John Bryant

Statistics New Zealand

Attrition in the Longitudinal Immigration Survey New Zealand

This is joint work with Frances Krsinich, Statistics New Zealand

The Longitudinal Immigration Survey: New Zealand (LisNZ) is a uniquely rich source of data on migrants’ labour market and settlement outcomes in New Zealand. But, like all longitudinal surveys, it is subject to attrition. Approximately 14 percent of respondents who were interviewed in wave 1 could not be re-interviewed in wave 2. We investigate whether this attrition leads to selection bias in simple cross-sectional models using LisNZ data. We apply two closely-related tests: (i) we examine whether attrition in wave 2 is related to outcomes in wave 1, after controlling for standard explanatory variables; and (ii) we examine whether the relationship between outcome and explanatory variables differs between non-attritors and the complete sample. Both tests suggest the existence of selection bias. These biases are nevertheless small compared with the size of the coefficients, especially when wages are used as the outcome variable. The small size of the biases means that, at current attrition rates, the LisNZ sample essentially remains representative of its target population.