The Cognitive Load of Financing Constraints: Evidence from Large-Scale Wage Surveys

In this paper, we take advantage of the implicit cognitive exercise available in standard Labor Force Surveys to propose a new indicator of financing constraints which is based on the cognitive load they generate (Mullainathan and Shafir, 2013). Survey respondents are requested to report their monthly wages, which we compare to their administrative, fiscal counterparts. We propose a well-defined index of worker-level uncertainty, which filters out their potential rounding behavior and reporting biases. We estimate it using unsupervised ML/EM techniques and find that workers tend to perceive their own wages with a degree of uncertainty of around 10%. Through the lens of a simple rational signal extraction model, this amounts to estimates of workers' attention ranging from 30% to 84% depending on their wage, education, tenure and gender. Most importantly, we show that the attention of the lowest paid 30% of workers is cyclical and increases steadily by 17 percentage points in the ten days preceding payday, before immediately dropping on that day, which, through the lens of a simple model, is indicative of end-of-month financing liquidity constraints. Furthermore, this pattern reveals that the cognitive cost induced by these financing constraints arises from the not too concave (or convex) costs of achieving high levels of attention, and the convex costs of maintaining it over time.