Asia-Paci c Journal of Rural Development

Vol. XXVI, No. 1, July 2016

Determinants of Rural Non-Farm Employment

-Empirical Evidence from Three Villages in
Northern Bangladesh
Ranjit Kumar Sarkar and M. A. Sattar Mandal

Abstract

The importance of rural non-farm employment is largely recognised because it is

growing as an increasing means to supplementing total income of the rural
households. While the contribution of rural non-farm income is measured using
mainly output approach, it is seldom appreciated how important it is to assess the
underlying factors determining the participation of rural households in RNF
activities with respect to who participate when and where, and how the personal,
family and locational attributes do influence individual’s capacity or opportunities
to get involved in RNF activities. To analyse the determinants of rural non-farm
employment, primary data were collected from 280 sample households selected from
three villages of Rangpur district in Bangladesh i.e. one near rural market, one near
rural towns and the other near peri-urban area. The participation of household
members in rural non-farm employment was analysed using logit model and it
revealed that the level of household member’s skill was the most powerful factor to
determine rural non-farm employment. The marginal effect for level of household
skills was 0.98, which was the highest, meaning that if the household acquired skills
by way of training, doing-by learning process and education, the probability to
participate in non-farm employment increased by 98 per cent. The other variables
like availability of cultivable land, total number of earning member, female earning
member and the distance of para from nearest hat/bazaar were also significantly
important factors To promote rural non-farm employment, more emphasis is needed
for skill development of the rural people in specific trade with provision of credit and
product marketing facilities. This would need upscaling the current training
initiatives of the government and non- government agencies working at the local
level i.e. near the rural growth centres and rural bazaars.

Keywords: Non-farm employment, Rural livelihood, Agriculture, Rural

Market.

Deputy Chief, Local Government Division, Ministry of Local Government, Rural

Development and Cooperatives, Bangladesh Secretariat, Dhaka.
Professor of Agricultural Economics, and former Vice-Chancellor of Bangladesh
Agricultural University, Mymensingh, Bangladesh.

127
 Sarkar and Mandal

Introduction
The contribution of rural non-farm (RNF) activities is increasing rapidly both
nationally and internationally. Non-farm activities include mainly rural trade,
agribusiness, manufacturing, construction, transport and agro-processing
services. Worldwide, rural households engage in a variety of non-farm
activities either as pull or push factor to generate income and employment In
Bangladesh, Labour Force Survey (LFS) data show that the employment in
agricultural sub-sector declined from 63 per cent in 1995/96 to 45 per cent in
2013 and that non-agricultural employment rose from 37 per cent to 55 per
cent during the same period (BBS 1996 2010). Recent research indicates that
the rural poor engage in non-farm activities, both as a complement to their
farm activities and as a substitute and also supplement for their farm incomes
(Lanjouw and Lanjouw 2001; World Bank 2003; Hossain et al. 1994, 2000;
Khandker 1996; Hossain 2004, 2005). Level of education of rural household
members have been underscored as the important factor determining
participation in various types of RNF activities (Mecharla 2002; Sanchez
2005; Daniel 2008). Engagement in RNF traits is also seen as ‘push’ factor
from low wage surplus labour market to high paid non- farm employment
(Woldehanna and Oskan 2001; Smith et al. 2001). Rural non-farm
employment also grows as backward and forward linkages of agricultural
mechanisation process, which facilitated spread of green revolution
technologies, e.g. pump irrigation, tillage operations (Mandal and
Asaduzzaman 2002). In some cases, non-farm employment may be a coping
strategy to deal with lack of access to sufficient land or with income shocks
in agriculture. In other cases, rural households may find it profitable to
reduce their dependence on farming activities and engage increasingly in
non-farm employment.

Although rural non-farm activities are regarded as a main driver to generate

income and employment opportunities, there is generally a lack of empirical
understanding on determinants of rural non-farm employment by duration of
time involvement. Besides, recent studies in rural non-farm employment
have underscored the importance of determinants of rural non-farm
employment using income and output data but very few studies investigate
into the determinants using time allocation by farm and non-farm households
engaged in different activities in the rural areas. Timing of work by the rural

128
Determinants of Rural Non-Farm Employment...

households considering status of households and their members is also

important. In this context, the frequently asked questions include: what
determines the participation in rural non-farm (RNF) employment, when and
how much time is allocated to RNF work and how opportunities for RNF
jobs can be increased?

Moreover, locations of households are also an important determinant of the

extent of RNF employment opportunities. The households who are in the
remote villages, participation in non-farm activities may not be the same as
the households who live near the urban and per-urban locations. This raises
the questions as to what factors are more important to determine participation
in non-farm activities and how these can be accelerated for the rural
households in different locations.

In this context, this study investigates into the main determinants or factors
that push or pull the household members to participate in non-farm activities
and how these factors act differently in different locations i.e. near rural
market, peri-urban, proximate to rural town.

The analysis was done using a logit model, which identified the important
determinants of rural non-farm employment. The first section of the paper
includes the introduction, while section II discusses the objectives of the
paper. Section III discusses the methodology and section IV incorporates
results and discussion. While the conclusions are drawn in section V,
recommendations are made in section VI. Limitations of the study are given
in the final section.

Objectives of the Study

The overall objective of the study is to highlight some major issues
concerning rural non-farm employment, determinants of rural non-farm
employment, factors determining non-farm employment participation. The
specific objectives are:
To find out determinants of rural non-farm employment;
To recommend policies and strategies to promote rural non-farm
employment in Bangladesh.

129
 Sarkar and Mandal

Methodology
The study was done on the basis of the primary data collected from three
villages of Pirganj upazila under Thakurgaon district of Rajshahi division.
Study area was selected following a stepwise approach giving emphasis on
poverty incidence, intensity of farm related non-farm enterprises, intensity of
rural households, communication and market linkages with growth centres,
rural town and rural bazaar.

Division wise poverty incidence data from Poverty Monitoring Survey 2004
BBS (2004b) showed that among the divisions of the country the highest
poverty was observed in Rajshahi Division in respect of all measures of the
three poverty measures such as head count ratio, poverty gap and squared
poverty gap which were 61.6%, 18.1% and 6.9% respectively . Rajshahi
Division was thus selected for this study.

Rajshahi Division consists of 16 districts. According to Income Poverty Index

(IPI) the districts were classified under different groups such as income
poverty index value up to 30, 30.1 to 35, 35.1 to 40, 40.1 to 45, 45.1 to 50 and
50.1 & above. Rajshahi, Bogra, Dinajpur, Joypurhat, Naogaon, Natore and
Pabna districts fall under the income poverty index class up to 45, Chapai
Nawabgonj and Thakurgaon districts fall under 45.1 to 50 and Gaibandha,
Kurigram, Lalmonirhat, Nilphamari, Panchagarh, Rangpur and Sirajgonj
districts fall under 50.1 and above classes. For district selection medium
poverty was considered. As Chapai Nawabgonj and Thakurgaon districts
remained 45.1 to 50 IPI classes, these two districts were selected among the 16
districts. Secondly, between the two districts, rural population was 90%
in Thakurgaon districts, which was higher than Chapai Nawabgonj district
of 81%. For this reason, Thakurgaon district was selected as the study area.

Thakurgaon district comprises of five Upazilas namely Baliadangi, Haripur,

Pirganj, Ranisankail and Thakurgaon Sadar. One Upazila was selected for
study considering the intensity of farm related non-farm enterprises. The
enterprises like tractor, power tiller, DTW, STW, fertiliser dealer, seed
dealer, insecticide dealer, husking mill, chira mill, poultry farm, dairy farm
and growth centre etc. were considered. First, Upazila wise data on the
availability of the enterprises were collected and then weight or point was

130
Determinants of Rural Non-Farm Employment...

given against each enterprise. By summing up the points of all enterprises

against each Upazila total score or point was found out. According to
weightage, Thakurgaon Sadar Upazila ranked highest and Pirganj stood
second. Although Thakurgaon Sadar Upazila ranked highest it was not
selected for study because Sadar Upazila is always urban biased. For this
reason, Pirganj Upazila was considered as the study area.

Pirganj Upazila comprises of ten Unions namely Bhomradah, Boirchuna,

Daulatpur, Hazipur, Jabarhat, Khongaon, Kosaranigonj, Pirganj sadar,
Sengaon and Syedpur. Among the unions two unions were selected for study
considering the intensity of farm related non-farm enterprises. The
enterprises like power tiller, DTW, STW, fertiliser dealer, seed dealer,
insecticide dealer, weeder, hand spray machine, drum seeder, husking mill,
poultry farm, dairy farm, growth centre and hat-bazar etc. were considered.
First, Union Parished wise data on the availability of the enterprises were
collected and then weight or point was given against each enterprise. By
summing up the points of all enterprises against each union total score or
points was found out. According to weight, Daulatpur union was ranked
highest and Jabarhat union ranked 2 nd highest. For this reason, Daulatpur and
Jabarhat union were considered as study area.

Daulatpur union comprises of two blocks namely Block-1 and Block-2. Each
Block consists of 9 villages. On the other hand Jabarhat union consists of 16
villages. Each Block has eight villages. Two villages from Daulatpur union
and one village from Jabarhat union were selected for the study. Three
villages were selected considering the following criteria:
1. Village Daulatpur was selected because it is located near urban area
or pourashava bazar;
2. Village Jabarhat was selected because it is located near a growth
centre or rural town; and
3. Village North Noyapara was selected because it is located near a
rural bazar.

Households of the Study Area

The selected villages were surveyed through a simple survey schedule. The
households of the villages were classified as farm and non-farm households.
Village wise households’ classification was given in the Table 1.

131
 Sarkar and Mandal

Table 1: Village-wise Household Classification

NFHH as
FHH as %
Villages FHH NFHH Total % of Total
of THH
THH

Daulatpur 68 178 246 28 72 100

Jabarhat 363 269 632 57 43 100

North Noyapara 74 90 164 45 55 100

Total Household 505 537 1042 48 52 100

Source: Sarker (2011).

In the study villages, there were 1042 households of whom 505 were farm
households and 537 were non-farm households which accounted for 48 and
52 per cent respectively.

Sample Household Selection Criteria

Selections of households depend on objectives of the study. For this survey,
households were selected as per following criteria:
To measure the determinants of rural non-farm employment,
proportionate samples from all classes of households were
considered. First, households were classified as farm and non-farm
households. Then, both farm and non-farm households were
classified as households having cultivable land and those having no
cultivable land. Again, non-farm households having cultivable land
were classified as pure and mixed non-farm households. Lastly, non-
farm households having no cultivable land were classified as farm
related and other non-farm households.

Sample Size Determination Criteria

Sample size was determined according to the objectives of the study.
Objective wise sample size determination criteria are given as follows:

To measure the determinants of rural non-farm employment, sample

size of farm and non-farm households were determined using sample
size determination technique.

132
Determinants of Rural Non-Farm Employment...

Sample Size Determination Approach

There are two alternative approaches for determining the size of sample. The
first approach is 'to specify the precision of estimation desired and then to
determine the sample size necessary to ensure it.' and the second approach
uses Bayesian statistics to weigh the cost of additional information against
the expected value of the additional information. The first approach is
capable of giving a mathematical solution, and as such is frequently used as a
technique of determining sample size ‘n’. For this reason first approach was
applied to determine the sample size.

To find the sample size for estimating a proportion, it is necessary to specify

the precision and the confidence level and considering that the sample size
was worked out as under:

Since the confidence interval for universe proportion, p is given by

pq
P = z.
n

Where, p = sample proportion

q = 1- P
z = the value of standard variation at a given confidence level
n = desired sample size
If we take the value of P=0.5, n will be maximum and the sample will yield
at least the desired precision (Kothari 2006).

With the given precision rate, the acceptable error “e” can be expressed as
under:
pq
e = z.
n
pq
e2 = z2.
n
n. e2 = z2. PQ
z2pq
N=
e2

133
 Sarkar and Mandal

The above formula gives the size of the sample in case of infinite population
in the universe. According to Kothari (2006), in case of finite population the
above stated formulae will be changed as under:
z2pqN
N=
e2(N-1)+ z2pq

Where p = Sample proportion, q = 1- p

z = the value of standard variation at a given confidence level and to be
worked out from the table showing area under normal curve.
e = Acceptable error (the precision)
n = size of sample
N = size of population

Selection of the Sample Households

From the total households, sample households were selected using sample
size determination technique. Following Kothari (2006), sample size was
determined by adopting the following procedure:

Let the value of p = 0.5 in which case ‘n’ will be the maximum and the
sample will yield at least the desired precision.
Therefore, q= 1- p = 1-0.5 =.5
z = 1.96
e= Acceptable error (the precision) = 0.05
N= 1042
n=sample size?
z2pqN
N=
e2(N-1)+ z2pq

(1.96)2 x 0.5 x 0.5 x 1042

N=
(0.05)2(1042-1)+ (1.96)2 x 0.5 x 0.5
3.84 x 0.25 x 1042
N=
0.0025 x 1041+ 3.84 x 0.25
0.96 x 1042
N=
2.60+ 0.96
999.14
N=
3.56
n = 280.54 = 280

134
Determinants of Rural Non-Farm Employment...

Analytical Techniques
To analyse the determinants of rural non-farm employment as main
occupation of the 280 sample households and to attribute a weight to these
variables logit model was used. In logit model, the dependent variable was a
dummy (i.e. a dichotomous variable which takes a value of 0 and 1)
(Mecharla 2002).

The LOGIT Model

The function used in logit is the natural log of the odds ratio (Gujarati 2003).

We know
if 1 2Xi +ui

i given Xi, E ( i |Xi), can be interpreted as the

conditional probability that the event will occur given Xi, that is, Pr=E
(
Assuming E (ui) =0, as usual (to obtain unbiased estimation), we obtain
E ( i|Xi 1 2Xi -------------------- (2)
If Pi i =1(that
is, the event occurs), and (1-Pi) = probability
i i will contain the
following probability distribution.
Yi Probability
0 (1- Pi)
1 Pi

Now, by the definition of mathematical expectation (Gujarati 2003), we

obtain
E ( i) = 0(1-Pi) + 1 Pi = Pi --------------- (3)
Comparing (2) with (3), we can equate
Pi = E( i=1|Xi 1 2Xi ------------- (4)
We can represent the equation as follows
1
Pi = E( i=1|Xi i) = 1 2 Xi ---------------- (5)
1 e
For ease of exposition we write (5) as
1 ez
Pi = = --------------------- (6)
1 e z
1 ez

135
 Sarkar and Mandal

1 2Xi, and
1
(1-Pi) = ---------------- (7)
1 ez
We can also write
Pi 1 ez
= z
= ez ------------------ (8)
1 Pi 1 e
Pi
Where, is simply the odds ratio
1 Pi
Now, if we take the natural log of (8), we obtain an interesting result, namely
Pi
Li = Ln = Zi = 1 2Xi ---------- (9)
1 Pi
Here, L, the log of the odds ratio, is not only linear in x, but also linear in the
parameter. L is called the logit, and hence the name logit model. In logit
models, the dependent variable is a dummy (i.e. a dichotomous variable
which takes a value of 0 and 1). Here, it takes the value 1 if the household
has main worker whose primary occupation is RNFE prior to the survey and
0 otherwise.

In logit model, the equation that used was as follows:

Yi = ao + b1X1i + b2X2i + b3X3i + b4X4i + b5X5i ....... b9X9i +b10X10i + ui
Where i denotes households, and Yi = A HH is engaged in non-farm activity,
if any working member has, as a primary occupation at least 183 days in a
year or more than 50% working time(in one or several activities);
a0 = Intercept;
X1 = Size of own cultivable land (in hectares);
X2 = Age of household head (in years);
X3 = Education of household head (number of years of schooling);
X4 = Number of literate in the family;
X5 = Total earning members in the family (number);
X6 = Female earning members in the family (number);
X7 = Level of household skill (if the household is skilled the dummy
takes 1, 0 otherwise);
X8 = Distance of para/house from nearest hat/bazaar (in kms);
X9 = Village location Dummy-1 =1 for Jabarhat (near rural town)
and 0 otherwise;

136
Determinants of Rural Non-Farm Employment...

X10 = Village location Dummy-2 =1 for North Noyapara (near rural

bazaar) and 0 otherwise;
ui = Disturbance term with the classical properties where i denotes
the household interviewed.

Results and Discussions

General Results of the Logit Model

Logit functions were generated and used for the purpose of analysing
determinants of rural non-farm employment. The logit model for
determinants of non-farm employment has been estimated separately for the
four situations like (1) pooled (three villages) village level data on all 280
households, (2) Daulatpur village level data on 66 households, (3) Jabarhat
village level data on 170 households and (4) North Noyapara village level
data on 44 households. The estimates of the parameters of the model are
presented in Table 2 and Appendix 1, 2 and 3.

From the analysis it was found that six out of ten variables were obviously
significant with 1% level of significance i.e. own cultivable land, total
earning member, female earning member, level of household skill, village
dummy 1 and village dummy 2 (Table 2). However, variable wise results
were discussed in turn as follows:

Size of Own Cultivable Land

The size of cultivable land owned by the household is taken for
measurement. For the pooled sample, the size of own cultivable land variable
was found to be negatively associated with RNFE and statistically significant
at 1 per cent level. The coefficient of the own cultivable land was negative.
This implies a negative correlation between the size of the own cultivable
land and the probability of being involved with non-farm employment. This
may be attributed to the predominance of small and marginal farmers or
landless households in these villages. As such the lower the cultivable land,
the higher will be the probability of being engaged in non-farm employment.
The reason is that if a household had a large farm, family members could
allocate more time on the agricultural production and divert less time in non-
farm activities.

137
 Sarkar and Mandal

The marginal effect of own cultivable land was -0.09 which implies that, at the
mean, if own cultivable land increased by one unit (one hectare in this exercise
the probability of participation in non-farm employment decreased by 9
percentage points or vice-versa. The negative relationship between own
cultivable land and non-farm employment suggested that the employment
diversification in rural areas was low or distress. The odds ratio of own
cultivable land was .695 implying that the probability of being employed in non-
farm employment is 30 per cent lower for cultivable land owners than others.

Age of the Household Head

The coefficient of age of the household head being employed in RNFE was
negative and insignificant, indicating rigidity in shifting of activities for the
elder person. The marginal effect of age of the household head was -.007,
implying that with one unit (1 year) increase in the age, the probability of
being employed in non-farm employment decreased by 0.70 percentage.

Education of the Household Head

Education is a potentially important determinant of RNFE. Education
improves an individual’s prospects for non-farm jobs as well as increases his
or her ability to allocate time to work efficiently among income producing
activities. However, in early development phases many rural non-farm
activities require levels of schooling. The coefficient for education of
households head was found 0.047. The result was positive and insignificant,
indicating a rather strong relationship with RNFE. Its marginal effect was
positive, suggesting that household head with higher education were more
likely to seek non-farm employment in rural villages. The marginal affect
was 0.116, implying that if education of the household head increases by one
year, the probability to participate in non-farm employment will be increased
by 11.60 percentage points.

No. of Literate in the Family

The coefficient of the no. of literate in the family was positive. This implies a
positive correlation between the no. of literate in the family and the
probability of being involved in non-farm employment. The marginal effect
of no. of literate in the family was 0.036. This implies that, at the mean, if no.
of literate in the family increases by one unit (one person in this exercise) the

138
Determinants of Rural Non-Farm Employment...

probability to participate in non-farm employment increased by 3.60

percentage points. The reason is that the parents in most households
encourage their children to be educated and employed in better non-farm
occupations.
Table 2: Determinants of Rural Non-Farm Employment-Estimates of
Logit Model for Pooled Data (Three Villages’ Altogether)
Whole Sample (N=280)

Variables

Cultivable land (in hectares) -0.364*** 0.11 -3.21 0.001 0.695 -0.089
Age of the household head (no.) -0.028 0.02 -1.52 0.129 0.973 -0.007
Education of the household head 0.047 0.07 0.72 0.472 1.048 0.116
No. of literate in the family (no.) 0.146 0.16 0.92 0.360 1.157 0.036
Total earning member (no.) 1.112*** 0.32 3.43 0.001 3.040 0.272
Female earning member (no.) -1.396*** 0.52 -2.67 0.007 0.247 -0.342
Level of household skill 3.996*** 0.49 8.17 0.000 54.395 0.979
House/Para distance from -0.691*
0.36 -1.90 0.058 0.501 -0.169
nearest hat/bazaar
Village location dummy-1 -2.276*** 0.81 -2.81 0.005 0.103 -0.558
Village location dummy-2 -1.577*** 0.53 -2.99 0.003 0.207 -0.386
Constant 0.537 1.18 0.46 0.648 1.710
- 2 Log likelihood 183.53
Number of observation 280
Cox & Snell R Square 0.512
Negelkerke R Square 0.691
Source: Sarkar (2011), *** Significant at the 1% level, *Significant at the 10% level.

Total Earning Member

The coefficient of the total earning members in the family was positive and
significant at 1% level. This implies a positive correlation between the no. of
total earning members in the family and the probability of being involved
with non-farm employment. The marginal effect of total earning members in
the family was 0.272. This implies that, at the mean, if numbers of total
earning members in the family increase by one unit the probability of
participation in non-farm employment increased by 27 percentage points.
The odds ratio of total earning members was 3.040, implying that the
probability of being employed in non-farm employment is 3.040 times higher
for one unit increase in earning members in the family, compared to others. If
total earning members increases, RNFE activities are randomly distributed

139
 Sarkar and Mandal

across persons, there are more persons in larger households, so there is a

relatively greater chance that at least one working member will be in non-
farm employment. Secondly it was observed during field work that once a
member of the household is engaged in RNFE, other younger members tend
to follow him/her.

Female Earning Member

The coefficient of the female earning members in the family was negative
and significant at 1% level. This implies a negative correlation between the
female earning members in the family and the probability of being involved
with non-farm employment. The marginal effect of female earning members
in the family was -0.342. This implies that, at the mean, if number of female
earning members in the family increased by one unit, the probability of non-
farm employment decreased by 34.20 percentage points. The odds ratio of
female earning members was 0.25, implying that the probability of being
employed in non-farm employment decreases 75 per cent for the one unit
increase female earning members. The implication is that female members of
the family are more illiterate than male members, their skill is low and for
this reason they had to work in the farm sector. From the field visit
observation it was found that, most of the poor female members had to work
in the field as agricultural wage labour.

Level of Household Skill

The coefficient related to skill had as expected, a positive sign and was
significant at 1 per cent level. The marginal effect was 0.979. This implies
that if the HH has acquired skills by way of training, learning- by doing
process and education, the probability of non-farm employment increased
97.90 percentage points. This variable was significant for all the villages.
The odds ratio was 54.395, implying that the probability of being employed
in non-farm employment is 54.395 times higher for the households members
having skill than others. People, who are trained for skilled occupations, will
generally have better opportunities of employment and their productivity is
said to be higher, which in turn is reflected in their relatively higher earnings.
It is expected that skilled household members will have a better chance of
taking up non-farm employment. So, the expected relationship between
skilled and RNFE is positive. Skill training includes training on mechanics,
automobiles, carpentry, masonry, tailoring, machine operators, electronics,

140
Determinants of Rural Non-Farm Employment...

surveyor, battery rechargers, driving, tonga making, chanachur making etc.

Those related with technical education includes - village doctor, diploma
engineers, and agricultural diploma.

House/Para Distance from Nearest Hat/Bazaar

For the pooled sample, the Para distance from urban hat/rural growth centre
and rural bazaar variable was found to be negatively associated with RNFE
and it was statistically significant. The coefficient of the variable was
negative. This implies a negative correlation between the para distance from
urban hat/rural growth centre and rural bazaar and the probability of being
involved with non-farm employment.

The marginal effect of a unit increase in para distance from urban hat/rural
growth centre and rural bazaar on non-farm employment at the mean level of
all variables was -0.169. This implies that, at the mean, if Para distance from
urban hat/rural growth centre and rural bazaar increased by one unit (one km
in this exercise) the probability of non-farm employment decreased by 16.90
percentage points.

Village Location Dummy 1

The village location dummy-1 (Jabarhat =1, and Daulatpur=0) was
statistically significant. This implies that there was significant difference
between the two villages. The model of rural RNFE spelled out here is that
households are more inclined to non-farm activities in Daulatpur, with the
coefficient being negative. Perhaps it is the nearness of the Pirganj
Pourashava which induces people of given skills and literacy in Daulatpur
village to move to Pirganj Pourashava. The odds ratio was 0.103, implying
that the probability of being employed in non-farm employment was 90 per
cent lower for the households of Jabarhat village than Daulatpur village.

Village Location Dummy 2

The village location dummy-2 (North Noyapara=1, Daulatpur=0) was
statistically significant. This implies that there was significant difference
between the two villages. The model of rural RNFE spelled out here is that
households are more inclined to non-farm activities in Daulatpur, with the
coefficient being negative. Perhaps it was the nearness of the Pirganj
Pourashava which caused people of given skills and literacy in Daulatpur

141
 Sarkar and Mandal

village to move to Pirganj Pourashava The odds ratio was 0.21, implying that
the probability of being employed in non-farm employment was 79 per cent
lower for the households of North Noyapara village than Daulatpur village.

Summary of the Logit Results

The logit regression results are summarised in Tables 3 and 4. The
conclusion for the logit regression analysis of the determinants of whether a
household participate in non-farm employment in Daulatpur, Jabarhat and in
the pooled data for the three villages revealed that it was significantly
negatively related to own cultivable land size but for North Noyapara village
it was insignificantly and positively correlated.
Table 3: Summary of the Coefficients Obtained in the Logit Model
Coefficients
N.
Independent variables Daulatpur Jabarhat Pooled
Noyapara
Own cultivable land (in hectares) (-)1.035** (-)0.303*** 0.309 (-)0.364***
Age of the household head -0.060 -0.043 0.045 -0.028
Education of the household head 0.073 -0.056 0.163 0.047
No. of literate in the family 0.753 0.155 -0.854 0.116
Total earning member (no.) (+)2.370** (+)1.341** -0.928 (+)1.112***
Female earning member (no.) -1.910 -2.451 1.418 (-)1.396***
Level of household skill (+)7.189*** (+)4.793*** (+)4.174*** (+) 3.996***
House/Para distance from nearest
-0.735 (-)1.180** -1.511 -0.691
hat/ bazaar
Village location dummy-1 (-)2.276***
Village location dummy-2 (-)1.577***
Constant -3.175 0.298 -1.511 0.547
- 2 Log likelihood 22.425 97.441 34.241 183.526
Number of observation 66 170 44 280
Cox & Snell R Square 0.565 0.546 0.451 0.512
Negelkerke R Square 0.818 0.734 0.603 0.691
Source: Sarkar 2011, PhD Thesis, *** Significant at the 1% level, ** Significant at the 5% level.

This relationship was stronger (i.e. the decrease is greater) in the Daulatpur
(nearest pourashava) village, compared to the other two villages. In either
case, big farmers were less likely to have a main RNFE, suggesting that farm
land actively reduced total-RNFE participation. Households with less farm
land had more RNFE; supporting the hypothesis that distress diversification
(DD) dominates growth linkages. In the pooled data set; for a one hectare
increase in land holding size, the probability of household having any non-
farm employment decreased by 3.6 percentage points.

142
Determinants of Rural Non-Farm Employment...

Wealthier villagers invest more in their children’s education, which increased

the likelihood of them taking non-farm employment; they also consumed
more goods and services, meaning more work for others. Except North
Noyapara village, total earning member was significantly and positively
related to non-farm employment in each village, and on the pooled data. The
marginal effect of total earning member on the probability of households
having non-farm employment was greater in the Jabarhat village, compared
to the other villages.
Table 4: Summary of the Marginal Effects Obtained in the Logit Model
Marginal Effects
Variables Daulatpur Jabarhat N. Noyapara Pooled
Own cultivable land (in hectares) -0.117** -0.073*** 0.073 -0.089***
Age of the household head -0.007 -0.010 0.011 -0.007
Education of the household head 0.008 -0.014 0.039 0.116
No. of literate in the family 0.085 0.038 -0.202 0.036
Total earning member (no.) 0.268** 0.324** -0.220 0.272***
Female earning member (no.) -0.216 -0.593 0.335 -0.342***
Level of household skill 0.812*** 1.159*** 0.987*** 0.979***
House/Para distance from nearest
0.083 -0.285** -0.357 -0.169
hat/ bazaar
Village location dummy-1 -0.558***
Village location dummy-2 -0.386***
- 2 Log likelihood 22.43 97.44 34.241 183.53
Number of observation 66 170 44 280
Cox & Snell R Square 0.57 0.55 0.451 0.512
Negelkerke R Square 0.82 0.73 0.603 0.691
Source: Sarkar 2011, PhD Thesis, *** Significant at the 1% level, ** Significant at the 5% level.

In North Noyapara village, female earning members were more likely to be

involved in non-farm employment than male members. The marginal effect
of female earning member on the probability of HH having non-farm
employment was 0.335, which indicated that female participation was greater
compared to the male earning members.

The dummy variable ‘level of Household skills’ (any marketable expertise,

e.g. motor mechanic, astrologer, photographer, radio repairs, tailors etc.) was
found to be significant in all the villages and also for pooled data i.e. three
villages altogether. A lack of skills provides an entry barrier into the wider
market place. It was very striking to see that the ‘skill’ dummy had the

143
 Sarkar and Mandal

expected signs a priori. A household with a main skilled worker had a 98 per
cent greater chance of being involved in the non-farm sector. Again the
requirement of special skills for entry into the non-farm sector is stressed.

In Jabarhat village distance of para from growth centre had significant effects
on participation of rural non-farm employment. Jabarhat village was big and
scattered long away from the Jabarhat growth centre. For Jabarhat village,
the marginal effects of Para distance from urban hat/rural growth centre and
rural bazaar was -0.285, which indicates that if distance of para increased by
1 unit (1 km) then participation of non-farm employment decreased by 28.5
percentage points.

Summary and Conclusion

Results from the study showed that lower the own cultivable land, the higher
was the probability of being engaged in non-farm employment. It was
expected because households owning large farm, could engage most of their
family members on agricultural production and divert less time in non-farm
activities. With one unit (one year) increase in the age, the probability of
being employed in non-farm employment decreased by 0.70 percentage
point, indicating rigidity in shifting of activities for the elder person. One
year of education increased the probability of non-farm employment by
11.60 percentage point. Education improves an individual’s prospects for
non-farm jobs as well as increases his or her ability to allocate time to work
efficiently among income producing activities. If the number of literate in the
family increased by one unit (one person in this exercise) the probability of
being employed in rural non-farm employment increased by 3.60 percentage
point. The reason is that the parents in most households encourage their
children to be educated and employed in better non-farm occupations. If the
number of total earning members in the family increased by one unit, the
probability to participate in rural non-farm employment increased by 27.20
percentage point. It implies that if total earning members increases, then
RNFE activities is randomly distributed across persons since there are more
persons in larger households, there is a relatively greater chance that at least
one working member would be in non-farm employment. Secondly, it was
observed during field work that once a member of the household is engaged
in RNF activities, other younger members tend to follow him/her. The

144
Determinants of Rural Non-Farm Employment...

marginal effect was 0.98 for level of household skills, implying that if
household acquired skills by way of training, learning by doing, education
and training, the probability to participate in rural non-farm employment
increased by 98 percentage points. The implication of it is that the people,
who were trained for skilled occupations, had generally better opportunities
of employment and productivity. As a result, they would be able to gain
relatively higher earnings. It is expected that skilled household members
would have a better chance of taking up non-farm employment. If house/para
distance from urban hat/rural growth centre and rural bazaar increased by one
unit (one km in this exercise) the probability to participate in rural non-farm
employment decreased by 16.90 percentage point.

Policy Recommendations
Since participation in rural non-farm employment mainly depends on the
skill level of the persons, emphasis should be given on skill development
of the rural people in specific trade. As the probability of an individual
entering a non-farm occupation as full-time economic activity increases
with level of skill it is necessary to invest in human capital development
training. Education and skill development training determines entry into
these jobs and these include both formal and non-formal education and
training.
It has been clearly identified in the study that most of the households
were poor, low literate and unskilled. Low level of education and skill of
the earning members hindered them to enter on non-farm employment.
The poor landless households of remote villages could not enrol their
children for school education because of poverty. Most of the guardians
had to engage their children in a certain enterprises as helper so that they
could earn money, although very little in most cases.

The young age children of poor families were forced to participate in

non-farm enterprises like tailoring, masonry, carpentry, rural mechanics
as helpers and apprentices. But the scope to participate as helpers in the
above mentioned activities was low because of limited enterprises and
capital available in the area. It is therefore necessary to establish formal
or informal skill development training centres in rural bazaar, rural
town and urban hat through government intervention, private sector and
NGO initiatives.

145
 Sarkar and Mandal

Besides, participation on rural non-farm employment also depends on

distance of house/para from nearest hat/bazaar. Road infrastructure is
particularly important for the non-farm sector development where the
distance between location of production and that of the market is long.
Improvement in road infrastructure will lead to improved transport
services and as a consequence trade and other business will increase. So,
increased investment in developing and maintenance of road
communications in rural areas is emphasised. For certain higher level
RNF activities such as power tiller operated farm machinery operation or
repairing electronic devices, provision of training with necessary tools
and operational support will also be needed.

Limitations
This study was conducted in a small geographical area in northern
Bangladesh with limited time and resource. Therefore, generalisation of the
findings for the whole country may not be possible. Other limitation of the
study is that the influence of rural remittance could not be ascertained due to
lack of data. Further research is also needed to study the dynamics of rural
income in the rapidly changing scenarios in the rural non-farm sector.

References

BBS. 1996. Report on Labour Force Survey 1995-96. Dhaka, Bangladesh:

Bangladesh Bureau of Statistics.
BBS. 2008. Report on Labour Force Survey 2005-06. Dhaka, Bangladesh:
Bangladesh Bureau of Statistics.

BBS. 2013. Report on Labour Force Survey 2013. Dhaka, Bangladesh: Bangladesh
Bureau of Statistics, Statistics.

Daniel, C. 2008. Determinants of Rural Non-farm Employment and Income in

Paraguay. Thesis submitted to Graduate School of Auburn University.
Auburn, USA: Auburn University.

Gujarati, D.N. 2003. Basic Econometrics (fourth edition), 580-635. New York:
McGraw-Hill Inc.

Hossain, M., M. Rahman and A. Bayes, 1994. 'Rural Non-farm Economy in

Bangladesh: A Dynamic Sector or a Sponge of Absorbing Surplus
Labor?' SAAT Working Paper. New Delhi: ILO..

146
Determinants of Rural Non-Farm Employment...

Hossain, M., B. Sen and H.Z. Rahman. 2000. 'Growth and Distribution of Rural
Income in Bangladesh: Analysis Based on Panel Survey Data.' Economic
and Political Weekly 35 (52-53): 4630-4637.

Hossain, M. 2004. 'Rural Non-farm Economy-evidence from Household Surveys.'

Economic and Political Weekly 39 (39): 4053-40.

Hossain, M. 2005. 'Growth of the Rural Non-farm Economy in Bangladesh:

Determinants and Impact on Poverty Reduction'. In Proceedings of
International conference Rice is life: scientific perspectives for the 21st
century, 436-439.

Khandker, S.R. 1996. 'Role of Targeted Credit in Rural Non-farm Growth.' The
Development Studies 24 (3-4): 181-193.

Lanjouw, J.O. and P. Lanjouw. 2001. 'The Rural Non-farm Sector: Issues and
Evidence from Developing Countries.' Agricultural Economics 26: 1-23.

Mandal, M.A.S. and M. Asaduzzaman. 2002. Rural Non-farm Economy in

Bangladesh: Characteristics and Issues for Development. BIDS-DFID
workshop paper.

Mecharla P.R. 2002. The Determinants of Rural Non-farm Employment in Two

Villages of Andhra Pradesh (India). Prus Working Paper 12. Palmer,
Brighton: Poverty Research Unit, University of Sussex.

Sanchez, V. 2005. The Determinants of Rural Non-farm Employment and Incomes in

Bolivia. A thesis submitted to Department of Agricultural Economics,
Michigan State University.

World Bank. 2003. Rural Poverty Alleviation in Brazil – Toward an Integrated

Strategy. Washington D.C.: World Bank.

Woldenhanna, T. and A. Oskam. 2001. 'Income Diversification and Entry

Barriers: Evidence from the Tigray Region of Northern Ethiopia.'
Food Policy 26: 351-36.

147
 Sarkar and Mandal

Appendices

Appendix 1: Determinants of Rural Non-Farm Employment: Estimates

of Logit for Daulatpur Village
Sample (N=66)

Marginal
Standard T Odds
Variables Coefficient Sig. / Impact
Error Ratio Ratio
Effects

Own cultivable land -1.035** 0.42 -2.45 0.014 0.355 -0.117

Age of the household head -0.060 0.07 -0.85 0.397 0.942 -0.007

Education of the household

0.073 0.23 0.32 0.751 1.076 0.008
head

No. of literate in the family 0.753 0.62 1.22 0.222 2.124 0.085

Total earning member 2.370** 1.16 2.05 0.041 10.698 0.268

Female earning member -1.910 1.59 -1.20 0.230 0.148 -0.216

Level of household skill 7.189*** 2.57 2.80 0.005 1324.835 0.812

House/Para distance from

-0.735 0.59 -1.24 0.214 2.086 0.083
nearest hat/ bazaar

Constant -3.175 3.53 -0.90 0.368 0.0418

- 2 Log likelihood 22.429

Number of observation 66

Cox & Snell R Square 0.565

Nagelkerke R Square 0.818

Sarkar, 2011, PhD Thesis, *** Significant at the 1% level, ** Significant at the 5% level.

148
Determinants of Rural Non-Farm Employment...

Appendix 2: Determinants of Rural Non-Farm Employment- Estimates

of Logit Model for Jabarhat Village
Whole Sample (N=170)

Standard T Odds Marginal

Variables Coefficient Sig.
Error Ratio Ratio Effects

Own cultivable land -0.303*** 0.13 -2.33 0.010 0.56 -0.073

Age of the household head -0.043 0.03 -1.61 0.808 0.99 -0.010

Education of the
-0.056 0.09 -0.62 0.926 1.01 -0.014
household head

No. of literate in the family 0.155 0.23 0.69 0.310 1.43 0.038

Total earning member 1.341** 0.51 2.63 0.023 7.80 0.324

Female earning member -2.451 0.87 -2.80 0.225 0.20 -0.593

Level of household skill 4.793*** 0.78 6.11 0.000 368.64 1.159

House/Para distance from

-1.180** 0.54 -2.20 0.012 0.08 -0.285
nearest hat/ bazaar

Constant 0.298 1.38 0.217 0.735 0.47

- 2 Log likelihood 97.441

Number of observation 170

Cox & Snell R Square 0.546

Nagelkerke R Square 0.734

Source: Sarkar 2011, PhD Thesis, *** Significant at the 1% level, ** Significant at the 5% level.

149
 Sarkar and Mandal

Appendix 3: Determinants of Rural Non-Farm Employment- Estimates

of Logit Model for North Noyapara Village

Sample (N=44)

Marginal
Standard T Odds
Variables Coefficient Significance / Impact
Error Ratio Ratio
Effects

Own cultivable land 0.309 0.45 0.69 0.489 1.362 0.073

Age of the household head 0.045 0.05 0.90 0.367 1.046 0.011

Education of the household

0.163 0.18 0.91 0.361 1.177 0.039
Head

No. of literate in the family -0.854 0.55 -1.54 0.123 0.426 -0.202

Total earning member -0.928 1.06 -0.88 0.380 0.395 -0.220

Female earning member 1.418 1.90 0.75 0.454 4.130 0.335

Level of household skill 4.174*** 1.42 2.95 0.003 64.987 0.987

House/Para distance from

-1.511 1.31 -1.15 0.248 0.221 -0.357
nearest hat/ bazaar

Constant -0.250 2.38 -0.10 0.917 0.779

- 2 Log likelihood 34.241

Number of observation 44

Cox & Snell R Square 0.451

Nagelkerke R Square 0.603

Source: Sarkar 2011, PhD Thesis, *** Significant at the 1% level.

150