This paper investigates whether the composition of a choice set alters the relative importance of different attributes. I derive distinguishing predictions of several prominent theories of choice-set-dependent preferences, including the focusing model of Kőszegi and Szeidl (2013), the salience theory of Bordalo, Gennaioli, and Shleifer (2013a), and the relative-thinking model of Bushong, Rabin, and Schwartzstein (2019). I test these predictions in a laboratory experiment in which I vary the prices of high- and low-quality variants of multiple products. The data provide clear evidence of choice-set dependence consistent with relative thinking: price increases that expand the range of prices in the choice set lead to more purchases. Structural estimates imply economically meaningful effect sizes: the average participant was willing to pay around 18% more when a seemingly irrelevant option is added to their choice set.

Participants in means-tested programs must periodically document continued eligibility through a recertification process. We find evidence that the administrative burden associated with SNAP recertification leads to decreases in program participation. Cases assigned to later recertification interview dates, which leave less time to reschedule missed interviews, are over 20 percent less likely to recertify than cases assigned to interviews earlier in the month. Cases that fail due to later assignments lose an average of $600 in benefits in the following year. These losses are highly skewed: many cases quickly re-enroll, while one quarter remain o SNAP for over a year post-recertification.

Two experimental studies tested consumers' ability to identify surpluses accurately when returns to attributes were non-linear. Participants made repeated, incentivised forced-choice decisions about whether products conferred surpluses, by comparing one or two plainly perceptible attributes against displayed prices. Surplus was determined by an objective function in which the linearity, scale and relative weight of returns were systematically varied. In contrast to other judgement domains, performance was unaffected by monotonic non-linearities. However, non-monotonic returns to attributes greatly reduced the accuracy of surplus identification. Regardless of the shape of returns, the absolute level of performance was consistently imprecise and systematically biased across all conditions. Surpluses were underestimated near the bottom of the range and overestimated near the top. The findings have implications both for models of judgement and choice and for consumer policy.