The Co-holding Puzzle: New Evidence from Transaction-Level Data

Using detailed and highly-disaggregated data on spending, income, bank account balances, and consumer credit, we examine the tendency of individuals to "co-hold", i.e., to simultaneously hold low-interest liquid deposit balances and high-interest debt in the form of overdrafts. The disaggregated nature of the data allows us to calculate co-holding at daily frequency, while prior studies have relied on more aggregated measures. Daily measures reveal that co-holding is less common than these prior studies have documented, occurring on approximately 15% of individual A? days in our baseline calculations. Most spells of co-holding are also short, lasting less than one calendar month. The detailed data allow us to examine the empirical relevance of the competing explanations for co-holding. When brought to the data, we find that co-holding appears to be driven by behavioral rather than rational forces. More specifically, we find evidence in support of explanations for co-holding based upon mental accounting while we find rational explanations for co-holding to be empirically much less relevant.