually cross is not automatic. It requires
additional assumption which says that there
investment project
is
is
some wealth
worthwhile. In the case where
E=
level at
0, this
an
which the
assumption
is
equivalent to assuming that the investment would take place in the absence of
any
risk, i.e.,
q'R
>
r.
This follows from the fact that our
utility functions
have built-in constant relative risk aversion which means that
for uj large
enough, the person must be almost risk neutral with respect to fixed absolute
risks.
The more rewarding
continue to dominate for
risky investment should then
all
dominate and should
larger wealth levels. In other words,
14
when
E=
0,
there
is
a
a;
such that those above
more general case where
E^
u
take the risk and the rest do not. In the
0, this is
not necessarily true: because richer
people also value their leisure more relative to extra income,
may be
it
may
the richest people do not want to invest. In other words the curves
what follows we
again. In
will ignore this effect,
that no one in our population
What
is
enough
the effect of an increase in
a change in V_ only affects
the project
rich
is
fails, i.e., if
u
if
V_.
u
However,
y*nsky curve can on iy cu ^ fa e y*safe curve
that V(ujr)
>
V. 15 Therefore,
if
effect
on u.
A fall
in risk aversion
implicit assumption
To answer
it is
possible to
rom below
f
u and
this note that
expects to end up at V_
the configuration
the V_ constraint can never binds at
cross
for this effect to matter.
V has on up.
the person at
V(ur) <
on the
that
is
as
when
show that the
if uj is
large
shown
in
enough
Figure 3a,
therefore changes in V_ has no
does however encourage people to take the
risky option as one might have expected. 16
15
Intuitively,
as V{uir)
<
V* T%shy grows more
V. In the case where
dV '^" v
condition
I(R'
16
— r) > W, V
The proof
option
is
=
is
slowly with respect to
u>
—
I
>
0, this follows
q'V'(I(R' -r)+ur) which
concave and
q'
<
1.
is
less
u>
compared
from the
dV
than
'Z''
less obvious, since
it
V* sa f e
=
fact that
as long
under this
V'(ur + W) because
In the alternative case a similar argument applies.
away from the
of the fact that increasing risk-aversion shifts people
more or
to
risky
amounts to saying that the certainty equivalent
has gone down. However, to do this exercise correctly, we must ensure that the fall-back
option remains
also change
E=
and
V
and the marginal rate of substitution between income and
when a changes. For
also that
when a changes A
more general case where
E ^ 0,
when we look
at this question
gets adjusted to ensure that
A
f_ a
similar effects are to be found but there
to control for the direct effect of
and
this reason,
a on the marginal
effort.
15
effort
is
=
do not
we assume
V_.
In the
no simple way
rate of substitution between income
However, this
of V_
V_
not the only possible configuration.
may be encouraged
the poor
,
is
to invest because the distance between
and what they would get otherwise,
V(W + cur), may
obviously true, for example, in the case where
it
must pay to
invest.
In this case,
For higher values
if
be small. This
V(W + uir) <
is
V, in which case
anyone does not invest
it
will
be the
V (W + cor) — V may still be
middle
classes,
who can
large).
This
the configuration shown in Figure 3b. In the case described
by that
is
still
lose a lot (for
two
figure, there are
them
critical values
-
cji
and
yj 2
below u\ and above U2 choose the risky option. In this
raise
u_\
without affecting
in the previous case),
(f° r
the same reason that
and therefore increases the share
In this case as well
risk.
u2
it
remains true that
Those who are
-
case, raising V_ will
it
does not affect
u
of people taking the
less risk aversion
encourages
investment.
To summarize:
Claim 2 For
very low values ofV_, the poor will not invest and the rich
For higher values of
middle class
i.
ii.
may
still
V_,
Once
some
sections of the
hold out for not investing. Also:
a higher V_ leads to
more
the poor will also invest but
will.
more investment;
risk aversion leads to less investment.
again, these are largely familiar points.
The argument
that risk
aversion leads to underinvestment and that insurance helps promote invest-
ment goes back
at least to Stiglitz (1969).
Morduch
(1990) provides
some
empirical evidence based on Indian agriculture suggesting that poorer people
do shy away from adopting profitable but risky technologies, such as
high-yielding varieties of crops.
16
We
There are also important caveats.
does not get insured away. Once
we
have not modelled why the
allow for such insurance,
who do not
that the poor will be the ones
it is
risk
not clear
invest, since, as pointed out in
Banerjee-Newman (1991) and more graphically
in
Newman
(1995), the poor
get the best insurance precisely because they are so vulnerable.
The evidence from many developing
countries shows that the poor are of-
17
ten well insured against the kinds of risk they face.
however, that our interpretation of this evidence
that
we only observe the kinds
words,
we cannot
Moreover, there
is
in order to limit
constrained by the fact
may have
foregone invest-
the risk they actually have to bear. 18
some evidence that there
kind of insurance that
must be emphasized,
of risk people have chosen to bear. In other
control for the fact that people
ment opportunities
is
It
is
substantial variation in the
available to people in different areas.
is
19
This
is
consistent with the fact that these informal systems of insurance have to be
self-enforcing,
It is
and
self-enforcing systems tend to
straightforward to look at
of this type. Focusing
enough
(i.e.,
have
uj t
>
all
be quite
fragile.
the wealth dynamics implied by a model
on the case where people invest only
u),
we
if
they are rich
have:
ur t+ i
=
j3[iu t r
+ W]
ut+
i
=
/3[(uj t
—
I)r(u>)
LUt+\
=
P[{oJ t
—
I)r(u)]
17
See Udry (1990), Townsend (1994),
18
As pointed out by Morduch
19
See Townsend (1995).
for u) t
+
<
IR'}
lu
with probability
with probability
Morduch
(1990).
17
(1995).
1
—
q'
q'
Since
u —
I can clearly be negative and
t
when u t — I
bequests, assume that
These dynamics are shown
that
f3r
<
is
in Figures
out negative
like to rule
negative the bequest
is 0.
4a and 4b under the assumption
1:
Insert Figures 4a
AA
The curve
curves
we would
BB
and 4b
here.
represents the dynamics for those below u, while the two
and B'B' represent the dynamics
above
for those
it
(the
two
curves represent the two outcomes).
Using standard techniques, the reader ought to be persuaded that
ure 4a, the poor eventually converges to the point
lu*
in Fig-
while the rich converge
to a distribution with support [t^,^]. In Figure 4b, the steady state
single distribution. Figure 4a, then,
fact that the
poor
happens when
4
might be
obviously more likely
low and a
The Two
Both views
it
V_ is
is
is
when u
is
high,
configu-
which
is
what
Poverties
inefficient.
different, consider the
1
The
high.
of poverty give reasons
3 and 4 of Claim
a
a poverty trap driven entirely by the
vulnerable and therefore underinvest.
feel
ration in Figure 4a
is
is
They
why
poverty
may be
persistent
are nevertheless very different.
statements of Claims
and Parts
1
and
2 of
1
and
2.
Claim
To
and why
see
In both of these (Parts
2)
we
relate V_
and the
extent of risk aversion to the extent of underinvestment, and through
18
how
it
to
the persistence of poverty. However, the effects go in exactly the opposite
When we
direction in the two cases.
of V_
and low
risk aversion
emphasized desperation, a high value
were both bad.
If vulnerability is
what we care
about, the same things are both good.
This ought not to surprise us: being vulnerable, after
literal
opposite of having too
little
to lose.
While
it
all, is
almost the
not impossible for
is
someone to be both vulnerable and desperate (because she
faces very different
patterns of risk in the two situations), emphasizing one aspect of poverty will
tend to make the other
one's vulnerability
is
less relevant.
To make matters
worse, the extent of
directly related to one's ability to
borrow to smooth
out short-run income fluctuations. 20 Since we have argued that what makes
one vulnerable
about who
is
may
different views
anti-poverty policy.
If
raises questions
cannot but suggest rather different views of
the poor are vulnerable, they will want to be protected
risk (a high value of
credit.
one better credit access, this
really vulnerable.
These two very
from
also give
V) but that could make
Conversely, lowering V_
makes borrowing
it
harder for them to get
easier,
it
also
makes them
more vulnerable.
Trade-offs, of course, are the bread-and-butter of economics.
this particular trade-off interesting
is
What makes
that while both the idea that the poor
are vulnerable and the idea that they have limited access to credit are very
much
20
in
in the literature, the trade-off
between
policies addressed
toward them
We have avoided this issue so far by having only very long-run fluctuations
our model
fluctuations
is
is
on the scale of a generation). But
an important part of vulnerability.
19
(everything
in the world, sensitivity to short-run
is
not discussed.
It is
not mentioned, for example, in the otherwise excellent
survey of the literature on poverty and anti-poverty policy by Lipton and
Ravallion (1995), despite the fact that the survey has a place for both the
idea that the poor
shy away from
may be
risk,
credit-constrained and the idea that the poor
may
and moreover has a discussion of the possible disincentive
effects of anti-poverty policies.
Of
is
possible that these concepts are ignored because they
are not very useful.
But both introspection and casual empiricism suggests
course,
it
otherwise. Moreover, both these ideas figure importantly in the fascinating
evidence on
how
the poor view themselves, emerging from the recent work
on participatory poverty mapping. 21
It
remains possible that there
people and the other to the
no
is
rest.
real conflict:
If this is true, it
one view applies to some
suggests that
it is
very
important that we find observable correlates of these poverty characteristics.
Alternatively, the conflict
may have
arisen from the specific formalization
have adopted here and a reconciliation may be possible.
This
is
we
what we
turn to now.
21
See Narayan et
ways
in
al (1999). It is,
however, also clear from this evidence that there are
which poverty gets conceptualized that are not
of this include the idea that poverty
is
voicelessness
powerlessness (lack of control over one's
own
leisure)
20
in
(i.e.,
destiny)
the formal literature. Examples
lack of control over public action),
and
stress (lack of ease, lack of
Conclusion: Reconciling These Views
5
The
close connection
fact that
both are related to
welfare level.
V_
between desperation and vulnerability comes from the
One way
V_,
which
to break the link
is
corresponding to the two views of what
there
one value of V which
is
which
tells
One
tells
why
minimum
to have two separate values of
it is
to be poor. In other words,
these outcomes
who simply had
may be
choice by the lender, while the misery of
not need anybody else's help.
the lender
costly,
after her with
is
amount
large. In other
expect to be pushed
In other words, the
all
of
V_ reflects
an active
someone whose investment has
If collecting
money from a
failed
defaulter
is
This will be especially true
money
if
the borrower
that can be extracted from her cannot
words, at least in expectation, the borrower will not
the
way down
to the
minimum
socially accepted level.
V that is relevant for the credit relationship is higher than
when we
focus on risk-taking behavior: the poor
when they
take on risks they have no control over
the V_ that comes into play
could be very vulnerable
and yet be protected from the wrath
it
by thinking
not have the incentive to go after her, or at least go
enough gusto.
poor, since the
be very
may
very bad luck.
different is suggested
about incentives. The fact that a defaulter ends up at
may
socially accepted
us what happens to a defaulter and another
us what happens to someone
reason
the
is
of the lender because the lender finds
too costly go after them.
A
related but less obviously economic
society
may
argument
stresses the fact that
take different views of misery that people bring upon themselves
and misery that others
inflict
upon them. Bankruptcy laws around the world
do not allow creditors to attach someone's
21
last
bowl of food, and yet many
of the
same
countries do not explicitly guarantee that no one will end
may
starving. This kind of moral schizophrenia
Both these arguments, however,
tinction between
bad luck and
not having enough
money
bad luck unfolds
through a
is
two
levels of V_.
on making an excessively sharp
rely
Bad
default.
also give us
up
dis-
luck for a farmer or a trader
to pay their suppliers.
series of defaults.
In other words, the
The two modes
is
way
of poverty
therefore remain connected, albeit perhaps less tightly.
Another approach to
this question
to recognize that people
is
different behavioral responses to the risk that
is
something they choose) and the
derestimate
of
how much
it
may
comes from defaulting (which
risk,
they
By
cost them.
may
have a tendency to un-
contrast,
where
it is
God, they may even overestimate the dangers that they are
result, despite
Where they
risk inherent in investments.
have control and actively choose the
may have
the real V_ being the same, the agent will act as
a pure act
facing.
if
As a
she had two
separate V's for the two decisions.
These more elaborate models do allow us to get away from the very stark
we
conflicts
outlined above. However, these models also have important pol-
icy implications of their
and the problem
is
uwn.
If,
for
example, the poor are actually vulnerable
commitment on the
lender's side, the natural policy re-
sponse would be to try to reduce the costs of collection.
This
may be an
advantage, for example, of group lending schemes that shift a part of the
burden of collection on to those who can
group).
If,
on the other hand, the problem
the important step
to repay.
easily
Dynamic
may be
is
do so (other members of the
in
the borrower's perceptions,
convincing the borrower that
it is
in her interest
incentives for borrowers like the ones built into
22
many
micro-credit schemes
may
All this, of course,
tell
is
serve this purpose.
necessarily speculative. In the end, only data can
us whether any of these models are right or even interesting. However,
without the theory being articulated, there cannot be data. Our hope
have taken the
first
is
to
step on this question.
References
Aghion,
P.
and
P.
Bolton (1997), "A Theory of Trickle-Down Growth and
Development", The Review of Economic Studies, Vol. 64
(2),
pp
151-
72.
Banerjee, A. and E. Duflo (1999), "Inequality and Growth:
What Can
the
Data Say?", mimeo, MIT.
Banerjee, A. and A.
Newman
(1991),
"Risk-Bearing and the Theory of
Income Distribution", The Review of Economic
Studies, Vol.
58
(2),
pp 211-35.
Banerjee, A. and A.
Newman
(1993), "Occupational Choice
of Development", Journal of Political
Economy,
and the Process
Vol. 101 (2),
pp 274-
98.
Banerjee, A. and A.
Newman
(1994),
"The Role of Poverty
in
Development
Theory", American Economic Review Papers and Proceedings, May,
pp. 211-215.
Banerji, S. (1997), "Financial Intermediary, Savings
and Investment
Macro Model with Imperfect Information" mimeo, Boston
,
23
in a
University.
Bernanke, B. and M. Gertler (1990), "Financial Fragility and Economic
Performance", Quarterly Journal of Economics, 105, pp 87-114.
Dasgupta,
trition
P.
and D. Ray
(1986), "Inequality as a Determinant of
and Unemployment: Theory", Economic Journal,
Vol.
Malnu96,
pp
1011-1034.
Galor, O. and
J.
Zeira (1993), "Income Distribution and Macroeconomics",
Review of Economic Studies, Vol. 60
(1),
pp
35-52.
Ghatak, M., M. Morelli and T. Sjostrom (1998), "Dynamic Incentives Occupational Mobility, and the American Dream", mimeo.
Hart, O. and
of
J.
Human
Hoff, K.
Moore
(1994),
"A Theory
of
Debt Based on
Inalienability
Capital", Quarterly Journal of Economics, 109,
and A. Lyon
(1995), "Non-leaky Buckets:
pp 841-879.
Optimal Redistributive
Taxation and Agency Costs", Journal of Public Economics, Vol. 58,
pp 365-390.
Jacoby, H. and E. Skcufias (1997), "Risk, Financial Markets, and
Human
The Review of Economic
Studies,
Capital in a Developing Country"
Vol. 64 (3),
Kanbur,
pp
,
311-35.
"Of Risk Taking and the Personal Distribution of
S. (1979),
come", Journal of Political Economy, Vol.
Kihlstrom, R. and
J.
Laffont (1979),
87,
pp 769-797.
"A General Equilibrium Entrepreneurial
Theory of Firm Formation Based on Risk Aversion" Journal of
,
ical
Economy,
Vol. 87,
In-
pp 719-48.
24
Polit-
Legros, P. and A.
Newman
(1996), "Wealth Effects, Distribution,
and the
Theory of Organization", Journal of Economic Theory, Vol. 70
(2),
pp
312-41.
Lipton,
M. and M.
Ravallion, (1995), "Poverty
and Policy",
Development Economics, Volume 3B. Behrman,
Amsterdam;
(eds.).
New York and
Oxford:
J.
in
Handbook of
and T. Srinivasan
Elsevier Science, North
Holland, pp 2551-2657.
Loury, G. (1981), "Intergenerational Transfers and the Distribution of Earnings", Econemetrica, Vol. 49,
Morduch,
J.
pp 843-867.
(1990), "Risk, Production
and Saving: Theory and Evidence
from Indian Households" mimeo, Harvard University.
,
Morduch,
J.
(1994), "Poverty
view, Vol. 84 (2),
Morduch,
J.
(1995),
and Vulnerability", American Economic Re-
pp 221-25.
"Income Smoothing and Consumption Smoothing"
Journal of Economic Perspectives, Vol. 9
(3),
pp
103-14.
Narayan, D., R. Patel, K. Schafft, A. Rademacher and
(1999),
Koch-Schulte
"Can Anyone Hear Us? Voices from 47 Countries",
of the Poor, Vol.
Newman,
S.
1,
Poverty Group,
PREM, World Bank
in Voices
(forthcoming).
A. (1995), "Risk- Bearing and 'Knightian' Entrepreneurship" mimeo,
,
Columbia
University.
Piketty, T. (1997),
terest
"The Dynamics of the Wealth Distribution and the
Rate with Credit Rationing"
25
,
In-
The Review of Economic Studies,
Vol. 64 (2),
Ravallion,
M.
ability",
pp
173-89.
(1988),
"Expected Poverty under Risk-Induced Welfare Vari-
Economic Journal,
Vol. 98 (393),
Sen, A. (1999), Development as Freedom,
Shultz, T. (1964),
New
pp 1171-82.
York: Knopf.
Transforming Traditional Agriculture, Yale University
Press.
Stiglitz, J. (1969),
"The
Effects of Income, Wealth,
ation on Risk- Taking"
(2),
,
and Capital Gains Tax-
The Quarterly Journal of Economics,
Vol. 83
pp 263-83.
Townsend, R. (1994), "Risk and Insurance
Vol. 62 (3),
in Village India", Econometrica,
pp 539-91.
Townsend, R. (1995), "Financial Systems
in
Northern Thai Villages", The
Quarterly Journal of Economics, Vol. 110
(4),
pp 1011-46.
Udry, C. (1990), "Credit Markets in Northern Nigeria: Credit as Insurance
in a
Rural Economy", World Bank Economic Review, Vol. 4
(3),
pp
251-69.
Walker, T. and
J.
Ryan
(1990), Village
and Household Economies
in India's
Semi-arid Tropics, Baltimore and London: Johns Hopkins University
Press.
26
o
y* dishonest
Q.
y*honest
V
/(«>)
/
Figure la
4-.
"W*honest
o
>>
a.
Y* dishonest
V
/
Figure lb
J
l+1
Figure 2a
CO
1+1
P[l(R-r) + cur]
Pco[X{R-r) + r
CO,
Figure 2b
00
1+1
Figure 4a
<°l+l
B'
-r) + wr]
B'/
\
I
j3[W+cor]J^
l
/
//
/
/v
5°
y
B
{0)-~l)r^
'
i
i
i
L
01
1
Figure 4b
(0,
561*5
'^25
Date Due
Lib-26-67
MIT LIBRARIES
3 9080 02246 2730