categories. The crude probability of access to international migration was 6.5% and
that of regular job was 23.5%.
Programme includes training, credit, information, veterinary services etc. It is
important to identify whether they contribute to the choice of any of the livelihoods.
To examine this we estimate equation (1) of asset-base framework using multinomial
logistic regression. Asset variables included in the model are human capital such as
education, household size and composition, age and training; natural capital such as
land and its quality; financial capital such as credit; physical capital such as business
assets, agricultural machineries; and social capital such as membership in the
programme and other organisations, etc. Market access and location variables are also
included in the model. The results are presented in Table 4. Statistically, model fit is
acceptable. Most of the results appear plausible. Definition of the variables along with
their mean and standard deviation are reported in Table A1 in the appendix.
Significant results are interpreted below.
Family education is a significant determinant of regular job based livelihood; the
more educated the family members the more they prefer regular job. An extra year of
median level of schooling of seven plus members in the household causes 1.3 times
higher probability of choosing a regular job than agriculture.
Households with higher dependency burden are associated less with wage labour.
They are more likely to choose agriculture and/or livestock as a major source of
livelihood than wage labour indicating that extra burden cannot be met with the low
paid wage labour income, instead dependent members could help raising extra unit of
livestock or could add extra unit value to agriculture and livestock. A households
having extra adult has a better chance of getting a regular job and less likely to enter
wage labour than agriculture with livestock. The likelihood of all non-farm
occupations except regular job is higher for larger families. Effective training in
poultry related activities reduces the likelihood of diversifying through non-farm
activities.
Land ownership is negatively associated with all non-farm routes but none is
significant at 5% level, only other category #5 is significant at 10%. In absence of soil
quality data, productivity of land in terms of log of per acre net income was used as a
proxy of land quality (it also includes other effects such as technology). This variable
is highly significantly negatively associated with three of the four non-farm routes.
This means that households are likely to stay with agriculture and livestock rather
than moving to non-farm occupations if better quality land and/or better technology
are available. Also, the likelihood of choosing agriculture is double or almost double
the all other routes with the increase in livestock asset by 1%. If the beneficiary
woman is single (unmarried or widow or divorces) the likelihood of non-farm
livelihood is much higher than agriculture. Longer stay with the programme is
negatively associated with regular job, other non-farm job and international migration.