“…Previous studies investigate how behavioral deficiencies, such as low program awareness, misperception of program features, and inattention affect program participation, with the conclusion that these contribute to incomplete take-up but cannot fully explain it (see, e.g., Madrian and Shea, 2001;Bhargava and Manoli, 3 2015;Kleven and Kopczuk, 2011;Liebman and Luttmer, 2015;Finkelstein and Notowidigdo, 2018). We exclude these factors by design to focus on stigma, which earlier studies typically aimed to detect from surveys (see, e.g., Baumberg, 2016) in the form of an observed positive correlation between benefit size and the probability of welfare take-up (e.g., Blundell et al, 1988;Blank and Ruggles, 1996;Currie, 2000;Whelan, 2010), or from the observation that non take-up rates tend to be higher in means-tested (income-based) than in universal (non-income based) programs (see, e.g., Andrade, 2002). These approaches suffer from identification problems.…”