DEMOGRAPHY© Volume 24, Number 2 May 1987 THE IMPACT OF ORIGIN COMMUNITY CHARACTERISTICS ON RURAL-URBAN OUT-MIGRATION IN A DEVELOPING COUNTRY Richard E. Bilsborrow Carolina Population Center and Department of Biostatistics, University of North Carolina, Chapel Hili, North Carolina 27514-3997 Thomas M. McDevitt International Statistical Programs Center, U.S. Bureau of the Census, Suitland, Maryland Sherrie Kossoudji Population Studies Center, University of Michigan, Ann Arbor, Michigan 48109 Richard Fuller Carolina Population Center and Department of Sociology, University of North Carolina, Chapel Hili, North Carolina 27514-3997 INTRODUCTION AND PURPOSE Considerable evidenee has aeeumulated during the past two deeades indieating that in developing countries, migration from rural areas is influeneed by high rates of rural natural inerease, inequitable land distribution, inadequate rural employment opportunities and ineomes, and large difIerenees in ineome and amenity levels between urban and rural areas (Bilsborrow, Oberai, and Standing, 1984; Firebaugh, 1979; International Development Research Centre, 1977; Peek, 1980; Shaw, 1974). Many households are "pushed" by eeonomie eonditions to leave the rural sector. Others are attracted by better eonditions in urban areas. Although migration deeisions are made in the eontext of prevailing institutional and struetural labor market eonditions, loeal wealth-property relationships, and geographie disparities in eeonomie opportunities and services, the eharaeteristies ofthe potential migrant and his/her household are also important (Bilsborrow, 198Ia,b; Oberai and Bilsborrow, 1984; Wood, 1982).1 As a result, the investigation of faetors influencing migration decisions is best earried out with a model that ineorporates faetors at both the micro, or individual/household, and the areal, or struetural, levels. Omission of either of the subsets of explanatory variables is thus seen as resulting in misspeeified equations and biased estimates of eausal relationships.? Moreover, sinee areal variables often closely approximate poliey instruments and are important for government planning, their omission, eustomary in many previous analyses of the determinants of migration decisions, limits the poliey implieations that ean properly be drawn. Most evidenee to date on the determinants of patterns of rural-urban migration in Latin Ameriea has been derived from macro-level studies based on aggregated eensus data.' Though household surveys have provided the basis for micro-level analyses of the determinants of migration in a few countries (see Balan, Browning, and Jelin, 1973; DaVanzo, 1976; Romero and Flinn, 1976; and others eited in Bilsborrow, Oberai, and Standing, 1984), there has been little systematie analysis ofthe efIeets of miero- and areal-level faetors based on merged household survey and eontextual data. In this paper we illustrate the use of merged miero and aggregate data to investigate the separate efIeets of individual, household, and eontextual faetors on migration. The particular applieation is to individuals moving from the rural Sierra to 191