Can survey data improve predictions?

Survey data can indeed improve predictions, primarily by incorporating large quantities and diverse types of auxiliary data.

1. Volume and Diversity of Data: Increasing the volume of both resume data and survey data enhances the quality of predictions. Surveys, although smaller in scale, are crucial, and their integration with extensive auxiliary data, such as resumes, scales the predictive accuracy significantly 1.

2. Transfer Learning: This technique leverages large-scale datasets to improve predictions on smaller survey datasets. For instance, using resume data for pretraining models has resulted in better predictions on subsequent survey data 2.

3. Combining Data Sources: Surveys, when combined with other forms of data (e.g., digital footprints, economic indicators), enhance the predictive power. For example, in econometric studies, predictive models utilizing this combination have been effective in estimating wage gaps and unemployment risks 3.

   • 

     1

     Improving Predictive Models

     00:34:31
   • 

     2

     Predictive Models and Small Survey Data

     00:07:19
   • 

     3

     Transfer Learning Applications

     00:37:01

However, it's also noted that there are inherent limitations and potential biases. It's crucial to audit and validate the data and model predictions to mitigate systemic biases 3.