The most interesting figure is the relationship between migration and land asset endowment,
which is downward sloping in cases of temporary and permanent migration, and inversely
shaped for international migration31. This is to say that land owned (i.e. household wealth) has
a negative, marginally increasing, effect on the household propensity to have a domestic
migrant (in particular, either temporary or permanent), and a positive, marginally decreasing,
effect on the probability to send a migrant abroad (Figure A.1 in Appendix illustrates the
predicted probabilities for the three migration outcomes according to household land
ownership). The relationship between migration and cattle endowment, instead, is mainly
negative across the three typologies of migration, marginally increasing in case of permanent
migration only: this seems to suggest that migration and cattle are substitute activities (or
cattle is sold to finance migration) and the reason for this may lie in the theoretical argument
that cattle is a liquid asset, playing the ‘same’ role as migration in the household risk
management32.
The education level of the most educated person in the household results now significant and
with opposite signs across migration types, as expected, since temporary migration is mainly
devoted to low-skill jobs, differently from the other two types of movements33.
Given that migration is a function of networks and contacts as well (and given that our survey
did not include much information on this), we use the presence of more than one migrant in
the household left more than three years prior to the survey year, as a proxy for (familiar)
social capital or chain migration (this is the ‘network’ variable).
Results show that the presence of another household member emigrated time ago is an
important factor in fostering especially permanent migration, and it is not significant for
temporary migration. This is consistent with the recent and scattered nature of international
migration, and with the more cultural and policy-driven permanent domestic migration in
Bangladesh.
Of course the migration history of the village (the proportion of the sample village labour
force out-migrated either temporary, permanently and internationally)34 is highly correlated
31 We find evidence of the first part of a U-shaped and hump-shaped land-migration relationship in case of
temporary/permanent and international migration respectively. We will discuss deeper these non-linear
relationships further on.
32 Nonetheless, if cattle and migration are similarly household saving-strategies to cope with risk and liquidity
constraints, the ‘scale’ of the strategy is different in that migration has sharply higher entry costs and subsequent
returns which can have a big impact on the long-run household welfare.
33 Ceteris paribus, a small increase in the educational level of the most educated household member decreases
the probability of having a temporary migrant by 3.9 percentage points, increases the probability of having a
permanent migrant by 2.7 percentage points and the one of having a international migrant by 0.3 percentage
points.
34 This variable represent the ‘emigration stock’ at village level and it is different from the ‘network’ variable,
which captures the ‘chain effect’ within the same household.
22