Using Contacting Information to Derive Employer Name in the Survey of Doctorate Recipients
Quentin Brummet
NORC at the University of Chicago
Wan-Ying Chang
National Center for Science and Engineering Statistics
Karen Grigorian
NORC at the University of Chicago
Carlann Unger
NORC at the University of Chicago
We demonstrate a new use of contacting information to derive employer name and employer characteristics in the Survey of Doctorate Recipients. A combination of external data sources on email domains and manual coding procedures was used to assign employer names to email address, work mailing address, and work phone numbers for a random sample of respondents. Our results show significant promise: using email addresses, employer names were coded for 77% of respondents, and 70% of these respondents have a coded employer that aligns with their survey reports. We then develop a least absolute shrinkage and selection operator (LASSO) model to predict the best contact information to use, which we show fits the data well and assists with selecting the most accurate pieces of information. We conclude with a discussion of setting an optimal error rate threshold that allows the model to be operationalized in future SDR operations.