Intergenerational Mobility and Interpersonal Inequality in an African Economy

To further explore the characteristics of those who inherit, Table A2 gives probits for any form of inheritance, while Table A3 gives probits for land inheritance in particular; in both cases we present the marginal effects. We also provide the breakdown by gender. We show two specifications, the second of which drops a number of variables that might be considered endogenous to inheritance. (Later we will use these pruned regressions as the first stage for an instrumental variables estimator.) While the causal interpretation of the first regression (including the endogenous variables) can be questioned, it is still of descriptive interest. We include a wide range of controls in these regressions (and those reported later), including: gender, age and age squared, age at first marriage, whether one is the first born of a given gender among siblings with the same mother and same father, whether one is the first born among all children with the same mother and father, whether the first born sibling from the same mother and father is a boy, number of brothers from the same father and mother, number of brothers from the same father only and same mother only, and the same three variables for sisters, ethnic group, being Muslim relative to other religions, having some formal education, whether fostered as a child, and whether fostered at a young age (prior to two years of age, which typically implies a permanent move for the child in the Senegal context). There are also controls for parental characteristics (education, occupation, place of residence, whether the father died in the last two years, and whether the mother did so) and some demographic variables describing the household (log household size) and the individual’s cell (log cell size, share of adults and share of children age 5 and under).


Intergenerational Mobility and Interpersonal Inequality in an African Economy Sylvie Lambert, Martin Ravallion and Dominique van de Walle
To further explore the characteristics of those who inherit, Table A2 gives probits for any form of inheritance, while Table A3 gives probits for land inheritance in particular; in both cases we present the marginal effects. We also provide the breakdown by gender. We show two specifications, the second of which drops a number of variables that might be considered endogenous to inheritance. (Later we will use these pruned regressions as the first stage for an instrumental variables estimator.) While the causal interpretation of the first regression (including the endogenous variables) can be questioned, it is still of descriptive interest.
We include a wide range of controls in these regressions (and those reported later), including: gender, age and age squared, age at first marriage, whether one is the first born of a given gender among siblings with the same mother and same father, whether one is the first born among all children with the same mother and father, whether the first born sibling from the same mother and father is a boy, number of brothers from the same father and mother, number of brothers from the same father only and same mother only, and the same three variables for sisters, ethnic group, being Muslim relative to other religions, having some formal education, whether fostered as a child, and whether fostered at a young age (prior to two years of age, which typically implies a permanent move for the child in the Senegal context). There are also controls for parental characteristics (education, occupation, place of residence, whether the father died in the last two years, and whether the mother did so) and some demographic variables describing the household (log household size) and the individual's cell (log cell size, share of adults and share of children age 5 and under).
We continue to find that men are more likely to inherit than women, even with the controls. Conditional on the included controls, being male adds 0.11 to 0.13 to the conditional probability of receiving any inheritance, while it adds 0.08 to the probability of inheriting land.
Unsurprisingly, the death of either parent increases the probability of inheritance, and the coefficients are considerably higher for paternal death. In the full sample, death of the father alone adds 0.67 to the probability of inheritance, while death of the mother adds only 0.13; with respect to land inheritance the probability is increased by 0.32 by a deceased father, but only by 0.03 by a deceased mother. This later estimate reflects the previously noted fact that at a woman's death, her land is returned to her husband (or brothers) first. By contrast, children inherit from their father at his death, whether or not their mother is still alive. These effects are significant across almost all strata and specifications, the only exception being that death of the mother is not a significant predictor of land inheritance by women. The effect of a father's recent death dampens the large "father dead" effect for any inheritance (bringing it down from 0.67 to 0.54, when the mother is still alive). This is consistent with our casual observations from interviews that inheritance, particularly of the house and non-land assets, is typically delayed. 1 The dampening effect is much lower for land inheritance and significant only for men.
There is a positive coefficient on education in the regressions for any inheritance, which suggests complementarity rather than substitution by parents between formal schooling and inheritance (whereby some children get some form of inheritance while others get formal schooling as hypothesized by Quisumbing et al. 2004). However, there is some sign of such substitution for land inheritance, though it is only statistically significant for women; those women with formal schooling are less likely to inherit land. Depending on the timing of parents' death, this might reflect the individual choice of an educated woman with a non-farm economic activity to give up her land inheritance to the benefit of her siblings, rather than a parental decision to substitute one form of transmission for another.
Men who were fostered as boys are more likely to inherit land unless they were fostered before age two. This pattern is plausible. Fostering out a very young child is suggestive of giving away the child (for example to a childless parent), which is an indication that inheritance is unlikely. By contrast, fostering an older child is in general less permanent and more suggestive of an investment in the child, which would also suggest that inheritance is more likely. 2 None of these effects are statistically significant for girls.
Having a mother active in the non-farm sector significantly increases the probability of any inheritance for men although not for women. For land inheritance, paternal activity in farm work has no effect, but maternal farm work has a positive correlation with men's, and less so women's, land inheritance. This could reflect the fact that a mother in farming suggests a greater availability of land for the parental household. The father's non-farm activity matters but negatively ─ significantly reducing the likelihood of inheriting land for both genders. In fact, 1    Table A4: Estimated effects of inheritance on log cell per capita consumption with and without controls for sample whose father is dead No controls

Rural location and department dummies
As in (2) Table A5: Estimated effects of inheritance on log cell per capita consumption with and without controls for sample whose mother is dead.

Rural location and department dummies
As in (2) Table 5 notes. The 'other' occupation drops out of the urban regressions as it is found only in rural areas.  Table 5.