key: cord-286975-id5dn795
authors: Carlitz, Ruth D.; Makhura, Moraka N.
title: Life Under Lockdown: Illustrating Tradeoffs in South Africa’s Response to COVID-19
date: 2020-08-28
journal: World Dev
DOI: 10.1016/j.worlddev.2020.105168
sha: 
doc_id: 286975
cord_uid: id5dn795

This research note sheds light on the first three months of the COVID-19 outbreak in South Africa, where the virus has spread faster than anywhere else in the region. At the same time, South Africa has been recognized globally for its swift and efficient early response. We consider the impact of this response on different segments of the population, looking at changes in mobility by province to highlight variation in the willingness and ability of different subsets of the population to comply with lockdown orders. Using anonymized mobile phone data, we show that South Africans in all provinces reduced their mobility substantially in response to the government’s lockdown orders. Statistical regression analysis shows that such mobility reductions are significantly and negatively associated with COVID-19 growth rates two weeks later. These findings add an important perspective to the emerging literature on the efficacy of shelter-in-place orders, which to date is dominated by studies of the United States. We show that people were particularly willing and able to act in the provinces hit hardest by the pandemic in its initial stages. At the same time, compliance with lockdown orders presented a greater challenge among rural populations and others with more precarious livelihoods. By reflecting on South Africa’s inequality profile and results of a recent survey, we demonstrate how the country’s response may deepen preexisting divides. This cautionary tale is relevant beyond South Africa, as much of the continent -- and the world -- grapples with similar tradeoffs. Along with measures to contain the spread of disease, governments and other development focused organizations should seriously consider how to offset the costs faced by already marginalized populations.

As COVID-19 began its global spread, it still appeared that some world regions might be spared --in particular, sub-Saharan Africa (Otu et al., 2020) . In short order, however, it became clear that such optimism was not merited. The virus quickly took a firm footing on the continent and as of August 7, 2020 the number of confirmed cases exceeded one million 1 --likely a significant underestimate given limited testing capacity.

The virus's impact has varied considerably across the continent, reflecting countries' varying degrees of global integration and capacity to respond . One country that stands out in both regards is South Africa. As seen in Figure 1 , the virus has spread faster in South Africa than in any of the continent's other large economies.

[ Figure 1 here] At the same time, South Africa has been recognized globally for its swift and efficient response (Kavanagh & Singh, 2020) . In this research note, we illustrate how the South African government's response affected the lives of ordinary people, focusing on the initial outbreak and subsequent lockdown.

First, we look at the impact of the government's strict lockdown orders on population mobility, which may be understood as a proxy for compliance. We leverage anonymized mobile phone data from Google's COVID-19 Community Mobility Reports, which chart trends over time, across different categories of places such as retail and recreation, transit stations, and workplaces. We look at changes in mobility by province to highlight variation in the willingness and ability of different subsets of the population to comply with lockdown orders.

We then analyze how changes in population mobility relate to the spread of disease, and conduct statistical regression analysis to show that mobility reductions are significantly and negatively associated with COVID-19 growth rates two weeks later. While this suggests the lockdown measures have been effective in achieving their goals, we also reflect on how the government's response and corresponding mobility reductions interact with existing inequalities, keeping in mind the country's status as the world's most unequal nation. 2 This research note contributes to a rapidly expanding literature on COVID-19's impact in the Global South, and in particular to studies that demonstrate and explain variation within countries (Okoi & Bwawa, 2020; Wenham & Smith, 2020; Brauer et al., 2020) . To our knowledge, ours is the first subnational analysis of population mobility in response to for South Africa. 3 We also contribute an important perspective to the emerging literature on the efficacy of shelter-in-place orders, which to date has been dominated by studies of the United States. 4 Finally, by discussing the tradeoffs inherent to stringent containment measures, we tell a story that resonates across the African continent and in other parts of the world where efforts to contain the spread of COVID-19 may have as many or more negative consequences as the disease itself (Coetzee & Kagee, 2020) .

This research note proceeds as follows. Section 2 details South Africa's efforts to contain COVID-19, and compares the country's response to others on the continent. Section 3 then presents our analysis of mobility trends and their relation to the growth rate of new infections. Section 4 describes the nature of inequality in South Africa in order to situate our results in context. Section 5 concludes.

The first case of COVID-19 in South Africa was confirmed on March 5th, 2020; subsequent cases were confirmed in the days that followed among citizens who had traveled to Italy on a ski trip. While the initial cases suggested the disease might be limited to the country's affluent, cosmopolitan population, President Cyril Ramaphosa announced broad measures to combat the spread of COVID-19 on March 15th. 5 As the timeline depicted in Table 1 shows, the response accelerated quickly from there.

[ Table 1 here.]

Formal regulations were published on March 18th, promoting social distancing at one person per square meter of floor space (RSA Government, 2020a). The regulations also closed schools, called for isolation of sick people, quarantining of asymptomatic people, limited gatherings to 100 people, and to 50 people at the premises where alcohol is sold and consumed.

These regulations were amended on March 25th, ushering in a strict lockdown phase (RSA Government, 2020b) . Beginning March 27th, all businesses were to remain closed, except those involved in the production and provision of essential goods and services. 6 Every person was to be confined to a place of residence unless performing essential services, obtaining essential goods or services, collecting social grants, emergency care or chronic medication attention. All places of work were to be closed except those providing essential services.

Movements between provinces, metropolitan areas and districts were prohibited, including commuter transport services, except when rendering essential services.

The lockdown was extended on April 16th to the end of the month (RSA Government, 2020c) to allow the government further time to prepare for management of the disease (Karim, 2020) . Revised regulations prohibited evictions from rental properties, permitted opening of refineries, and allowed mining to operate at reduced capacity. These were seen as laying the groundwork for re-opening the economy (RSA Government, 2020d). At the end of April, the lockdown was relaxed to allow transition into Alert Level 4 beginning May 1st, 2020 (RSA Government, 2020e). 7 The country transitioned to Alert Level 3 on June 1st (RSA Government, 2020g), which provided for movements of school children across provinces and limited religious gathering to 50 people.

Compared to many of its neighbors on the continent, South Africa's response has been swift and extensive, as shown in Figure 2 . As of April 1st, 2020, South Africa scored an 88 out of 100 on the "Stringency Index," as coded by the Oxford COVID-19 Government Response Tracker . Only 10 of 54 other countries on the continent were coded as having more stringent responses as of that date.

[ Figure 2 here.]

As we show in the next section, the country's residents acted accordingly to reduce spread of the disease by dramatically reducing their mobility.

This section examines how South African citizens responded to the government's strict containment measures, drawing on Google's COVID-19 Community Mobility Reports. These reports are based on aggregated, anonymized data from users of Google Maps, and show how visits and length of stay at different places change compared to a baseline. 8, 9 We examine four categories:

1. Workplaces: Mobility trends for places of work.

Mobility trends for places like restaurants, cafes, shopping centers, theme parks, museums, libraries, and movie theaters.

3. Transit stations: Mobility trends for places like public transport hubs such as subway, bus, and train stations.

Mobility trends for places like grocery markets, food warehouses, farmers markets, specialty food shops, drug stores, and pharmacies. Figure 3 shows that South Africa 10 consistently "outperforms" its peers in terms of reduced mobility relative to the stringency of government response with the exception of grocery and pharmacy visits reflecting the exemption of such businesses from the lockdown order.

[ Figure 3 here.]

The national average reduction in mobility masks considerable variation within the country. Figure 4 illustrates mobility trends by province 11 in terms of retail and recreation from February 14-May 15, 2020. 12 Although the provinces exhibit differences, in each case we can observe a substantial dip corresponding to the beginning of the strict lockdown period. Table 2 depicts the overall average percent change in mobility reductions during the most stringent lockdown period (27 March-30 April 2020) by provinces for the different categories.

[ Table 2 here.]

Western Cape registers the largest average decrease in mobility for all categories. On the other side, Limpopo registers the smallest decline in retail/recreation; Mpumalanga the smallest declines in grocery/pharmacy and transit station visits; and Northern Cape the smallest declines in workplace visits and smallest increase in residential mobility.

We also calculate the average mobility reduction for the three most highly correlated categories: retail/recreation, workplace mobility, and transit stations (see correlation matrix in Appendix Table A4 ). This is depicted graphically in Figure 5 .

[ Figure 5 here.]

In order to account for variation in mobility trends across provinces, we consider a number of economic and demographic factors. We also consider each province's caseload (number of confirmed cases) at the beginning of the lockdown period. These features are summarized in Table 3 .

[ Table 3 here.]

We use qualitative comparative analysis (QCA) to identify the features shared by the provinces experiencing the greatest mobility reductions. 13 Our analysis suggests that the main factors associated with substantial mobility reductions are the number of confirmed cases prior to the lockdown period and the relative size of the provincial economy. The provinces containing South African's metropolitan hubs --Gauteng (containing Johannesburg) and Western Cape (containing Cape Town) also tended to experience larger mobility reductions. This may reflect the challenge of reducing mobility in rural areas among populations that are more likely to be food insecure (Tibesigwa and Visser, 2016) . Emerging research on the determinants of compliance with social distancing and other measures suggests that concerns about income losses play an important role in determining compliance (Wright et al., 2020; Bodas and Peleg, 2020) .

In contrast to the United States, where partisanship has been shown to be a key determinant of mobility reductions and other efforts to contain the spread of disease (Grossman et al., 2020; Adolph et al., 2020) , party politics do not seem to feature prominently when it comes to explaining variation in mobility in South Africa. Both Gauteng (controlled by the ruling African National Congress party) and Western Cape (the only province held by the main opposition party, the Democratic Alliance) exhibit similar trends when it comes to mobility reductions. That said, such unity appears to be waning. For example, the Democratic Alliance filed a legal challenge against some coronavirus lockdown rules in mid-May (Democratic Alliance, 2020). The Economic Freedom Fighters (the second largest opposition party) subsequently issued a statement calling for prolonged stringent lockdown (Economic Freedom Fighters, 2020).

South Africa's strict lockdown policies --and corresponding reductions in mobility by the country's citizens --were put in place with the aim of reducing the spread of COVID-19. This leads us to ask: how effective have they been? Before we attempt to answer this question, we first present the trajectory of disease by province in Figure 6 up until May 24th, 2020. Although the first cases were confirmed in KwaZulu-Natal, the figure indicates how the disease has taken hold primarily in Western Cape. 14 [ Figure 6 here.]

In order to determine how the spread of disease has changed as a consequence of the mobility reductions discussed above, we estimate a series of regression models with average weekly exponential growth in confirmed cases as the dependent variable. 15 The independent variables are average weekly mobility reductions for each of the three main categories discussed above (workplace, retail/recreation, and transit stations) for the preceding two weeks. 16 All models include province fixed effects and standard errors clustered by province. We also include a time trend to account for any other factors changing over time within each province. The results, depicted in Table 4 , suggest that people's mobility reductions have indeed helped to play an important role in reducing the rate of new infections. However, as we discuss in the next section, the benefits and costs of lockdown are unlikely to be distributed equally.

[ Table 4 here.]

As noted above, South Africa bears the unflattering distinction of the world's most unequal nation in terms of income inequality. Furthermore, the nature of inequality in South Africa extends beyond economic well-being (Leibbrandt et al., 2012; Tibesigwa and Visser, 2016) . We 3. Inequality in the social domain: Whereas access to basic education is high and fairly even across the country, access to health care is characterized by greater inequality. More than 80% of Black Africans use public health facilities and fewer than 20% use private health facilities. There are considerable differences in access to private medical care across provinces, with Limpopo reporting private coverage at less than 10%, whereas Gauteng and Western Cape reported rates of 25.0% and 24.8% in 2017, respectively. 4. Gender inequality: Women were less likely to participate in the formal labor market as compared to men and also experienced higher unemployment (29.6%) compared to men (25.7%) in 2017. Gender inequality is also observable in food security (Tibesigwa and Visser, 2016) . This multidimensional view of inequality is important to keep in mind when it comes to identifying tradeoffs associated with the country's response to COVID-19.

The dramatic reductions in population mobility documented above have come at a cost for many households, particularly those who are no longer able to work. Income from the labor market has been the main source of household income in South Africa, accounting for over 70% of overall income (Statistics South Africa, 2019). Fears of losing such income can reduce compliance with measures to mitigate the spread of COVID-19, especially in low income areas (Wright et al. 2020; Coetzee and Kagee, 2020) .

A web-based survey 18 conducted between April 29th -May 6th 2020 (Statistics South Africa, 2020b) paints a picture of the pandemic's impact on employment, income, and hunger, highlighting the potential for deepening inequality. While 60.2% of the respondents were employed on a permanent basis during the national lockdown, just under 2.0% lost their jobs and 5.2% had to close their businesses. Further, while 89.5% of those who were employed before the national lockdown remained employed during this period, 8.1% lost their jobs or had to close their businesses, 1.4% became unemployed and 0.5% were out of the labor force. For those who stayed employed during the lockdown, 21.3% indicated reduced income. Given the voluntary, web-based nature of the survey, these and other estimates are likely significant underestimates. 19

The survey also showed that while a majority (75.4%) of respondents who had businesses before national lockdown were white, among black Africans and the coloured population, the share among those who had to close businesses were larger than their share of business ownership (19.9% vs. 14.6% and 6.4% vs. 4.6% respectively).

While social grants and remittances have played a crucial role in reducing income inequality over the years in South Africa, the survey shows other coping strategies. For example, 74.9% of respondents reduced their spending to compensate for the loss of income, while about half of respondents had to access their savings to close the income gap. Some respondents (36.8%) relied on extended family members, friends and/or their communities for support, while 14.6% relied on claims from the Unemployment Insurance Fund. Disaggregated analysis of these claims

is not yet possible but would shed further light on the extent to which lockdown has furthered pre-existing inequalities.

As noted above, the more rural provinces and black Africans on average have tended to lag further behind in access to basic services (Statistics South Africa, 2019). The COVID-19 pandemic appears to have further deepened these inequalities. The survey shows that the majority of those able to work from home are in suburban areas (88%). In contrast, just 5.4% of township residents reported being able to work from home, followed by just 3.9% of those residing on farms and 0.9% of respondents in rural areas.

The survey also revealed considerable food insecurity and income losses. Since the start of lockdown, the proportion of respondents who reported experiencing hunger increased from 4.3% to 7.0%. The percentage of respondents who reported no income increased from 5.2% to 15.4% by the sixth week of lockdown. Again, these figures are likely significant underestimates.

Notably, the government of South Africa has provided a number of relief measures, including the release of disaster relief funds, emergency procurement, wage support through the Unemployment Insurance Fund, and funding to small businesses. On April 21, 2020, the President announced a massive social relief and economic support package of R500 billion ($30,50 billion), amounting to around 10% of GDP (RSA Government, 2020h) . This was complemented by The South African Reserve Bank easing monetary policy with reduced interest rates (SARB, 2020) and subsequent loan of about $4.5 billion from IMF (RSA Government, 2020i). These resources have been deployed to prepare health infrastructure, provide food and income support, and provide financial relief to businesses and individuals. The effectiveness and efficiency of such efforts are still yet to be determined.

This research note paints a picture of life under lockdown in South Africa, the world's most unequal nation. We present evidence of swift and effective action by the government -mirrored in substantial reductions in mobility among the population. People were particularly willing and able to act in the provinces hit hardest by the pandemic in its initial stages (Gauteng and Western Cape). At the same time, compliance with lockdown orders presents a greater challenge among rural populations and others with more precarious livelihoods. By reflecting on South Africa's inequality profile and results of a recent survey conducted during lockdown, we demonstrate how South Africa's response may deepen preexisting divides. This cautionary tale is relevant beyond the country's borders, as much of the continent --and the world --contemplates similar tradeoffs. Along with measures to contain the spread of disease, governments and other development focused organizations should seriously consider how to offset the costs faced by already marginalized populations.

Data from European Centre for Disease Prevention and Control (ECDC) in Roser et al. (2020) .

Data from Oxford COVID-19 Government Response Tracker . -0.14 *** -0.14 *** (0.03) (0.03) L14.Average weekly reduction in mobility (retail/recreation) -0.14 *** -0.12 *** 

We supplement our analysis of the Google COVID-19 Community Mobility Reports with Mobility Trends Reports published by Apple Maps (https://www.apple.com/covid19/mobility). These reports present data on the relative volume of directions requests per country/region, subregion or city compared to a baseline volume on January 13th, 2020. Higher proportions indicate smaller mobility reductions.

In addition to being available at the country and province level, these reports are available for the cities of Johannesburg and Cape Town, allowing us to compare trends in these cities to the greater provinces. Note that these reports are available for both driving and walking for the two cities and South Africa as a whole; and for driving only at the greater province level. Figure A1 depicts city/province comparisons for driving trends over the same period considered in the manuscript (February 14-May 15, 2020; see Figure 4 ).

We see that the city-level trends appear to mirror the province-level trends. In order to confirm this and relate to the analysis presented in the paper, we also calculate average mobility reductions for the most stringent lockdown period (27 March-30 April 2020). These are depicted in Table A2 below. As with our analysis of the Google Mobility Reports, Western Cape and Gauteng exhibit the largest mobility reduction (fewer directions requests compared to baseline, suggesting people are moving around less), and Limpopo and Mpumalanga province the smallest (a higher proportion of directions requests in comparison to the baseline, indicating less of a change in mobility).

We also see that the city-level mobility changes mirror the province level for Cape Town/Western Cape and Johannesburg/Gauteng, confirming the results suggested by the figures above. It is notable that in Cape Town, driving requests and walking requests are on par with each other whereas in Johannesburg the lockdown appeared to have a greater impact on people's driving habits compared to walking. This may be due to the fact that Cape Town and Western Cape are mostly tourist areas and thus when the lockdown instructions and impact of the disease set in, both driving and walking habits were similarly affected. On the other hand, the driving and walking populations of Johannesburg and Gauteng tend to be distinct. Most of the drivers are wealthier, while the walkers tend to be poorer. When the lockdown instructions set in, it was easier to enforce compliance with driving (using the existing traffic enforcement framework), while it was more difficult to enforce walking restrictions (due to the lack of a coherent enforcement framework and insufficient resources).

All figures in this section depict mobility trends from February 14-May 16, 2020, using data from Google Mobility Reports 

In order to conduct crisp qualitative comparative analysis (QCA) we begin by creating sets -dichotomizing the variables of interest to designate membership within a given set. The sets are defined as follows:

• M = large average mobility decrease (defined alternatively as provinces in the top quintile of the distribution and provinces in the top two quintiles) • I = 100 or more confirmed infections as at March 26, 2020 (start of lockdown) • D = densely populated provinces, i.e. those with an average of 100 people or more per km sq. • G = Provinces whose contribution to national GDP is greater than 10% • P = Provinces with a multidimensional poverty rate greater than 10% • A = Provinces with a proportion of agricultural households exceeding 25%

We have reproduced a crisp set version of the relevant data in Table A2 as a data matrix. We see that for either definition of M, the set also contains members of I and G --that is, provinces with at least 100 confirmed cases pre-lockdown and those that contribute significantly to national GDP. The dependent variable is the average weekly reduction in mobility to workplaces, retail and recreation, and transit stations. All models include province fixed effects and robust standard errors clustered by province. * p < 0.10, ** p < 0.05, *** p < 0.01 Figure A5 . Confirmed Cases by Province, March 1-May 24, 2020 (excluding Western Cape)

Pandemic politics: Timing state-level social distancing responses to COVID-19

Self-Isolation Compliance In The COVID-19 Era Influenced By Compensation: Findings From A Recent Survey In Israel: A cross sectional study of the adult population of Israel to assess public attitudes toward the COVID-19 outbreak and self-isolation

Global access to handwashing: implications for COVID-19 control in low-income countries

Structural barriers to adhering to health behaviours in the context of the COVID-19 crisis: Considerations for low-and middle

Strong Social Distancing Measures In The United States Reduced The COVID-19 Growth Rate: Study evaluates the impact of social distancing measures on the growth rate of confirmed COVID-19 cases across the United States

Shaking Stability: COVID-19 Impact on the Visegrad Group Countries' Financial Markets

Economic Freedom Fighters statement on the COVID-19 developments

Preparedness and vulnerability of African countries against importations of COVID-19: a modelling study

Political partisanship influences behavioral responses to governors' recommendations for Covid-19 prevention in the United States

Blavatnik School of Government. Data use policy: Creative Commons Attribution CC BY standard

State of the ICT Sector Report

#ListenToTheExperts Is South Africa's epidemic trajectory unusual? Professor Salim Abdool Karim

Democracy, Capacity, and Coercion in Pandemic Response-COVID 19 in Comparative Political Perspective

Describing and decomposing post-apartheid income inequality in South Africa

Modelling the potential impact of social distancing on the COVID-19 epidemic in South Africa

How health inequality affect responses to the COVID-19 pandemic in Sub-Saharan Africa

Tackling COVID-19: Can the African continent play the long game

Coronavirus Pandemic (COVID-19

Regulations issued in terms of Section 27(2) of the Disaster Management Act

Disaster Management Act 2002: Amendment of Regulations issues in terms of Section 27(2). (Government Gazette

Disaster Management Act 2002: Amendment of Regulations issues in terms of Section 27(2). (Government Gazette

Disaster Management Act 2002: Amendment of Regulations issues in terms of Section 27(2). (Government Gazette

Disaster Management Act 2002: Amendment of Regulations issues in terms of Section 27(2). (Government Gazette

Disaster Management Act 2002: Amendment of Regulations issues in terms of Section 27(2). (Government Gazette

President Cyril Ramaphosa: Additional Coronavirus COVID-19 economic and social relief measures

Social distancing to slow the US COVID-19 epidemic: an interrupted time-series analysis

South African Reserve Bank (SARB)

Inequality Trends in South Africa: A multidimensional diagnostic of inequality

Quarterly Labor Force Survey

Results from Wave 2 survey on the impact of the COVID-19 pandemic on employment and income in South Africa

Assessing gender inequality in food security among smallholder farm households in urban and rural South Africa

Discovery 20200305 First case of COVID-19 confirmed in KwaZulu Natal Province 20200307 Second case of COVID-19 confirmed in KwaZulu Natal Province 20200311 First case of COVID-19 confirmed in Western Cape Province 20200312 First case of local transmission confirmed in Free State Province Early Response 20200315 President announces measures to combat COVID-19 20200316 Government declares State of National Disaster 20200316 Ports of entry closed (Namibia

Testing capacity increased with 60 new mobile lab units launched 20200408 Critical personal protective equipment secured for frontline healthcare workers 20200409 Lockdown extended until end of April (with 7 days travel grace across provinces for relocation) 20200418 Government postpones May/June Exam rewrites 20200420 President announces interventions to address livelihoods of the vulnerable groups 20200421 President outlines expanded COVID-19 economic & social relief 20200421 Government recommends wearing of a cloth non-medical face-mask when in public 20200423 President announces risk-adjusted strategy to respond to COVID-19 pandemic

Oxford COVID-19 Government Response Tracker, Blavatnik School of Government

Census 2011 Agricultural households Key Highlights

Mid-Year Population Estimates

Estimate as of August 10, 2020 from the European Centre for Disease Prevention and Control (ECDC) cited in Roser et al

According to the World Bank's World Development Indicators, South Africa's Gini Index is 63/100, the highest in the world according to available data

The paper that bears the most similarity to ours is Nyabadza et al. (2020), which models the impact of social distancing on the transmission dynamics of COVID-19 in South Africa

These included travel restrictions, encouraging social distancing, limiting contact between persons who may be infected, and working to strengthen the public health response

The prohibitions included the sale, dispensing, or transportation of alcohol

The government has devised five Coronavirus Alert Levels, in line with a risk-adjusted strategy that seeks to slow down the rate of infection and flatten the curve. Level 5 entails "drastic measures are required to contain the spread of the virus to save lives

Further details about these reports and other data sources analyzed in this research not can be found in Table A1 in the Appendix

According to the World Bank's World Development Indicators, the country had 159.9 mobile cellular subscriptions per 100 people

We are very grateful to Clara Tomé for excellent research assistance and to the organizers of the 2019 Sustainability and Development conference, which led to the authors' meeting and inspired our collaboration on this article.

1. We analyze subnational variation in population mobility as a response to COVID-19 in South Africa.2. We leverage anonymized mobile phone data to capture mobility reductions across provinces.3. People tend to reduce mobility substantially in response to government's initial lockdown orders.4. Mobility reductions are significantly and negatively associated with COVID-19 growth rates. 5. We illustrate how the government's response and corresponding mobility reductions can exacerbate existing inequalities. 10 South Africa is labeled by its 3-letter country code, "ZAF." 11 The province is the lowest level at which comparable data on mobility trends is available. We supplemented our analysis with data from the Mobility Trends Reports published by Apple Maps (https://www.apple.com/covid19/mobility), which allow us to compare trends for Cape Town and Johannesburg with the surrounding provinces. As shown in the Appendix, this analysis depicts largely similar trends to those captured by the Google Reports.

12 Additional province-level comparison charts are shown in the Appendix. 13 This process is described in detail in the Appendix.14 An alternative version of this figure, excluding Western Cape, is presented in the Appendix for better visualization of the other provinces. 15 We first calculate the daily exponential growth rate and then take the weekly average. Daily exponential growth is calculated as the natural log of cumulative confirmed cases minus the log of cumulative confirmed cases on the prior day. As in other recent studies ), we use this functional form because epidemiological models predict exponential growth in the absence of intervention. In computing exponential growth, we follow recent studies and add one for province-dates with zero cases to avoid dropping observations. 16 We take weekly averages given fluctuations in mobility, e.g. in workplace trends where the average reduction in mobility is considerably lower on weekends when people are typically less likely to go to their workplaces. 17 This section draws heavily on a recent report profiling trends since 1994 conducted by the country's statistical agency (Statistics South Africa, 2019). 18 The survey drew on a non-probability, convenience sample of 2,688 South African residents. 19 The most recent Afrobarometer survey, conducted between August-September 2018, indicates that nearly 50% of all rural residents in South Africa never use the Internet compared to 29.5% of urban residents (http://afrobarometer.org/online-data-analysis/analyse-online).