2.1 Data collection

Fieldwork for the 2020 OTUS was carried out in two waves. Wave one ran from 28th March to 26th April 2020, and wave two commenced on 5th September and ended on 11th October 2020. The first wave took place during the first COVID-19 lockdown and the second wave took place during the subsequent restrictions. The survey used a multi-stage stratified probability (random) sample. The Scottish component of the 2020 OTUS drew on the ScotCen Panel (NatCen in the rest of the UK) for responses. This panel is made up of 4,000 people aged 16 and over living in England, Scotland and Wales who were invited to take part after completing the British or Scottish Social Attitudes surveys.

The Scottish sample was made up of 581 respondents and 1,100 diary days, made up of the following:

• 336 females (632 diary days) and 245 males (468 diary days)

• 194 respondents aged 65+ (380 diary days), 231 respondents aged 45 to 64 (436 diary days), 141 respondents aged 25 to 44 (258 diary days), and 15 respondents aged 18 to 24 (26 diary days)

• 121 disabled respondents (236 dairy days) and 460 from non-disabled respondents (864 diary days)

• 8 ‘Asian, Asian Scottish or Asian British’ respondents (16 diary days), 8 ‘other’ minority ethnic respondents (14 diary days) and 559 ‘White’ respondents (1,056 diary days)

• 304 ‘Christian’ respondents (584 diary days), 251 respondents with no religion (468 diary days), and 15 respondents from ‘other’ religions (28 diary days)

• 535 heterosexual or straight respondents (1,017 diary days) and 38 LGB+ respondents (68 diary days)

• 163 respondents with a household income of £1700 p.m. and under (310 diary days), 181 respondents with a household income of £1700 to £3300 (342 diary days), and 110 respondents with a household income of over £3300 p.m. (205 diary days)

Data was weighted to be representative of the Scottish population, taking into account age, ethnicity, sex, employment and tenure. Weighting also factored in differences between workdays and weekends.

Within a participating household, each respondent was asked to complete two pre-allocated diary days, including one weekday and one weekend day. Respondents filled in what they were doing at ten minute intervals during the day, using pre-coded options provided to them via an activities list. Diary entries were then recorded by the participants online where possible, or participants were contacted over the phone by interviewers who recorded diary information on participants’ behalf. Research by Reg Gatenby found that pre-coded diaries were comparable to self-completion diaries, with minor issues only arising at the detailed code level, due to differences in interpretation of the codes.

The activities reported here combine a number of different codes under single headings. For example, the activity ‘housework/cooking’ combines a large number of codes concerned with domestic work, i.e. ‘making food and drinks, cooking or washing up’, ‘cleaning, hoovering, tidying house’, ‘washing up’ and ‘ironing, washing or mending clothes’. For a full list of the codes used for each activity, please see the data tables accompanying this report.

2.2 This report and interpreting the results

When interpreting the results of the 2020 OTUS it is important to keep in mind that the fieldwork took place during the UK’s first national lockdown and thereafter during periods of restrictions as a result of COVID-19. These restrictions are likely to have had an impact on how time was used in Scotland. Any future comparisons with the 2020 OTUS should situate findings in this context, always keeping in mind that differences might be due to the COVID-19 lockdown and restrictions, and not necessarily due to larger societal trends.

The findings presented here relate to the average number of minutes per day spent on an activity. This is worked out as (the average time per day for all people divided by the proportion of people who participated in the activity) multiplied by 100. The amount of time spent per day on an activity might be lower than expected when compared to a hypothetical person’s activity. For example in Scotland in 2020 the average amount of time spent on paid work was 152 minutes, or 2 hours 32 minutes. However, this must be understood as the amount of time spent across the whole Scottish sample, not just the proportion of the Scottish sample participating in paid work on a given day.

The sum of all the averaged activities when broken down by the relevant characteristics add up to 24 hours (or 1,440 minutes). However, when looking at developmental and non-developmental childcare amongst the proportion of the sample with children there were some cases where a respondent was doing both forms of childcare at the same time (e.g. helping with homework whilst making lunch for their child). In these cases the activity was double counted in both the developmental and non-developmental categories, and has resulted in these categories totalling to just over 24 hours.

The number of diary days where an activity takes place is provided as a percentage. This percentage is included for an activity where it appears useful and to add further clarification. Full data for the diary day percentages for each category can be found in the data tables accompanying this report.

When analysing the relationship between the characteristics included in this report and an activity, confounding variables should be borne in mind. Confounding variables could be contributing to any correlation, for example we know that disabled and Christian respondents are more likely to be older, therefore age may also be contributing to these respondents’ likelihood to do or not do an activity. Hence, there might be contextual reasons for differences in time use across different groups, however, we are not able to speculate on these reasons in the report.

2.3 Establishing significance

Statistical significance testing was used to determine how certain we are that differences seen in the survey are due to real-world differences between the groups and not differences in our sample or due to chance. Significance testing was done at the 95% level, which means that there is a less than 1 in 20 chance to observe a difference between groups that does not actually exist and is just a product of chance. Where our results show a difference that is not statistically significant, there may still be a real-world difference that was not reflected in our current sample. It is not possible to draw any conclusions about the differences that are not significant. It is worth noting that a difference being significance does not necessarily mean that this difference is meaningful, as very small numerical differences could be statistically significant if consistent across the sample. In addition, it should be noted that while there will be contextual reasons for differences in time use across groups we are not able to speculate on these reasons in our analysis.

We used two statistical tests to determine significant difference between groups. A Mann-Whitney U test was performed to compare characteristics with two groups (sex, disability, and sexual orientation). Kruskal-Wallis H test was performed to compare characteristics with three or more groups (age, ethnicity, religion, income). A significant Kruskal-Wallis H test only shows that there are differences between some groups but not where these differences originate. For this reason, significant Kruskal-Wallis tests were followed up with a pairwise test to establish which groups differed from each other. These tests compare the median time spent on an activity by each relevant group to determine whether the difference between these groups is significant at the 95% level. Non-parametric tests were chosen for the analysis due to the data being non-normally distributed as average times being influenced by both the time spent on an activity and the proportion of people doing the activity. Both of these tests formally require a sample of at least 5 independent observations per group, therefore. This requirement has been satisfied in all comparisons.

2.4 Sample size

This refers to the total number of diary days completed for each breakdown. Small sample sizes increase the margin of error and standard deviation, which means that the estimates are further away from the true median and the chance is higher that they are affected by random variation in the sample and not due to a real effect. This can distort significance estimates and therefore affect the robustness of statistics. Thus, users are advised to exercise caution where estimates are based on sample sizes of 50 or fewer diary days.

In some cases, groups might have less than 10 diary days, for example when looking at breakdowns with small sample sizes (e.g. other religions or minority ethnic groups) for only those respondents who participated in an activity. In such cases, comparisons with less than 10 diary days have been suppressed, to ensure disclosure control and statistical robustness.