COVID-19 has exacerbated people’s reluctance to board crowded trains and the effects of in-vehicle crowding on perceived travel time. This study examines COVID-19 effects of crowding perception in the New York City subway system using 1939 observations from a stated preference (SP) choice experiment with attributes of travel time, travel cost, ratio of mask usage, and passengers per square meter. Discrete choice models in preference space and crowding multipliers (CM) space were built to understand user behavior. Multinomial logit (MNL) models were used to estimate these coefficients, assuming the same estimates for all respondents. Mixed multinomial logit (MMNL) and latent class logit (LCL) were built to account for heterogeneity among individual preferences. MNL results show that passengers perceive travel time, travel price, and crowding negatively and that mask usage positively impacts the alternative. Travelers are willing to pick longer trips in uncrowded conditions in exchange for reducing travel time in crowded conditions. Vaccination pass requirement has no significant effect on subway line choice. MMNL and LCL proved that no significant heterogeneity of preferences exists among respondents.Estimation in the CM-space supported by post-estimation analysis in the preference space show that CM for 0% of mask usage on the train range from 1.4 to 3.34 for 1 to 6 pax/m2, respectively. Similarly, for 50% of mask usage, CM range from 1.28 to 2.72. Finally, for 100% of mask usage, CM range from 1.16 to 2.01. A 100% of mask usage mimics the pre-pandemic scenario. For 50% mask usage, CM is similar to those during the pandemic. For 0% mask usage, CM increases up to values 30% greater than those during the pandemic.