Can models incorporating nontraditional, behavioral elements into the classical, expected-utility framework help explain seemingly dominated choice behavior of participants in centralized matching markets? Can they help in redesigning matching mechanisms to reduce such behavior? Investigating the widely used deferred-acceptance (DA) algorithm, we answer positively, both in theory and using laboratory experiments. We use an off-the-shelf news-utility model where preferences include expectations-based reference dependence (EBRD): individuals care not only about actual consumption but also about news (relative to previously held expectations) regarding consumption. Each participant in our laboratory experiments (N = 500) plays 10 simulated large-market school-assignment problems of varying competitiveness, in one of four ([Formula: see text]) different DA variants. They vary by whether individuals submit their rankings prior to the matching process (a static implementation) or make sequential decisions (a dynamic implementation), and whether they are on the proposing or receiving side: [Formula: see text]static, dynamic[Formula: see text]student proposing, student receiving[Formula: see text]. Whereas a traditional, reference-independent model predicts the same straightforward behavior across all problems and variants, a news-utility EBRD model predicts stark differences across them. Intuitively, to avert losses from bad news, EBRD individuals may downrank, or even avoid applying to, competitive positions that, when offered to them (with certainty), they will accept. Consistent with our predictions, we find that (i) across DA variants, dynamic student receiving leads to significantly fewer deviations from straightforward behavior; (ii) across problems within the other three variants, deviations increase with problem competitiveness; and (iii) in those three variants, the specific deviations predicted by EBRD are indeed those commonly observed empirically. We discuss practical implications. This paper was accepted by Itai Ashlagi, revenue management and market analytics. Funding: This work was supported by United States-Israel Binational Science Foundation [Grant 2022417], the Israel Science Foundation [Grants 1796/22 and 2968/21], the Maurice Falk Institute, and the Johnson School at Cornell University. Supplemental Material: The online appendix and data files are available at https://doi.org/10.1287/mnsc.2024.05446 .