Sie sind auf Seite 1von 20

The Relative Efficiency of Hired and Family Labour in Bangladesh Agriculture

Nasima Tanveer Chowdhury


Department of Economics
Gothenburg University
Gothenburg
Sweden.
nasima.chowdhury@economics.gu.se

April 15, 2010


Abstract

In Bangladesh labour supply is abundant but land for cultivation is shrinking mainly due
to growing human settlement and urbanisation with a very high population density. The
aim of this study is to examine whether labour use is efficient in Bangladesh agriculture.
The study analyses the relative efficiency of hired and family labour for 3 crop seasons. It
uses a production function approach, estimates marginal returns of the inputs and
compares these with their prices. Production functions are estimated using data collected
by International Rice Research Institute (IRRI) on expenditures of agricultural inputs and
returns on investment from a nationally representative sample of 1928 farm households
from all 64 districts of Bangladesh. Results show that family labour is more productive
than hired labour in Bangladesh agriculture which has important policy implications for
the rural labour market.

Key words: relative efficiency, hired labour, family labour, marginal product, production
function, labour market, Bangladesh agriculture.

JEL Classification: J21, J24, J43, Q12, Q18

1. Introduction

Being a country of 150 million inhabitants, agriculture still faces a tremendous challenge
in feeding a growing population where at least 40 million people do not have a square
meal. In the face population growth, land is getting scarcer. Almost all the agricultural
land is double or triple cropped. Only labour is abundant. About 52 percent of the total
labour force is engaged in agriculture (Labour Force Survey, 2002-03, Bangladesh
Bureau of Statistics 2006). Employment opportunities or the labour absorption capacity
in the formal sector of the economy is inadequate in comparison with the size and growth
of the workforce (high percentage of working age population) in Bangladesh. It is
established in the literature that in labour surplus societies, family labour is paid the
average product where hired labour is paid the marginal product (the efficiency wage).
Therefore, I would like to examine whether family labour use is efficient in Bangladesh
agriculture for households who are using both family and hired labour. Given the
development of various off-farm activities like the agro-processing industry and the
2
expansion of micro credit, farmers always have an option to move to non-farm sectors.
Against this backdrop the present study attempts to examine the relative efficiency of
hired and family labour use in household agriculture. As food security and employment
generation are still major development challenges for Bangladesh, this study would have
important policy implications for off-farm employment in rural Bangladesh.

The specific research objectives are to estimate the marginal value product of family
labour and compare this with the wage rate. I compare the marginal return of family
labour with the wage rate (marginal return on hired labour) in order to find out if the use
of family labour is efficient in Bangladesh agriculture. I would also compare the marginal
return of farmers who are using both family and hired labour with those using only hired
labour. The marginal return estimates of family and hired labour would indicate the state
of farm labour allocation in Bangladesh agriculture spatially across farm size for 3 crop
seasons. Further the marginal return estimates of family and hired labour and their
comparison with the market wage rate will give important insights into the efficiency of
agricultural labour use in the rural labour market of Bangladesh.
The foregoing remarks serving as the introduction (section 1), the rest of the paper
proceeds as follows. In section 2 the literature on productivity of labour in agriculture is
reviewed. Section 3 develops research methods, econometric specification of the function
to be estimated and the hypotheses to be tested. Section 4 gives a brief description of the
study area and the data. Section 5 presents methods for data analysis and the variables
used. Section 6 presents and discusses the estimated results. Finally the paper is
concluded in section 7 with a summary and some brief policy recommendations.

2. Productivity of Labour in Agriculture
There is a dearth of relevant empirical work in the literature. In the Bangladesh context,
as far as known to the author, there has been only one study testing the allocative
efficiency of input use in agriculture. In the developing country context one study was
done by Linde-Rahr (2005) who tested the relative efficiency of input use in Vietnamese
agriculture. He assumes a risk neutral household and examines the ability of households
to allocate factor inputs efficiently within the household. His results suggest that the
households are efficient in this regard.
Jacoby (1992) estimated the productivity of men and women in peasant agriculture of the
Peruvian Sierra using household data from the Peruvian Living Standards Survey. He
estimated a Cobb-Douglas pseudo-production function where he regressed the
logarithm of the value of crop output on the logarithm of all input expenditures in Peru.
He also derived the OLS estimates of a restricted translog production function. His results
suggest that marginal productivity of men and women are significantly positive and that
men are more productive than women.

In another study, Jacoby (1993) estimated an agricultural production function with
different types of labour (men, women and children) as distinct inputs in the Peruvian
highlands. He estimated the opportunity cost of time or shadow wages. The marginal
product of labour is calculated for each household. He estimated both Cobb-Douglas and
translog production functions and derived OLS and instrumental variable (IV) estimates
3
of the Cobb-Douglas production function. Since the correlation between marginal
products for male and female labour was very high the labour supply parameter estimates
from the OLS and instrumental variable method did not differ much although the OLS
estimates were more efficient. Further since the Cobb-Douglas production function
imposes strong separability between male/female labour and other inputs he estimated the
translog (OLS) flexible functional form to derive the shadow wages.
Thapa (2003) used a production function approach to test for heterogeneity between
family and hired labour as inputs in crop production in the southern lowland of Nepal. He
followed the same estimation strategy of Jacoby (1993) described above. His study found
that family labour is more productive than hired labour. The present study considers
family and hired labour as distinct inputs since in Bangladesh hired labour is only used
when there is a shortage of family labour in the farm.
Earlier Bardhan (1973) estimated a production function with farm level data, Deolaikar
and Vijverberg (1983) used aggregate district level data for India to test for labour
heterogeneity. The latter authors found substantial productivity differences between hired
and family labour and family labour was seen as more productive. In a subsequent
analysis Deolaikar and Vijverberg (1987) used household level data from India and
Malaysia and found that hired labour is more productive than family labour. In all these
cases a Cobb-Douglas production function was found to be the appropriate form.
Squires and Tabor (1994) estimated a translog production function using Indonesian data.
Their results suggest family and hired labour are both substitutes and complements for
some crops and regions. Frisvold (1994) used very detailed plot level production data
collected by ICRISAT in southern India and found that the productivity of hired labour
increased with the supervision intensity of family labour and approached the productivity
of family labour at a high level of supervision. While these diverse studies demonstrate
difference in productivity between hired and family labour, issues such as higher average
and marginal productivity of family labour over hired labour are still not so well
documented. In order to better understand the differential impacts of hired and family
labour on agricultural productivity, the present study estimates both Cobb-Douglas and
translog production functions with Bangladeshi household data.
3. Methods
3.1 Hypotheses to be tested
The major issue is whether families that employ both family and hired labour engage an
efficient level of labour meaning the marginal returns to the family labour is equal to the
returns on hired labour. The underlying assumption is the household first pays the hired
workers their MVP (i.e., the going wage rate) and then divides the residual among the
family members supplying labour that is according to the average product of labour. I test
this hypothesis by comparing the marginal return of hired labour with the marginal return
of both family and hired labour. The details follow later.



4
3.2 Econometric Specification
Let
j
Y be a concave production function for activity- j as a function of labour ( L ), and
all other inputs
i
x , . ,..., 1 n i =
| ) 1 ) ,..., (
, jn jI j j j
x x L Y Y =
Where
j
Y denotes the relevant output and j denotes the crop type. Unit of labour
j
L is
defined in days per season. There are three crop seasons in Bangladesh.
From| ) 1 , one can define all marginal product of labour neatly:
| ) 2 ,
j
j
jL
L
Y
MP
c
c
=
while the corresponding marginal value product is | )
|
|
.
|

\
|
c
c
j
j
j
L
Y
p and
j
p denotes the market
price per unit of crop- j .
Similarly, given the production function, one can define total costs from crop- j as

| ) 3 | ),
1

=
+ =
n
i
ji i j j
x q L C e
where e is the wage rate and
i
q denote all other relevant input prices. Formulation | ) 3
presumes the functioning of competitive input markets. The cost function, relating the
minimum costs of production for a given level of output, may also be written as:

| ) 4 | ). ,
, j i j j
Y q C C e =

From (4) one may easily derive the demand for inputs by using the Shepherd's lemma:
| ) 5 | ) ; , ,
e
e
c
c
=
j
j i j
C
Y q L and

| ) 6 | ) . , ,
i
j
j i i
q
C
Y q x
c
c
= e
A number of different flexible functional forms can be estimated including the translog
form. The translog form provides a greater variety of substitution possibilities than those
restricted by constant elasticity of substitution. Therefore this form is widely used in the
empirical analyses of production technology and factor markets.

The translog production function can be written as:

| ) 7
j
n
i
n
l
jl ji il
n
i
ji i j
x x x Y u | | | +
)
`

+ + =

= = = 1 1 1
0
ln ln
2
1
ln ln

Where =
j
Y ln log of total production of crop- j
5
=
ji
x ln log of input n i ,..., 1 = including
j
L
The elasticity of production with respect to each factor of production is calculated by
taking the partial derivative of output with respect to the factor under consideration. For
labour, the elasticity can be derived as:

| ) 8
l
i l
il i ii i
j
x x
L
Y
ln ln 2
ln
ln

=
+ + =
c
c
= | | | c

The marginal productivity of labour in crop- j is then:
| ) 9
L
Y
L
Y
L
Y
L
Y
j j j j
c p =
c
c
=
c
c
= *
ln
ln


If Y is the total value of agricultural output equation | ) 9 then gives the marginal value of
labour for agricultural use. A price elasticity of labour can also be derived if the wage
ratee is assumed to equal the marginal value of labour. For a profit maximising farmer
the marginal value of an input is equal to its marginal cost. One can define as price
elasticity of labour which is:

| ) 10 .
ln
ln
ln
ln
p e

c
c
=
c
c
=
L L


An efficient household equates the marginal value product of inputs with its price and
also equates MRTS among production functions. The marginal return on labour and input
i
x in terms of rice and other crop production functions is
(11) e =
c
c
=
c
c
other
other
other
rice
rice
rice
L
f
p
L
f
p

i
other
i
other
other
rice
i
rice
rice
q
x
f
p
x
f
p =
c
c
=
c
c
.
where
other rice
p p , are output prices,
other rice
f f , are production functions,
other rice
L L , are
labour used in rice and other crop cultivation respectively;
other
i
rice
i
x x , represent all other
inputs and e and
i
q s are labour and other input prices. I develop marginal product
conditions for hired and family labour and compare these with their respective returns
and wage rate.

4. Study Area and Data
In Bangladesh agriculture comprises 21.11 percent of GDP in 2006-07 and employs
about 52 percent of the countrys labour force (Bangladesh Bureau of Statistics, 2008).
The crop calendar is based on the temporal distribution of rainfall and temperature
throughout the year. Three cropping seasons are pre-monsoon, monsoon (wet season) and
6
winter or dry season. Major crops are rice, wheat, jute, sugarcane, oilseeds, pulses,
potato, onion, spices and vegetables. For testing efficiency of labour use in agriculture I
use the whole dataset for 3 crop seasons.

Data

International Rice Research Institute (IRRI) conducted an Agricultural Household Survey
in Bangladesh in 2000 for 3 crop seasons and collected data on expenditures of inputs
(including irrigation water) and returns on investment from a nationally representative
sample of 1880 farm households from 62 villages belonging to 57 of 64 districts in the
country. In this study I use an updated and more recent version (2004) of the same dataset
from IRRI. In the context of developing countries this dataset collected by IRRI is very
rich particularly in terms of information on returns of different types of agricultural
labour and other inputs.
The 2004 survey collected expenditure and returns data only for the largest plot of the
land portfolio of farmers. Hence the sample size is larger because some households have
been transformed into several due to separation within the extended family (Hossain,
2006). A few households migrated out and they were replaced with new entries from the
same "wealth ranking group". The sample in this survey consists of 1928 households.
There are 1104 rice farmers.
The dependent variable is output in BDT (Bangladesh Taka
1
). Land is measured in ha
and includes both own and sharecropped land. This is essentially a sample of small
farmers, which is highly appropriate since the average farm size in Bangladesh is 0.68 ha.
The plot size varies from 0.01 ha to 1.74 ha in this sample. Labour is used mainly for
ploughing, sowing, weeding, harvesting and threshing. Labour days consist of both
family and hired labour. Agricultural households can be categorised into 3 groups on the
basis of type of labour use. Some households use only family labour, some only hired
labour and majority uses both family and hired labour as input. Irrigation is the total
irrigation expenditure measured in BDT and we do not know the price per unit of water
or the number of hours used for pumping water. The expenditure on irrigation is basically
the energy (electricity or diesel) expenditure of pumping water and salary for the pump
mechanics. Farmers mainly use low lift pumps for pumping water from surface water
sources and shallow and deep tube wells from aquifers and groundwater. Seeds and
fertilisers are measured in kilogramme (kg). Farmers mainly use urea and TSP besides
phosphate and other chemical fertilisers. Only expenditure data measured in BDT are
available for pesticides and manure use.

Table 1 Descriptive Statistics for Family Labour
------------------------------------------------------------------------------------------------------------
Variable Mean Std. Dev. Min Max
------------------------------------------------------------------------------------------------------------
Output (BDT) 2904.19 3134.54 0 26000
Land cultivated (ha) 0.11 0.09 0.01 0.80
Family labour (days) 14.14 16.11 1 203

1
1 USD = 58 BDT (2004)
7
Ploughing expenditure (BDT) 142.69 161.84 0 878
Seeds (kg) 9.14 13.66 0 120
Irrigation (BDT) 234.28 498.96 0 4875
Fertilisers (kg) 23.45 36.72 0 567
Manure (BDT) 56.83 197.37 0 2400
Pesticides (BDT) 29.08 53.69 0 420


Table 1 contains descriptive statistics for farmers those who are only using family labour.
There are 456 such households in the dataset. Except manure use the mean output and all
other input use are lowest for this category of farmers dependent on only family labour.

Table 2 Descriptive Statistics for Hired Labour
------------------------------------------------------------------------------------------------------------
Variable Mean Std. Dev. Min Max
------------------------------------------------------------------------------------------------------------
Output (BDT) 7556 17933.4 520 247000
Land cultivated (ha) 0.19 0.14 0.03 0.81
Hired labour (days) 21.21 15.61 3 86
Ploughing expenditure (BDT) 399.17 342.55 0 2040
Seeds (kg) 47.6 184.11 0 2000
Irrigation (BDT) 638.87 913.18 0 4600
Fertilisers (kg) 61.33 79.85 0 700
Manure (BDT) 74.53 196.38 0 1000
Pesticides (BDT) 128.14 256.89 0 2020

-----------------------------------------------------------------------------------------------------------
Table 2 presents descriptive statistics for farmers those who are only using hired labour.
There are only 256 households in this dataset for this category. Except labour use all
other input and mean output are highest for this category of farmers.

Table 3 Descriptive Statistics for Hired and Family Labour
------------------------------------------------------------------------------------------------------------
Variable Mean Std. Dev. Min Max
------------------------------------------------------------------------------------------------------------
Output (BDT) 5024 5555 100 86000
Land cultivated (ha) 0.17 0.14 0.01 1.74
Family labour (days) 8.14 8.7 1 136
Hired labour (days) 13.56 14.07 1 164
Ploughing expenditure (BDT) 298.46 297.7 0 2720
Seeds (kg) 19.66 83.81 0 2000
Irrigation (BDT) 535 962.68 0 11000
Fertilisers (kg) 45.39 54.56 0 700
Manure (BDT) 54.98 165.78 0 2500
Pesticides (BDT) 79.01 146.58 0 3000

8
------------------------------------------------------------------------------------------------------------
Table 3 presents descriptive statistics for 1223 households who are using both family and
hired labour for cultivation. Farmers use of hired labour on average is more than that of
family labour.

5. Methods for Data Analysis

I estimate a production function, equation (7) as above with an exception of two forms of
labour, namely hired and family labour. I consider their marginal product conditions on
two alternative hypotheses.
First Hypothesis: That the two types of labour are treated and rewarded the same (i.e.,
marginal value products equal the return to each).
Second Hypothesis: The farm first pays the hired workers their MVP (i.e., the going farm
wage) and then divides the residual among the family members supplying labour. This
assumes that the farm uses the traditional method for rewarding own members according
to the AP of labour. I would like to see which hypothesis is more relevant for rural
Bangladesh.
First step: 256 households are using only hired labour. I find the marginal productivity of
labour and compare with the wage.
Second step: 456 households are using only family labour. I find the marginal
productivity of labour and compare with the average return of family labour and average
daily market wage.
Third step: 1223 households are using both family and hired labour. We find their returns
to labour and check whether the return to family labour is less than the return to hired
labour. I find the average returns to family labour as a residual = revenue - imputed value
of hired labour - other input expenditures.
I estimate production functions and the marginal productivity of labour for each type of
household and assess whether marginal value product varies significantly among them.
Hypotheses

In this section I discuss the expected signs on the coefficients of the explanatory variables
included in the estimation of production function. A summary is presented in Table 4.

Land, ploughing expenditure, both family and hired labour, seeds, fertiliser, irrigation,
manure and pesticides all these inputs are expected to have positive effects on output.
Increasing the amount and expenditure of all the inputs are also expected to have positive
impacts on output. However I am not sure about the sign of the complementary input
terms on output.
Table 4 List and Description of Variables

Variable Definition Expected sign
loutput log of value of output in BDT dependent variable
lland log of land in ha (+)
9
lplow log of ploughing expenditure in
BDT
(+)
llabour log of labour days (+)
lfam log of family labour days (+)
lhlab log of hired labour days (+)
lseed log of seeds in kg (+)
lfert log of fertilisers in kg (+)
lircst log of irrigation expenditure in
BDT
(+)
lmncst log of manure expenditure in
BDT
(+)
lpest

log of pesticide expenditure in
BDT
(+)
l2land log of land* log of land (+)
l2lplow log of ploughing expenditure *
log of ploughing expenditure
(+)
l2labour log of labour days * log of
labour days
(+)
l2lfam log of family labour days * log
of family labour days
(+)
l2lhlab log of hired labour days * log of
hired labour days
(+)
l2seed log of seeds * log of seeds (+)
l2fert log of fertilisers * log of
fertilisers
(+)
l2ircst log of irrigation expenditure *
log of irrigation expenditure
(+)
l2mncst log of manure expenditure * log
of manure expenditure
(+)
l2pest log of pesticide expenditure *
log of pesticide expenditure
(+)
landplow log of land * log of ploughing
expenditure
(?)
landlab log of land * log of labour days (?)
landfam log of land * log of family
labour days
(?)
landhlab log of land * log of hired labour
days
(?)
landseed log of land * log of seeds (?)
landfert log of land * log of fertilisers (?)
landircst log of land * log of irrigation
expenditure
(?)
landmncst log of land * log of manure
expenditure
(?)
landpest log of land * log of pesticide
expenditure
(?)
plowlab log of ploughing expenditure *
log of labour days
(?)
plowfam log of ploughing expenditure *
log of family labour days
(?)
plowhlab log of ploughing expenditure *
log of hired labour days
(?)
plowseed log of ploughing expenditure *
log of seeds
(?)
10
plowfert log of ploughing expenditure *
log of fertilisers
(?)
plowircst log of ploughing expenditure *
log of irrigation expenditure
(?)
plowmncst log of ploughing expenditure *
log of manure expenditure
(?)
plowpest log of ploughing expenditure*
log of pesticide expenditure
(?)
famhlab log of family labour days* log
of hired labour days
(?)
labseed log of labour days* log of seeds (?)
famseed log of family labour days* log
of seeds
(?)
hlabseed log of hired labour days* log of
seeds
(?)
labfert log of labour days* log of
fertilisers
(?)
famfert log of family labour days* log
of fertilisers
(?)
hlabfert log of hired labour days* log of
fertilisers
(?)
labircst log of labour days* log of
irrigation expenditure
(?)
famircst log of family labour days* log
of irrigation expenditure
(?)
hlabircst log of hired labour days* log of
irrigation expenditure
(?)
labmncst log of labour days* log of
manure expenditure
(?)
fammncst log of family labour days* log
of manure expenditure
(?)
hlabmncst log of hired labour days* log of
manure expenditure
(?)
labpest log of labour days* log of
pesticide expenditure
(?)
fampest log of family labour days* log
of pesticide expenditure
(?)
hlabpest log of hired labour days* log of
pesticide expenditure
(?)
seedfert log of seeds* log of fertilisers (?)
seedircst log of seeds* log of irrigation
expenditure
(?)
seedmncst log of seeds* log of manure
expenditure
(?)
seedpest log of seeds* log of pesticide
expenditure
(?)
fertircst log of fertilisers* log of
irrigation expenditure
(?)
fertmncst log of fertilisers* log of manure
expenditure
(?)
fertpest log of fertilisers* log of
pesticide expenditure
(?)
irmncst log of irrigation expenditure*
log of manure expenditure
(?)
irpest log of irrigation expenditure* (?)
11
log of pesticide expenditure
mncstpest log of manure expenditure * log
of pesticide expenditure
(?)

6. Results and Discussions
This section first outlines the procedure followed in identifying the appropriate
production function specification for agricultural output with 3 different types of labour
before discussing the estimated results.

6.1 Model Selection
The dependent variable is the value of output in BDT. First I estimate an unrestricted
translog function. Several of the coefficients in this unrestricted translog function were
not significant at conventional levels. Therefore I impose restrictions and estimate the
restricted function without the second order and cross product terms

0 = = =
li il ii
| | | i, l.

This in essence is a Cobb-Douglas production function:
(12)
j
n
i
ji i j
x Y u | | + + =

=1
0
ln ln
Then I conduct F-test to determine the appropriate function.

(13)
K N
R
j
R R
F
u
r u

=
2
2 2
1
~
K N j
F
,
,

where =
2
u
R R
2
from the unrestricted (translog) function
=
2
r
R R
2
from the restricted (Cobb-Douglas) function
j = the number of restricted parameters
N = number of observations
K = number of estimated parameters in the translog

(a) Hired Labour
The calculated F statistic = 11 . 3
4 256
73 . 0 1
3
74 . 0 73 . 0
=


The critical value F value is somewhere between 3.84 and 3.92 at the 5-percent level of
significance. Since this critical value is greater than the calculated F statistic the
hypothesis that all the coefficients on the second order and cross product terms are zero is
12
accepted. The preferred specification for households who use only hired labour therefore
is:
(12)
j
n
i
ji i j
x Y u | | + + =

=1
0
ln ln
The results from the restricted Cobb-Douglas production function are reported in Table 5
and are discussed below. These are corrected for heteroscedasticity.

(b) Family Labour
The calculated F statistic = 91 . 22
3 452
51 . 0 1
4
61 . 0 51 . 0
=


Since the calculated F statistic is higher than the critical value the 5-percent significance
level the hypothesis that all the coefficients on the second order and cross product terms
are zero is rejected. The preferred specification is:
| ) 14
j
n
i
n
l
ji ii
n
i
ji i j
x x Y u | | | +
)
`

+ + =

= = = 1 1
2
1
0
ln
2
1
ln ln

This is a restricted form of the translog production function where there are no cross
product terms. This thus rules out the complementarities of input variables in the
production function for family labour. The results from the restricted translog function
are reported in Table 6 and are discussed below. The results are corrected for
heteroscedasticity.

(c) Hired and Family Labour

The calculated F statistic = 44 . 23
12 1223
69 . 0 1
5
66 . 0 69 . 0
=



The critical value F
5, 1211
is somewhere between 2.21 and 2.29 at the 5-percent level of
significance. Since the calculated F statistic is greater than this critical value the
hypothesis that all the coefficients on the second order and cross product terms are zero is
rejected. The preferred specification is:
| ) 7
j
n
i
n
l
jl ji il
n
i
ji i j
x x x Y u | | | +
)
`

+ + =

= = = 1 1 1
0
ln ln
2
1
ln ln
The results from the unrestricted translog function are reported in Table 7 and are
discussed below. The results are corrected for heteroscedasticity.

6.2 Discussion of Results

An increase in land, hired labour, seeds, fertiliser and irrigation would increase
agricultural output for households using only hired labour (Table 5). Land has highest
13
elasticity (0.56). A 10 percent increase in cultivable land would increase the value of
output by 5.6 percent. The elasticity of hired labour indicates that a one day additional
hired labour would lead to an increase of 17 percent of value of output but this is
significant only at the 10-percent level. Fertiliser has a marginally higher elasticity (0.24)
than hired labour. There is a high correlation between labour and fertiliser use (0.72) in
farms that are dependent only on hired labour (Table 12). The ploughing expenditure has
a negative impact on output.
Table 5 Restricted Cobb-Douglas Production Function Estimates for Hired Labour
Variable Coefficient t- statistics
lland 0.56*** 5.29
llabour 0.17* 1.68
lplow -0.11** -2.17
lseed 0 .15** 2.28
lfert 0.24*** 3.83
lircst 0.02** 2.26
Constant 8.29*** 14.44
Adj. R
2
0.73
Number of
observations
256
***, ** and* represent significance of parameters at 1, 5
and 10 percent respectively.

An increase in family labour for households using only family labour however indicates a
stronger impact on output (Table 6). Here the elasticity of family labour is 0.67; thus one
family labour day increase would increase the value of output by 67 percent. Doubling
the amount of fertiliser, on the other hand, increases the value of output by 21 percent.
Table 6 Restricted Translog Production Function Estimates for Family Labour
Variable Coefficient t- statistics
llabour 0.67*** 11.02
l2fert 0 .04*** 6.91
Constant 5.66*** 45.58
Adj. R
2
0.51
Number of
observations
452
*** represents significance of parameters at 1 percent.
An increase in both family and hired labour would increase output for households using
both hired and family labour. However hired labour (0.17) has a higher elasticity than
14
family labour (0.13). The correlation between hired and family labour is negative (Table
13). The elasticity of substitution
2
between family and hired labour is -0.26, there is a
limited substitution between family and hired labour. Hired labour cannot perform all the
tasks done by family labour when it comes to supervisory and managerial role.
Land has diminishing returns on output as demonstrated by the negative coefficient of the
land squared term. However doubling the amount of seed, fertiliser, irrigation, manure
and pesticides will have positive effects on output. A simultaneous increase in hired
labour and irrigation expenditure would increase output. However a joint increase in
seeds and irrigation expenditure shows a negative impact on output.
Table 7 Unrestricted Translog Production Function Estimates for Hired and Family
Labour
Variable Coefficient t- statistics
lfam 0.24*** 4.11
lhlab 0.19*** 3.61
l2land -0.12*** -11.18
l2fert 0 .02*** 5.51
l2seed 0.05*** 4.05
l2ircst 0.02*** 6.07
l2mncst 0.002** 2.25
l2pest 0.003** 2.05
hlabircst 0.02*** 3.37
seedircst -0.05*** -5.79
famhlab -0.05** -2.3
Constant 7.2*** 41.47
Adj. R
2
0.69
Number of
observations
1223
*** and **represent significance of parameters at 1 and 5 percent
respectively.


2
Elasticity of substitution between family and hired labour for a translog production function is
) ( ) (
) 2 ( ) ( ) ( ) ( ) ( 2 ) 2 ( ) (
)) ( ) ( (
)) ( ) ( (
2 2
lhlab lfam
lfam l lhlab famhlab lhlab lfam lhlab l lfam
lhlab lfam
lhlab lfam
| |
| | | | | | |
| |
| |
o
+
+ +
+
=

Alauddin et al (1993).
15
(d) Marginal Products and Wages
Table 8 Marginal Products at Sample Mean for Households using Hired Labour
Input Marginal Product
Land 22269.81
Hired
labour
60.56
Seeds 23.81
Fertiliser 29.57
Irrigation 0.24

Land has the highest marginal value product, one ha increase in plot size would increase
output by BDT 22,270 for households using only hired labour. Hired labour is more
productive than seeds, fertiliser and irrigation. However there is a high correlation
between labour and fertiliser use (0.72) meaning households dependent on hired labour
use more fertiliser to complement low productivity of hired labour. Farmers are least
efficient in irrigation water use. However since irrigation is expressed in total
expenditure, we do not know the per unit price of irrigation water. There is no
information in the dataset on how much water the farmers are using to irrigate their rice
fields or for the numbers of hours irrigation pumps used.
From the estimates presented above it follows that marginal productivity is uniformly
higher for family labour, particularly when there is no hired labour on the farm. For
instance family labour earns a marginal product of BDT 138 and BDT 80 respectively in
farms without and with hired labour. By contrast hired labour registers marginal products
of BDT 63 and 61, respectively, when working alongside family labour and when alone.
It will be instructive to compare the above pattern of the marginal products of labour with
their respective wage rates and/or average productivity (for family labour as appropriate).
Tables 9 and 10 estimate the average daily wage from the wages of different kinds of
agricultural labour available from the IRRI data used in this analysis.
Table 9 Daily Wage Rate for Hired Labour
------------------------------------------------------------------------------------------------------------
Sample Size Mean SD Min Max
Daily wage for ploughing (BDT) 17 81 24 50 130
Daily wage for sowing (BDT) 256 70 21 40 160
Daily wage for weeding (BDT) 244 62 20 30 150
Daily wage for harvesting (BDT) 255 82 28 20 250
Daily wage for threshing (BDT) 233 83 32 40 300

16
Average (unweighted) daily wage rate comes to BDT 76 (an average of the daily wages
for 5 types of labour) in farms exclusively employing hired labour. Marginal value
product of hired labour (BDT 61) is therefore much lower than this average wage.
Similarly average daily wage is calculated for households using both family and hired
labour in Table 10. Here the average (unweighted) daily wage rate is found to be BDT
82 for this category. The above evidence on wages and productivity calls for an
investigation into the workings of the rural labour market. First, why is family labour so
much more productive when there is no hired labour? Second, is hired labour more
productive while working alongside family labour? A related issue concerns the
productivity of family vs. hired labour. Finally the key issue is why rural labour earns in
excess of their respective marginal products.
Table 10 Daily Wage Rate for Households using both Hired and Family Labour
----------------------------------------------------------------------------------------------------------
Sample Size Mean SD Min Max
Daily wage for ploughing (BDT) 27 83 25 8 150
Daily wage for sowing (BDT) 966 79 24 5 163
Daily wage for weeding (BDT) 853 70 23 1 150
Daily wage for harvesting (BDT) 1080 90 28 2 250
Daily wage for threshing (BDT) 560 89 30 12 255
---------------------------------------------------------------------------------------------------------
Several alternative hypotheses, not all mutually exclusive, may explain some of these
phenomena, which we shall review presently. The higher observed productivity of family
labour (in the absence of any hired labour) may be symptomatic of liquidity constraint
faced by very small farms.
3
There may be a lack of resources in these farms to employ
hired labour. In effect while there is further room for employment these remain unfilled
resulting in higher marginal product for those employed. By contrast farms that employ
hired labour in addition to family labour carry total employment to the point where the
expected marginal product matches the market wage. In such cases family labour may be
engaged so long as their productivity also matches the market wage. In other words the
hypothesis of allocative efficiency would be met here. Accordingly we find the marginal
product of family labour in such farms (BDT 80) to be close to the average wage (BDT
82). Over and above these observations, very small farms may also be endowed with
higher soil quality although we have no direct evidence on that.
The second stylised fact that we noted above relates to the presumed higher productivity
of hired labour when working alongside family labour. In general one would expect that
supervision by family labour would result in higher productivity than in its absence.
However, the evidence here is inconclusive as the productivities are rather close to each
other (BDT 61 vs. BDT 63). On the relative productivity of hired and family labour there
are several plausible explanations. The foremost of these is the hypothesis of adverse
selection, namely the less capable self-select themselves for wage employment as
opposed to, e.g., seeking self employment. Another possible explanation is moral hazard

3
These are the smallest of farms in the sample. For instance the size of their cultivable land is 58 percent of
that of farms engaging exclusively hired labour.
17
namely the exertion of low effort on the part of hired labour. However the observed
insignificance of the differential productivity (BDT 61 vs. BDT 63) in the presence of
supervision suggests that moral hazard need not play a dominant role.
Finally we come to the contentious issue of the compensation of rural labour. As seen
above, hired labour earns between 25 and 30 percent in excess of their marginal product
(whether working alone or working alongside family labour). Ordinarily wage premiums
exist in tight labour markets which can only be true during the harvesting season in
Bangladesh. Indeed the data confirm the peaking of wages during the harvesting and
threshing cycles. Given that we are using the average wage as the reference value the
latter may suffer from an upward bias due to large spike in employment in harvesting and
threshing periods. Additionally, employers may have incomplete knowledge about the
quality and the capability/effort of the hired labour. There might be a mismatch of skill in
various jobs like ploughing, sowing, weeding, and threshing leads to shirking behaviour.
Such a phenomenon may be further compounded by an inadequate monitoring regime.
Further the output price may not be as accurate due to its fluctuations over the harvesting
and post-harvest period. In any event, while exceeding the marginal product, the farm
wage pattern in the present sample matches the wage structure in the broader rural labour
market of Bangladesh (BBS, 2008). Hence the average daily wage figure of BDT 80 in
2004 may well be close to the subsistence wage below which the worker may be
physically unable to perform the tasks required of them. Insofar as the low productivity
per se is concerned this may arise due to the environment in which they operate. Hired
labour may be engaged in plots of marginal quality, operating with inadequate tools and
implements, and being not as knowledgeable as family labour.
7. Conclusions and Policy Recommendations
This study finds that both hired and family labour are each productive in Bangladesh
agriculture. But family labour is much more productive than hired labour and there is a
limited substitution between family and hired labour. The high correlation between hired
labour and fertiliser use for households using only hired labour demonstrates the fact that
these two inputs are complementary; the productivity of hired labour is being
compensated by using more fertiliser. Hired labour performs slightly better when there is
supervision in terms of involvement of family labour.
However the marginal return on family labour or shadow wage rate is uniformly higher
for households using only family labour, being nearly double the shadow wage rate of
hired labour used by households dependent on hired labour only. The shadow wage rate
for family labour is higher than that of hired labour used by farms using both family and
hired labour. This study also finds that the average product of family labour (BDT 94) is
higher than the efficiency wage of hired labour. Family labour needs less motivation and
less supervision. Therefore family labour is always more productive than hired and wage
labour in Bangladesh agriculture both in terms of average and marginal products.
Therefore contrary to previous studies on labour productivity in agriculture, this study
finds that family labour is more productive than hired labour in Bangladesh agriculture.
This finding is similar to the Nepalese study done by Thapa (2003). These results suggest
that it is profitable for the farm to use family labour as much as possible since there is no
cost of monitoring. In terms of relative efficiency of various inputs, households that use
only hired labour are more efficient in land use than labour, and labour use in turn is
18
more efficient than seeds, irrigation and fertiliser. Land is very scarce and is the most
expensive input and hence has the highest marginal product.
One question that remains unresolved is why hired workers earn more than their marginal
product. Several plausible answers have been reviewed above. One answer would be that
the market for hired labour is inefficient since the day labourers present an indifferent
skill distribution such that the employers are unable to observe the true marginal product,
and the equilibrium wage extracts a risk premium from the true marginal product. There
is also seasonality in agricultural labour market of Bangladesh. Therefore simple
rationing of jobs does not operate well, which would have restored the equality of the
going wage to the marginal product. It would thus be in the interest of the high quality
workers to learn how to signal their higher productivity vis--vis the going wage, e.g., by
undertaking some vocational training in farming methods (knowledge of fertilizer use,
learning how to operate the irrigation pumps etc). More generally, hired labour would be
better off by acquiring more human and physical capital. Consequently, those unwilling
to train would be better off if they move to the non-farm sector. Agricultural extension
can play a very important role in this context. Some of these phenomena require a
detailed model of labour demand at work (of various types) in this context which is a
subject of further research.
Acknowledgements

This paper has benefited from many comments and suggestions of Professor Syed M.
Ahsan. I am also grateful to Professors Thomas Sterner and Gardner Brown for their
invaluable comments and advice.
References
Bangladesh Bureau of Statistics (2008): Labour Force Survey, 2006-07.
Government of Bangladesh (2005): Unlocking the Potential National Strategy for
Accelerated Poverty Reduction, General Economics Division, Planning Commission, and
Government of the Peoples Republic of Bangladesh.
Greene, W.H. (2000): Econometric Analysis, 4
th
edition, Prentice Hall, New Jersey.

Jacoby, H. G. (1992): Productivity of men and women and the sexual division of labour
in peasant agriculture of the Peruvian Sierra, Journal of Development Economics, 37,
265-287.

Jacoby, H. G. (1993): Shadow Wages and Peasant Family Labour Supply: An
Econometric Application to the Peruvian Sierra, Review of Economic Studies, 60, 903-
921.

Linde-Rahr, M. (2005): "Differences in agricultural returns: an empirical test of
efficiency in factor input allocation using Vietnamese data", Agricultural Economics, 32,
35-45.

Thapa, P. J. (2003): Modelling the Efficiency of Family and Hired Labour Illustrations
from Nepalese Agriculture, Ashgate.
19
Wooldridge, J. M. (2003): Introductory Econometrics A Modern Approach, 2
nd
edition,
Thomson, South western.
World Bank (2000): Bangladesh: Climate Change and Sustainable Development,
Bangladesh Report No. 21104-BD, Rural Development Unit, South Asia Region,
Document of the World Bank.


Appendix
Table 11 Correlation between Inputs for Households Using Family Labour
--------------------------------------------------------------------------------------------------
lland lplow llabour lseed lfert lmncst lircst lpest
-------------------------------------------------------------------------------------------------
lland 1
lplow 0.09 1
llabour 0.53 -0.01 1
lseed 0.43 0.22 0.07 1
lfert 0.23 0.25 0.44 0.18 1
lmncst 0.03 -0.07 0.29 -0.06 0.26 1
lircst 0.02 0.15 0.28 -0.01 0.5 0.34 1
lpest 0.03 0.12 0.23 0.13 0.39 0.24 0.32 1


Table 12 Correlation between Inputs for Households Using Hired Labour
-------------------------------------------------------------------------------------------------
lland lplow llabour lseed lfert lmncst lircst lpest
-------------------------------------------------------------------------------------------------
lland 1
lplow 0.41 1
llabour 0.83 0.32 1
lseed 0.45 0.42 0.5 1
lfert 0.62 0.50 0.72 0.65 1
lmncst 0.27 0.09 0.24 0.12 0.18 1
lircst 0.14 0.18 0.33 0.29 0.48 0.2 1
lpest | 0.32 0.19 0.46 0.42 0.54 0.22 0.42 1
--------------------------------------------------------------------------------------------------

Table 13 Correlation between Inputs for Households Using Hired and Family
Labour
-----------------------------------------------------------------------------------------------
lland lplow lfam lhlab lseed lfert lmncst lircst lpest
-----------------------------------------------------------------------------------------------
lland 1
lplow 0.14 1
20
lfam 0.26 -0.2 1
lhlab 0.59 0.25 -0.16 1
lseed 0.54 0.18 0.11 0.41 1
lfert 0.43 0.25 0.18 0.47 0.47 1
lmncst 0.05 -0.15 0.08 0 -0.03 0.13 1
lircst 0.05 0.22 0.04 0.24 0.17 0.48 0.03 1
lpest 0.13 0.10 0.07 0.26 0.26 0.51 0.15 0.37 1

Das könnte Ihnen auch gefallen