It's one thing to infer aspects of someone's personality from a public Facebook profile or Twitter feed. But it's another thing altogether to understand what makes people tick using no more than a record of digital expenses. New, however, brings us one step closer to this unexampled reality.
Next, the researchers obtained participants' bank records that included all digital purchases made from a checking account or credit card over the last 12 months. They classified purchases into 279 categories, such as"supermarkets,""insurance policies," and"furniture stores." When possible, they also kept track of where people made their purchases .
They note, however, that some traits showed a higher prediction accuracy than others. They write,"Across all psychological traits measured in our study, the average correlation between actual and predicted scores was .19. However, this aggregated measure of accuracy masks considerable variation across individual traits, with openness having the lowest accuracy and materialism having the highest.
The researchers also tested to see how rapidly prediction accuracy would decline when using less data. When they recalculated their models using only one month of spending data, they noticed a marginal drop in prediction accuracy. This suggests that while more data increases model accuracy, stable predictions can still be made from relatively thin slices of personal spending data.