3. Class project for ALP301 2020, IPA Philippines

>>>>> gd2md-html alert: ERRORs: 0; WARNINGs: 0; ALERTS: 13.

• See top comment block for details on ERRORs and WARNINGs.
• In the converted Markdown or HTML, search for inline alerts that start with >>>>> gd2md-html alert: for specific instances that need correction.

Links to alert messages:

alert1 alert2 alert3 alert4 alert5 alert6 alert7 alert8 alert9 alert10 alert11 alert12 alert13

>>>>> PLEASE check and correct alert issues and delete this message and the inline alerts.

---

To: Innovations for Poverty Action

From: Zelin Li, Vivan Malkani, Niki Panich, Maria Fernanda Porras

Date: May. 25th 2020

Re: COVID-19 Mental Health Impact Evaluation in the Philippines

---

Executive Summary

The COVID 19 pandemic has increased stress levels and exacerbated mental health issues worldwide. We worked with Innovations for Poverty Action (IPA) in the Philippines and developed a Facebook (FB) Chatbot survey with three main goals:

1. Test the utility of an FB chatbot survey for future surveys and interventions.
2. Better understand mental health impacts of COVID - 19 in the Philippines
3. Evaluate communication methods and trusted sources of information for IPA

Key Recommendations

Based on the feedback received from IPA staff and insights gleaned from the current study, we propose the following for future consideration:

• The Facebook chatbot survey was** inexpensive** and **simple to deploy, but specific consideration of how to **include **older **and less educated respondents in future survey populations is important
• IPA should consider developing peer-to-peer social media communication strategies in the future. **Peer-to-peer information sharing **(respondents sharing FB posts) achieves far higher traction than centralized information dissemination (respondents clicking on information resource links).
• Consider **COVID- and mental health- resource communication strategies across multiple media **(TV, Social Media, News) since survey respondents indicated that they received COVID-related information and updates from government and non-government sources via television and social media.
• Consider different methods of increasing study power (run study longer or provide incentives for participants) and use the proposed chatbot study design to identify effective communication strategies for population subgroups - age, level of education, etc. - and tailor strategies accordingly.
• Conduct a follow-up study to **evaluate the impact of different mental health resources **in the Philippines, using FB post-sharing to maximize visibility of resources.

Research Question and Experiment Design

Our initial discussion with the IPA Philippines provided us with a better understanding of one of the less visible impacts of COVID-19 in the Philippines - increased mental health burden on health care workers and the general population. While our first instinct was to conduct an impact evaluation of specific mental health resources (eg. online chat options with mental health organizations like Philippine Mental Health Association, Department of Health mental health hotline, etc.), the limited timeframe of this class project pushed us towards understanding how to best communicate mental health awareness and resources, focusing on trusted sources of information and how resources were presented.

We recruited participants to our chatbot survey via Facebook advertisement. We developed an English/Tagalog chatbot script with embedded user engagement prompts and the option for future participation reminders, although our study did not focus on these components of the survey. The first part of the survey contained nine demographic questions, identical to the Philippines Census, followed by seven mental health questions that ask participants about their level of stress, challenges encountered, coping strategies, thoughts on enhanced quarantine measures, and ways of sourcing information during COVID-19 pandemic. (See Appendix for the survey questions.) These questions were used to qualitatively assess the impact of COVID-19 on the mental health of participants.

The second part of the survey is experimental, with interventions focusing on engagement with mental health awareness resources, for different sources and presentation methods. Both treatments described below were randomly assigned.

The first part of the experiment aims to find out what source of information participants trust the most. Participants are presented with one of the three Facebook posts from the following sources: Department of Health (Government), Tanglaw Mental Health (NGO), or In Touch Community Services (Mental Health Services Provider) and are asked if they would like to share the post with someone else. Each of the posts contains a visually appealing infographic that includes information about a mental health hotline in the Philippines and raises awareness of mental health resources. We measure the** post share rate** for each of the sources to compare which source has the most engagement.

The second part of the experiment aims to compare engagement across different presentation styles. The control group receives links to three mental health resources (government, civil society organization and mental health service provider), and the order of link presentation is randomized. The two treatment groups receive additional information in addition to the randomized links. Treatment group 1 receives a picture from the World Health Organization (WHO) that conveys the message, “It is okay to be worried”; Treatment group 2 receives an infographic from the WHO that has specific recommendations on how to cope with increased mental health burden during the COVID-19 outbreak. We measure whether participants click any of the presented links and use the click through rate to determine which presentation style participants prefer the most.

Results

Descriptive Results:

The chatbot proved to be an inexpensive, simple method to survey people across the Philippines. The Facebook ad click through rate of 1.11% was above the 0.9% average click through rate for facebook ads (Wordstream, 2018). Of the 1.1% of viewers clicked the ad, 76% consented to the study, and 74% of the consentees completed the study; meaning that, overall, 56% of people who clicked on the ad, completed the study in its entirety. The average study completion time was 12 mins and 45 seconds and, although our study did not focus on this, we found that 37% of respondents requested reminders from the chatbot to reach out to a friend at a later date. The total cost of the study was $267.70. Overall, we received ~6080 total responses over an approximate three week time frame. Of the total responses, 280 were used as pilot data, and the remaining 5800 observations were filtered for outlier values and non-consent responses to leave ~2610 observations for the analysis below.

To ascertain if our study sample was representative of the broader Philippines population, we conducted a series of demographic comparisons, using national census data from 2015. Analysis of the geographic distribution of respondents demonstrated that the study population mirrors the regional population distribution of the Philippines. Our sample also seemed to be representative in the dimension of household size (used as a proxy for household income). Our analysis of the age of study participants showed that our study sample skews towards a younger age distribution when compared to the Philippines, which we attribute to the disproportionate age distribution of FB users in the Philippines . The study, further, captures an increased response rate from individuals with higher education attainment (especially greater than highschool level) compared to the general population. A gender imbalance was also noted with 58% of female respondents, compared to an expected 50% based on the most recent census. Overall, our study method was effective in recruiting participants from across the Philippines with a slight bias towards younger, female respondents with higher levels of education attainment.

Mental Health Survey Findings

The following is a summary of the most pertinent findings from the mental health portion of the survey:

• COVID-related information was sought from a broad range of sources, in our study population, with a nearly equal proportion coming from Government TV, Government Social media, Non-Government Social Media and Non-Government News (Fig. 7).
• The two most significant COVID induced challenges for our study participants were the fear of contracting COVID and the financial stress (reduction in hours, fear of job loss) associated with COVID. Unemployment was the third most significant concern (Fig. 8).
• Study participants deployed a wide - range of coping mechanisms to reduce their stress, with Movies/TV, prayer and sleep as the three most commonly named (Fig. 9).
• Overall, respondents supported enhanced community quarantine (ECQ) in their communities, although twice as many respondents were frustrated by it than relieved by it (Fig. 10).

The findings from the mental health portion of the survey are important for IPA to consider in future COVID-specific initiatives. IPA may, for example, choose to leverage some of the effective coping strategies we identified, in future outreach.

Experiment Results

We ran several regressions to compare the average treatment effect for each experiment and compute the respective confidence intervals of our estimates both with and without control variables. Two different methods were used to calculate confidence intervals. The confidence intervals for regressions with control variables were calculated assuming a normal distribution and the ones without control variables were calculated using 95% quantiles of bootstrapped sample distributions of the estimates. The control variables we chose to include are: age, gender, marital status, current student status, highest achieved school level, household size, and region of residence.

Infographic Post-Share Rate - Different Sources

Approximately 25% of study respondents shared the post with someone else, with the mental health service provider treatment group at 26.19%, the Department of Health treatment group at 25%, and NGO treatment group at 24.94%. The comparison result is shown in Figure 1, which indicates no significant difference (ɑ = 0.05) across different sources, which we interpreted to mean that respondents are eager to engage with this topic regardless of the source. This high peer-to-peer share rate warrants further exploration and is a recommended topic of future study for IPA, especially when considering methods of information dissemination.

>>>>> gd2md-html alert: inline image link here (to images/image1.png). Store image on your image server and adjust path/filename/extension if necessary.
(Back to top)(Next alert)
>>>>>

Figure 1: Difference in share rates across treatments. DOH - Department of Health, MED - Mental health service provider, NGO - non-governmental organization.

Information Resource Click Through Rate - Different Presentation Styles

In our survey, the click through rate was relatively low across all groups. It is measured by recording if the respondent clicked on any of the three links presented (0.79% for government, 0.98% for NGO, and 0.87% for mental health service provider). The control group has a click through rate of 2.73%, with image group at 2.26% and infographic group at 1.95%. Image and infographic both seemed to suggest a tendency towards lower click through rates compared to the link alone. We suspect the reason could be the overload of information within chatbot, as discussed with members of the IPA Philippines team, although this was difficult to definitively establish since all values were very low, and the study was underpowered. This low click-through-rate is an informative result because it could inform IPA’s future communication strategies especially when compared to the high peer-to-peer post share rate obtained in the previous treatment.

>>>>> gd2md-html alert: inline image link here (to images/image2.png). Store image on your image server and adjust path/filename/extension if necessary.
(Back to top)(Next alert)
>>>>>

Figure 2: Difference in clickthrough rate across treatments and model specifications

Heterogeneous Treatment Effect Analysis:

Several additional regressions were conducted with individual covariates of interest, with the goal of capturing whether some covariates would contribute to higher post share rate / click through rate. The covariates were split into subgroups (as outlined below) to analyze differences among different groups of respondents while retaining more observations for each sub-group: Age was divided into “Young” (18-25), “Middle Age” (26-49), “Old” (50+); Marital Status was divided into “Single” and “Partner”; Region was divided into “Urban” and “Rural” (constructed based on whether the region urbanization level is higher than national average) (Psa.gov.ph., 2020); Education was divided into “Post Secondary or higher” and “High School or less”; Household size was split into “Solo” (size = 1) and “Group” (size > 1).

Most of the regressions returned results without any significant treatment effects because of the limited number of observations in each subgroup, although some of the covariates were significant across all three treatment groups for post share rate.

**Age **(Fig. 11): The young group (18-25) is less likely to share than the middle age group (26-49). This result only holds for two of the treatment groups (government and mental health service provider) but not the NGO treatment group if we compare the young group against the old group. We believe if we increase the sample size, this would be a good question to explore further to better understand which organizations different age groups are more willing to engage with, and can guide IPA outreach strategies.

School & School Level (Figs. 12 and 13): Those who are out-of-school are more likely to share compared with current students. Those with at least a post-secondary degree are more likely to share as well. These results are not unexpected as people who are out-of-school and have higher levels of education attainment could have had more exposure to the topic of mental health.

Discussion

The Facebook ad clickthrough rate (1.1%) was above the industry average of 0.9% (Wordstream, 2018), so users were attracted to the study by the ad. However, only 56% of respondents completed the study in its entirety and nearly one quarter of respondents dropped out prior to completing the consent. Further analysis into why respondents chose NOT to complete the study could be meaningful. The FB ad was effective in recruiting a study population that was representative of the population in terms of geographic distribution, but male, older, and less educated individuals were under-represented in the sample. Strategies to specifically target these under-represented groups will be important for future IPA studies.

The peer to peer post share rate (~25%) far exceeded the information resource click through rate (~2%). This may be interpreted to mean that people are more likely to share resources with others, than seek care for themselves. This is a useful finding for IPA to inform future communication strategies - if IPA seeks to spread awareness about specific resources or best practices, about mental health or any other dimension of social wellbeing, condensing the information to an aesthetically appealing social media post will likely get an order of magnitude more engagement with respondents. The question as to why is one worth exploring. Perhaps some of the participants noticed the onset or worsening of mental health symptoms in those they chose to share the resource with. Existing mental health studies support this and reports that partners of depressed patients are often aware of the signs of depression weeks before the patient recognizes them. Alternatively, some of the sharers could be aware of their own mental health needs and are attempting to open a dialogue with the individual they shared with. **The explanatory mechanism of these results cannot be confirmed with our findings but the fact that one-quarter of participants chose to share mental health resources others gives credence to the importance of this issue and communication style, and warrants further exploration. **

Our study results show that there was no significant difference in share rates based on source. One explanation is that our result was expected as it was underpowered to detect a difference. Our discussions with IPA Philippines team suggested that local governments are the most trusted information source for Filipinos. Perhaps if we had included local resources we would have seen a difference in resource selection with our study size. IPA could consider a future study design that presents a local government mental health resource as an option to test the hypothesis that local government resources are the most trusted, along with other recommendations described in the next section to obtain a precise estimate of trust in specific sources.

Next Steps and Conclusion

Our facebook ad and chatbot survey were effective in recruiting, engaging, and soliciting information from participants. The following next steps could be considered to further refine the study methodology:

• Test for heterogeneity across sub-samples of the population, particularly to guide IPA communication strategy (For example, consider age, income or region specific communication tactics)
• Use larger, representative sets of posts with varying characteristics to obtain a more precise estimate of engagement/trust with that specific source.
• Present clear answer options for respondents for open-response questions for increased respondent retention
• Consider strategies to increase the number of study participants (run study longer, offer incentive to participants) and to target under-represented populations (male, older, less educated)

The following suggestions would develop IPA’s understanding of mental health engagement in the Philippines:

• Explore the effect of user engagement treatments (committing respondents to take mental health awareness pledges, using identifiable victim hypothesis, etc.)
• Consider using the chatbot survey design for clinical studies (For example, inclusion of PHQ-9 questionnaire for depression screening)
• Use FB posts to spread awareness about IPA-verified mental health resources and conduct a follow-up survey with those who access each resource (using Facebook would be helpful here because of the ease of tracking clicks) to assess self-reported improvement after a mental health professional-recommended period of time (6-8 weeks).
• Consider focusing on increased mental health burden in health care workers
• Test the effects of different kinds of images and infographics on clickthrough rates. For example, test if images with specific themes, family, community, religion, positivity have different effects

The COVID pandemic is going to continue for the foreseeable future and having effective communication strategies to quickly disseminate reliable information to vast populations is critical for organizations and governments alike. Our chatbot survey proved to be a meaningful engagement strategy and considered for future IPA initiatives in the Philippines.

**Analysis Code README: **https://docs.google.com/document/d/196GRWRu2TISZE6a8Du7-jqVozNnSe_DMf71xqx9DzFA/edit?usp=sharing

References

Lally J, Tully J, Samaniego R. Mental health services in the Philippines. BJPsych Int. 2019;16(3):62‐64. doi:10.1192/bji.2018.34

Lauber, Christoph; Rössler, Wulf (September 20, 2006). "Stigma towards people with mental illness in developing countries in Asia". International Review of Psychiatry. 19 (2): 157–178. doi:10.1080/09540260701278903. PMID 17464793.

MIMS News. 2020. Mental Health In The Philippines: By The Numbers. [online] Available at: https://today.mims.com/mental-health-in-the-philippines--by-the-numbers. [Accessed 27 April 2020].

Ncmh.gov.ph. 2020. [online] Available at: http://www.ncmh.gov.ph/. [Accessed 24 April 2020].

Psa.gov.ph. 2020. Urban Population In The Philippines (Results Of The 2015 Census Of Population) | Philippine Statistics Authority. [online] Available at: http://www.psa.gov.ph/content/urban-population-philippines-results-2015-census-population [Accessed 26 May 2020].

Who.int. 2020. [online] Available at: https://www.who.int/mental_health/evidence/philippines_who_aims_report.pdf. [Accessed 25 April 2020].

Wordstream.com. 2020. Facebook Ad Benchmarks For YOUR Industry [Data]. [online] Available at: https://www.wordstream.com/blog/ws/2017/02/28/facebook-advertising-benchmarks> [Accessed 24 May 2020].

Appendix A: Executive Summary of the State of Mental Health in the Philippines

The COVID 19 pandemic has increased stress levels and exacerbated mental health issues worldwide. Interviews with the IPA Philippines team revealed that mental health awareness and resourcing are **unmet needs in the Philippines **that could be worsened by the COVID-19 pandemic.

Further discussion and research revealed that there is a widespread tendency to **stigmatize and discriminate **against people with mental illness in the Philippines (Lauber, 2006). People with mental illness, in the Philippines, are often considered to be dangerous and aggressive. Social disapproval and devaluation of families with mentally ill individuals is common (Lauber, 2006). The literature further suggests that **supernatural, religious and magical approaches to mental illness treatment are prevalent **as patients are skeptical of mental health services and treatments offered by medical providers.

Resources for mental health treatments in the Philippines are also lacking. In 2006, the World Health Organization (WHO) reported 3.5 mental health providers per 100,000 with approximately 5% of the health budget dedicated to mental health in the Philippines and a strong recommendation was made for a national mental health act (Who.int, 2020). A Philippines National Mental Health Act was passed in 2018 (Lally, 2019). Despite this, awareness and resourcing have not improved, and, in fact, the Philippines have the lowest mental health resourcing in all Pacific Rim Countries (Mental Health in the Philippines: By the numbers, 2020). For example, the ratio of psychiatrists to residents in the Philippines is 0.5 / 100,000 compared to 125/ 100,000 in the USA - a difference of 300%. At this level, the Philippines remains well below the WHO recommendation of 10 psychiatrists per 100,000 (Who.int, 2020).

Given the underlying deficiency in mental health resources, and increased burden due to the COVID pandemic, assessing the mental health burden, alleviating factors and effective communication tools could provide meaningful insights for IPA go forward.

Appendix B Survey Questions

Demographic Questions

1. How old are you?
2. Which gender do identify with?
   1. Male
   2. Female
   3. Other: Please Specify
3. Are you single, married, widowed, divorced/separated, or in a common-law/live-in arrangement? 4. Single 5. Married 6. Widowed 7. Divorced/separated 8. Common-law/live-in 9. Don’t know/prefer not to answer
4. What is your religious affiliation?
5. Are you currently in school? 10. Yes 11. No 12. Don't know/No answer
6. What is the highest level of education you have completed? 13. None 14. Preschool 15. Elementary grad 16. High school grad 17. Post secondary grad 18. College grad 19. Post baccalaureate 20. Don't know/No answer
7. How many people live with you in your household (including yourself)?
8. Which country do you live in?
9. What region of the Philippines do you live in? 21. NCR 22. I Illocos 23. Cagayan Valley 24. III Central Luzon 25. IV-A Calabarzon 26. Mimaropa 27. V Bicol 28. VI Western Visayas 29. VII Central Visayas 30. VIII Eastern Visayas 31. IX Zamboanga Peninsula 32. X Northern Mindanao 33. XI Davao Region 34. XII Soccsksargen 35. XIII Caraga Region 36. XIV Cordillera Administrative Region 37. BARMM 38. Other/N/A"
10. Which city do you live in now?

Mental Health Impact Questions

11. Has COVID-19 led to an increased frequency of the following feelings (choose all options that apply): 39. Nervous 40. Anxious or on edge 41. Trouble relaxing 42. Becoming easily annoyed or irritable 43. Afraid that something awful might happen 44. Trouble falling or staying asleep 45. None 46. Other 47. Don't know/No answer
12. Have you faced any specific challenges during the lockdown period? Please select as many as apply from the list below: (ENTER THE LETTER(S)) 48. Fear of unemployment 49. Fear of contracting COVID-19 (self or family or friend) 50. Other health issues 51. Inability to access healthcare service 52. Challenging/Problematic relationships 53. Financial stress and pressure 54. Violence 55. None 56. Other (please specify)
13. If you resided in a region that was/is placed under enhanced-community quarantine (ECQ), how did/do you feel about it? 57. Frustrated and felt ECQ was unnecessary 58. Frustrated but felt ECQ was necessary 59. Indifferent 60. Felt relieved because of ECQ 61. Did not experience ECQ 62. Don't know / No answer
14. During the lockdown, I have communicated with others by doing one or more of the following (select as many as apply to you): 63. Chatting to the people I live with (in person) 64. Video calling 65. Groups / group chats (e.g. Facebook, WhatsApp) 66. Posting / commenting on Social media 67. None 68. Other (please specify) 69. Don’t know / No answer
15. During the lockdown, I get my Covid-19 updates and information from the following sources (select as many as apply): 70. Televised speeches from the government 71. Government social media 72. Press, media, news (non-government) 73. Social media (non-government) 74. Friends and family 75. None 76. Other (please specify) 77. Don’t know / No answer
16. During the lockdown, I feel better after I have: (select as many as apply) 78. Exercised 79. Watched a movie or television show (not the news) 80. Kept up to date with the latest news (online or on television) 81. Took a nap 82. Chatted to someone 83. Done some remote work 84. Written in my journal 85. Done housework 86. Shared a meal with family members 87. Prayed / meditated 88. Stayed away from problematic people I am in lockdown with 89. None 90. Other (please specify) 91. Don’t know / No answer
17. If a family member or a friend approached you about their increased level of stress, which of the following resources are you most likely to recommend? 92. Department of Health (National Center for Mental Health) 93. Non-governmental organization 94. Mental health service professional 95. None 96. Other (please specify) 97. Don’t know / No answer

Action Encouragement Questions

Following Questions (18-20) are factorial; Respondents may receive none, 1, 2, or all

18. Do you think of yourself as someone that other people rely on for information or support? 98. Yes 99. Not Really
19. Will you join us in making a pledge to take action and raise awareness around mental health during the Coronavirus pandemic? 100. Yes 101. Not Right Now
20. Given the challenges of the Coronavirus pandemic and community quarantine, can you think of someone you care about who feels alone and might be struggling right now? 102. Yes 103. No 104. Don’t know / No Answer
21. If not Yes to Question 20, Is there anyone in your neighborhood, family, or community who feels alone and might be struggling right now? 105. Yes 106. No 107. Don’t know / No Answer

Treatments

22. You can help spread awareness by sharing information about mental health with your friends and family. If you would like to share the below post to your timeline, click "Share." (Shown a randomized post from one of three sources: Department of Health, NGO, Mental health service provider)
23. If you are interested in learning about more resources to support mental health in the Philippines, you can find the links below (shown three links from different sources in a randomized order: Department of Health, NGO, Mental health service provider)

Reminder

24. Would you like to set a reminder to reach out to a friend? We can send you a reminder every day, starting tomorrow, for 1, 2, or 3 days. How many reminders would you like us to send you? 108. Choices: 0, 1, 2, 3

Note: The full script is also available in google sheet [LINK], including introductions and transitional sentences between questions.

Appendix C Demographic/Mental Health Questions Visualizations

>>>>> gd2md-html alert: inline image link here (to images/image3.png). Store image on your image server and adjust path/filename/extension if necessary.
(Back to top)(Next alert)
>>>>>

Figure 3: Geographic Distribution

Analysis of the Geographic distribution of respondents demonstrates that the study population matches the actual population geographic distribution in the Philippines.

>>>>> gd2md-html alert: inline image link here (to images/image4.png). Store image on your image server and adjust path/filename/extension if necessary.
(Back to top)(Next alert)
>>>>>

Figure 4: Age Distribution

Our study sample skews towards a younger age distribution when compared to the Philippine national age distribution. It also had a gender imbalance with 58% female, as compared to 50% female on the national census.

>>>>> gd2md-html alert: inline image link here (to images/image5.png). Store image on your image server and adjust path/filename/extension if necessary.
(Back to top)(Next alert)
>>>>>

Figure 5: Level of Education

The study captures an increased response rate from individuals with higher education (greater than highschool level) compared to the general population.

>>>>> gd2md-html alert: inline image link here (to images/image6.png). Store image on your image server and adjust path/filename/extension if necessary.
(Back to top)(Next alert)
>>>>>

Figure 6: Household Size

Survey Respondents came from a range of household sizes. The average household size in the study, 4.4, matched the national average of members per household in the census. Note that household size can be used as a proxy for income levels.

>>>>> gd2md-html alert: inline image link here (to images/image7.png). Store image on your image server and adjust path/filename/extension if necessary.
(Back to top)(Next alert)
>>>>>

Figure 7: COVID Information Sources

COVID related information is sought from a broad range of sources in the study population with a nearly equal proportion coming from Government TV, Government Social media, Non-Government Social Media and Non-Government News.

>>>>> gd2md-html alert: inline image link here (to images/image8.png). Store image on your image server and adjust path/filename/extension if necessary.
(Back to top)(Next alert)
>>>>>

Figure 8: COCID Induced Challenges

The two most significant COVID induced challenges are :

1. The FEAR of Contracting COVID
2. Financial Stress

Unemployment is a distant third.

>>>>> gd2md-html alert: inline image link here (to images/image9.png). Store image on your image server and adjust path/filename/extension if necessary.
(Back to top)(Next alert)
>>>>>

Figure 9: COVID Coping Mechanisms

Study participants deploy a wide - range of coping mechanisms to reduce stress, with Movies & TV, Prayer and Sleep as the three most common. Work and journaling with the two least popular of the modalities surveyed.

>>>>> gd2md-html alert: inline image link here (to images/image10.png). Store image on your image server and adjust path/filename/extension if necessary.
(Back to top)(Next alert)
>>>>>

**Figure 10: ** Sentiment towards Enhanced Community Quarantine

The overall understanding of the necessity for enhanced community quarantine (ECQ) was clear, although twice as many respondents were frustrated by it than relieved by it.

Appendix D Conditional Average Treatment Effect (CATE) Regression Outputs

>>>>> gd2md-html alert: inline image link here (to images/image11.png). Store image on your image server and adjust path/filename/extension if necessary.
(Back to top)(Next alert)
>>>>>

Figure 11: Conditional Average Treatment Effect on Age Segments

We have three groups of age segments: young population between 18 and 25, adults between 25 and 50, and elders, above 50 years old. The above regression shows that people between 18 and 25 are less likely to share than people between 25 and 49.

>>>>> gd2md-html alert: inline image link here (to images/image12.png). Store image on your image server and adjust path/filename/extension if necessary.
(Back to top)(Next alert)
>>>>>

Figure 12: Conditional Average Treatment Effect on School Segments

We have two groups in school: one that it is participating currently in school and the other that it is not. The regression above shows that current students are less likely to share than non-students.

>>>>> gd2md-html alert: inline image link here (to images/image13.png). Store image on your image server and adjust path/filename/extension if necessary.
(Back to top)(Next alert)
>>>>>

**Figure 13: **Conditional Average Treatment Effect on School Level Segments

We have two groups at school level: one that has at least a post-secondary degree and other that does not. The regression above shows that respondents with a higher education level are more likely to share.