Abstract |
Graduate Destination Studies have the potential to provide detailed information about graduate transitions to work that cannot easily be collected in household surveys. However, response rates are typically very low and raise the concern that the nonresponse is not random and inferences using data on those who respond will be inaccurate. This study examines response rates in the Western Cape Graduate Destination Study where 22% of all 2010 university graduates from the four Western Cape universities were successfully interviewed in 2012. We examine differences in observable baseline characteristics, assess the extent of non-response bias for a labour market participation analysis, compared rates of continued study to those in the HEMIS database and implement a selection correction methodology that uses type of email address as an exclusion restriction. We find that those who successfully responded to the survey are more likely to be studying in 2012 and have some systematically different baseline information that signals that response is not random. Our selection correction methodology however finds limited impact for an equation of employment. This study provides important input into plans for a national destination study. We recommend that focus be taken in preparing and standardising the sampling frame and that detailed records of the survey process be kept. In addition, we illustrate the potential benefits of linking graduate destination study data with administrative resources to assess bias and supplement the survey information obtained. |