Recent developments in machine learning allow for flexible functional form estimation beyond the approaches typically used by survey researchers and social scientists. Support vector machines (SVMs) are one such technique, commonly used for binary classification problems, such as whether or not an individual decides to participate in a survey. Since their inception, SVMs have been extended to solve categorical classification and regression problems. Their versatility in combination with the fact that they perform well in the presence of a large number of predictors, even with a small number of cases, makes them very appealing for a wide range of problems, including character recognition and text classification, speech and speaker verification, as well as imputation problems and record linkage. In this article, we provide a non-technical introduction to the main concepts of SVMs, discuss their advantages and disadvantages, present ideas as to how they can be used in survey research, and, finally, provide a hands-on example, including code, as to how they can be used in survey research and how the results compare to a traditional logistic regression.
No takes yet. Share an insight, caveat, or question.
Kirchner et al. (2018) studied this question.