As populations increasingly become less willing to participate in survey research studies, many researchers have moved to multiple modes of data collection in an effort to maximize response. The Survey of Earned Doctorates, an annual census of research doctorate recipients, is one such survey. Though the majority of survey responses arrive via PAPI questionnaires, survey staff also implemented o web survey several years ago, followed by a shortened CATI component more recently. In the most recent survey round, the CATI component was reserved for respondents form whom survey staff were unable to get completed questionnaires using the other two modes. This paper will! answer the question of whether non-response bias is likely to be reduced by extra efforts at the end of data collection, which can be addressed by examining how similar the survey's CATI respondents were to other respondents in the Survey of Earned Doctorates. Implications for the utility of 'last-ditch' response modes will be discussed.