Introduction

Hong Kong’s major political event in 2019: the Anti-Extradition Law Amendment Bill Movement (Anti-ELAB)

In March 2019, the Hong Kong government has proposed a bill on amending the Fugitive Offenders Ordinance and Mutual Legal Assistance in Criminal Matters Ordinance, commonly known as the extradition bill that allowed possible extradition to China. This bill was highly controversial and not well-received by the Hong Kong residents, who expressed their strongest opposition ever to the Hong Kong government. Before the second reading of the bill that was scheduled on 12th June 2019, more than 1 million Hong Kong people protested and voiced their concerns on 9th June 2019. Since then, protests on the extradition bill were organized nearly every week (Hong Kong Democracy, 2020) with no signs to end even when the government withdrew the bill on 4th September 2019. The protests went on until the emergence of the novel coronavirus, knows as COVID-19, when mass gatherings were discouraged to prevent the transmission of disease. This political movement was named the Anti-Extradition Law Amendment Bill Movement (Anti-ELAB). Until April 2020, the five demands of the protestors are yet to be addressed by the Hong Kong Government (Shek, 2020). A lot of protests turned into violence and more than 7000 protestors were arrested. A total of 16,000 tear gas, 1400 pepper sprays, 10,000 rubber bullets, 1800 sponge rounds, 2000 bean bag rounds, and 19 live bullets were used during the period of anti-ELAB, as reported by the Hong Kong Police Force (Hong Kong Democracy, 2020).

Two characteristics of the Anti-ELAB were their long duration and high participation rates. Before 2019, the longest civil event was the Occupy Central Movement, or so-called the Umbrella Movement, which lasted for 79 days. If we reckon the protest of the anti-ELAB began on 9th June 2020, the protests had lasted for more than ten months. Besides the length of the protests, their number of participants was also record-breaking huge. In the history of Hong Kong protests after its handover in 1997, the largest protest was of a size of 500,000 protestors. It was the 1st July 2003 protest (Wong, 2004) and most protestors demonstrated their fear of legislature of Hong Kong Basic Law Article 23 that limited the freedom of speech of Hong Kong people. After this protest, all subsequent 1st July protests were only less than 100,000 (Wikipedia, 2020). During the anti-ELAB era, protests with one million people were not uncommon, including the 16th June protest (2 million (The New York Times, 2019), which set new records for Hong Kong political protests (Ku, 2020), the 18th August protest (1.7 million (Radio Television Hong Kong, 2019a), and the 8th December protest (0.8 million (Radio Television Hong Kong, 2019b). Besides real-time counting of protestors, post-hoc surveys also supported the finding that as much as 36–46% of Hong Kong residents had participated in at least some protests (Cheng et al., 2022; Lee, 2019; Ni et al., 2020). The political environment also raised the voting incentive of Hong Kong residents. The 2019 Hong Kong District Council election had a record-breaking 4.1 million registration and 71% turnout rate (Hong Kong SAR Government Registration and Electoral Office, 2019; Lam, 2021), representing a 50% increase relative to the previous record of 47% in 2015 District Council election (Hong Kong SAR Government Registration and Electoral Office, 2019; Lam, 2021).

Effects of anti-ELAB protests on Hong Kong residents

During the period of extradition bill controversy, protests were carried out in nearly all weekends and in some of the weekdays (Chong, 2020; Hong Kong Democracy, 2020). At the occurrence of these protests, many journalists, professional or amateur, conducted live broadcasts. Therefore, most of these events were filmed and videos recordings were readily available and free of charge. Besides the high participation rate, a high proportion of Hong Kong people who had not participated were still able to witness these events by watching the live streams. For example, the 4-hour broadcasts of HK Apple Daily on the 31st August, displaying the Prince Edward Mass Transit Railway (MTR) station attack, accumulated 860k views on YouTube (HK Apple Daily, 2019). In fact, live streams had become the most important source of information about anti-ELAB protests, and it provided users an experience of collective witnessing (Fang & Cheng, 2022). Given the violent nature of most Anti-ELAB protests, participants and witnesses were having a traumatic effect which made them stressed and depressed (Lee, 2019; Ni et al., 2020). A population-based survey showed that 9.1% of respondents reported suicidal ideation during anti-ELAB protests (Hou et al., 2021).

Difficulties in examining the effects of anti-ELAB protests

Given the high visibility and aggressiveness of both the protestors and the Police Force during the anti-ELAB protests, daily activities of Hong Kong people have been significantly impacted, for example, the transpiration system, including MTR, traffic lights, and paving blocks, had been repeatedly interrupted or damaged. The anti-ELAB was exerting a territory-wide impact on evert Hong Kong residents, not only to those who participated in the protests (Li et al., 2021). Surveys are traditional methods to investigate the effect of political events, however the nature of anti-ELAB protests is spontaneous, unplanned, leaderless, and unpredictable, making any pre-planning of scientific studies nearly impossible (Yuen et al., 2022).

While we are still able to examine long-term impact of anti-ELAB protests such as depression and psychological distress (Lee, 2019; Ni et al., 2020), their immediate impact remains difficult to investigate. Some time, routinely collected data could be used to examine the immediate consequences of such kind of unplanned political events, for example a 14-day diary study conducted in the US happened to coincide with the presidential election and allowed the investigators to examine its day-by-day effect (Roche & Jacobson, 2018). In the big data era, huge amount of data are being collected every day and they offer new opportunities for research that could not be imagined decades ago. For instance, text mining the contents posted on social network sites and online forums can inform us about what the users were discussing and thinking. A text-mining study during anti-ELAB showed a positive association between protests and psychological distress expressed online (Lam et al., 2021).

Political events and sleep

Sleep is essential for our body to rest and restore, and sleep can be described by sleep duration where the American Academy of Sleep Medicine and Sleep Research Society suggested adults to sleep at least 7 h per night (Watson et al., 2015), and sleep quality which can be defined as “an individual’s self-satisfaction with all aspects of the sleep experience” (Nelson et al., 2022). Sleep is influenced by numerous factors, including socio-demographic characteristics, personality, environmental factors, and health condition (please see (Ksiazkiewicz, 2022) for a review). However, the relationship between politics and sleep remains poorly understood and requiring further research. Previous studies have concluded that about one-fifth to one-third adults’ sleep was affected by politics (Smith, 2022), adolescents living in a war-affected area reported a higher frequency of nightmare (Harb & Schultz, 2020) and US residents reduced their sleeping time at the day of Presidential Election (Anýž et al., 2019; Cunningham et al., 2022; Roche & Jacobson, 2018). In addition to sleep, political events can also affect mental health and other psychological functioning, which may in turn affect sleep. A longitudinal study during the Palestinian political turmoil found that trauma-induced difficulties in falling sleep would exert further negative impacts on one’s mental health, in particular post-traumatic stress disorder (Gerhart et al., 2014). To sum up, while there is some evidence that political events affect sleep, the link between politics and sleep is still largely unknown (Ksiazkiewicz, 2022).

The current study

Sleep disturbances such as nightmares, night terrors, and difficulties to fall asleep are commonly found among those experiencing politically-induced trauma across all age groups as these symptoms represent their strategies to internally cope with severe stress (Hjern et al., 1991). Hong Kong residents are well-known of their poor sleep quality (a 40% prevalence of insomnia and only 50% of them achieved the recommended sleep duration of > 7 h per night, (Wong & Fielding, 2011). However, it is still unclear how the anti-ELAB protests affected the sleep of Hong Kong people, due to the difficulties in answering this research question. Previous research also showed that national political events including Brexit poll and US Presidential election had an immediate effect on the residents’ sleeping pattern and they would recover the next day (Anýž et al., 2019; Cunningham et al., 2022; Roche & Jacobson, 2018). As mentioned above, the nature of anti-ELAB protests is unplanned, meaning that a prospective study design to examine its sleep effect is not feasible (Yuen et al., 2022). To overcome this limitation, this study examined the sleep pattern of Hong Kong residents during the anti-ELAB from June to December of 2019 using crowdsourced sleep data obtained from Sleep as Android app (https://sleep.urbandroid.org/).

Based on previous studies that people had reduced sleep duration (Anýž et al., 2019; Cunningham et al., 2022) and lower sleep quality (Cunningham et al., 2022; Roche & Jacobson, 2018) at the day of major election, we postulated that the anti-ELAB movement would have similar effects of sleep pattern and sleep quality on Hong Kong people, and have made the following hypotheses:

  • H1: Hong Kong people would have shorter sleep duration at the dates of these six events.

  • H2: Hong Kong people would have delayed sleep onset time at the dates of these six events.

  • H3: The subjective sleep quality of Hong Kong people at the dates of these six events will be reduced.

Methods

Sleep as Android

The Sleep as Android smartphone application (https://sleep.urbandroid.org/) is developed for the Android operating system as a smart alarm clock that wakes the users according to their sleep cycles and can ensure the users has wakened up. This app was launched in 2011 and is still ongoing, thus allowing us to detect the longitudinal change in sleep patterns during any time starting from 2011.

Sleep onset time can be identified using two methods, namely subjective manual assignment and objective detection. Nearly all (99.6%) of the users selected the first method, where the sleep onset was defined as the moment when they pressed the “start sleep tracking” button in the app. For the remaining 0.4% users, they used an automatic sleep tracking function that detects the sleep onset of the users via different sources, including external wearable wristbands, the built-in accelerometer in the smartphone, sleep sounds such as snores, and contactless ultrasonic tracking (Chaudhry, 2017). Although Sleep as Android app has not been validated against the gold measurement of sleep, the polysomnography, we believe it would have a similar validity and reliability of sleep diary, which is validated against polysomnography and reliable (Rogers et al., 1993). The awake time was defined as the moment when they terminated the tracking manually, interacted with the phone, talked near the phone and detected by the app, or the light sensor of the phone showed > 60 lx after the sunset time of the geographical location. It has accumulated more than 300,000 reviews in Google Play and is one of the most reviewed and highest rated apps for sleep analysis (Choi et al., 2018; Ong & Gillespie, 2016).

Data retrieval

Only users who agreed to share their data were retrieved. Sleep as Android began to collect data in 2010, however, the sample size in 2010 was too small and we decided to analyze the sleep data from 1st January 2011. The end date was on 31st December 2019. A total of 33,786,868 sleep records from 141,509 users were retrieved. Variables available in the dataset including the date of the sleep records, sleep duration, sleep onset time, self-rating of sleep quality (from 1 to 5 by 0.25 step), and time zone obtained from the smartphone’s system information.

Data processing

Outliers and impossible records of sleep duration (17,218,464 records) were removed from the analysis. These included sleep duration of < 3 h or > 13 h, and sleep onset time between 6:00 and 18:00. Furthermore, we treated all sleep records with sleep onset time between 0:00 and 6:00 as the sleep occurred at its previous day (for example, sleep onset on 2 January 1:00 was treated as sleep occurred on 1 January).

All sleep records from Hong Kong, classified as time zone of Asia/Hong_Kong, were included. Similar to a previous report using data from Sleep as Android, the system time zone, instead of geographical coordinates, was used to classify the user location due to missing data (Anýž et al., 2019). A total of 15,693 sleep records from 274 users were available for analysis (mean age 33.3, SD 9.6). The sex distribution was not reported given the high amount of missing data (93.6%). A total of 123 participants (44.9%) provided at least 30 days of data.

The sleep records during the 2019 anti-ELAB were compared with the following three sleep records, namely (1) sleep records in Hong Kong in Year 2011–2018 (55,165 sleep records from 751 users), (2) sleep records outside Hong Kong in Year 2019 (3,951,014 sleep records from 39,208 users), and (3) sleep records outside Hong Kong in Year 2011–2018 (12,033,709 sleep records from 98,902 users). In total, our data analysis comprised of 16,055,581 sleep records from 114,514 users (rating of sleep quality was available in 5,442,005 sleep records from 76,600 users). Sleep records among users from outside Hong Kong were used as a comparison to illustrate the change in the sleep pattern during anti-ELAB was a local rather than a global effect.

Besides the sleep pattern of Hong Kong residents from June to December of 2019, Six specific dates with events that were regarded as the signature of the anti-ELAB protests (Chong, 2020) were analysed individually, including the 12th June protest, the 21st July Yuen Long attack, the 31st August Prince Edward station attack, the 1st October National Day protest, the Chinese University of Hong Kong (CUHK) conflict, and the 2019 Hong Kong District Council election. These events received high attention in the research community (Chong, 2020; Hu, 2022; Lam, 2021; Leung & Fang, 2022; Purbrick, 2019). To help readers understand how these events affected sleep, the detailed timelines of these events will be described in the Results section below, together with the proportion of participants sleeping across the corresponding period.

Statistical analysis

Three sleep parameters, namely sleep duration, sleep onset time, and subjective sleep quality, were analyzed. We compared these sleep parameters in Year 2019 to their expected values. The expected value of these variables was computed using linear regression, controlling for time zone, the month of the record, and the day of the record (S/M/T/W/T/F/S). These variables were associated with sleep patterns in crowdsourcing sleep data (Anýž et al., 2019; Walch et al., 2016). The deviations between the actual sleep parameters and the expected sleep parameters, as well as the standard error of these deviations, were produced in regression analysis. SPSS version 25 was used to conduct the data analysis.

Analysis of specific events

For the analysis of the sleep pattern at the days of the aforementioned six events, we compared these sleep patterns with those at the reference dates, i.e., the closest date with the same day of the week in the year 2011–2018. The proportions of participants sleeping on major event dates (listed in Table 1) and their reference dates from 20:00 to 16:00 of the next day (corrected to 1 s) were plotted for comparison. Details of these events were described in the “Results” section, with important time points for each of the events highlighted in the figures for better interpretation of the results.

Table 1 Date of the major events during the anti-extradition bill protest in Hong Kong (1st June 2019–31st December 2019) and sleep parameters at these dates
Full size table

Results

Figure 1 shows the Hong Kong participants’ actual and expected sleep durations. For six out of the seven events (except 1st October protest), the participants reduced their habitual sleep duration by 8 to 23 min (Fig. 1C; Table 1), although the differences were insignificant. We concluded that hypothesis H1 was confirmed as the amount of reduction was comparable to those found for Brexit vote (16 min) and 2016 US presidential election (13 min) (Anýž et al., 2019). Figure 2 shows the actual and expected sleep onset time of the participants in Hong Kong. A 34-min delay in sleep onset was observed on 31st August 2019, while the participants slept around 23 min earlier on 23rd and 24th November 2019 (Fig. 2C; Table 1). Again, hypothesis H2 was confirmed. Figure 3 shows the actual and expected sleep rating of the participants in Hong Kong who provided their self-reported rating. On 11th and 23rd of November 2019, they had a rating of 0.40 and 0.45 lower than the expected ratings (Fig. 3C; Table 1). While these differences corresponded to Cohen’s d effect size of 0.40 to 0.45 (the rating had a standard deviation of 1.0 in our sample), we could not detect statistical significance which may be due to small sample sizes of < 10. Hypothesis H3 could not be confirmed.

Fig. 1
figure 1

Sleep duration in Hong Kong before and after the anti-extradition bill controversy (continuous lines: actual sleep duration; dotted lines: expected sleep duration)

Full size image
Fig. 2
figure 2

Sleep onset in Hong Kong before and after the anti-extradition bill controversy (continuous lines: actual sleep onset; dotted lines: expected sleep onset)

Full size image
Fig. 3
figure 3

Sleep rating in Hong Kong before and after the anti-extradition bill controversy (continuous lines: actual sleep rating; dotted lines: expected sleep rating)

Full size image

Figures 4 and 5 show the proportions of participants who were sleeping across the timeline of 20:00 to 16:00 on the day following the six major anti-ELAB protests compared to the 2011–2018 reference dates. On 12th June 2019, protesters surrounded Legislative Council to object the second reading on the extradition bill (Purbrick, 2019). After 18:00, an increasing number of Hong Kong residents got off work and joined the protest. The Admiralty MTR station was shut down at 22:00 to stop people from going there, but there were still more than a thousand protestors present at the site (Purbrick, 2019). The protesters at Admiralty began to leave and completely dispersed at around 02:00 of 13th June. During this period, the proportion of participants who were sleeping was smaller than the one on the 2011–2018 reference dates (Fig. 4A).

Fig. 4
figure 4

Proportion of users sleeping (12th June, 21st July, 31st August, and 1st October) (continuous lines: Year 2019; dotted lines: Year 2011-2018)

Full size image
Fig. 5
figure 5

Proportion of users sleeping (12th November, 23rd November, and 24th November) (continuous lines: Year 2019; dotted lines: Year 2011-2018)

Full size image

At 22:00 of 21st July 2019, a group of > 100 armed assailants wearing white shirts (which was believed to be chosen to represent the opposite side of the democracy protesters whose dress code was black) entered Yuen Long MTR stations and attacked passengers and passers-by (Chong, 2020; Purbrick, 2019). Police arrived at the MTR station at 01:00 after the attack ended. Due to the huge delay of the Police Force, many people went to Yuen Long Police station to protest until 02:30. In 2019, the proportion of participants who were sleeping was lower during the occurrence of the incidence from 22:00 to 02:00, and the proportion was higher during 04:00 to 09:00, relative to the reference proportions in 2011–2018 (Fig. 4B).

At around 22:45 on 31st August 2019, the Police arrived at Prince Edward MTR station at 23:00 and conducted arrest of protesters in the MTR platform and the train. Many passengers were beaten by the Police (Wikipedia, 2019). Immediately after the incident, many people surrounded Mong Kok Police station to protest until 03:00 of 1st September 2019 (Wikipedia, 2019). Figure 4C shows a large discrepancy of sleep patterns between 31st August 2019 and its reference dates that the proportion of sleeping participants reduced for 10–20% from 00:00 to 02:00.

During the National Day protest on 1st October 2019, at 16:00, the police shot an 18-year-old man with a handgun. He was the first protestor to be shot during the anti-ELAB protests (Chong, 2020). No major events occurred at night and the sleep pattern on 1st October 2019 presented a slight left shift when compared to the one on the reference dates of 2011–2018 (Fig. 4D).

At 22:00 of 12th November 2019, a water cannon from Police arrived the Number 2 Bridge of CUHK where hundreds of protesters guarding there to stop Police entering the university (Leung & Fang, 2022). To divert the manpower of the Police, territory-wide protests were carried out at night of 12th November 2019 (Leung & Fang, 2022). Most of these protests ended at 03:00 to 04:00 of 13th November 2019. In comparison to the reference dates, participants had lower sleep proportions during the period of 22:00–00:00 and 01:00 to 06:00, and a higher sleep proportion during the period of 08:00–16:00 (Fig. 5A).

The 2019 Hong Kong District Council election was conducted on 24th November. The voting time of the election was between 07:30 and 22:30. A total of 4.1 million Hong Kong citizens aged 18 or above were registered as voters and 71% voted that day (2.9 m) (Hong Kong SAR Government Registration and Electoral Office, 2019; Lam, 2021). This turnout rate broke the record since the handover of the Hong Kong Special Administrative Region, as this election was regarded as a chance to voice out for many anti-ELAB protestors (Lam, 2021). Figure 5B shows that the sleep proportion from 04:00 to 09:00 on 23rd November 2019 was lower by around 10% than that of the reference dates. After the completion of the voting at 22:30, the votes were immediately counted and the results were announced after the counting was confirmed. At 12:45, all results were officially announced (HK01, 2019). Figure 5C shows that the sleep proportion was higher from 21:00 to 01:00 and 08:00 to 11:00, and lower at 02:00–05:00.

Discussion

This study demonstrated the impact of major events during the anti-ELAB protest period. Some events analyzed in this study happened unexpectedly, including the Yuen Long station attack on 21st July 2019 and the Prince Edward MTR station attack on 31st August 2019. For these events, live broadcasts were available within minutes and the agreement between the time of the resolution of these events and the proportion of participants who sleep supported the argument that many Hong Kong people stayed awake past their usual bedtime to obtain the latest information of these events. A possible and tentative explanation for the hypothesized shorter sleep duration and onset during the anti-ELAB protests may be a “Fear of missing out” (FOMO). FOMO is a worry about missing out something that other people are having, that can be amplified at times of during social unrest (Tang et al., 2020). This phenomenon was evidenct in the Arab Spring and the Occupy Central movement (Tang et al., 2020). Thus, although not directly measured, we can expect that FOMO would also be prevalent among Hong Kong residents during the anti-ELAB controversy. FOMO increased the use of social network (Riordan et al., 2022) and would have a replacement effect on sleep duration (e.g., spending bedtime on smartphone news reading instead of sleeping (Koban et al., 2022). Further research about FOMO on anti-ELAB protests or other political events and its effect is warranted.

On the other hand, the schedule of the 2019 Hong Kong District Council election was fixed and Hong Kong people could plan their sleep time to accommodate the election. There were concerns that protestors would surround the voting stations, and the Government had announced that any protests during the election period might result in postponement or cancellation of the election (Ming Pao, 2019). Therefore, many people decided to vote within the first few hours of the voting period (Hong Kong SAR Government Registration and Electoral Office, 2019) and this was supported by our data that more people chose to wake up earlier than their usual schedule (Fig. 5B). To compensate for the reduced sleep time on the day before the election, they slept earlier on the night of the election (Fig. 5C), and some of them woke up from 02:00 to 04:00 probably to check the real-time voting results online. As the announcement progressed to 04:00 that the number of pro-democracy camps secured more than half of the seats (HK01, 2019), nearly all participants went to sleep. It was also observed that a higher proportion of participants woke up between 08:30 and 11:00 when most voting results were officially announced (HK01, 2019).

Our results showed that the sleep pattern of Hong Kong people followed the time trend of most major events that occurred during the period of anti-ELAB controversy, but their subjective sleep quality remained the same. This was a surprising finding as several studies showed that Hong Kong residents were more depressive and anxious during the period of anti-ELAB controversy (Lee, 2019; Ni et al., 2020), and we should expect that the sleep quality of Hong Kong residents should have been negatively affected. A longitudinal study during the Palestinian political turmoil found that trauma-induced difficulty in falling asleep would exert further negative impacts on one’s mental health, in particular post-traumatic stress disorder (Gerhart et al., 2014).

Our results showed that people would sleep less at the day of major political events. Short-term sleep loss induces stress, emotional distress, and affects mood, memory, and cognitive performance (Medic et al., 2017). While the effect only lasted for a single day, the impact is territory-wide and it may cause significant public health consequences. Besides sleep loss, other negative reactions reported from a group of US undergraduates a day after presidential election (anger, fear, marginalised, and being discriminated) were likely to occur in our sample (Roche & Jacobson, 2018), however these variables were unmeasured, thus the associations could not be confirmed, in the current study.

The strength of our study lied in the continuous monitoring of sleep patterns from 2011 onwards. This continuous monitoring system enabled us to observe the longitudinal change in sleep pattern during 2019 where traditional survey design studies could not do so, except in a rare case that sleep pattern of a group of participants had already been collected regularly before 2019. Nonetheless, there were several limitations in this study. The major limitation was that we could not confirm whether the users of Sleep as Android had participated (or the degree of participation) in anti-ELAB protests. Given that our sample from Hong Kong was young (4,832 sleep records from Hong Kong in the second half of 2019 included birth date information and their mean age was 33.3 (SD 9.6), together with the data obtained from an on-site survey showing that 60% of the protestors were below 30 years old (Lee et al., 2020), we could expect that many Sleep as Android Hong Kong users should have participated in and being affected by some protests. Another limitation was that we could not detect the sleep pattern of the Hong Kong population subgroup who did not use Sleep as Android app. Although not assessable, we can expect that the users of the Sleep as Android app would be younger and wealthier than non-users (Krebs & Duncan, 2015). We can expect that there would be very few elderly and young children in our sample. Among participants from Hong Kong, at the dates of the six examined events only 17% of the sleep records (87/514, Table 1) were provided with a sleep quality score, which limited the power and the generalisability of the analysis related to H3. Finally, the reduction of sleep duration might be a result of the displacement effect that people delayed their sleep onset for watching live broadcasts or browsing social media websites, or the traumatic and depressive effects of these violent events, or both. Our results could only identify the change in sleep duration but not the underlying reason.