A Nested Logit Migration Model with Selectivity
构建了一个多选择迁移模型，利用1971年委内瑞拉人口普查数据，修正工资方程中的选择性偏差，并采用较宽松的离散选择建模技术，研究个体在23个州之间的迁移决策。
In this paper we specify a multiple choice migration model with selectivity. We estimate this model using observations on individuals drawn from the 1971 census of Venezuela. There are two main innovations in this paper: we apply a correction for the presence of selectivity in wage equations in the context of a multiple choice migration model; and we apply a less restrictive technique for modelling multiple discrete choices than in previous studies. We study the choices of individuals among 23 states of Venezuela. Each individual is observed to have chosen one state, so one wage is observable per person. Our hypothesis is that individuals choose where to locate on the basis of an evaluation of costs and benefits associated with choosing each state. We measure the benefits associated with choosing a given state by means of the wage a person could earn in this state. Since we observe only one wage for each person, we have to impute the unobserved wages for the states this person did not choose in order to study the role of wages in migration. However, it is possible that unobserved factors which affect both an individual's choice of location and his wages may result in sample selection bias if we try to estimate state-specific wage equations by OLS on the basis of the self-selected samples. We apply an estimator proposed by Lee [1983] to estimate state-specific wage equations free from sample selection bias. The wage equations permit us to investigate whether individuals who choose particular states earn different wages from observationally identical individuals drawn from the population at random. In the context of a developing country where internal migration flows are large, evidence on the selectivity of migration will enhance our understanding of an important labor market adjustment mechanism. Evidence that people who choose particular locations earn more in those locations than would observationally identical individuals drawn at random from the population (positive selection) will suggest that migration decisions are made according to the principle of comparative advantage. In modelling the choice of location, we test whether subsets of states are more highly similar to each other than to other states. Unobserved similarities between states may arise because of similar amenities or economic conditions which are not otherwise explicitly controlled for in the model. We implement the