The lowest line (temporary) plots the probability of having a temporary migrant for a given
size of land owned. The upper line (permanent) plots the sum of the probability of having a
temporary migrant or permanent migrant (i.e. the area between the two lines is the probability
of having a permanent migrant only). And top line (international) plots the summed
probability of migrating either temporary, permanently or internationally against the option of
non-migrating (see also plotted probabilities in Figure A.2. in Appendix). The shape and areas
of predicted probability clearly illustrate the non-linear relationship between land asset
ownership and the household participation to the migration process.
The differently-shaped non-linear relations can be explained as follows. Overall, at low level
of wealth a marginal increase of land-ownership decreases the propensity to migrate. In
particular, in cases of temporary and permanent migration, at low levels of land owned a
marginal increase of land decreases the propensity to migrate for these specific destinations
(more rapidly in case of temporary migration than permanent migration). This is consistent
with the idea that these migration typologies are low-cost and low remunerative in terms of
remittances, thereby also little ‘preferred’ by households. In case of international migration
instead, at low levels of wealth the propensity to migrate abroad is close to nil because of high
entry barriers. After a threshold level, though, (around 1,2 acre of land), a marginal increase
in household wealth increases the propensity to migrate, specifically to migrate abroad. In
other words, land ownership becomes a pushing factor that releases household constraints to
migrate and favour outflows overseas. This is consistent with other findings on historical
migration (such as Italian hump-shaped migration in early last century; see Faini and
Venturini, 1994) and with the fact that Bangladesh is a very poor country38 (therefore, we
capture the first part of a hump-shaped wealth-international migration relationship).
Along with reporting the model coefficients, we also show effects on log-odds ratios, which
are useful to illuminate the dynamics among outcomes. They provide perhaps more
illustrative information on the migration probabilities of household members with different
characteristics than the marginal effects.
Table 9 shows the (marginal and unit) effect on the odds ratio of main determinants of
migration decision.
38 In addition, our dataset is even more skewed towards smallholder poor families, being large holders less
represented than at national level (see note 17).
27