With pupils in England set to return to school in the next couple of weeks (fingers crossed), there is widespread concern about the toll that lockdown has taken not only on their learning but also on their broader wellbeing and mental health. As we reported in May, SchoolDash has been collaborating with EduKit to monitor the collective experiences and life satisfaction of pupils during lockdown using a free survey offered to schools between late April and early July 2020. We reported preliminary results in May. This post updates and adds to those findings with a larger sample size and additional analyses.
We find that:
Feelings of loneliness were widespread, especially among older pupils. Up to Year 8, around a third of respondents reported feeling lonely "a bit" or "very often", but for Years 12 and 13 it was more than half, with 20% or more saying that they felt lonely "very often"
Secondary pupils were much more likely than primary pupils to say that their schools had not provided sufficient wellbeing and mental-health support, with around 20% of respondents in Years 9-12 saying "not really" or "not at all". Around 30% of secondary pupils (compared to 50-60% of primary pupils) said that they received plenty of mental-health support.
Among secondary pupils, girls reported lower average levels of life satisfaction than boys of the same age. However, this effect preceded school closures and, if anything, the gender gap narrowed during lockdown.
Schools with high proportions of poor pupils tended to have more respondents who reported spending only a short time on homework, or unhappiness about homeschooling. However, self-reported happiness among these schools tended to increase during lockdown while among schools with low proportions of poor pupils it fell.
Among primary pupils, feeling comfortable at home was the most significant factor contributing to overall happiness. Among secondary pupils, it was (absence of) loneliness.
We analysed 12,917 responses collected between 27th April and 7th July from pupils at 77 different schools across England, representing all nine regions and all year groups from 1 to 13. These consisted of 47 primary schools, 54 secondary schools and one all-through school. Response rates were similar across both phases (25.5% in primary, 26.4% in secondary), but because secondary schools usually have much bigger pupil populations, 83% of the responses came from that phase.
Response rates varied considerably by pupil type, as shown in Figure 1. For example, when viewed by gender, girls (30%) were considerably more likely to respond than boys (21%). There were even larger differences by year group, with very low rates for infant pupils (Years 1 and 2) and for GCSE and A-level cohorts (Years 11 and 13). Within each phase, response rates tended to rise slightly with age, though as already mentioned above, there was little overall difference in response rates between phases.
Although we did not ask for respondents' free-school-meals (FSM) status, we can examine response rates by the proportions of FSM pupils in each school. Schools with low proportions of FSM pupils showed almost double the response rates (35%) of those with high proportions of FSM pupils (19%).
(Use the menu below to view response rates by these different pupil groups. Hover over the columns to see data values and sample sizes.)
Notes: Since the survey was anonymous, the numbers of responses shown do not necessarily correspond to unique pupils: obvious duplicate submissions have been filtered out, but multiple responses from the same pupil are possible and at least one school is known to have surveyed its pupils twice during the lockdown period. A high proportion of FSM pupils means 15% or more, while a low proportion means below 9%.
Sample size: 12,917 responses.
Figure 2 shows how self-reported life satisfaction (on a scale of 0 to 10) varied by age for all pupils (grey line), boys (blue) and girls (red). Years 1, 2 and 13 are omitted because the sample sizes for those groups were relatively small.
During lockdown, mean life satisfaction declined with age, the score for girls falling more than that for boys, at least for secondary schools (Years 7 to 11). This is consistent with previous surveys, but compared to those earlier surveys, average life satisfaction was lower during lockdown, with an average change of -0.20 across the year groups shown. This decline was larger for boys (-0.22) than for girls (-0.08), so although girls tended to report lower life satisfaction than boys – an effect that predates this year's educational disruption – lockdown appears to have had a bigger overall effect on boys, causing the gender gap to shrink.
Figure 3 shows mean life satisfaction score by pupil type during May and June (the two months during which the majority of responses were collected). This allows different pupil groups to be compared based on their overall levels of life satisfaction as well as how (if at all) this changed during the course of lockdown. In doing this, it is important to note that the pupils who responded in May (blue columns) were not necessarily the same as those who responded in June (red columns). However, they shared similar overall characteristics in terms of age, gender and school FSM rate1, and with thousands of responses collected in both months, we would expect each sample to be similarly representative.
Across all pupils, the mean life satisfaction score remained roughly constant from May to June at about 6.7. Furthermore, there were no significant differences in month-on-month changes by phase. Although there were differences by year group, these didn't fall into any obvious pattern. As already described above, there are differences by gender, with boys reporting higher average scores than girls (albeit by a smaller margin than in other surveys conducted before lockdown). Furthermore, the mean score for boys was essentially unchanged between May and June, while that for girls fell slightly.
Even more interesting were the trends by school FSM rate. In May, mean life satisfaction at schools with high proportions of FSM pupils was very similar to that for schools with low proportions. Furthermore, schools with greater proportions of poor pupils showed an overall increase in average life satisfaction during June. This seems counterintuitive, but might arise for a variety of reasons. First, as we saw above in Figure 1, pupils at these schools showed lower response rates, so it could be that the pupils who replied to the survey are not completely representative of the school as a whole. Second, as we shall see below, some aspects of the lockdown experiences of these pupils (such as having enough time to relax) may have contributed positively to their life satisfaction. Finally, it is possible that these pupils are disproportionately unhappy when attending school and so grew happier as lockdown went on.
Figure 3: Average life satisfaction score (0-10) by pupil type
Notes: See notes to Figure 1. Year groups with fewer than 100 respondents in either month have been omitted.
Sample size: 12,917 responses.
The survey also included questions about specific aspects of life under lockdown. Figure 4 shows how answers to these varied by year group. Results for this full sample are broadly similar to those seen in our preliminary analysis in May, albeit with one or two changes.
Some questions showed relatively little variation by age. Around 20-30% of respondents across all year groups said that they felt generally upset about the situation, while around 10-15% said that they lack sufficient personal space and about 5-10% reported being uncomfortable at home. Around 70% of all year groups reported feeling bored "a bit" or "very often", though the latter category increased with age, rising from around 15% in Years 1 and 2 to over 30% in Years 8-12. Similarly, the proportion of pupils who reported feeling unhappy or very unhappy about doing school work at home tended to increase with age.
Most concerning of all, feelings of loneliness were widespread, especially among older pupils. Up to Year 8, around a third of respondents reported feeling lonely "a bit" or "very often", but for Years 12 and 13 this proportion was more than half, with 20% or more saying that they felt lonely "very often". In a similar vein, secondary pupils were much more likely to say that their school has not provided sufficient wellbeing and mental-health support, with around 20% of respondents in Years 9-12 saying "not really" or "not at all". This compares to around 30% who say that they get plenty of support.
Finally, some questions showed clear exceptions for Years 11 and 13 (for whom GCSEs and A-level exams were cancelled). When asked about time spent on school work, well under 10% of those in Years 10 and 12 said that this came to less than an hour a day, while the proportions for Years 11 and 13 were around 40-55%. Consistent with this, the same year groups reported having time to relax, with over 60% of Year 11 respondents and nearly 75% of those in Year 13 saying that they had "plenty" of time. This compares to only around 25-30% of respondents in Years 10 and 12. Note also that response rates for Years 11 and 13 were low, so it's possible that those answering the survey were more engaged with school matters than the average pupils in those year groups.
(Use the menu above Figure 4 to explore different questions. Click on the legend to turn different groups on or off. Hover over the graph to see corresponding data values and sample sizes.)
Figure 6 shows the relationship between these questions about pupils' experiences and their overall life satisfaction scores. A value of 1.0 indicates perfect correlation between answers to a given question and overall life satisfaction, while a value of zero indicates no correlation at all. (Answers to questions were scored in such a way that high values always corresponded to 'good' answers, so negative correlations were not seen.)
Across all pupils, loneliness and general feelings about being at home correlated most strongly, followed by other factors such as personal space, mental-health support, healthy food, relaxation time and absence of boredom. Time spent on school work was not a substantial factor.
For primary pupils, correlations were generally lower and tended to emphasise factors such as comfort at home and not feeling lonely or bored, while secondary pupils seemed to care much more about loneliness. There was relatively little difference between boys and girls, with the former placing a slightly higher emphasis on boredom and the latter on healthy eating.
It is important to note that the relationship between these two indicators is not merely a result of older pupils tending to feel more lonely and also having lower average life satisfaction: even within each age group, reports of loneliness tended to correlate highly with life satisfaction, especially for older pupils.
(Use the menu below to view different pupil groups. Hover over the chart to see corresponding data values.)
Figure 6: Correlation of responses to questions with overall life satisfaction
As befits an analysis of human experience during difficult times, the results presented here are nuanced, confirming some preconceptions while calling others into question. They also give cause for both concern and hope. It is deeply worrying that loneliness has been rife, and that so many teenagers felt that they did not receive adequate mental-health support. On the other hand, there is relatively little evidence from the poorer pupils who participated that they had worse lockdown experiences than their more affluent peers, at least from the point of view of self-reported wellbeing.
Of course, none of this is to say that there aren't more concerning cases hidden among hard-to-reach groups of pupils that may have been under-represented in this survey, nor that the academic progress of disadvantaged groups didn't suffer during lockdown – there surely are and they surely did. We will only be certain about these things when children finally return to school.
The proportion of female respondents fell from 59% in May to 56% in June. The proportion of secondary respondents rose from 81% to 85%. The proportion of respondents from schools with low or medium FSM rates rose from 67% to 68%.
For the last couple of years, we have provided summaries of school-related Twitter trends in the UK (see previous posts from 2019 and 2018). This is partly for edification (some of the trends are interesting), but also for a bit of light summertime entertainment (because, let's face it, Twitter is mostly concerned not with deep insight but with the spasms of our collective id).
This year, however, the COVID-19 pandemic has given these analyses a more sober significance because we can use them to track changing patterns of concerns over time. To that end, we posted an interim analysis in late March, soon after lockdown. This post serves as both an annual review of trends during the 2019-2020 academic year and an update on activity over these last few extraordinary months.
Figure 1 shows relative numbers of tweets from the UK mentioning schools since the beginning of January1. During the early part of the year, activity was relatively stable, with regular weekly fluctuations. From late February, however, things started to pick up, rapidly reaching a peak of more than 5 times usual activity on 18th March, which coincided with the announcement of school closures in England.
Activity returned to more normal levels during the initial part of lockdown in April, but picked up again in early May upon the announcement that some children would be able to return to school from early June. There were further small spikes throughout early June, also coinciding with news about pupils returning to school (or not). These culminated in a final peak on 16th June, which was provoked by the government's U-turn on free school meals in response to a campaign led by the footballer Marcus Rashford. Since then, activity has faded and now appears to have entered a summer lull at around half of normal levels.
Figure 1: Number of daily tweets from the UK mentioning schools
Sources: Twitter; SchoolDash analysis.
Figure 2 shows word clouds of terms that have characterised each month of the school year just past. These are calculated relative to the same months in 2018 and 2019, which downplays perennial topics and tells us instead about what was particular to this year.
Back in that distant age known as July 2019, there was a focus on private schools amid calls from within the Labour Party for them to be abolished, while in the state sector there were demands for investigations into bullying and asbestos. August saw a mix of reflections on the school year just past, including GCSE results, and preparations for the year to come, notably the purchase of school uniforms. Come September, not only were schools back, but so was that earlier ding-dong about state versus private. Campaigns against cuts to nursery-school budgets briefly stole the headlines in October, with the twin political topics of Brexit and the general election also on the rise. By November, private schools were firmly back in the frame, along with the general issue of school budgets – and of course the general election. The latter inevitably continued to dominate into December too.
(Use the menu below to select a month to view.)
Figure 2: Common terms in UK tweets about schools
Sources: Twitter; SchoolDash analysis.
Following a resounding Conservative victory, January saw a series of petitions on sign language, sanitary products for girls and halal meat. Ominously, February saw the first significant uptick in mentions of coronavirus, amid other talk of Greta Thunberg and Pancake Day. And by March, there was really only one topic of conversation: school closures. In April, attention turned to homeschooling and (optimistically for most) school re-openings. This continued into May, joined in June by campaigns to extend provision of free school meals and to eliminate racism. As schools closed at the end of the academic year in July, most of the online chatter was still, ironically, about when and how they might open up again. Such are the topsy-turvy times in which we live.
All very interesting (we hope), but what more can we do to quantify these different subjects, especially ones related to the pandemic? Figure 3 shows the prevalence of these selected topics over time. Note that because individual tweets can mention more than one topic, the topics can sum to more than 100%. (For further details please see Footnote 2 of our previous post.)
As we pointed out in our earlier analysis, tweets about schools have tended not to talk much about the disease itself or the government's advice – these subjects accounted for no more than a few percent of messages. Mitigation strategies such as hand washing and social separation have been more prevalent, especially during the most stringent lockdown period in March, April and May. School closures were predictably huge in March, accounting for almost half of all tweets about schools, but have since subsided (save for a brief resurgence in the last few days). School re-openings has since been a hotter topic, especially during May. This is not to say that everyone has been in favour: there were tweets both for and against and there's some evidence to suggest that the balance were opposed (see Figure 4 below).
(Click on the figure legend to turn individual data sets on or off; double-click to show one data set on its own. Hover over the graph to see corresponding data values.)
Figure 3: Topics mentioned in UK tweets about schools
Sources: Twitter; SchoolDash analysis.
Key workers, a hot topic when schools first closed, has since faded. Home schooling has shown more sustained interest, though partly because homework is a perennial favourite (notice the relatively high levels of activity even before lockdown). A remarkable, though short-lived, story was the campaign over free school meals. On 16th June, when the government announced their U-turn, this accounted for almost all tweets about schools, but then went away as quickly as it had come. Another interesting trend – though more modest in scale and not as directly linked to coronavirus – is the increase in posts about racism in the days following George Floyd's death in the US on 25th May.
This all seems rather harrowing, though amid the death and disruption there are also tales of triumph in adversity and efforts to build a better future out of the ashes of despair. What can we say about how this has left people – or at least Twitter users – feeling? Broadly speaking, there are two ways to determine this using data analysis:
Build an AI capable of parsing human sentences and inferring the emotions of the people who wrote them.
Look at the emojis.
We enjoy overwrought technical solutions as much as the next band of data nerds, but since it's August we decided to go with Option 2.
Figure 4 shows a 'Net Emoji Score' for each day since the start of January. To generate this, we categorised the most commonly used emojis (such as 😀, 😩 and 😐) into 'positive', 'negative' and 'neutral' categories, then calculated the difference between the prevalence of positive and negative emojis on each day. This metric tends to jump around a lot, so the data shown below have been smoothed by applying a 7-day moving average.
With some notable exceptions, such as the end of the Christmas and New Year holiday at the very start of the year, it would appear that the natural state of the Twitter users analysed here is closer to 😀, though there have been some deep troughs of 😩 too. The biggest of these coincided with the announcement of school closures across England in mid-March. There was then a small dip in mid-April, coinciding with increased mentions of school re-openings and COVID-19 mitigation strategies, and a bigger fall in mid-May, after the government announced that partial school re-openings would commence in June.
Interestingly, Marcus Rashford's free-school-meals triumph in mid-June was accompanied by a downturn in sentiment, suggesting that the thought of hungry kids weighed more heavily than any feel-good factor arising from the campaign to help them. The gradual decline in Twitter activity during July makes the sentiment analysis more difficult to interpret (because it might make the scores more volatile for purely statistical reasons). However, it probably doesn't surprise parents of school-age children to learn that the onset of the summer holiday appears to have been accompanied by a greater number of whimpers than cheers, at least in this particular corner of the Twitterverse.
Figure 4: 'Net Emoji Score' of daily tweets from the UK mentioning schools (7-day moving average)
This uses a relative scale in which 100 is set to the mean daily activity during January; we don't know the absolute numbers because this analysis is based on a sample of tweets provided by the Twitter API. We further limit the sample to tweets geolocated within or close to England that mention schools.
This is the SchoolDash blog, where we write about some of our projects and other things that spark our interest.
Blue links take you to other parts of the page, or to other web pages. Red links control the figures to show you what we're talking about.