A study on migration is of key importance in social, economic particularly in population studies. The present study is conducted to pay attention on causes and consequences of international migration at individual or household level of Bangladesh and also tried to find out different socio-economic and demographic factors that initiate international migration. It was also intended to determine the model of statistical relationship between remittance flow and different migrant characteristics and related socio-economic factors.

The study is based on primary data collected from a village of Munshiganj district of Bangladesh.

In the study, data on different socio-economic and demographic factors from both migrant and non-migrant households have been considered and Discriminant Analysis has been performed to identify those factors which account the most for separating the two groups (migrant households and non-migrant households). Analysis shows that the covariates, total adult male members, gender of the head of the households, land ownership, age of the head of the households and year of formal schooling are the significant variables that separate mostly between the two groups.

For searching significant factors and evaluation of risk, Multiple Logistic Regression model has been used. Applying stepwise multiple logistic regression analysis, the covariates, Gender, Year of schooling, Occupation, Land owner, Adult male members Age of the household head; - all are found significantly associated with the migrant household.

In order to determine the impact of different socio-economic and demographic characteristics of the migrant on remittance flow, the Multiple Regression Model in log-lin form (Semilog model) is considered. After performing stepwise multiple regression analysis, we found that, Gender, Marital status, Year of formal schooling, Duration of stay and Working status of the migrants were the most significant variables associated with the remittance flow.

Econometric validation of the multiple regression model, that is, normality assumption, zero mean property of the residual, multicollinearity, auto-correlation and heteroscedasticity has been checked by appropriate statistical method and no violation of these assumptions has been detected.