This study of teleworking during the COVID-19 pandemic avoids a common self-selection bias by controlling for preferences and whether the introduction of teleworking was voluntary or mandatory. In a survey (n = 1,500) conducted in December 2020, we inquired about working arrangements and agent preferences in January, April, and December 2020. Productivity levels initially did not significantly differ for newly teleworking individuals after correcting for age, gender, marital status, breadwinner and management roles, childcare gap, area, job type, and general awareness, but were negatively affected by a preference mismatch (PM; around −3pp) and cognitive dissonance (not going with one’s preference in spite of having a chance to do so, adding another −5pp). By December 2020, individuals with a PM lost as much as 9pp of their productivity levels, but the net effect for those in partially remote work was only about −2pp. Similarly, stress levels initially did not differ for newly teleworking individuals using largely the same controls as before. By December 2020, stress levels for individuals newly undertaking fully remote telework were significantly lower than for the control group of individuals with unchanged working arrangements, but the PM contributed significantly to higher stress levels.