Use of Survey Rotation Patterns in the Estimation of Population Flows: An Application to International Migration to/from Mexico
Victor A. Bustos¹, Sara Mera²
¹General Directorate for Strategic Liaisons, Research and Development, INEGI, Aguascalientes, Mexico; ²General Directorate for Socio-Demographic Statistics, INEGI, Aguascalientes, Mexico

Mexico is working on the implementation of a three-pronged approach towards the study of migrant populations: their size, measured through censuses and administrative registers, notably birth certificates, from destination countries; their flows, through data on household composition changes collected along with the quarterly Labour Force Survey (LFS); and their characteristics, through harmonized joint multinational household surveys. In particular, the LFS rotating panel sample design, in which households remain in the sample for five consecutive rounds of interviews one quarter apart with an 80% quarter-to-quarter sample overlap, has proven particularly useful in determining quarterly change patterns in household composition. First interviews record the number of residents in the dwelling as well as their age and sex. Successive interviews check for new arrivals to, and departures from, the household over the most recent quarter. In the first case, their place of origin is recorded; in the second, their destination is. The optimum rotation pattern is not a simple matter since it depends on analytical, methodological and operational factors which may pull in different directions. Even though the rotation pattern seeks to prevent undue respondent burden which may result in attrition for households that are selected for the survey, non-response on either interview has to be faced. Therefore, weights which take non-response into account are developed linking the data bases for two successive rounds for the above mentioned four-fifths sample overlap. This paper deals with both the methodology outlined above and the results it has yielded so far: a declining trend in the number of out migrants from 2006, apparently related to the beginning of the burst of the housing bubble in the US, which contrasts with a stable number for the returns, and a distinct seasonal pattern exhibited by both out- and in-migrants over the same period.

Biography: After obtaining his Ph.D. at LSE, Alfredo Bustos became lecturer at Mexico's National University.

Since 1991 is affiliated with INEGI, Mexico's NSO. Under his leadership, INEGI ICT-unit introduced new technologies for analysis and dissemination of information. Examples are: relational-database engines for on-line systems like SIMBAD; or Internet. His proposal for using multivariate-analysis during processing of economic census data was implemented and used also in agricultural censuses.

He works now seeking to bridge the gap between official and mathematical statistics. His leading indicator system for the Mexican economy, using multivariate time-series canonical-correlation analysis, accurately forecasted the onset of the 1994 credit crisis. He is interested on design-related properties of sampling units for consideration in small-area estimation.

He was part of the UNSD and UNECE expert groups which updated recommendations for the 2010 Population and Housing Census Programmes. Markovian estimation of population growth led him to study migration statistics, and to the Suitland group.