MIT LIBRARIES
3 9080 02246 1187
'1
H^*
Digitized by the Internet Archive
in
2011 with funding from
Boston Library Consortium IVIember Libraries
http://www.archive.org/details/isliteracysharedOObasu
DEWEY
HB31
.M415
r\o
ol-
Technology
Department of Economics
Working Paper Series
Massachusetts
Is
Institute of
Literacy Shared Within Households?
Theory and Evidence from Bangladesh
Kaushik Basu
Ambar Narayan
Martin Ravallion
Working Paper 01-44
October 2001
Room E52-251
50 Memorial Drive
Cambridge, MA 02142
This paper can be downloaded without charge fronn the
Social Science Research Network Paper Collection at
http://papers.ssrn.com/abstract id=xxxxx
•^
Is literacy
shared within households?
Theory and evidence from Bangladesh
Kaushik Basu', Ambar Narayan^, and Martin Ravallion''
May, 2000; revised October, 2001
Abstract:
A member of a collective-action
of literacy with others
offset
by
we
may
or
may
not share the benefits
household; the shared gains from doing so
a shift in the balance of
for Bangladesh
Holding
in that
household
power within
the family.
may
well be
Using household survey data
find strong external effects of education on individual earnings.
a range of personal attributes constant, an illiterate adult earns significantly
in the
non-farm economy when living
These
effects are strongest,
ruled out, but
would
family with
at least
one
literate
member.
and most robust, for women. Omitted-variable bias cannot be
also be consistent with an intra-household externality of literacy.
Key words: Household
JEL classifications:
in a
more
behavior, literacy, externalities, earnings, Bangladesh
DIO, J24, J31
'Department of Economics, Massachusetts Institute of Technology, Cambridge,
02142, and Department of Economics, Cornell University, Ithaca, NY 14853
MA
2,3
World Bank, 1818
H Street NW, Washington DC, 20433
Acknowledgements: These are the views of the authors, and should not be attributed to the World
Bank. The paper is an output of Kaushik Basu's World Bank research project, 'Literacy and Child
Labor'
(RPO
683-07).
1.
Introduction
This paper asks whether an ilHterate worker's labor earnings are affected by the educational
members of that worker's household. The
attainment of other
very
little
large literature on literacy says
about intra-household externalities of literacy and education. Yet one might well
expect that one person's literacy will have spillover benefits for others within the same
household. In their study of the propensity to innovate
and Nesman (1985) observe
how having
among Guatemalan
a literate family
member conveys many
of being literate oneself. Likewise, Dreze and Saran (1995) note
can spread to others
in the
farmers, Green, Rich
how
of the benefits
the advantages of literacy
household by virtue of certain kinds of decision-making on behalf of
the household shifting toward the literate.
It
person
can be argued that an
is literate
will
illiterate
be better off than an
(Basu and Foster, 1998). For certain
does not become
person
literate this
who
illiterate
activities, all
way, but
it
lives in a
person
household
who
one needs
is
which
in
lives in a
one
at least
household of illiterates
access to a literate person; one
eases the handicap. For instance, an illiterate farmer
new
occasionally receives pamphlets from extension workers or needs to read the label on a
fertilizer
An
packet could benefit enormously from easy access to a literate person.
worker whose work requires some knowledge of what
plenty of 'commonsense'
— which comes from
within the family?
empirically,
is
One
namely
in the
world
there that literacy, or education
literate
more
urban
at large
knowledgeable people
Of course, employers who
pay more for someone coming from a
What evidence
happening
interacting with
benefit a lot from having a literate person at home.
will be willing to
is
who
or just
— might
are aware of this,
household.
generally, has an external effect
external effect of education within families has been well researched
the effect of parental education on child welfare. There
is
a large
body of
evidence indicating that children's health, nutritional status and educational attainments are
enhanced by having better educated parents, particularly the mother.' However, sharing
knowledge with children
is
more obvious than amongst
adults,
which
will also
depend on how
sharing knowledge affects other aspects of decision making within the family.
There
is
also evidence that a farm-household's total
income depends on the highest
education level reached, rather than the mean, or the education level of the household head
(Foster and Rosenzweig, 1996; Yang, 1997a,b). This does not, however, imply an external effect
of one person's
—
it
human
may simply
More
capital
reflect gains
direct tests can be
work on earnings
it
is
on the productivity of another person within
that
same household
from a division of labor within the family.
found by looking
commonly assumed
that
individual labor market earnings. In empirical
at
own wages depend
on
own
characteristics
and not
family characteristics directly. Indeed, the exclusion restriction that such external effects on
earnings do not exist has been used to
regressions (Alderman et
1996).
al.,
positively to wife's schooling in the
and parents-in-law
in Brazil
external effect or the
background
'
An
effects
test for the
However, there
US (Benham,
(Lam and Schoeni,
outcome of marital
sorting.
on public sector wages
important early contribution was
endogeneity of human capital
in
wage
evidence that men's earnings respond
is
1974) and to the schooling of the wife, parents
1993).
It is
unclear whether this
is
a direct
Krishnan (1996) finds strong family
in Ethiopia,
which she
attributes to a selection effect.
Behrman and Wolfe (1984) who demonstrated
the importance of
family background (including parental schooling, notably of the mother) to children's schooling in
Nicaragua. In a recent example, Gibson (2001) finds strong effects of adult literacy on children's
nutritional status in
Papua
New
Guinea. In the same Bangladesh data
set that is
analyzed here there
is
evidence that children of better educated parents are more likely to attend school (Ravallion and Wodon.
2000). For a survey of the evidence on these effects see Strauss and
^
If there are
Thomas
(1995).
indeed external effects on earnings then family characteristics are not valid instruments for
endogenizing individual human
capital.
whereby family background influences access
the
wage
to public sector jobs; correcting for selection bias,
effect vanishes.
We aim here to test for external effects for both
turn deliberately to a developing country setting in
men and women
which there
educational attainments, in the expectation that this will
effects of education.
We examine the efficacy of men
externality of literacy.
It is
within the family, and there
is
been any research on whether
women
in the
women
illiterate
men?
it
easier to identify any external
women
as recipients of the
worse as the recipients of
externalities.
Does
household yield significantly greater externalities on
Before turning to the formal analysis,
larger debate
are large disparities in
are better at generating positive externalities
are better or
member
than on
versus
We
Bangladesh.
supportive evidence.^ However, there does not appear to have
the presence of a literate
illiterate
women
believed that
make
in rural
we
should clarify that
on the inter-relationship between the concepts of
we
are staying
literacy,
knowledge. These are of course distinct concepts, though sharing a
away from
the
schooling and
common
core.
Once one
goes into the distinctiveness of these concepts, one could argue that a minimal level of education
enables a person to receive externalities better. But these are complexities that
scope of this paper. Here literacy
is
equated with the
ability to read,
though
lie
we
beyond the
will look at the
effect of different years of schooling.
We begin by examining a key theoretical question:
Given
so chooses, prevent others from benefiting from her literacy,
could
'
It
be that
men and women have
A number of studies
others in the household
Strauss and
Thomas
why does
different propensities to share?
have suggested that
more than do men;
(1995).
that a literate
women
member can,
if
she not do so? Moreover,
The following
section
share their education and other advantages with
for further discussion
she
and references to the
literature see
model
constructs a theoretical
to
answer these questions. The paper then
within households. Section 3 describes our data for Bangladesh. Section 4 describes the
regressions that
selectivity bias
regressions,
we
we
estimate across
by standard means.
illiterate
sharing
tests for literacy
wage
sub-samples of the workforce, correcting for sample
In addition to reasonably standard vanables for earnings
include measures of the highest education level in the worker's household.
stratify the regressions
by gender and employment sector
We
(subject to data limitations) in the
expectation that external effects will be stronger for certain types of jobs. Section 5 presents the
results,
and section 6 consists of concluding remarks.
available
2.
from the authors.
Why
It is
A detailed Statistical Addendum is
might there be external
effects of education within
households?
not obvious on a priori grounds that a literate person will share the advantages of
members of
literacy with the illiterate
large, non-rival but excludable.
brochure for an
illiterate
They
the
same household. The
benefits of literacy are, by and
member
are non-rival because for a literate
or to give the illiterate a piece of information
is
to read a
virtually costless. In a
household one can, for instance, exchange such information over dinner or while doing
household chores. They are excludable because a
part with
any information.
It is
true that
some of
watching, for instance, what he does with the
can block a
with other
In
lot
of the benefits
he wants
to.
member can,
if
he so chooses, refuse to
the benefits can be acquired
new
fertilizer.
by osmosis
But nevertheless the
So why would we expect him
literate
to share
— by
person
knowledge
members of the household?
answering
this,
matter. Interestingly,
making
if
literate
it
we
is
(see, for instance,
could directly appeal to altruism and
possible to go deeper.
The new
treat that as the
literature
end of the
on household decision-
Udry, 1996; and Quisumbing and Maluccio, 1999) has
lots
of disputes.
but one proposition generally agreed upon
in
most households. There
are the
from anybody's nse
benefit
in
there
is that
is
some pooling of resources and earnings
household public goods
like the
common
kitchen, which
income. Then there are the shared tasks, such as one person
belonging to the household or hired by the household going to the market to shop for everybody.
This means that
if
individual
/
can raise individual/s income by conferring some of his literacy
advantage loj, a part of the gain will
come back
to
/
through the sharing process. However, one
has to balance against this the fact that a higher income will
may
expenditure pattern of the household.
alter the
and
literate
literacy
illiterate
If
is
In our
members and
people then a shift in power within the household associated with shared
advantage can work against the interests of the
it
illiterate
preferences differ sufficiently between
literate
naturally be disinclined to share the advantages of literacy.
effects
empower the
To
family member,
who
get a grip on these conflicting
useful to construct a model.
model, the household collectively chooses
are free to restrict illiterate
members choose
utility gains,
members from enjoying
to help illiterate
which
in turn
its
consumptions, but
literate
members of the same household depends on
depend on how sharing the
literacy
advantage
a positive
income
potential of the illiterate
view of the
literate
power within
effect of sharing literacy
members. The second
is
members
Whether
the advantages of literacy.
literate
their personal
alters the level
composition of household consumption. This depends on two potentially opposing
first is
will then
and
effects.
The
on consumption, through higher earning
a potentially negative effect (from the point of
members) on the composition of consumption,
the household, given differences in preferences.
via effects
How
these
on the balance of
two
effects balance out
determines whether there will be an external effect of literacy within the household.
Let us, for simplicity, assume that a household has two members, one literate and one
whose
illiterate,
two goods.
literate
If
and
.r
utility
functions aref/^(.) and
the vector of
is
members
illiterate
U ^,(.)
.
respectively.
We
assume
goods consumed by the household, then the
are given
by
U ^(x)
and f/y
(.v)
It is
.
that there are
utilities
assumed
of the
that both utility
functions are strictly increasing and quasi-concave.
The two goods can
tobacco and
skirts.
who consumes
1
it.
Or
interpreted in
the
We may think of the two
two goods may be the same commodity
Thus we could think of apples consumed by
woman
and apples consumed by the
private goods.
more than one way.
The
literature
as
good
2.
man
of the household as good
on household decision making often allows
on individual
utility
by the person
Either way, both goods are assumed to be
goods
the household. In the presence of household public
transferability restriction
the
differentiated
goods as
it
is
for public
goods within
possible to place a kind of
functions under which the household provision of
public goods ceases to be sensitive to the balance of power (see Bergstrom and
Comes,
1983).
For reasons of simplicity, we stay away from public goods.
We
shall take this to
modeled and analyzed
be a 'collective-action household', both because this has been
in recent
times (see, for example, Bourguignon and Chiappori, 1991,
1994; Moehling, 1995; Basu, 1999) and because of
its flexibility.
The
latter
enables the model to
capture households that meet the hard requirements of the unitary model, as one polar extreme,
and also of households which
Our model assumes
are deeply divisive.
that the
household consumption maximizes
and Uo, the weight X being a function of the
person's income, y^
think of
A.
.
Since the
as a function of y„
.
latter is
It
will be
illiterate
assumed
to
assumed
a
person's income,
weighted average of Ul
y^^
and the
literate
be constant throughout the analysis,
that
A(y^ ) e
[0,1]
and A'(y^
)
^
.
we can
Hence
the household's
problem
is
choose x so as to maximize:
to
X{y^)U^{x)^\\-X{y^WL^A
subject to the budget constraint p.x
(1)
<y^^^-yQ where p =
[Pi
,
P2 ]
'^ '^he
price vector for the
two
goods. Let us denote the solution of this problem by:
^=
4yo)
(2)
(While X depends also on p and y^ these are assumed to be constant and so are ignored.)
,
The
literate
interesting question
is:
what
will
member's decision whether or not
be the value of y^
When
Jo ~
will the literate not
yp'^ ^"^^
^^y
be the period-2 problem. In period
or y,. Clearly, he will set yg
x(j^
)
are given
literate
=yp
and only
member
to share the benefits of literacy.
U ^(x)
member
will the literate set
household's problem described above to
member has
if
when
to
choose between setting y^ = yp
t/^(x(yp)) > ^/^(^(y,)) where x(yp) and
problem such
It is
that
it
evident that
will not exclude the illiterate
will
if
be
in the interest of the
preferences do not differ
member from
the benefits. If
then the household's consumption choice problem reduces to maximizing
subject to p.x
Therefore
the externalities
(2).
identify special cases of this
Uq(x) =
In other words,
to think of the
the literate
if
on the
illiterate if she, respectively, gets the
illiterate?
One can
then the literate
U ^{x)
by
1
is
member from
in part
Of course, yp>y,.
it.
exclude the
of answering this
This will depend
to exclude the illiterate
of his education. Let y p and y^ be the income of the
benefit of literacy and if she does not get
?
< y^ + y^ and
,
U i^{x{yp)) > U ^{xiy^ ))
.
the
maximum
level of t/^ {x) will be non-decreasing in
Another special case
in
which the
literate
member
yQ
will not
.
exclude the
for
illiterate
member from
the benefits
is
when both goods
y^ Then the consumption possibilities of the
all
.
unambiguously higher when knowledge
what
Intuitively,
the illiterate family
member when
U ^(x)
consider a special case:
towards
two
0, the
-
)
and
-
f/f,(x)
x^
-\-
Ax"
assume P2 =i- Thus the household's (period-2) problem
+
Ax^_)
subject to p,X|
This
is
the first-order condition
Inspecting (3)
A{yp ) - A{yj
zero, x^
it
[\
,
be
is
that they
,
To
with
see this
« <
1
.
have similar
much
in
favor of
more formally,
Note
that as
A
let
us
goes
Xj
is
to
maximize
~ Aiyo)]x,
<
+ y^
v^
changes
it
at
is
must
-
PiXX
have:
Pi
obvious that changing v^ from y, to Vp will increase
small or
A
is
sufficiently small.
Of course,
.x,
if
as long as
either of these terms
rise.
that if
most a
means
we
>'o
Pi
is sufficiently
)
Note now
small
+
+
+Aiyo)A{y^ +
'
is
-A
equivalent to the problem of maximizing:
X,
From
)
functions tend to converge. Further, and, without loss of generality,
utility
A{yo){x,
will
does not change too
literacy benefits are shared.
x^
member
to share their literacy
power (captured by A
preferences, or that the bargaining
family
normal and Aiy^
shared within the household.
members
necessary for
is
is
literate
are
little
A(yp)-A{y,)
is
small
it
means
that the structure of
decision-making
m response to changes m the earnings of different members. And if A is
that both
members have
similar preferences; preferences being identical
if
A -0.
From
these interpretations
household.
we
may
It
we
clear that these are conditions that one
is
not be 'unitary', but
have just shown
the hypothesis
it
that
is
proximate
is
it
illiterates will
From
are about to test.
not too divisive either.
If that
within households, that
is
cannot
see
tell
how
us
how
the above discussion
i:
literate
the exact value of
literacy
of
e
[0,1]
,
Otherwise
where
not.
what
illiterates.
This
is
we
are testing for is a certain
expect to see external effects of education
While we hypothesize
yp > y,
that
So an important purpose of our empirical
s,
can be shared.
men and women.
it
is
=
5
means not
to share at all
5.
That
is,
easy to see that a
y^ = y^ (s) with y^ (0) =
,
literate
,
theory
exercise
is
to
=
y,
,
yg
1
means
to fully
where there
(1)
is
= Jp and
how much
to share, that
whom
depend on the nature of the
It is
reasonable to expect basic differences between the utility functions
In that case,
s,
and therefore, y^
externality of literacy, note that if there
is
there can be large natural differences
woman
5
will
utility
(s) will
man
the household. In anticipation of our empirical finding that
Suppose the
and
person's decision on
the benefits of literacy are to be shared with an illiterate
—
that holds,
should be evident that the
the incom.e of an illiterate person in a household
be a function of
From our model
is,
.
we would
normal
is.
Then we would have
someone
y,
right
it
in a
easy to adapt the above model and alter the sharing of literacy from being a 0-1
variable to a variable
share.
>
is
large the gap will be.
large this gap
It is
yp
view
And when
be better off than isolated
hypothesis meets the criterion of falsifiability. In a sense, what
view of the household.
would expect
function of the person with
be different depending on whether
or an illiterate
women
belonging to
are better "recipients" of the
a third person in the household
between men and women
woman
—
a child, for instance
as recipients of externality.
uses increases in her bargaining power in the household to spend in a
10
way
that benefits the child but the
However,
benefits
if
go
the
man
same
in the
man (woman) does
place uses the
power
it
is
own
advantage.
not get benefits for himself, he (herself, she) prefers that the
to the child rather than the other adults in the household.
assumptions,
to his
more willing
natural to expect that a literate person will be
benefits of his or her literacy flow to a
woman
rather than a
Given these plausible
man, since
to let the external
that helps direct
more
advantage towards the child.
For similar reasons,
s
and ypis) can also depend on other factors
that determine the utility
functions of the individuals. These factors include whether the person works in the agricultural
or the non-agricultural sector, and whether he or she belongs to the urban or the rural sector.
What we mean
rural
that there
is
may
be systematic differences
and urban households or households
in the utility
in different sectors
functions that prevail in
we
of the economy. In that case
m_ay find that the externality of literacy depends not only on the gender of the participants but
also on the sector to
which they belong professionally. Theory of course cannot
nature of the differences in 5 and y'pis)
question.
The
definitive
The aim of
show why
.
All theory says
is
that this
is
potentially an interesting
was not
to derive
some
testable hypothesis, but simply to
the sharing of literacy is neither a foregone conclusion (as the unitary
household suggests) nor obviously false (as
is
us the exact
answer has to come from empirical investigation.
this theoretical section
making). Hence, there
tell
in a
model of
the
completely individualistic model of decision
indeed an open empirical question of
how much
"literacy sharing"
occurs within households.
3.
Data
Our data
are
from the Household Expenditure Survey for Bangladesh, 1995-96. This
is
a
reasonably comprehensive nationally representative socio-economic survey of 7420 randomly
11
sampled households (comprising 39,043 individuals)
in
22 regions comprising of 63
including rural, urban metropolitan and urban municipal areas.
61.1% of people
in
Bangladesh over seven years of age are
females). In rural areas, the percentage
is
higher,
The survey
illiterate
districts,
results indicate that
(55.7% of males; 66.6% of
namely 65.4% (60.2% of males; 70.8% of
females). Bangladesh Bureau of Statistics (1998) provides a detailed description of the survey
and summary tabulations of the
results.
complete tabulations of descriptive
As
a
first
step in the analysis,
statistics
it
is
interested in, given that the data reveal
reasons
we
activity.
The
Statistical
Addendum
to this paper contains
on the variables used below for
all
important to consider the kind of earnings
incomes from
are interested in only that part of
income
a
that is
So we exclude incomes derived from personal and
income from own farms and businesses. The reasons
objectives as well as with the nature of data.
most cases should be seen as accruing to the
we
are
wide range of sources. For obvious
earned from some form of productive
individual wealth and assets, from
the categories of incomes to be considered. For reasons that are less obvious,
for this exclusion
have
we
to
also exclude
do with our
Income from own farms and businesses, we
entire
household rather than
of the household. In such a scenario, education of one or more
likely to
sub-samples.
feel, in
members
to individual
members of the household
is
have a direct effect on household earnings and should not be considered a part of the
externality effect that
we
are interested in. Also
due
not possible to apportion them between individual
way, which makes
earnings.
From
apportioned
it
members of the household
it
is
often
in a satisfactory
very difficult to measure the effect of household education on individual
eyeballing the data
among
to the nature of these earnings,
we
various household
get the impression that farm
members somewhat
been added to the income of the head of the household.
12
incomes have been
arbitrarily.
For instance,
it
has often
For these reasons we focus on the following types of earnings: salary and wages received
from regular and/or irregular/seasonal employment,
remuneration. Given our data
To
anticipate our results,
farm labor earnings, and so
type of
we found
that the external effects of literacy are
what we
this is
find
it
agricultural workers will be unaffected
will focus
Addendum. (There
To
we
are too
schooling.
where
illiteracy is
The next
step
is
manual non-farm
For
activities.
sample for
we
this to
identify
be possible.)
all illiterate
defined as being unable to read, irrespective of the level of
to identify the
household each
is
illiterate
that takes the value
1
person belongs
the household
education level of the households. The variable for (a)
dummy variable
illiterate
agricultural laborers in the Statistical
are interested in,
construct variables that measure (a) whether anybody
maximum
following discussion. Given the
agricultural laborers in the
we
confined to non-
a literate person in their household.
is
in less
male
separate out the sub-samples or groups
individuals,
in the
by whether there
also give results for
few female
on
implausible that the productivity of an
Returns to family literacy will presumably appear
completeness, however,
both cash and kind, and professional
quarterly figures for these categones of earnings are used.
set,
work involved, we do not
m
is
is literate
and
and
(b) the
constructed by defining a
one member of the household
if at least
to,
is literate.
Variables for (b) are constructed in the following manner: Using the available figures on
individual educational attainment,
we
define the
educational level attained by the most educated
maximum
to
education level of each household,
maximum
education level of a household as the
member of the
we
household. Once
are in a position to define the
we denve
dummy
the
variables
proxy the educational attainment of each household.
Accordingly, the four educational levels
Class
I
we
consider are being able to read, completed
and above. Class IV and above, and Class VII and above. The
13
dummy
variable takes the
value one
illiterate
persons
into rural
the highest education attainment in the household.
if that is
is
then sub-divided into males and females.
and urban sub-groups, with the
latter
The
entire
The male group
municipal or metropolitan areas. However, the female group
further divided
is
who
group comprising of males
group of
live in
urban
not sub-divided further because
is
of concerns about the sample-sizes.
To summarize, amongst
sector
we
illiterate
consider three sub-groups:
wage
(i)
earners
whose
urban and rural females;
earnings from the non-agricultural sector, and
(iii)
work
principal
(ii)
is in
the
non-farm
males with primary
rural
urban males.
Testing for external effects of education on earnings
4.
There are many ways that
literacy
might be shared with an
could well be contingent factors. For example,
it
illiterate
worker, and there
might be conjectured that the extent of literacy
sharing depends on the precise relationship between the illiterate worker and the literate family
member.
main
We will
test
whether the gender or age of the
however,
analysis,
we
illiterate
worker
is literate,
to the
is
a
member
=
dummy
C(LITj
+ Xjfi +
is
Wj, of
(4)
£j
variable taking the value one
if at least
X,is a (IxAT) vector of control variables for person;
to allow for an intercept.)
The parameter a measures
on the worker's earnings, andy5
£} is
For the
We assume that the labor earnings,
one other person
compnsing other
worker's productivity and hence wages, including experience. (The
error term
matters.
are determined by:
log W^
where LITj
family
focus on whether the presence of literacy in the household
associated with higher earnings for illiterate workers.
an
literate
assumed
to
is
first
in the
household
factors relevant
element of X,
the external effect of the family's literacy
a (Kxl) vector of parameters on worker characteristics.
The
be normally distributed with mean zero and standard deviation o
14
is 1
.
There are three potential problems
classic
problem of estimating wage equations, namely
individuals
who have chosen
We assume
that
where Z^is
M-1
a
for
j-
a
is
for person
a from
assumption
f,.
any selection
then aLlT^ + X^/3 + po^A(dLlT^
tests will exploit this fact.
+Z ^y)
S even
if
age and a
/.
dummy
and only
if
adding extra controls for the
To
element
1
in
is
1
and other elements
model
(5) is interpreted as a
error term »^is standard normal with a
the
maximum-likelihood Heckman
The expected value of observed log earnings
is
the usual inverse Mills ratio.
A
is
a non-linear function.
results will also
make
We
One
is
can
of our
a widely-used identifying
with selectivity, namely that certain household
1
if
X include
age of the earner, square of the
the earner has been to school.
maximum
level of education in
We also test
each household
allow for geographic effects, a series of regional
dummies
are also
depending on the region a particular earner belongs to completes
of independent variables in the
A small number of persons
The
the variables in the vector
included that take the value
first
do not influence earnings given participation.
variable that takes the value
(discussed further below).
list
if
whose
X, =Z, given that
in estimating earnings regressions
sensitivity to
the
bias.
However, our main
wage equation,
the
Here we follow standard practice.
Then we can use
where A
characteristics influence participation, but
In the
force.
(Mxl) vector of parameters. Equation
to coirect for
separately identify
first is
(5)
non-zero correlation coefficient (p) with
model
The
+u^>Q
of the individual's labor force participation decision.
selection
in (4).
one only observes wages for those
that
person/ are observed
(IxM) vector of regressors
regressors and
work
to enter the (paid)
wage earnings
SLIT^+Z^y
are
estimating the test regression
in
wage equation.
our sample have been to school but remain
15
illiterate.
In the selection equation, the vector
Z includes
all
the independent variables in the
equation, along with variables that represent household characteristics that are
determine participation in the workforce. This
latter
deemed
wage
to
group of variables includes dummies for
land holdings, and demographic characteristics including characteristics of the household head.
In the case of females with primary earnings
whether the household
urban males, a
is
rural or
the variables in Z,
The second
in
some
common
to
problem
is that
become
we
way;
literacy is
is
no bias under the
null hypothesis of
this source of selection bias in the
The
third
problem
is
household.
in
We can
literate.
X all
Sample
so
selection bias
not think
it
normally acquired through schooling. And, of
that there
is
no sharing.
(very) effective sharing of literacy;
We choose to ignore the possibility of
following analysis.
our data
— which
due
to possible
endogeneity of
In the present context,
— observed
are correlated with the presence of a literate person
suggest two possible reasons
make
included
may be
On a priori grounds we do
We believe this to be a more serious concern.
recognizes that workers can
is
ratio for identification.
whether there are omitted determinants of an individual's earnings
by the employer but not
in the
become
in estimating our test equation (4) is bias
literacy within the family.
a key question
is
in the case of
area or not
the intra-household sharing of literacy
select illiterate workers.
literate this
for
robustness of our results to including in
cases that previously illiterate workers
course, a necessary condition for this bias
there
test
and exploiting the non-linearity of the inverse Mills
potential
dummies
urban are also included as explanatory variables;
However, we also
could arise from the fact that
is
the non-agricultural sector,
dummy for whether the household is from a metropolitan
as an explanatory variable.
complete
from
why
this
might happen. The
first
reason
a choice about the extent of schooling in their family, such as
through the choice of marriage partner or the schooling choices parents
16
make
for their children.
This will bias our results
if
the choices
made
are correlated with the omitted determinants of
earnings. For example, an illiterate person with high latent ability
be better able to
Note
that,
worker derives
wages
yield higher
to
utility
own
from having
right.
a literate
to the
employer) might
educate her children more.
a
while such sorting effects will lead us to over-estimate
interpreted as literacy externalities in their
illiterate
may choose
spouse, or
attract a literate
(known
The choices made
,
they can also be
reveal that a high
spouse or child, even
if this
wage
effect does not
for that worker.
Another reason
to suspect
endogeneity bias
the presence of literacy
is that
correlated with other measures of the extent of schooling in the family, which
the earnings of an illiterate worker. Possibly
it is
may
may
be
also matter to
not literacy that gives the external benefit, but
schooling more generally. This would be consistent with the proposition that schooling has
external effects on earnings, though
There are obvious
data.
limits to
In pnnciple, panel data
in the error
it
how
would
far
bias our estimates of the effects of literacy per se.
one can address such concerns with cross-sectional
might help, by permitting a
time-invanant individual effect
term of the earnings regression, correlated with family
from the time means, our estimate of the wage
would then be robust
practice,
latent
changes over time
effect of
to the existence of this individual effect
family literacy
literate
member,
In
are also
slow moving, and even time invariant given the relatively short duration of most
household panel data
sets available.
external effects of family literacy.
literacy that
in
deviations
under standard assumptions.
however, family characteristics such as the presence of a
likely to be
Takmg
literacy.
might be found
in
Standard panel data
It is
sets
may have
little
power
for detecting
also unclear whether the short term fluctuations in family
— when
— would
panel data
simply time varying measurement error
a literate uncle visits briefly, for example, or
entail
17
any
real
scope for passing on the
external effect, even
member over the
when
there are large long term external benefits in having a literate family
longer term. There could then be a large attenuation bias in the within
estimator on panel data.
We can, however, perform two tests of the robustness of our estimates of
of LITJ in our cross-sectional data
on the assumption
that
marriage
is
set.
the
In the first,
main way
we
in
influence the extent of literacy in her or his family.
restrict the
which a high wage
on the unmarried people belonging to each group. This
test will
family;
5.
we
test,
endogeneity
illiterate
worker can
in
each case
now based
only
not however be robust to other
illiterate
worker may choose
to
parent or sibling so as to assure that the external gains from
literacy are captured; higher earning ability
In the second
to
unmarried workers,
that the estimates are
forms of selective family formation. For example, an unmarried
literate
to
The wage and probit equations
have the same specifications as before, with the difference
continue living with a
sample
a
we add
may
help assure that this
is
possible.
extra control variables measuring the extent of schooling in the
use the highest years of schooling reached.
Results
We estimate equations
(i)
Women
variable in the
(4)
(5) for the following sub-groups:
with primary earnings from the non-agricultural sector Here the dependent
.
wage equation
estimated on the
and
full
the log of earnings of individuals in this group. In the probit
is
sample of
illiterate
individual's earnings are observed and
sector,
and
(ii)
females, the dependent variable takes the value
if
1 if
an
her primary earnings are from the non-agricultural
otherwise.
Rural males with primary earnings from the non-agricultural sector The dependent
variable in the
.
wage equation
is
the log of earnings of individuals in this group. In the probit
'
estimated on the entire sample of rural males, the dependent variable takes the value
mdividual's earnings are observed and
sector,
and
if
his
pnmary earnings
if
an
from the non-agncultural
are
otherwise.
Urban males The dependent variable
(iii)
1
.
wage equation
in the
is
the log of earnings of
individuals in this group. In the probit estimated on the entire sample of urban males, the
dependent variable takes the value
In each of these cases, three
if
1
an individual's earnings
models
household-education variable, a second
and a
- one with
which the sample
is
summarizes the
1
results.
and participation probits used
We
restricted to
to
(The
Addendum
conect for selectivity
gives
full details
unmarried workers,
women, and
in a
household w ith
the effect
earnings to be
65%
is
illiterate
member adds 0.50
highly significant. This implies that
higher
the sample to unmarried
a literate
if
someone
women;
is literate in
we
this case the
find the earnings gain
this group.
In the probit
takes the value
1
if
which
in the
is
at least
is
is
one
is
can expect their
robust to restncting
literate
member has
even greater
wage equation
is
about double the
also illiterate.
in the
sample of unmarried
It is
women
not clear, however,
from the agricultural
sector; in
estimated on the entire sample of rural males, the dependent variable
otherwise.
19
if his
is
why
the log of earnings of individuals belonging to
an individual's earnings are observed and
agricultural sector, and
women
This effect
also gives results for rural males with primary earnings
dependent variable
working
to log earnings of illiterate
not suggestive of a bias due to selective sorting through marriage.
The Addendum
illiterate
indeed, the gain in log earnings rises to 0.74. This implies that
female worker from a family with
fact that
wages of
illiterate
their family.
earnings as one from a family where everyone else
The
on the wage regressions
bias.)
find a large and significant effect of family literacy on the
women. Living
'
literacy as the only
third with controls for other levels of education.
Table
an
in
are estimated
observed.
is
primary earnings are from the
would be stronger
the effect
for illiterate unmarried
of one possible explanation.
power
in the
from
the gains
effect
remains and so the
literacy with the illiterate
Women's
worker
women may
unmarried
literate
family
well have very
member
sectors of living in a household
they
and the
is
where
at least
one other person
a family
where
at least
one person
illiterate
is literate.
to restricting the sample to unmarried workers; the effect
For urban males the
men
illiterate rural
household, although the effect
is
literate
by living
tests
variables
was
female (the
for
left
non-farm
in log
15%
on average
However,
no longer
rural
higher
if
this effect is not robust
statistically significant.
is
is
significantly higher if they live in a
barely significant
at
the
5%
level.
illiterate
Nor
men
is
is
this effect
not changed
in a literate family.
do not distinguish the external
family member.
gains from literacy,
are
and
The gain
is literate.
found for unmarried men. The labor-force participation of urban
The above
and unmarried women.
the urban
men
illiterate
effect vanishes.
Labor-force participation of
significantly
bargaining
when an
significantly lower
effect is similar for married
about 0.14-0.15, implying that earnings of
come from
literate
suggestive
has a strong incentive to share
We find significant gains to illiterate males working in both
is
little
is
unmarried woman.
labor force participation, by contrast,
lives in a literate household,
earnings
model
theoretical
household even when they bring extra wage earnings into the family. Thus, only the
income
positive
Illiterate
women. Our
we
To
effect of having a
test the possibility that different
re -ran the tests including
whether the
out category
literates are
was
two extra
male versus female
genders generate different external
dummy
variables.
One
of these
only male, and one was for whether they are only
the case in
family members). The coefficients on these
which there are both males and female
dummy
20
literate
variables were small and (individually and
jointly) insignificant in almost
all
wage
the
variable for "male-only" literates in the
was -0.361 suggesting an
only
men
are literate;
Similarly,
we
The one exception was
regressions.
womens' wage
however, the coefficient was only significant
tested for whether age of the literate
variables for whether
women
to
were under
all literates
16, or
These were mdividually and jomtly insignificant
member
workers
at
dummy
which there the coefficient
regression, for
gam
attenuation of the external
the
the
matters.
m households where
8%
level (t=1.76).
We added dummy
over 16 (again the mixed case was omitted).
in all
of the
wage
regressions.
We also tested how robust these results are to the specification of the variables
included as controls
m the wage regression
we have imposed
the most contentious restriction
earnings via
effect
its
on the correction term
not have a direct effect on labor earnings.
include
dummy
(comprising the vector
is
that
equation
The
final
(4)).
Possibly
household landholding only influences
for labor force participation,
column
variables for landholding class in the
family literacy are similar to the
X in
be
to
wage
in
Table
1
regression.
i.e.,
landholding does
gives the results
The estimated
when we
effects of
model.
first
Table 2 presents the main results from the wage and participation regressions which also
included controls for other schooling
same
Table
as
used for Table
2.)
The
1.
The Addendum
literacy effect
on earnings
effect of high education levels.
labor-force participation
For both
still
^
present
rural
is
Similarly to Table
The
found
household. (Otherwise the estimation method
is still
strong for
1,
was
And
there
is
the
a very strong
on female
due to higher schooling rather than literacy per
controls for other family
the effect
women.
significant negative effect of family literacy
to be
is
also gives full details on the regressions underlying
and urban male workers, the positive
when one adds
Addendum from
in the
effect of family literacy
members' schooling, but
also stronger for unmarried
the authors.
21
women;
it
is
se.
on earnings
not
details available in the
is
significant.
Instead
we
get a significant positive effect of family schooling. For rural males,
participation in non-farm
work
is
not changed significantly by family literacy; for urban men,
higher levels of family schooling attenuate their labor-force participation.
we
So, while
earnings, this
is
find
some
signs that literacy in the family increases an illiterate
not robust to adding controls for higher levels of education in the family. This
not of course inconsistent with external effects of education. However,
whether
it
The
that
we
is
man's
the fact of literacy that
is
it
is
does lead us to question
driving the results for males, or something else.
external effects of family literacy and schooling on the earnings of illiterate
find are very strong indeed, and
much
stronger than for men.
Why
might
this
women
be?
We
can suggest the following explanations.
One might
conjecture that
women have
less
power
to influence the literacy of their family
than do men. Social sanctions on women's' behavior might limit their ability to choose a
partner, while
men
are less constrained.
Or
possibly
men have
greater influence over whether
the children are sent to school. If there are in fact pecuniary gains to an illiterate
having
at least
one
literate
men may
person in the family then
external effects of literacy through family formation than are
worker from
well be better able to capture the
women.
We would then find less
variance in family schooling for males than females.
This explanation
of family literacy for
this explanation
were
is
not,
illiterate
however, easily reconciled with our finding that the external effect
males drops out when we confine attention to unmarried males.
right, then
we would
surely expect a stronger effect to
emerge
for
unmarried males than married ones.
We believe that a more plausible explanation
much more dependent on
is
that illiterate
women
in
Bangladesh
their family for assistance with tasks that require reading
22
are
and writing
If
than are illiterate men. There are sources of help in reading and writing for illiterate workers,
including paying for the services of a scribe. However, cultural and religious sanctions on
women's
social activities
may
well entail that they have a harder time than
men
m
accessing
would be no weaker we suspect, and
these sources of help outside the family. This effect
possibly stronger, for unmarried than married men.
Another possible interpretation of our
allows
illiterate
workers
results
is
that
to avoid exploitation (in sectors
having a
where there
power), or simply allows a better matching of the worker's other
(Even
if all
workers receive
their
marginal product,
this
worker
literate
is
family
some monopsony
skills to the
may depend on
in the
jobs available.
the kind of firm that a
person joins.)
Finally,
women may
well be better able to absorb the benefits of household literacy.
They
may, for instance, be unencumbered by pride or other defenses which could render one unable
learn easily
from
others.
Conclusions
6.
Intuition
and piecemeal evidence suggest
of others in the household. That
or her literacy with other
in a
is,
that literacy has positive externalities
a literate person
is
likely to share
members of the household. This means
household where there
is
someone
literate
than
We have tested this hypothesis in two ways.
in a
that
some of the
it
is
Firstly,
benefits of his
illiterates.
we have asked whether
Our
sense for agents to share their education with others in the household. Secondly,
up
to systematic scrutiny in earnings data for
23
it
can be
theoretical analysis
of the household, based on the collectivist model of the household, confirms that
the hypothesis stands
on earnings
better to be an illiterate
household of only
derived from a consistent set of assumptions about household behavior.
if
to
it
can make
we have
Bangladesh.
We
tested
found
that
it
not only stands up,
observed personal
when
it
does so with a considerable amount of robustness. Controlling for
attributes,
an
illiterate
living in a family with at least
one
worker earns appreciably more
literate
member. This finding
in the
is
non-farm sector
robust to the tests
we
perform for selection bias and marital sorting.
The empirical
investigation has yielded insights deeper than
the basis of theory and intuition.
greater benefits
It is,
on the children than a
and female recipients!
Is
for instance, widely noted that a literate
literate father.
benefits of literacy in the household than a
gives no hint as to which
it
women
efficient recipients.
are
more
male
illiterate?
way
this
24
mother confers
But what about differences between male
there any reason to believe that a female
have differences,
what we had anticipated on
may
is
better able to absorb the
While theory suggests
go.
Our empirical
that they can
results suggest that
Table
1:
Illiterate
Effects of Family Literacy on the Earnings
and Labor Force Participation of
Non-Farm Workers
Unmarried
sample only
Full sample
Women
Earnings (Q
)
Participation {S
Number
)
of
With land
effects
on earnings
0.504
0.737
0,487
(4.577)
(4.477)
(4.368)
-0.256
-0.290
-0.253
(-3.512)
(-2.353)
(-3.527)
295
139
295
observations
Men
Earnings (a
)
(rural)
Participation
Number
(
S
)
of
0.146
0.143
0.143
(2.769)
(1.127)
(2.636)
0.110
0.113
0,111
(1.980)
(1.125)
(1.991)
582
157
582
observations
Men
Earnings (a
0.144
)
(urban)
Participation
Number
(
6
)
of
-0.031
0.136
(2.274)
(-0.251)
(2.405)
-0.023
-0.023
-0.151
(-0.153)
(-0.153)
(1.800)
550
132
550
observations
Note: Gain
in log
earnings of an
illiterate
worker attributed
to literacy within the household.
Log earnings
regressions include correction for sample selectivity bias as well as controls for worker characteristics
(age, age squared,
the
if
any school attendance, and regional
STATA 6 command
for the
effects), t-ratios in parentheses
MLE version of the Heckman
corrected for heteroskedasticity and clustering.
25
.
We
have used
selection model, with standard errors
Table
2: Effects
and Schooling on the Earnings and Labor Force
of Family Literacy
Participation of Illiterate
Non-Farm Workers
There
is
literate
the
Women
Earnings
Participation
Men
Earnings
(rural)
Participation
Highest schooling
another
person
Completed
in
household
Class 1+
Earnings
(urban)
Participation
Note:
Log earnings
household
Completed
Class
VII-i-
0.462
0.186
-0.311
0.757
(2.578)
(1.510)
(-1.628)
(4.154)
0.102
-0.249
-0.223
-0.196
(0.866)
(-2.678)
(-1.820)
(1.705)
0.104
0.152
-0.042
-0.053
(1.252)
(2.225)
(-0.495)
(-0.613)
0.016
-0.069
(-0.949)
(0.189)
Men
in the
Completed
Class IV+
0.171
0.012
(1.953)
(0.145)
0.048
0.056
-0.020
0.271
(0.493)
(0.743)
(-0.205)
(2.736)
0.005
0.001
0.026
-0.460
(0.004)
(0.008)
(0.187)
(-4.075)
regressions include correction for sample selectivity bias as well as controls for
worker characteristics (age, age squared,
parentheses. Also see Notes to Table 1.
if
any school attendance, and regional effects),
26
t-ratios in
References
Aldeman, H.,
J.R.
Behiman, DR. Ross and R. Sabot, 1996, The Returns
Capital in Pakistan's Rural
Wage Labor Market, Oxford
to
Bulletin of
Endogenous Human
Economics and
Statistics 58(1), 29-55.
Bangladesh Bureau of
Bureau of
Statistics,
1998, Household Expenditure Survey 1995-96, Bangladesh
Statistics, Statistics Division,
Dhaka, Bangladesh.
Basu, K., 1999, Child Labor: Cause, Consequence and Cure with Remarks on International
Labor Standards, Journal of Economic Literature 37, 1083-1
Basu, K., and
Behrman,
Foster, 1998,
J.
J.R.,
On Measuring
Literacy,
1
19.
Economic Journal
108, 1733-49.
and B.L. Wolfe, 1984, The Socioeconomic Impact of Schooling
in a
Developing
Country, Review of Economics and Statistics 66, 296-303.
Benham,
L.,
1974, Benefits of
Economy
Women's Education Within
Marriage, Journal of Political
82(1), S57-71.
Bergstrom, T.C. and R.C. Comes, 1983, Independence of Allocative Efficiency from
Distnbution
Bourguignon,
F.
in the
and
Introduction,
Bourguignon,
F.
Theory of Public Goods, Econometrica 51, 1753-65.
P. A.
Chiappon, 1991, Collective Models of Household Behavior:
European Economic Review 36, 355-364.
and P. A. Chiappon, 1994, The Collective Approach
R. Blundell,
An
I.
Preston and
I.
to
Household Behavior,
Walker, eds.. The Measurement of Household Welfare
(Cambridge University Press, Cambridge).
Dreze,
J.,
and M. Saran, 1995, Primary Education and Economic Development
India
eds..
:
Overview and
Two
Case Studies,
in:
China and
K. Basu, P. K. Pattanaik and K. Suzumura,
Choice, Welfare and Development (Clarendon Press, Oxford).
27
in
in:
Foster, A.D.
and M.R. Rosenzweig, 1996, Technical Change and Human-Capital Returns and
Investments: Evidence from the Green Revolution, American Economic Review 86, 931-
53.
Gibson,
Green,
J.,
2001, Literacy and Intrahousehold Externalities, World Development 29: 155-166.
S. E.,
T.A. Rich and E.G. Nesman, 1985, Beyond Individual Literacy: the Role of Shared
Literacy for Innovation in Guatemala,
Krishnan,
P.,
Human
Organization 44, 313-21.
1996, Family Background, Education and
Bulletin of
Economics and
Lam, D. and R.F. Schoeni, 1993,
Employment
CM.,
Effects of Family
Background on Earnings and Returns
The Intrahousehold Allocation
1995,
Economy
to
101(4), 710-740.
of Resources and Participation of
Household Decision-Making Evidence from Early Twentieth Century
Children,
in:
America,
mimeo Northwestern
Quisumbing, A.R. and
New
Urban Ethiopia, Oxford
Statistics 58, 167-183.
Schooling: Evidence for Brazil, Journal of Political
Moehling,
in
:
:
J. A.
University.
Maluccio, 1999, Intrahousehold Allocation and Gender Relations:
Empirical Evidence, mimeo: International Food Policy Research
Institute,
Washington, DC.
Ravallion,
M. and Q. Wodon, 2000, Does Child Labor Displace Schooling? Evidence on
Behavioral Responses to an Enrollment Subsidy
in
Bangladesh, Economic Journal
1
10,
158-176.
Strauss,
J.,
and D. Thomas, 1995,
Family Decisions,
in: J.
Human
Resources: Empirical Modeling of Household and
Behrman and T.N.
Srinivasan, eds..
Economics Volume 3 (North -Hoi land, Amsterdam).
28
Handbook of Development
Udry,
C,
1996, Gender. Agricultural Production and the Theory of the Household, Journal of
Political
Economy
104(5), 1010-46.
Yang, D.T., 1997a, Education
in
Production: Measuring Labor Quality and Management,
American Journal of Agricultural Economics 79, 764-772.
Yang, D.T., 1997b, Education and Off-farm Work, Economic Development and Cultural Change
45(3), 613-632.
29
8
^ 7
67
a
Date Due
MIT LIBRARIES
3 9080 02246
DUPL
187
'fsil
i
m
.
iVi
.
ipj^af
m:^
'
I.