key: cord-117800-jzokod4q authors: Umer, Hamza; Khan, Muhammad Salar title: Evaluating the Effectiveness of Regional Lockdown Policies in the Containment of Covid-19: Evidence from Pakistan date: 2020-06-04 journal: nan DOI: nan sha: doc_id: 117800 cord_uid: jzokod4q To slow down the spread of Covid-19, administrative regions within Pakistan imposed complete and partial lockdown restrictions on socio-economic activities, religious congregations, and human movement. Here we examine the impact of regional lockdown strategies on Covid-19 outcomes. After conducting econometric analyses (Regression Discontinuity and Negative Binomial Regressions) on official data from the National Institute of Health (NIH) Pakistan, we find that the strategies did not lead to a similar level of Covid-19 caseload (positive cases and deaths) in all regions. In terms of reduction in the overall caseload (positive cases and deaths), compared to no lockdown, complete and partial lockdown appeared to be effective in four regions: Balochistan, Gilgit Baltistan (GT), Islamabad Capital Territory (ICT), and Azad Jammu and Kashmir (AJK). Contrarily, complete and partial lockdowns did not appear to be effective in containing the virus in the three largest provinces of Punjab, Sindh, and Khyber Pakhtunkhwa (KPK). The observed regional heterogeneity in the effectiveness of lockdowns advocates for a careful use of lockdown strategies based on the demographic, social, and economic factors. The world is struggling to combat the Covid-19 pandemic, which has spread from Wuhan China to over 190 countries. Unless there is a viable treatment or vaccine to treat Covid-19, the world is taking possible preventive measures to minimize the spread of the disease. Lockdown (complete or partial) is one of the most evident and widely used preventive measures. The effectiveness of lockdown in controlling the spread of Covid-19, however, is not a well-established outcome for a few reasons. First, so much is unknown about the novel coronavirus and that the situation is evolving, which gives policymakers little time to think through, implement, and properly investigate or foresee the effectiveness of any policy, such as a lockdown. Second, while many countries-for instance, New Zealand and Germanyhave already implemented complete or partial lockdown (McFall-Johnsen et al., 2020) , it is just hard to conduct a long-term ex-ante analysis of the repercussions of policy concerning once-in-a-century pandemic. Still, some preliminary studies tried to investigate the effect of lockdown strategies on the spread of Covid-19. For instance, Walker and Colleagues (2020) predicted that interventions and lockdown strategies in almost all countries (precisely, 202 countries in their analysis) would reduce infections and deaths by nearly half. Mushfiq Mobarak and Colleagues (2020) , contrarily argued that while lockdown strategies are viable Umer & Khan 3 options in high-income countries, low-income countries such as Nigeria and Pakistan cannot afford to have a fruitful lockdown. Because of their weak capacity in enforcing lockdown strategies, these countries may witness counterproductive effects if such strategies make workers and migrants migrate back from heavily populated urban areas and spread the disease to remote rural areas, the researchers argue. Similarly, we have a few studies that show different outcomes on the country-level. For example, research (Dowd et al., 2020 ) on the Italy outbreak shows the effectiveness of early lockdown. In Italy, the Covid-19 was first detected in the Lodi province, which placed restrictions beginning February 22. As opposed to this, Bergamo province, which started with fewer cases but did not impose restrictions until March 7, far surpassed the number of cases in Lodi (Stancati, 2020) . In a similar vein, a study (Kumar and Nataraj, 2020) on Indian regions shows a regional differential in the spread of Covid-19 in the face of regional policy variation. Overall, the world is divided regarding the use and effectiveness of lockdown policies. On one side, we see countries like Japan (Du and Huang, 2020; Ian, 2020) and Sweden (Karlson et al., 2020) that have used mild lockdown or no lockdown and yet effectively contained the spread of the virus. On the other side, we see countries like New Zealand and Australia using strong lockdown policies to flatten the spread of the virus (Fifield, 2020) . This article offers a systematic contribution to the aforementioned debate by examining the effectiveness of lockdown policies in the containment of Covid-19 in the context of Pakistan. Specifically, the article explores how effective the lockdown strategy has been in combating Covid-19 outcomes in the country. This evaluation of lockdown policies is based on the econometric analysis (such as regression discontinuity and negative binomial regressions) performed on official data from Pakistan. Pakistan is selected because it offers a valuable opportunity to analyze the effects of both complete lockdown and partial lockdown policies on the spread of the Covid-19 virus. Moreover, the regional use of lockdown policies in Pakistan is heterogeneous and hence enables us to perform cross-regional analysis as well. We find that in comparison to no lockdown, complete and partial lockdowns have been ineffective in the containment of the virus in the three largest provinces Punjab, Sindh, and Khyber Pakhtunkhwa (KPK). On the other hand, complete and partial lockdowns have been Umer & Khan 4 very effective in the containment of the virus in the province of Balochistan and the three administrative territories/regions of Gilgit Baltistan (GB), Islamabad Capital Territory (ICT), and Azad Jammu and Kashmir (AJK). The observed regional heterogeneity in the effectiveness of lockdowns advocates for a careful use of lockdown based on the demographic, societal, and economic factors. "One size fits all" approach for lockdown could be counterproductive in some regions of the world and subsequently make the spread of virus more acute, as demonstrated by researchers in the context of Africa as well (Mehtar et al., 2020) . Pakistan-home to about 220 million and wobbly health infrastructure that has close to 1.5 million hospital beds (Khan & Latif, 2020) -reported the first case of the Covid-19 on February 26, a returning pilgrim from neighboring Iran (Hashim, 2020) . On the same day, the Pakistan Federal Ministry of Health confirmed another positive case in Islamabad (Ali, 2020) . Since then, the virus has diffused quickly. By March 18, all the administrative regions of Pakistan, including four provinces (Punjab, Sindh, KPK, and Balochistan), the two autonomous territories (AJK and GB), and the federal territory of Islamabad registered positive cases. The entire country reported over eighty-five thousand confirmed cases and 1,770 deaths, as of June 4, 2020. 1 In terms of the total number of cases and deaths, Pakistan ranks 17 th and 21 th worldwide, respectively. 2 While now the virus has entered the community transmission stage, initially, all the confirmed cases in Pakistan had recent travel history from Iran, Syria, London, and Saudi Arabia. The Covid-19 pandemic has spread unevenly across regions within Pakistan, with the four regions (provinces) making up more than 95 percent of the cases as of June 4, 2020. Sindh registered the most cases at over 39,900, followed by Punjab (31,104), KPK (11, 373), and Balochistan (5, 224) . 3 The province of Punjab reported 607 deaths, the most in the country, followed by Sindh (555) and KKP (500) and Balochistan (51). The situation in the three special regions or territories is not that bad, with Islamabad, GB, and AJK reporting 3,544, 1 This data was obtained from Worldometers: https://www.worldometers.info/coronavirus/ The data was cross-verified here: https://www.cdc.gov/coronavirus/2019-ncov/global-covid-19/world-map.html 2 Ibid. 3 Official government website for data on Covid-19 was consulted: http://covid.gov.pk/stats/pakistan Source: http://covid.gov.pk/stats/pakistan As a nascent federalist country, 5 when the Interior Ministry of Pakistan announced a lockdown on March 23 to combat the spread of the virus, all the seven administrative regions also implemented their regional lockdown measures at or around March 23. Army troops were deployed throughout the country to help the divisions in tackling the spread of the virus. 6 Initially, the regions implemented a full or complete lockdown, with ICT shutting down as early as on March 12, Punjab on March 22, and all other divisions on March 24. 7 4 Ibid. 5 Pakistan is a federalist country, with provincial governments having the right to decide on important issues, according to the Eighteenth Amendment of the Constitution of Pakistan. 6 https://www.geo.tv/latest/278812-government-calls-in-pakistan-army-troops-amid-coronavirus-outbreak 7 Complete lockdown and partial lockdown information is extracted from several newspapers and online sources. These include: Business Recorder, Dawn News, Radio Pakistan, Technology Times. Also please refer to the report (Covid-19 Legislation and Measures in Pakistan) published by Zafar Kalanuri (2020). Complete lockdown refers to complete shutdown of socioeconomic, religious activities and mobility pathways while partial lockdown refers to controlled opening of the aforementioned. Complete description of lockdown types is mentioned on page 6. The duration of this full lockdown also varied, with ICT and KPK observing a short duration lockdown for less than a week, Punjab, Sindh, and Balochistan observing a medium duration lockdown for about two weeks, and AJK and GB observing a long duration lockdown for almost a month. Later, the divisions moved to a partial or controlled lockdown (please see Table 3 for details). Policy-wise, Pakistan acted quickly and formulated a National Action Plan for Covid-19 early in February 2020 (Mukhtar, 2020) . The Ministry of National Health Services, Regulation & Coordination Pakistan presented the Plan, that was supposed "to provide (a) policy framework for federal, provincial, and regional stakeholders for building capacity to prevent, detect and respond to any events due to COVID-2019 in Pakistan." 8 Along with the health ministry, National and Provincial Disaster Management Authorities, National Command and Operation Center (NCOC), and National Coordination Committee (NCC) have been formulating, coordinating, analyzing, and implementing policy efforts about Covid-19. With federal directives, regions have been managing outbreaks according to their circumstances. Table 1 provides details about the total number of positive cases, deaths caused by and total tests performed in the seven regions of Pakistan. workplaces, parks and other public places, ban on social gatherings and social events, closing of land and air transport and people are restricted to stay home unless they need medical help or require grocery shopping. Partial lockdown refers to the controlled opening of economic activities for a specific time every day, limited resumption of land and air transport, and maintaining social distance during outdoor activities. Most of the educational institutes, however, remain closed. 11 The data covers a time period of 60 days (March 12 to May 11, 2020) and encompasses no lockdown, complete lockdown, and partial lockdown phases. The entire data variables are described in Table 2 . It is a time variable that represents day. In Table 3 , the summary of the variables is reported. Punjab, Sindh, and KPK are three provinces witnessing the highest number of average daily positive cases and average daily deaths. ICT had the longest complete lockdown (34 days), while KPK had the shortest one (5 days). 3/12--5/11 3/12--5/11 3/12--5/11 3/12--5/11 3/12--5/11 3/12-- Lockdown 12 Note: n = Observations; m = Mean. Standard deviations are in parentheses. *AJK did not report any death during the data duration specified in this paper. All the econometric analyses are performed in STATA 16. The outcome variables include the number of daily deaths due to Ccovid-19 and the number of people testing positive for Ccovid-19. The main explanatory variables are lockdown and partial lockdown dummies. The starting point of the analysis is a visual representation of the impact of lockdown policies on the outcome variables. This is achieved by using regression discontinuity (RD) with the date as running variable and the lockdown and partial lockdown dates as multiple cutoffs. In STATA and KPK) cumulatively representative of 96% of Covid-19 cases are in Figure 2 . As these four provinces account for the majority of the positive cases, we discuss them at length here, while binned scatter plots for ICT, AJK, and GB regions are reported in appendix A. The binned scatter plots reported above indicate the lockdown policies induced heterogeneous regional effects on Covid-19 outcomes. We further analyze the impact of lockdown policies on Covid-19 outcomes systematically by using different regression techniques and estimate the following two equations. In the above equations ∑ 7 15 LR test is used for the evaluation of alpha. Null hypothesis alpha =0 is rejected at 1% for both equations. and indicates Poisson regressions had overdispersion (conditional variance exceeds conditional mean). Hence, we use the output from Negative Binomial regressions for analyzing the effects of lockdown policies in Table 4 . Robust standard errors are in parentheses. ***p<0.1; **p<0.5; *p<0.1 Regressions 1 While Holding all other variables constant, the effect of daily tests performed is significant and positive on daily positive cases, however, the magnitude is very small (difference in logs of expected counts increases by 0.0002 in regression 1). Next, we turn to main explanatory variables. In comparison to no lockdown, complete lockdown had a significant and negative discussion section 7, we explore the possible reasons leading to regional heterogeneity in the outcomes of lockdown policies. In this section we turn to another proxy for lockdown-the daily stringency score estimated by Hale et al. (2020) . The score varies from zero to 100; a higher value indicates more stringent controls to contain the virus spread. The stringency score is based on the cumulative value of restrictions imposed on schools, workplaces, public events, social gatherings, public transport, stay home orders, domestic and international travel, public information campaign, testing policy, and contact tracing (Hale et al., 2020) . Essentially, lockdown and stringency measures are two different ways of quantifying restrictions in an economy. The stringency score by Hale et al. (2020) is for the whole country, and resultantly, we are unable to do regional analysis using this score. Therefore, we pool the regional data to obtain country-level data and subsequently perform analysis using daily stringency score as the main explanatory variable. The stringency score has 55 observations; its value ranges from 34 to 97, with a mean value of 86.29 and a standard deviation of 18.51. Using the stringency score following two equations are estimated. Negative Binomial regression is used; the value of alpha is significant 17 , indicating Poisson regression had over -dispersion. The regression output is reported in table 5. In regression 2 and 4, the variable for daily tests is excluded to check the robustness of the effect of the stringency score on the two outcome variables. This exclusion, however, does not alter the significance of the stringency variable. The effect of daily tests on daily deaths is insignificant in the main regression (1) level results show that stringency measures appear to be ineffective in the control of damage caused by Covid-19 and in its spread as well. The regional results show heterogeneity in the effectiveness of lockdown measures while country-level results point towards the ineffectiveness of stringency measures. To reconcile the regional results with country-level results, we need to focus on the three important regions of Punjab, Sindh, and KPK. Together these three regions account for 90% of the total positive cases and 96% of the total deaths during the time period considered in this study. As these three regions represent a huge share of Covid-19 outcomes, they are playing a significant role in driving the country-level statistics. Resultantly, the ineffectiveness of lockdown policies in these regions is reflected in the ineffectiveness of stringency measures at the country-level. As discussed earlier, the effect of regional lockdown policies on Covid-19 outcomes is heterogeneous, with very few regions effectively using the lockdown policies to contain the spread of the disease. From a policy perspective, we identify here the factors that have contributed jointly to the ineffectiveness of lockdowns, specifically in the three largest regions of Punjab, Sindh, and KPK. The central government in Pakistan has never been unified over the imposition of lockdown measures. The country's Prime Minister (PM) adamantly opposed lockdown fearing economic impacts on daily wagers that comprise a significant proportion of the country's labor force. On the other hand, several ministers in the PM's cabinet proposed strict lockdown measures to contain the virus. As a result of this confusion prevalent in the government circles, the potential risks of Covid-19 were downplayed, the public did not observe the lockdown restrictions seriously, and resultantly the lockdown proved to be ineffective in three large provinces 18 . A significant proportion of Pakistan's population inhibits in Punjab (110 million), Sindh (47.9 million) and KPK (30.5 million) regions (Wazir & Goujon, 2019) . Together these three regions account for approximately 89% of the country's population, and a major proportion of this population resides in rural areas or slums in large cities (for example Sindh's capital city Karachi has the world's largest slum population approximated to be 2 million 19 ). In rural and specifically slum areas, social distancing or keeping one restricted to home are almost nonexistent, issues of cleanliness are acute, and poverty rates are high. All these factors, when combined together, provided an ideal habitat for the sharp spread of Covid-19, even in the presence of lockdown measures. The informal labor force makes up 75% of the total labor force (48.75 million). About 40% of this informal labor force (19.5 million) is employed by the agriculture sector, while 60% (29.25 million) works for the industrial and service sector. 20 The majority of the informal labors are paid on a daily basis and unfortunately forced to leave homes in an attempt to earn money for food and subsequently violate and undermine the effectiveness of lockdown restrictions by serving as a potential source of virus spread within their workplace and residences. Pakistan is a religiously homogeneous country with Muslims making up 96.28% of the population. 21 In Islam, daily five congregational prayers in the mosque are a vital part of worship. Congregations and gatherings could be favorable grounds for viral transmission. In Pakistan, the religious, as well as political leaders, were divided on the closing down of mosques. Resultantly, government in coordination with religious leaders set up protocols for congregational prayers in mosques, however in most cases these protocols were not strictly followed, possibly leading to rapid transmission of virus even during the lockdown phases. 22 As an entirely unprecedented situation for masses, the heterogeneous behavioral response of the public to lockdown was evident. In the case of Pakistan, poor handling of the situation by the government over using force to implement lockdown strategies in the big provinces led to fear in the minds and created a social stigma around Covid-19. Subsequently, people with symptoms kept on living with family members, and whole families got infected. These numbers eventually started showing up post lockdown in the largest regions. A lack of literacy and misinformation about the disease and its treatment further complicated the outcomes. On the contrary, in AJK and GB, where people were willing to observe lockdown, showed better outcomes. One reason why they showed this willingness is likely their high literacy rate than the national average rate. From a policy perspective, it is clear that lockdown effectively used by rich countries, including Germany, Japan, and USA, is to a large extent, ineffective in controlling the spread of virus in a poor country like Pakistan. In fact, several researchers have already discussed the possibility of the lockdown being ineffective in poor countries (for example Barnett-Howell & Mobarak, 2020; Cash & Patel, 2020) . The country-level results from this article support the predictions of these studies. We think contextualized strategies would be way more effective in the control of the virus in poor countries. In the case of Pakistan, the socio-economic and political conditions and religious norms are some of the important contextual elements that should be considered while making future lockdown strategies. The purpose of the paper was to examine the impact of lockdown strategies in different regions ▪ Federal and local governments should involve local religious scholars and community elders in teaching the public the importance of social distancing, hygiene, and prevention. They should also work on dispelling and addressing the stigma around Covid-19 so that people can test themselves without anxiety and fear from society. ▪ Government should allow shorter working hours, limit number of people at religious and social gatherings, and implement universal masking using cloth masks for the community. (1) (2) (3) Robust standard errors are in parentheses. ***p<0.1; **p<0.5; *p<0.1 Pakistan Confirms First Two Cases of Coronavirus, Govt Says 'No Need to Panic Should Low-Income Countries Impose the Same Social Distancing Guidelines as Europe and North America to Halt the Spread of COVID-19? What Japan Can Teach the World About the Pandemic Has COVID-19 Subverted Global Health? A practical introduction to regression discontinuity designs How Does Your Kindergarten Classroom Affect Your Earnings? Evidence from Project Star Demographic Science Aids in Understanding the Spread and Fatality Rates of COVID-19 Did Japan Just Beat the Virus Without Lockdowns or Mass Testing? Umer & Khan 24 New Zealand Isn't Just Flattening the Curve. It's Squashing It Variation in government responses to COVID-19. Blavatnik School of Government Working Paper The Curious Case of South Asia's 'low' Coronavirus Deaths Covid-19 Legislation and measures in Pakistan Sweden's Coronavirus Strategy Will Soon Be the World's The Coronavirus and Pakistan: Why People Must Immediately Begin Social Distancing?" The National Interest The Impact of COVID-19 Mobility Restrictions in India: Comparing State and Central Responses Assessing the 2017 Census of Pakistan Using Demographic Analysis: A Sub-National Perspective Countries around the World Are Reopening -Here's Our Constantly Updated List of How They're Doing It and Who Remains under Lockdown Limiting the Spread of COVID-19 in Africa: One Size Mitigation Strategies Do Not Fit All Countries Preparedness and Proactive Infection Control Measures of Pakistan during COVID-19 Pandemic Outbreak National Action Plan for Corona virus disease (COVID-19) Pakistan Lockdown of Recovering Italian Town Shows Effectiveness of Early Action The Global Impact of COVID-19 and Strategies for Mitigation and Suppression