Content area
Full text
EXTENDED ABSTRACT
The standard theory in economics posits that an individual's consumption should not change in response to predictable income, like a salary pay check (Carroll 1997; Deaton 1991; Hall 1978). However, researchers have continued to document a "payday effect" whereby they consistently find that individuals spend more immediately following receipt of predictable income (Olafsson and Pagel 2018; Parker 1999; Souleles 1999). While these studies provide evidence that there appears to be a consistent effect of income receipt on consumer spending, the findings regarding why this effect occurs are inconclusive. Some research suggests that this effect seems to be present for consumers who are financially illiquid (Kaplan and Violante
2014). However, others suggest that the payday effect occurs even for those with substantial liquid assets (Olafsson and Pagel 2018). A natural question emerges: Why are consumers deviating from standard economic theory (e.g. Hall 1978) and failing to smooth consumption on paydays?
To gain greater insight into the underlying mechanism of the payday effect, we used real income and spending data (N=67,360) provided by a popular financial application in the United Kingdom, Money Dashboard, to examine whether certain groups were more prone to this effect. The structure of the data allows us to observe the timing and amount of each user's pay checks and discretionary spending from January 2013 to December 2018 before individuals began using the budgeting application, as well as demographic information for each user. Our analysis follows the conventions set by Gelman et al. (2014) for investigating payday effects with big data. First, to measure the payday effect we perform panel regression analysis for which the econometric specification is:
(ProQuest: ... denotes formula omitted.) (Eq. 1)
Where is the spending ratio of user i on day k in category c, Paid is a dummy equal to 1 if the user receives a pay cheque on day t, and is a vector of control variables. We compared the coefficients generated by Equation 1 using between-subject t-tests to determine if the size of the effect differs between groups (e.g., men and women). All consumers in our sample display significantly greater spending compared to their average spending across product types, however, we find that this effect is significantly larger amongst women (=0.15, =0.08; p< .001), younger1...