This study was carried out by the Department of Biology, Eberhard Karls University Tübingen, Tübingen, Germany. Data were collected in accordance with the Declaration of Helsinki for experiments involving humans approved by the Eberhard Karls University’s ethics committee (Faculty for Economics and Social Sciences: nr. A.Z.: A2.5.4-124_kr).

We started our anonymous online survey on May 18, 2020, and continued until June 17, 2020. We therefore collected data during the most restrictive phase in Germany. Participants were informed about the study via an electronic mailing list (employees and students of the Eberhard Karls University Tübingen; >20,000 mails) and postings on different social media platforms (Facebook/Instagram). The recruitment text included an online link to the questionnaire. The survey was hosted on an online platform (SoSciSurvey) to fulfil the European Union’s data privacy rules and took an average of 12 min ± 5 min (standard deviation, SD) to complete. The theoretical background and study goals but not the hypothesis were declared. We explicitly informed about the voluntariness of the participation, the option to stop the data collection at any point without consequences, and that participation would not be remunerated. The recruitment text was available in German only and formal consent was inquired in advance. The total number of evaluable cases amounted to 681.

Age, sex, household size, number of children in the household, profession, and the option to work in flexitime were asked for. Profession was later dichotomized into student (N = 400) or non-student (N = 281). 197 participants were male, 484 were female. Mean age was 28.63 years, SD 10.49 years. N = 545 participants noted that there were no children in their household, while N = 136 reported one or more children. We explicitly asked for the number of children in the household and not the number of own children, because, for example, students may have travelled back home to their parents during the restriction phase and lived with younger siblings. Thus, children in the household is a better measure than own children, because regardless of relationship (own children/siblings/other cases), children in general may have an impact on sleep during the pandemic.

The questionnaire used was composed of validated questionnaires concerning chronotype and sleep duration as well as additional questions to quantify changed circumstances during the pandemic. Examined characteristics were chronotype/midpoint of sleep, sleep duration, and COVID-19-induced changes in sleep/work hours.

The Morningness–Eveningness Stability Scale improved (MESSi [18, 19]) and the corrected midpoint of sleep (MSF corrected) were used as separate measures to determine the chronotype. The MESSi is composed of three subscales: the morning affect subscale (MA), the eveningness subscale (EV), and the distinctness subscale (DI). Five items in a 1–5 Likert-format represent each scale. The MA is concerned with the affective facet of the morningness–eveningness trait (M/E; e.g., alertness in the morning: “How alert do you feel during the first half hour after having awakened in the morning?”), while the EV queries feeling/mood, energy level, and learning capacity in the evening (e.g., “In general, how is your energy level in the evening?”). The DI shows the subjectively felt amplitude of diurnal active phases (e.g., “There are moments during the day where I feel unable to do anything” with response options ranging from “totally” to “not at all”). Higher MA or EV scores represent higher morning and evening orientation, respectively, while higher DI values indicate higher daytime fluctuations. MESSi’s factorial invariance, structure, and reliability have already been confirmed repeatedly in different languages [18, 20,21,22,23]. In addition, actigraphy data corroborated the validity of the MESSi [24]. Cronbach’s α in the current study sample was 0.899 for MA, 0.889 for EV, and 0.775 for DI.

We asked for bed and wake times during the week and at weekends to assess sleep duration and the midpoint of sleep, both during and before the COVID-19 restriction phase. Furthermore, a correction algorithm [25] was used to measure the sleep/wake time differences on work-free days due to social jetlag and to calculate a corrected midpoint of sleep (MSF corrected) for both periods. Average sleep duration was calculated: five times the weekday sleep duration plus two times the weekend sleep duration divided by seven.

To assess the sleep phase delay, we subtracted the prior clock times from the clock times during the COVID-19 restriction phase. This resulted in four clock time differences, which were subjected to a factor analysis (principal component). All loaded onto the same single factor, labelled “delayed sleep phase” (58.9% of the variance explained). Week bedtime delay loaded with 0.833, weekend bedtime delay with 0.770, wake week delay with 0.757, and wake weekend delay with 0.705 onto the factor.