Policy Research Working Paper 10223 Firm Entry, Exit and Suspension Evidence from Household Businesses in Vietnam Ergys Islamaj Duong Trung Le Thanh Minh Pham East Asia and the Pacific Region Office of the Chief Economist November 2022 Policy Research Working Paper 10223 Abstract Household businesses make up the majority of firms in exceeding 2.5 months. The suspension rate spiked to 40 developing economies. This paper uses a novel tax census percent during the onset of the COVID-19 pandemic in database that covers the universe of tax-registered household 2020. Third, the findings show that the pandemic-related businesses to analyze the entry and exit of owner-operated effects were more pronounced for businesses dependent firms in Vietnam during January 2018 to August 2020. It on face-to-face interactions with customers and suppliers. documents new stylized facts about the survival dynamics of However, these effects were short lived, and activity and informal businesses. First, the entry and exit rates were about earnings rebounded by August 2020. The findings may 5–6 percent a year for tax-registered household businesses reflect the relatively short COVID-19 distress in Vietnam during the pre-pandemic period. Second, an additional 25 during the first phase of the pandemic, but they illumi- percent of household businesses suspended their activity nate both the vulnerabilities and resilience of the household in a year on average, with the annual suspension duration business sector. This paper is a product of the Office of the Chief Economist, East Asia and the Pacific Region. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/prwp. The authors may be contacted at dle6@worldbank.org and eislamaj@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Firm Entry, Exit and Suspension: Evidence from Household Businesses in Vietnam ∗ Ergys Islamaj † Duong Trung Le ‡ Thanh Minh Pham § Keywords : informal sector, household business, entry, exit, suspension, COVID-19, Vietnam JEL Classifications : O12, O17, D22 ∗ We thank Aaditya Mattoo, Francesca de Nicola, Ivan Korolev, Solomon Polachek, David Slichter, Jonathan Timmis, and the Labor Group Meeting participants at Binghamton University for helpful comments and suggestions. The findings, interpretations, and conclusions are entirely those of the authors. They do not necessarily represent the views of the World Bank Group, its Executive Directors, or the governments they represent. † The World Bank. East Asia and Pacific Chief Economist Office. Email: eislamaj@worldbank.org. ‡ Corresponding author. The World Bank. East Asia and Pacific Chief Economist Office. Email: dle6@worldbank.org. § Mekong Development Research Institute. Email: thanhpham@mdri.org.vn. 1 Introduction Household businesses comprise the majority of firms in developing countries (Hsieh and Olken, 2014; La Porta and Shleifer, 2014). 1 They employ a large share of the workforce, and matter significantly to the livelihoods of households in low-income countries (Gollin, 2002, 2008; Nataraj, 2011; McCaig and Pavcnik, 2021). Their performance and dynamics are therefore key to understanding the livelihoods of poor households, generation of jobs, and aggregate labor productivity in low-income countries. In this paper, we document novel stylized facts about entry and exit of household businesses in a developing country context. We utilize a unique tax census data disclosed by the General Department of Taxation that covers the universe of tax-registered household-owned businesses in Vietnam between January 2018 and August 2020, a period spanning the COVID-19 pandemic. The data include monthly information on (permanent) closure filings and (temporary) suspensions of household businesses across all communes in the country. 2 We find that tax-registered household businesses exhibit rates of entry and exit of about 5-6 percent a year. Existing estimates in the literature of the rate at which small firms die (exit) in developing countries range from as low as 3 percent per year (Frazer, 2005; Davies and Kerr, 2018) to over 30 percent annually (Fajnzylber et al., 2006; e, 2017). McKenzie and Paffhausen (2019) find that the rate of firm exit in developing economies Nagler and Naud´ averages 8.3 percent per year. McCaig and Pavcnik (2021) document an entry and exit rate of 14-18 percent for informal sector firms in Vietnam. These estimates are higher than the ones documented in this paper and the difference is likely due to by the nature of the data, which includes household businesses whose larger revenue scales subject them to register and pay annual taxes by law, and are thus less prone to turnover. 3 Our unique dataset allows us to document that an additional 25 percent of household businesses suspend their activity for part of the year, with the average annual suspension duration exceeding 2.5 months. The informal economy in developing economies has been able to weather shocks and even expand employment during past economic downturns (Loayza and Rigolini, 2011). However, informal business, many of which operate in the services sector, were hit hard during the COVID-19 shock in 2020, and the suspension rate spiked to 40 percent of registered household businesses. We analyze the effects of the initial outbreak of COVID-19 on household businesses in Vietnam utilizing two high-frequency data sources: the tax census data, which provides information on the extensive margin of household 1 Household business refers to household-owned, or family run businesses, and it is a good approximation of informality in developing economies. (Loayza and Rigolini, 2006). 2 Commune is a third-tier administrative unit in Vietnam, subordinated to district (second-tier) and province (first-tier). There are over 11,000 communes in the country. 3 In Vietnam, a household business must register and are responsible to file and pay one annual lump-sum tax within a fiscal year if their annual revenue exceeds 100 million Vietnam Dong (VND), or approximately USD4,300. According to the 2017 Vietnam Enterprise Census (VEC), registered household businesses comprise about one-fifth of total informal businesses in Vietnam. Further discussion on household business characteristics is presented in Section 2. 1 business performance, and the Vietnam Labor Force Survey (LFS), which provides data on the intensive margin of household business performance. Measures to curb the spread of the virus, such as mobility restriction, business closures, and border shutdowns, weighed heavily on businesses in developing economies. Many businesses were forced to close, and many did so permanently due to their inability to sustain prolonged shutdown-related losses.Household businesses, were highly vulnerable to the pandemic-related shocks given their limited resources, and likely suffered disproportionately. We first estimate the impact of the pandemic on household businesses using an event study approach. Overall, we find substantial reduction across all household business performance indicators, including business suspension, business closure, work hours, and earnings. During April and May 2020—the period immediately following the imposition of a national lockdown, suspension and closure filings surged over three times pre-COVID-19 levels, and weekly work hours and monthly earnings declined by more than a quarter of their pre-pandemic levels. However, household business activity rebounded relatively quickly; both suspensions and closures reduced by over 50 percent in June, while work hours and earnings had made steady recovery by September. We then show that the impact of the pandemic on household businesses is uneven across industry sectors and firm size. Consistent with our expectation, firms operating in transportation, accommodation, food, and entertainment industries, the sectors most vulnerable to social-distancing orders, witnessed the greatest reduction in business activities. Further, firms with larger revenue—i.e., those with the highest opportunity cost of staying open during lockdowns—were more likely to close permanently. Further, we find that losses were larger in areas dominated by businesses relying on face-to-face interactions. We follow Avdiu and Nayyar (2020) and construct a commune-level face-to-face (F2F) index to highlight geographical differences across communes in Vietnam. F2F is a continuous measure that captures the average intensity of human-to-human interaction required to conduct business in a commune based on its existing sectoral composition of household businesses. An array of empirical estimates find that household businesses in communes with a higher F2F index suffered higher suspension rates, higher closure rates, lower weekly working hours, and lower monthly incomes. This paper contributes to the literature in several ways. First, our analysis adds to a literature that studies the performance of informal businesses in developing countries by using high-frequency, tax-registration information covering all tax-registered Vietnamese household businesses. Previous work has relied on information from national household and labor force surveys. See, for example, Maloney (2004); Jayachandran (2020); Banerjee (2013); Banerjee et al. (2015); de Mel et al. (2019); Bruhn and McKenzie (2014); McKenzie and Woodruff (2013); McKenzie and Paffhausen (2019); Li and Rama (2015). Specifically related to the business cycle dynamics of informal firms, McKenzie and Paffhausen (2019) examine the characteristics of, and motivations behind, small firm exit, using a 2 sample of 14,000 small businesses in 12 developing countries. McCaig and Pavcnik (2021) use multiple rounds of the Vietnamese nationally representative household surveys–conducted once every two years—to examine the entry and exit dynamics of family-owned firms over a long horizon. The tax-census data allow us to document exit and entry of household businesses at a monthly interval, better capturing business cycle fluctuations. Second, our study is among the first to document the extent of (temporary) suspensions of household businesses in a developing country context. Earlier literature has documented a significant dormancy rate of registered enterprises in developing countries. For example, Sengupta and Singh (2019) show that about 20 percent of all registered Indian firms are dormant. High exit barriers associated with governments’ costly and uncertain legal procedures—such as those related to labor retrenchment and bankruptcy—are often the major reasons behind formal enterprises’ prolonged inactive status. Household businesses in Vietnam, however, are not subject to significant exit barriers; most owner-operated firms do not employ any contracted workers, and the business closure does not usually involve lengthy legal procedures. 4 Their motivation to suspend operation, therefore, could be considered more purely as a copying mechanism to disruptions in business activities, such as those caused by the COVID-19 pandemic lockdowns. Third, our analysis contributes to the literature that examines the impacts of COVID-19 on small and/or informal businesses, especially during the onset of the pandemic in early 2020. Several studies have explored the initial COVID-19 effects on small businesses in developed economies such as the United States (Fairlie, 2020; Crane et al., 2021), Canada (Beland et al., 2020), and Western European countries Kraus et al. (2020). In a developing- country setting, Shafi et al. (2020) investigate the pandemic effects on micro, small, and medium-sized enterprises in Pakistan and find that most firms were severely affected and faced with financial and supply chain disruptions, and decrease in demand. Several other studies have found adverse effects of COVID-19 on self-employment outcomes (for instance, (Dhingra and Machin, 2020) for India, (Mahmud and Riley, 2021) for Uganda, (Dang and Nguyen, 2020) for Vietnam, and (Mathew et al., 2020) for Zambia). Our paper adds to these studies and, to the best of our knowledge, is the first to document the effects of COVID-19 on household business suspension and exit in a developing country. 5 Lastly, this paper contributes to a strand of literature that studies the cyclical behavior of self-employment. Previous work has found that self-employment expands during downturns, led by an increase in transitions from unemployment to self-employment during recessions (Bosch and Maloney, 2008; Loayza and Rigolini, 2011; Shapiro, 2014, 2018). The nature of the COVID-19 recession, which was accompanied by reductions in demand and restric- tions in mobility, have affected the self-employed disproportionally. Nevertheless, we find that household businesses 4 According to Decree 78/2015/ND-CP, a Vietnamese household business wanting to dissolve its business needs to submit (i) an An- nouncement of Business Closure Letter and a (ii) Certification of Non-tax Obligations Letter (obtained from the local Tax Department), to the District’s People Committee where the firm is operating. 5 Crane et al. (2021) show the business exit of firms during COVID-19 in the United States. 3 in Vietnam recovered quickly. The rest of the paper proceeds as follows. Section 2 introduces the background of household businesses in Vietnam and the main data sets employed in this paper. Section 3 presents our methodology to estimate the impacts of COVID-19 on household business survival. Section 4 discusses the results and Section 5 concludes. 2 Household Businesses in Vietnam In Vietnam, a household business (HB) is a firm run by a household or family that employs fewer than ten workers and operates in a fixed location (usually at their own residential premises). All HBs must register with the Department of Taxation, which operates under the Ministry of Finance, for a tax identification (tax ID) and pay taxes if their annual revenue exceeds 100 million Vietnam dong (VND), or approximately 4,300 USD. 6 Household businesses do not face stringent financial oversight like enterprises which operate at a larger scale and are required to house an accountant as well as to prepare official accounting documents. Its annual lump-sum tax payment, usually remitted at the beginning of a fiscal year in January, is often calculated as a percentage of annual revenue, and comprises annual license tax, value added tax (VAT), and personal income tax (PIT). 7 2.1 Household business tax data We analyze a high-frequency tax census that comprises all tax-registered household businesses in Vietnam since January 2018. We retrieved historical tax filing records of household businesses from the official online portal of the Department of Taxation. 8 Pending reporting delays, this information from the Tax Department allows us to tabulate a complete set of tax filings in the country at the commune-month level, between January 2018 and August 2020. 9 Each household business is responsible to file and pay one annual lump-sum tax within a fiscal year. We use the (change in) tax filing numbers as a proxy for the (changes in) number of tax-registered household businesses each month. For brevity, we refer to this data set as the “household business tax census” (“HBTC”). Beyond the regular annual filings reported in January each year, the tax portal also keeps separate retrievable records of several types of irregular filings. We focus on two such indicators directly related to firm exit: the number of tax fillings for (i) permanent business closure, and (ii) temporary suspension. While the former provides a measure of firm survival 6 See Article 1, Decree 92/2015/ND-CP. 7 Thelicense tax is a fixed proportion based on revenue brackets. VAT and PIT rates are set by the central government and vary by revenue and industry. See Le et al. (2020) for further details on household business tax information. 8 The link to the online portal is http://www.gdt.gov.vn/wps/portal/home/hct (accessed January 2022). The records can be retrieved by querying an administrative location up to the commune level (a third-tier administrative unit). 9 While several provincial tax offices have published household businesses tax records until the end of 2021, there are currently missing records for many provinces after August 2020. To our knowledge, each provincial tax office is responsible for updating the database with monthly tax filings in their province. This results in a heterogeneous lagged reporting times across provinces. 4 in line with most existing studies on informal firms’ entry and exit, the latter is a novel and important indicator for studying how household businesses cope with economic shocks in the short term. To examine sectoral impacts, we also rely on data on operating industry and location of the business. 10 2.2 Entry and exit By the nature of the dataset, this analysis focuses only on tax-registered household businesses. While we do not observe HBs that are not tax-registered, data from the latest Vietnam Enterprise Census (VEC) that is collected every five years, the last time in 2017, shows that there are approximately 5 million HBs operating in the country. 11 Table 1 shows the total count of tax filings retrieved from the DoT portal for the January 2018-August 2020 period. The first three columns show records for regular annual filings. Column 1 reports the total number of (once-a-year) annual lump-sum filings, the majority of which usually take place each year in January (approximately 90 percent of records). The number of registered HBs increased slightly from over 908,679 firms in 2018 to 917,917 in 2019. The total registered HBs was on the rise in 2020, with 907,472 filings having been recorded as of August (note that since most filings happen in January including for the year 2020, the existing statistics mostly capture pre-COVID-19 tax filings). Overall, these numbers represent approximately one-fifth of the total household businesses operating in the country. 12 It is, therefore, noted that the majority of household businesses in the country are not tax-registered. 13 In each year, we distinguish between incumbents and new businesses using available identification information (tax ID). “Repeated” (column 2) is the number of filing records in a given year that also appeared in the previous year, and “New” (column 3) is the number of filing records that appeared in the dataset for the first time in the current year. The data suggest entry and exit rates of household businesses of about 5-6 percent a year during 2018-2019. 14 The category of “irregular filing”, housed in separate databases from the regular annual filing database, is of particular interest. “Closure filings” defines the number of firms that filed to (permanently) close their business. “Suspension filings” defines the number of firms that filed to (temporarily) suspend the business in at least one month of a given year. 15 While the former provides a measure of exit in line with earlier studies, we believe the latter is a new indicator that captures how household businesses cope with idiosyncratic or economy-wide shocks 10 The data on operating industry is only available for 2018 and 2019. 11 The VEC is collected by the General Statistical Office of Vietnam (GSO) once every five years, the last round in 2017. The survey includes data on both formal enterprises and household businesses operating in the country (Le et al., 2020). 12 This number also aligns with data from the 2017 VEC, which shows that there are over 850,000 HBs reported in the census that possess a tax ID, approximately 135,000 of which located in Ho Chi Minh city and 71,000 in Ha Noi. 13 See Le et al. (2020) for a discussion of pull and push factors related to HB registration incentives. 14 Certain HBs might have closed the business without officially filing for business closure. The repeated and new filings are subject to errors due to provincial tax offices’ tax-ID reporting timeline and accuracy. In provinces where we do not obtain complete tax-ID information of HBs, we approximate the incumbency/entry rates using the corresponding rates observed in the previous year. 15 Note that Table 1 reports the total number of firms filing for suspension for at least one month in a year. The corresponding total number of suspension filings in each year (i.e., inclusive of multiple filings per firm-year) are 655,244; 570,933; and 776,934 in 2018, 2019, and January-to-August 2020, respectively. 5 in the short-term. The suspension rate increased sharply from 25 percent in 2019 to over 40 percent in 2020. The large difference between exit and suspension rates warrants the need to empirically explore the causal impact of the pandemic and government lockdowns separately on permanent closures and temporary suspensions among HBs. We do so in the next section. Table A.1 provides detailed summary statistics of the main HBTC analysis sample (commune-month level) (Panel A). There were, on average, 6.7 temporary business suspension filings and 0.5 permanent closure filings in a commune per month during January 2018-January 2020 (pre-COVID-19 period). For the statistics related to suspension filings, note that here we count all filings observed in the data, which captures both singular and multiple filings per firm-year, instead of the annual aggregation of the number of firms filed for suspension at least once a year reported in Table 1. By extension, the average dormancy duration of firms that filed for suspension was 2.8 months in 2018 and 2.5 months in 2019, respectively. Compared with the average number of regular annual filings per commune of 92.5 in January 2019 (s.d. is 169.45), the business suspension rate equates over 7 percent of all HBs each month, while the closure rate is over half a percent, as confirmed in Table 1. There are strong heterogeneities in both business closure and suspension rates across sectors and firm size. On average, HBs exiting from the wholesale and retails industry accounts for more than half of the total business suspension and closure filings. The data also suggest a decline in the suspension rate as operating revenue grows (the smallest firms by revenue are most likely to file for business suspension). But the largest firms by the revenue quartile (above 75th percentile) are the most likely to file for business closure. 2.3 The Labor Force Survey The Labor Force Survey (LFS) is a nationally representative survey that is conducted on a quarterly basis by the General Statistical Office of Vietnam. The survey collects detailed employment information on a representative sample of workers in the country. In this paper, we utilize monthly data collected from three LFS surveys, covering the period between January 2018 and September 2020. Our sample is truncated to include only working-age indi- viduals (15 years or older), who were identified as either being self-employed or working in a household business. 16 For abbreviation, we refer to this group uniformly as household business workers, or “HB workers”. We focus on two indicators that capture the intensive margins of COVID-19 impact: worker’s reported working hours during the week prior to the survey, and worker’s reported income during the month prior to the survey. To correct for idiosyncratic interference and measurement biases, we exclude individuals who reported to have reduced workload due to sickness, natural disaster, or holidays. Our analysis sample includes 419,323 individuals interviewed in 2018, 2019 and the first three quarters in 2020. 16 “Self-employment” and “working in a household business” belongs to occupational categories 2 and 3 in the LFS. 6 The descriptive statistics of the main LFS outcome variables are presented in Table A.1, Panel B. All statistics are presented as the means of the pre-COVID-19 period (i.e., all months in 2018 and 2019). On average, an HB individual worked for over 47 hours per week, earning approximately 5.8 million VND per month (approximately 250 USD). While average workload is mostly homogeneous across industries and firm sizes, average monthly income varies, between 5.3 million VND in the accommodation, food services, and entertainment sector (lowest) to over 7.2 million VND in transportation. 2.4 Face-to-Face Index The Face-to-Face Index measure (F2F) measures the importance and intensity of F2F interactions with other people for more than 900 occupations based on 4-digit International Standard of Industrial Classification (ISIC) codes (Avdiu and Nayyar, 2020). The measure utilizes the Occupational Information Network (O*Net) database for the United States, which provides scores on the importance of four work activities: (1) establishing and maintaining personal relationships; (2) assisting and caring for others; (3) performing for or working directly with the public; and (4) selling to or influencing others. 17 Table A.2 provides illustrative examples of F2F index scores and their closest 4-digit industry matches that appear in our data. Generally, manufacturing and wholesaling activities receive smaller F2F scores relative to accommodation or retails, reflecting the lower face-to-face intensity required to conduct business. We calculate a commune-level baseline F2F measure by taking the average 2019 F2F index of all household businesses in a commune. Figure A.1 provides a visualization of the geographical distribution of the F2F mea- sure. Visually, there is significant variation in the intensity distribution of the index, or equivalently, significant geographical variation in business lockdown exposure. 18 3 Estimation Strategy To identify the impact of COVID-19 on Vietnamese household businesses, we estimate the following equation: yit = α0 + α1 P ost + Xi Θ + ci + mt + it (1) 17 Following Oldenski (2012), the score from each of the four work activities is aggregated for each firm in the O*Net database. The arithmetic average of the scores of all firms in each occupation then denotes the F2F index for that occupation. 18 In Figure A.1, we aggregate our F2F indices to the district level for visualization purposes. There are 670 districts in the country, which comprises of over 11,000 communes. 7 where yit is the household business outcome of interest recorded for month t. In the HBTC regressions, yit represents commune i’s total suspension (temporary) and closure (permanent) filings. P ost is a dummy variable indicating the time period after January 2020, the month the first COVID-19 case was detected in Vietnam. In the LFS regressions, yit represents individual i’s workload (weekly work hours) or monthly income. We use monthly income residuals as the main dependent variable, which is computed as the difference between the reported absolute income and the predicted income level after accounting for pre-COVID linear trend and individual observables, in- cluding age, gender, education level, working industry, and occupational group. The temporal trend and the linear predictive model are presented in Figure A.2 and Table A.3. 19 For LFS regressions, the model further controls for a covariate vector Xi which represents a set of individual characteristics including age, gender, education level, occupation level (ISIC-08 major groups), and work industry. ci and mt are the commune fixed effects and month fixed effects, respectively. In Eq.1, coefficient α1 captures the average changes in household business outcomes for the period between February and September 2020 relative to before. We also study how the pandemic affected household businesses over time by estimating a month-by-month event study model. Specifically, 8 yit = β1 + {β2 }[monthD ummies]t + Xi Θ + ci + mt + it (2) t=1 where all elements remain the same as in Eq.1, except for the newly-introduced summation term which represents month-specific changes in household business outcomes. This equation allows us to capture COVID-19 effects for each month post-COVID-19 in 2020 (February (t = 1) to September (t = 8)). We also control for differences in the intensity of face-to-face interactions at the commune-level, i.e., the F2F index. The F2F index is interacted with time fixed-effects under a difference-in-differences (DiD) framework to estimate the differential effect of COVID-19 across communes. The DiD model to estimate post-lockdown aggregate impact is as follows: yit = α0 + α1 · F 2Fi · P ost + Xi Θ + ci + mt + it (3) Likewise, the corresponding model to estimate impact observed in each month is 19 All of our results are robust to using the raw reported income from LFS. 8 8 yit = β1 + {β2 } · F 2Fi · [monthDummies]t + Xi Θ + ci + mt + it (4) t=1 4 Empirical Results 4.1 COVID-19 and economic activity Vietnam was relatively successful in controlling the infections early on during the COVID-19 pandemic. It reported only a little over 300 cases and zero COVID-19-related deaths by the end of July 2020, despite sharing a border and trading extensively with China. 20 A subsequent and more severe surge in infections started at the end of July 2020 and was largely contained by September 2020. At the end of October 2020, Vietnam registered a total of 1,200 cases and 35 deaths. Both infection and mortality rates were kept constantly low in the country relative to the rest of the world, until at least the Delta-variant-fueled surge in June 2021. The initial success in curbing the spread of COVID-19 was largely attributed to early and decisive containment measures such as border closures, strict quarantine rules, well-coordinated testing-tracing-isolating strategies, and the nationwide lockdowns imposed (Islamaj et al., 2021). Figure 1 shows monthly-average plots of four commune-level household business outcomes: number of temporary suspension and closure filings (panels A and B), individual work hours per week (Panel C), and individual monthly income (Panel D). Suspension and closure filings surged immediately after the first cases of COVID-19 infections, peaking right after the government ordered a nationwide lockdown. Suspensions and closures per commune per month increased from 6.7 and 0.5, respectively, before the pandemic to over 37 suspensions in April 2020 and 2.2 closures in May 2020, or by more than 400 percent. Likewise, both household business work hours and income plummeted. Work hours dropped from the pre-pandemic mean of 47 hours/week to 35 hours/week by April 2020, whereas monthly average income dropped by over a quarter, from 5.8 million VND to 4.5 million VND, by May 2020. 4.2 Impact of COVID-19: event study analysis We estimate 1 and 2 to analyze the impact of the pandemic on household businesses. The results are presented in Table 2. Consistent with graphical evidence in Figure 1, the estimates in Panel A indicate statistically significant increases in the filings for suspension (column 1) and closure (column 2), and reductions in the work hours (column 1) and income (column 2) of household businesses in the period since the first corona-virus cases were discovered in 20 Vietnam documented the first COVID-19 case as early as January 23, 2020. 9 the country in January 2020. Panel B further shows the temporal breakdown of the event-study estimates, for the months between February and September. Three key findings emerge. First, the greatest detrimental effects took place in April and May, the period during and immediately following the national lockdowns. 21 On average, the average business suspensions in a commune surged to 30 firms in April 2020 (over four times the pre-pandemic mean), while the average number of business closures was 1.6 firms (3.2 times the pre-pandemic mean). Further, household business workers reported a 13 hour loss in weekly work hours and 1.9 million VND in monthly earnings in April. 22 Second, even though of lesser magnitude, the significant labor-market impacts of the pandemic were observed as soon as the first cases were reported in the country, i.e., from February onward. This evidence coincides with the gradual increase in February and March in the stringency of government lockdown index (Figure A.3) and the respective mobility reduction reported by the Google Mobility Index (Figure A.4). 23 Finally, despite the sharp decline in business activities and earnings in April and May 2020, household businesses rebounded quickly. Both the average numbers of suspensions and closures reduced by over 50 percent in June 2020 and remained stable in the subsequent months. Similarly, HB work hours had made almost a full recovery by September 2020. Nevertheless, income stayed stagnantly below baseline, with the average monthly earnings of a household business worker 13 percent less than the average predicted level. Overall, while the dramatic impacts observed almost immediately after the economy-wide lockdown suggest the financial fragility of many household businesses, 24 the observed rapid recovery indicates the resilience of this sector. More broadly, the result also speaks to the success of the Vietnamese government’s containment strategy in introducing early and effective domestic and international restrictions to suppress infections. By being able to contain the disease, the country was able to transition quickly to the reopening phase. 4.3 Heterogeneities by sector and business revenue size We next analyze the impact of the pandemic on household business activities and earnings across sectors and firm sizes. Table 3 shows that household businesses operating in the services industries were the most affected from the pandemic. These results are consistent with sector-specific quarterly output growth in Vietnam in 2020 A.5. The most detrimental impacts are observed in Transportation and Accommodation, Restaurants, and Entertainment (ARE) (panel A). There was an average 164-percent increase in the number of ARE firms filling for suspension, and 108 percent increase in business closure (percentage-change statistics shown in square brackets). The average work 21 The government imposed a national lockdown on April 1, 2020. The lockdown was lifted after 22 days, on April 22, 2020. There were scattered targeted lockdowns imposed in different parts of the country since April 2020 in response to local outbreaks. 22 Monthly income (column 2) is reported with a one month lag. For income, the LFS asks respondents for income earned during the last month. Thus, the “May X Post Covid” estimate reflects the income effect of the pandemic in April 2020. 23 Data on the daily government lockdown stringency index (Figure A.3) was obtained from the Oxford COVID-19 Government Response Tracker (OxCGRT). Data on mobility (Figure A.4) is obtained from the Google’s COVID-19 Community Mobility Reports. 24 This evidence is consistent with the finding in Bartik et al. (2020) for small businesses in the United States. 10 hours dropped by over 13 percent and incomes by 21 percent from pre-COVID-19 levels for household businesses operating in Transportation. Further, firms with larger revenue—those that would suffer the highest opportunity cost of staying open during the lockdown—were the most likely to go out of business (panel B). Appendix Tables A.4, A.5, and A.6 provide supplementary results on the month-by-month distributional impacts by industry and revenue. We observe a consistent temporal pattern across these disaggregated results; on average, the negative consequences of the pandemic were felt as early as February, the greatest damages came in April and May, but firms across the distribution rebounded rapidly after infections were curbed. 4.4 Difference-in-differences analysis A difference-in-differences estimation framework (DiD) complements the event-study estimates presented earlier. The DiD estimator utilizes a commune-level F2F index to measure treatment intensity, i.e., the degree of household business exposure to the pandemic and government policies. In Eq. 3 and 4, the interaction between the temporal indicator and our constructed F2F index allows us to estimate the differential impact on firms operating in location where the baseline average business practice requires greater human-to-human interaction. Table 4 presents the DiD estimation result. Overall, the estimates are consistent to those from the event-study estimates. For HBTC outcomes (columns 1 to 4), communes with higher F2F index, i.e., those endowed with a greater proportion of businesses operating in industries that require more human-to-human interaction at baseline, experienced significantly higher number of suspensions and closures (a standard deviation increase in F2F Index is associated with an additional of 2.831 temporary suspension filings and 0.304 permanent closure filings per month in the commune.) For the LFS outcomes (columns 5 to 8), HB workers who operated in more-exposed community— proxied by the location’s baseline F2F index—experienced significantly greater aggregate reduction in workload and earnings. Specifically, for individuals residing in a commune, a standard deviation increase in the commune’s F2F index score leads to, on average, a reduction of 0.86 weekly working hours, 83,000 VND in monthly income residuals. The month-to-month result presented in Panel B continues to offer a similar pattern of effects on business activities. Across both the extensive (suspensions and closures) and intensive (total work hours) margins, we find significant effects starting two months prior to the de-jure imposition of mobility restrictions in April 2020. At the peak of the period in April and May following the national lockdown, an additional standard deviation increase in the commune’s F2F index is associated with as much as 9.7 more suspensions and 1.7 more closures per month. However, in contrast to the event-study result, the post-lockdown effect seem to be significantly less persistent when estimated with our DiD model. We find sharper declines of the labor-market damages starting in June, with the monthly workload and suspension estimates being no longer statistically significant in July. Similarly, the DiD 11 coefficients on earnings are mostly indistinguishable from zero except for the month of May, further suggesting a swift recovery of household business performance. 25 Indeed, this clear reversion to the pre-COVID-19 level speaks to the resilience of the informal businesses in Vietnam. 5 Conclusion In this paper, we document novel stylized facts about entry and exit of household businesses in Vietnam between January 2018 and August 2020, a period covering the onset of the COVID-19 pandemic. The data include monthly information on a universe of household businesses’ (permanent) closure and (temporary) suspension filings. We find that tax-registered household businesses exhibit rates of entry and exit of about 5-6 percent a year. Our unique data set allows us to also document that an additional 25 percent of household businesses suspend their activity during part of a year. In 2020, the suspension rate spiked to 40 percent of registered household businesses. We combine this high-frequency statistics of household-business closures and suspension with individual working hours and earnings data from the Vietnamese Labor Force Survey to measure the pandemic impacts on the survival of informal businesses—those accounting for over 30 percent of the labor force. Our two econometric methods, an event-study and a difference-in-differences strategy that employs a sectoral face-to-face intensity index as a proxy for businesses’ lockdown exposure, indicate significant damages to both business activities and earnings as soon as the first COVID-19 cases were reported in the country in February 2020. The greatest damage took place in April and May, following the mandatory national lockdown. Suspension and closure fillings surged by over three times the pre-COVID-19 level in April, and workload and income plummeted by more than a quarter. Across sectors, the greatest impacts were observed in accommodation, restaurant, entertainment, and transportation. Firms with a larger baseline in terms of revenue were more likely to shutdown. Despite enduring such a dramatic impact on business activities and earnings in April and May, household businesses recovered quickly. We find an immediate rebound across both intensive and extensive margins, with the number of suspensions and closures reduced by over 50 percent as soon as June, and HB work hours and earnings making almost a full recovery by September. Overall, while the dramatic impacts observed almost immediately after the economy-wide lockdown suggest the financial struggle faced by many household businesses in the country, the observed rapid recovery indicates the resilience of this sector during the pandemic. More broadly, the paper introduces a unique high-frequency dataset of tax registration and filings, which provides comprehensive and novel measures for different household business’ operating decisions, namely entry, exit, and 25 To check for the robustness of our DiD findings, we estimate the same regressions using an alternative commune-level F2F index set that is constructed using the 2018 sectoral composition of HBs (instead of 2019). The results, presented in Table A.7, are robust and consistent to our main findings in Table 4. 12 suspension. By doing so, it lays a foundation for future research that seeks to better understand informal sector activities and performance via high-frequency tax reporting information. In particular, future work should try to better understand both the extent and the factors that lead small firms, and especially household-owned firms, in developing countries to suspend their businesses when faced with shocks that could be idiosyncratic or aggregate. Understanding the performance and dynamics of small firms is key to improving the livelihoods of poor households, generating jobs, and boosting labor productivity in low-income countries. References Avdiu, Besart and Gaurav Nayyar, “When face-to-face interactions become an occupational hazard: Jobs in the time of COVID-19,” Economics Letters, 2020, 197, 109648. Banerjee, Abhijit, Dean Karlan, and Jonathan Zinman, “Six Randomized Evaluations of Microcredit: Introduction and Further Steps,” American Economic Journal: Applied Economics, January 2015, 7 (1), 1–21. Banerjee, Abhijit Vinayak, “Microcredit Under the Microscope: What Have We Learned in the Past Two Decades, and What Do We Need to Know?,” Annual Review of Economics, 2013, 5 (1), 487–519. Bartik, Alexander W, Marianne Bertrand, Zoe Cullen, Edward L Glaeser, Michael Luca, and Christopher Stanton, “The impact of COVID-19 on small business outcomes and expectations,” Proceedings of the National Academy of Sciences, 2020, 117 (30), 17656–17666. Beland, Louis-Philippe, Oluwatobi Fakorede, and Derek Mikola, “Short-Term Effect of COVID-19 on Self-Employed Workers in Canada,” Canadian Public Policy, 2020, 46 (S1), S66–S81. Bosch, Mariano and William Maloney, “Cyclical Movements in Unemployment and Informality in Developing Countries,” World Bank Policy Research Working Paper No. 4648, 07 2008. Bruhn, Miriam and David McKenzie, “Entry Regulation and the Formalization of Microenterprises in Devel- oping Countries,” The World Bank Research Observer, 03 2014, 29 (2), 186–201. Crane, Leland D., Ryan A. Decker, Aaron Flaaen, Adrian Hamins-Puertolas, and Christopher Kurz, “Business exit during the COVID-19 pandemic: Non-traditional measures in historical context,” Journal of Macroeconomics, 2021, 72, 103419. Dang, Hai-Anh and Cuong Viet Nguyen, “Did a Successful Fight against the COVID-19 Pandemic Come at a Cost? Impacts of the Outbreak on Employment Outcomes in Vietnam,” IZA Discussion Paper No. 13958, 2020. Davies, Elwyn and Andrew Kerr, “Firm survival and change in Ghana, 2003–2013,” Journal of African Economies, 2018, 27 (2), 149–171. de Mel, Suresh, David McKenzie, and Christopher Woodruff, “Labor Drops: Experimental Evidence on the Return to Additional Labor in Microenterprises,” American Economic Journal: Applied Economics, January 2019, 11 (1), 202–35. Dhingra, Swati and Stephen Machin, “The Crisis and Job Guarantees in Urban India,” IZA Discussion Papers 13760, Institute of Labor Economics (IZA) 2020. Fairlie, Robert W, “The Impact of Covid-19 on Small Business Owners: Evidence of Early-Stage Losses from 13 the April 2020 Current Population Survey,” Technical Report, National Bureau of Economic Research 2020. Fajnzylber, Pablo, William Maloney, and Gabriel Montes Rojas, “Microenterprise dynamics in developing countries: How similar are they to those in the industrialized world? Evidence from Mexico,” The World Bank Economic Review, 2006, 20 (3), 389–419. Frazer, Garth, “Which Firms Die? A Look at Manufacturing Firm Exit in Ghana,” Economic Development and Cultural Change, 2005, 53 (3), 585–617. Gollin, Douglas, “Getting Income Shares Right,” Journal of Political Economy, 2002, 110 (2), 458–474. , “Nobody’s business but my own: Self-employment and small enterprise in economic development,” Journal of Monetary Economics, 2008, 55 (2), 219–233. Hsieh, Chang-Tai and Benjamin A. Olken, “The Missing ”Missing Middle”,” Journal of Economic Perspec- tives, September 2014, 28 (3), 89–108. Islamaj, Ergys, Young Eun Kim, and Duong Trung Le, “The Spread of COVID-19 and Policy Responses,” World Bank. https://openknowledge.worldbank.org/handle/10986/35010, 2021. Jayachandran, Seema, “Microentrepreneurship in Developing Countries,” Working Paper 26661, National Bureau of Economic Research January 2020. Kraus, Sascha, Thomas Clauss, Matthias Breier, Johanna Gast, Alessandro Zardini, and Victor Tiberius, “The economics of COVID-19: initial empirical evidence on how family firms in five European countries cope with the corona crisis,” International Journal of Entrepreneurial Behavior Research, 2020. Le, Duong Trung, Edmund Malesky, and Anh Pham, “The impact of local corruption on business tax registration and compliance: Evidence from Vietnam,” Journal of Economic Behavior Organization, 2020, 177, 762 – 786. ın Rama, “Firm Dynamics, Productivity Growth, and Job Creation in Developing Countries: Li, Yue and Mart´ The Role of Micro- and Small Enterprises,” The World Bank Research Observer, 01 2015, 30 (1), 3–38. Loayza, Norman and Jamele Rigolini, Informality trends and cycles, Vol. 4078, World Bank Publications, 2006. Loayza, Norman V and Jamele Rigolini, “Informal employment: safety net or growth engine?,” World Devel- opment, 2011, 39 (9), 1503–1515. Mahmud, Mahreen and Emma Riley, “Household response to an extreme shock: Evidence on the immediate impact of the Covid-19 lockdown on economic outcomes and well-being in rural Uganda,” World Development, 2021, 140, 105318. Maloney, William F, “Informality Revisited,” World Development, 2004, 32 (7), 1159–1178. Mathew, Nyashanu, Ikhile Deborah, Thamary Karonga, and Chireshe Rumbidzai, “The impact of COVID-19 lockdown in a developing country: narratives of self-employed women in Ndola, Zambia,” Health Care for Women International, 2020, 41 (11-12), 1370–1383. PMID: 33030978. McCaig, Brian and Nina Pavcnik, “Entry and Exit of Informal Firms and Development,” IMF Economic Review, 2021, (69), 540–575. McKenzie, David and Anna Luisa Paffhausen, “Small Firm Death in Developing Countries,” The Review of Economics and Statistics, October 2019, 101 (4), 645–657. 14 and Christopher Woodruff, “What Are We Learning from Business Training and Entrepreneurship Evalua- tions around the Developing World?,” The World Bank Research Observer, 07 2013, 29 (1), 48–82. e, “Non-farm entrepreneurship in rural sub-Saharan Africa: New empirical Nagler, Paula and Wim Naud´ evidence,” Food Policy, 2017, 67 (C), 175–191. Nataraj, Shanthi, “The impact of trade liberalization on productivity: Evidence from India’s formal and informal manufacturing sectors,” Journal of International Economics, 2011, 85 (2), 292–301. Oldenski, Lindsay, “Export versus FDI and the communication of complex information,” Journal of International Economics, 2012, 87 (2), 312–322. Porta, Rafael La and Andrei Shleifer, “Informality and Development,” Journal of Economic Perspectives, September 2014, 28 (3), 109–26. Sengupta, Rajeswari and Manish Singh, “Firm formation in India: The last 40 years,” Ideas for India, 04 2019. Shafi, Mohsin, Junrong Liu, and Wenju Ren, “Impact of COVID-19 pandemic on micro, small, and medium- sized Enterprises operating in Pakistan,” Research in Globalization, 2020, 2, 100018. Shapiro, Alan Finkelstein, “Self-employment and business cycle persistence: Does the composition of employ- ment matter for economic recoveries?,” Journal of Economic Dynamics and Control, 2014, 46, 200–218. , “Labor force participation, interest rate shocks, and unemployment dynamics in emerging economies,” Journal of Development Economics, 2018, 133, 346–374. 15 Figure 1: Household business outcomes: monthly averages (a) Monthly Suspension Filings (HBTC) (b) Monthly Closure Filings (HBTC) Average Commune Monthly Suspension Filings Average Commune Monthly Closure Filings 40 Apr Number of Suspension Filings Number of Closure Filings May 30 2 1.5 20 Aug 1 Aug 10 .5 0 0 2016 2017 2018 2019 2020 2016 2017 2018 2019 2020 Month Month Average Suspension Filings per Commune Average Closure Filings per Commune Pre-Covid Mean Pre-Covid Mean (c) Weekly Work Hours (LFS) (d) Monthly Income (LFS) Average Weekly Working Hours Average Monthly Income (Inflation Adjusted) 6500 Average Monthly Income (thousands) 50 Average Weekly Working Hours 6000 45 Sept Sept 5500 40 5000 35 Apr 4500 May 2018 2019 2020 2018 2019 2020 Month Month Average Weekly Working Hours Average Monthly Income (Individual) Pre-Covid Mean Pre-Covid Linear Trend Note: This figure shows monthly-average plots of four commune-level household business outcomes: monthly suspension filings (panel A), monthly closure filings (panel B), weekly work hours (Panel C), and individual monthly income (Panel D). Blue/solid lines indicate monthly average values. Red/dashed lines represent pre-COVID means (2018-19). Data on household business suspension and closure filings are obtained from the Household Business Tax Census. Data on individual work hours and income are obtained from the Vietnamese Labor Force Surveys. 16 Table 1: Annual tax filings, and filings for business closure and temporary suspension Regular tax filings (annually) Irregular filings Year Total Repeated (approx.) New (approx.) Closure filings Suspension filings (1) (2) (3) (4) (5) 2018 908,679 - - 33,673 227,205 2019 917,917 872,467 45,450 52,042 225,230 January-August 2020 907,472 851,863 55,609 53,957 372,895 Note: “Annual regular filings” is defined as the number of (once-a-year) annual lump-sum household business tax filings each year (filed mainly in January each year). “Repeated filings” is an approximation for firm incumbency, calculated as the number of filing records that also appeared in the previous year. “New filings” is an approximation for firm entry, calculated as the number of filing records that did not appear in any previous years. “Closure filings” defines the number of firms filed to permanently close their business. “Suspension filings” defines the number of firms filed to temporarily suspend their business in at least one month in a given year. The corresponding total number of suspension filings recorded for each year (i.e., inclusive of multiple filings per firm-year) are 655,244; 570,933; and 776,934 in 2018, 2019, and January-to-August 2020, respectively. 17 Table 2: Event-study estimates of COVID-19 impacts on household business performance Monthly Suspensions Monthly Closures Weekly Work Hours Monthly Income (1) (2) (3) (4) [Panel A] Aggregate effects Post Covid 7.380*** 0.553*** -3.279*** -945.9*** (0.364) (0.0473) (0.142) (42.82) [Panel B] Effects by month February -1.988*** -0.250*** -3.583*** -511.4*** (0.309) (0.0196) (0.342) (90.09) March 2.646*** 0.453*** 0.864*** -384.6*** (0.383) (0.0434) (0.295) (82.65) April 29.70*** 0.402*** -13.07*** -1,313*** (1.151) (0.0410) (0.355) (74.84) May 11.48*** 1.596*** -4.116*** -1,885*** (0.553) (0.288) (0.314) (106.8) June 3.487*** 0.442*** -0.835*** -750.8*** (0.372) (0.0603) (0.257) (73.14) July 1.886*** 0.905*** -0.602*** -801.1*** (0.261) (0.0649) (0.230) (86.55) August 4.454*** 0.320*** -1.806*** -812.6*** (0.458) (0.0707) (0.310) (121.7) September - - -0.962*** -858.9*** - - (0.317) (74.17) Pre-COVID average 6.67 0.50 47.18 6362.68 Observations 355,080 371,635 419,252 418,824 Observation level Commune Commune Individual Individual Month fixed effects Yes Yes Yes Yes Commune fixed effects Yes Yes Yes Yes Individual controls - - Yes Yes Note: Panel A shows the event-study estimates of the aggregate effect of COVID-19 on household business outcome variables. The coefficients are estimated following Equation 1. Panel B presents the estimated effect for each month after January 2020. The coefficients are estimated following Equation 2. Data on household business suspension and closure filings are obtained from the Household Business Tax Census. Data on individual work hours and income are obtained from the Vietnamese Labor Force Surveys. Monthly income is detrended from pre-COVID linear trend and residualized from individual observables including age, gender, educational level, and occupation level. Standard errors are clustered at the commune level and are presented in parentheses. *** p<0.01, ** p<0.05, * p<0.1. 18 Table 3: Event-study estimates – heterogeneous effects by sector and firm size Monthly Suspensions Monthly Closures Weekly Work Hours Monthly Income (1) (2) (3) (4) [Panel A] Heterogeneity by sector Manufacturing-Construction -0.00772 0.0221*** -1.354*** -737.9*** -0.0274 -0.0027 -0.173 -39.43 [-] [+74.00%] [-2.86%] [-11.87%] Wholesale-Retail 3.499*** 0.367*** -3.630*** -947.8*** -0.254 -0.0388 -0.199 -84.48 [+93.81%] [+135.93%] [-7.58%] [-14.38%] Transportation 0.132*** 0.0206*** -6.402*** -1,676*** -0.0255 -0.00327 -0.376 -100.3 [+23.16%] [+51.50%] [-13.32%] [-21.40%] Hotel-Food-Entertainment 1.119*** 0.0977*** -4.634*** -1,125*** -0.0547 -0.0107 -0.295 -97.1 [+164.56%] [+108.56%] [-9.99%] [-19.51%] [Panel B] Heterogeneity by firm size Revenue less than 25th percentile 2.195*** 0.0551*** - - -0.0883 -0.00493 - - [+105.53%] [+50.10%] - - Revenue between 25th and 50th percentile 1.838*** 0.0459*** - - -0.0792 -0.00415 - - [+112.75%] [+51.00%] - - Revenue between 50th and 75th percentile 1.516*** 0.0380*** - - -0.0844 -0.00369 - - [+104.55%] [+47.50%] - - Revenue greater than 75th percentile 1.775*** 0.409*** - - -0.211 -0.0438 - - [+125.00%] [+204.50%] - - Note: This table shows the event-study estimates of heterogeneous effects across sector and firm’s operating size. Disaggregated sectors (Panel A) include manufacturing & construction; wholesale & retails; transportation; and accommodation, food & entertainment. Firm’s operating size is broken down by revenue quartiles. The coefficients are estimated following Equation 1. Data on household business suspension and closure filings are obtained from the Household Business Tax Census. Data on individual work hours and income are obtained from the Vietnamese Labor Force Surveys. Monthly income is detrended from pre-COVID linear trend and residualized from individual observables including age, gender, educational level, and occupation level. Square brackets show the equivalent of estimated coefficients in percentage changes from pre-COVID means. Standard errors are clustered at the commune level and are presented in parentheses. *** p<0.01, ** p<0.05, * p<0.1. 19 Table 4: difference-in-difference estimates with face-to-face (F2F) index Monthly Suspensions Monthly Closures Weekly Work Hours Monthly Income (1) (2) (3) (4) (5) (6) (7) (8) [Panel A] Aggregate estimated effects Post Covid x F2F Index 2.831*** 2.831*** 0.304*** 0.304*** -0.882*** -0.855*** -52.42 -83.02* (0.616) (0.616) (0.0622) (0.0622) (0.160) (0.157) (47.81) (45.17) [Panel B] Effects by month Feb x F2F Index 1.718** 1.718** -0.0404** -0.0404** -0.684 -0.643 22.34 6.209 (0.683) (0.683) (0.0200) (0.0200) (0.417) (0.414) (108.7) (102.8) Mar x F2F Index 1.992*** 1.992*** 0.131*** 0.131*** -0.567** -0.533** -69.19 -102.9 (0.592) (0.592) (0.0422) (0.0422) (0.251) (0.253) (89.79) (87.18) Apr x F2F Index 9.765*** 9.765*** 0.0951* 0.0951* -1.765*** -1.739*** -0.876 -27.97 (1.359) (1.359) (0.0500) (0.0500) (0.414) (0.414) (81.81) (80.48) May x F2F Index 2.847*** 2.847*** 1.655*** 1.655*** -1.240*** -1.274*** -165.6* -198.5** (0.743) (0.743) (0.399) (0.399) (0.331) (0.328) (98.91) (93.61) Jun x F2F Index 1.075* 1.075* 0.122*** 0.122*** -0.641** -0.638** -21.81 -84.83 (0.589) (0.589) (0.0457) (0.0457) (0.256) (0.252) (75.14) (72.32) Jul x F2F Index 0.775 0.775 0.173** 0.173** -0.366 -0.372 -48.58 -57.49 (0.478) (0.478) (0.0820) (0.0820) (0.246) (0.242) (78.87) (74.75) Aug x F2F Index 1.653** 1.653** -0.0203 -0.0203 -0.438 -0.487 35.89 -0.937 (0.643) (0.643) (0.0657) (0.0657) (0.315) (0.309) (101.8) (97.52) Sep x F2F Index - - - - -0.650** -0.667** -112.1 -161.1** - - - - (0.271) (0.272) (70.48) (68.83) Observations 330,330 330,330 341,165 341,165 372,204 372,203 371,814 371,813 Observation level Commune Commune Commune Commune Individual Individual Individual Individual Individual controls - - - - No Yes No Yes Month fixed effects No Yes No Yes No Yes No Yes Commune fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Note: Panel A presents the difference-in-difference estimates of the aggregate effect of COVID-19 on household business performance, following Equation 3. Panel B presents the estimated effect for each month after January 2020, following Equation 4. Data on household business suspension and closure filings are obtained from the Household Business Tax Census (HBTC). Data on individual work hours and income are obtained from the Vietnamese Labor Force Surveys. Data on the face-to-face (F2F) index is derived from Avdiu and Nayyar (2020), constructed using the 2019 HBTC industrial composition in each commune. Monthly income is detrended from pre-COVID linear trend and residualized from individual observables including age, gender, educational level, and occupation level. Standard errors are clustered at the commune level and presented in parentheses: *** p<0.01, ** p<0.05, * p<0.1. 20 A Appendix Figure A.1: Spatial distribution of the sectoral face-to-face intensity index Sectoral F2F Intensity 41.1 - 49.8 49.9 - 52.6 52.7 - 54.9 55.0 - 57.3 57.4 - 62.0 0 100 200 ¹ 400 Km Note: This figure shows the geographical distribution of baseline sectoral F2F index measure. The F2F indices are aggregated to the district level for visualization purposes. Larger score indicates greater F2F intensity (more vulnerable to lockdown). Blank (white) represents districts with no F2F score (no household business data). There are approximately 670 districts in the country, comprising of over 11,000 communes. 21 Figure A.2: Average Monthly Income Residuals Note: This figure shows monthly-average plots of household business income residuals, i.e., the main measure of earnings used in the analysis. Monthly income residuals are calculated as the difference between the reported income and the predict income after accounting for pre-COVID linear trend and individual observable factors including age, gender, education level, occupation level, and working industry. Data obtained from the Vietnamese Labor Force Surveys. Figure A.3: Government Stringency Index in 2020 (monthly average) Note: This figure shows the monthly average COVID-19 government response stringency index in 2020. Data on the daily government lockdown stringency index was obtained from the Oxford COVID-19 Government Response Tracker (OxCGRT), which systematically collects information on several different common policy responses that governments have taken to respond to the pandemic on 20 indicators such as school closures and travel restrictions. 22 Figure A.4: Google Mobility Index in 2020 (monthly average) Note: This figure shows the monthly average mobility index in Vietnam in 2020. Data obtained from Google’s COVID-19 Community Mobility Reports which provides a measure for the changes in human activities relative to the pre-COVID baseline by geography, across different categories of places such as retail and recreation, groceries and pharmacies, parks, transit stations, workplaces, and residential. pre-COVID baseline was calculated as median value, for the corresponding day of the week, during the 5-week period from Jan 3 to Feb 6, 2020. Figure A.5: Sector-wide quarterly growth rates (year-on-year) Note: This figure shows the year-on-year quarterly growth rates in aggregate outputs of the main sectors in Vietnam. Quarterly sectoral outputs data is obtained from Haver Analytics. “Accomo-Restaurant-Tran” refers to the accommodation, restaurant, and transportation sectors. “Info-Comm-Finance” refers to the information, communication, and financial sectors. 23 Table A.1: Descriptive statistics of main household business variables (pre-COVID-19) Mean S.D. Observations [Panel A] Household Business Tax Census (commune-month average; Jan 2018-Jan 2020) Business Suspension Filings (businessses per commune) 6.67 17.61 316,344 Industry: Manufacturing-Construction 0.90 3.75 316,344 Industry: Wholesale-Retail 3.73 12.38 316,344 Industry: Transportation 0.57 2.53 316,344 Industry: Hotel-Food-Entertainment 0.68 2.33 316,344 Industry: Other 0.70 2.49 316,344 Revenue: <=25th percentile 2.08 6.24 316,344 Revenue: 25-50th percentile 1.63 4.92 316,344 Revenue: 50-75th percentile 1.45 4.88 316,344 Revenue: >75th percentile 1.42 7.65 316,344 Total Monthly Closure Filings (businessses per commune) 0.50 3.16 331,093 Industry: Manufacturing-Construction 0.03 0.36 331,093 Industry: Wholesale-Retail 0.27 1.63 331,093 Industry: Transportation 0.04 0.56 331,093 Industry: Hotel-Food-Entertainment 0.09 1.57 331,093 Industry: Other 0.07 0.71 331,093 Revenue: <=25th percentile 0.11 0.74 331,093 Revenue: 25-50th percentile 0.09 0.57 331,093 Revenue: 50-75th percentile 0.08 0.52 331,093 Revenue: >75th percentile 0.20 2.51 331,093 [Panel B] Labor Force Survey (individual level; January 2018-January 2020) Weekly Working Hours 47.18 12.23 289,243 Industry: Manufacturing-Construction 47.41 11.44 104,179 Industry: Wholesale-Retail 47.88 12.10 88,214 Industry: Transportation 48.05 11.55 18,640 Industry: Hotel-Food-Entertainment 46.40 13.07 43,866 Industry: Other 44.32 14.35 26,980 Business Size: <5 employees 45.41 13.16 52,384 Business Size: >=5 employees 46.33 12.41 65,832 Monthly Income (’000 VND) 5,841.44 5,859.15 288,876 Industry: Manufacturing-Construction 5,627.77 4,695.29 103,812 Industry: Wholesale-Retail 6,017.72 6,828.50 88,214 Industry: Transportation 7,224.92 4,939.61 18,640 Industry: Hotel-Food-Entertainment 5,270.75 6,262.67 43,866 Industry: Other 5,851.17 5,203.40 26,980 Business Size: <5 employees 5,789.70 3,861.84 52,384 Business Size: >=5 employees 6,083.95 6,993.23 65,832 Face-to-Face Index (F2F) (2019 baseline) 54.59 4.64 9,096 Note: summary statistics of the main variables used in the analysis. Averages are shown over the baseline (pre-COVID) period between January 2018 and January 2020. Panel A shows commune-averaged household business statistics obtained from the Household Business Tax Census, including monthly suspension and closure filings. Panel B shows individual-level employment and socio-demographic statistics for self-employed households and those running family business obtained from the Vietnamese Labor Force Surveys, including reported weekly work hours and monthly income. 24 Table A.2: Examples of Face-to-Face (F2F) index for different industries Percentiles F2F scores Closest equivalent 4-digit industry Smallest 31.17455 Manufacture of veneer sheets and wood-based panels 1% 39.56975 Manufacture of articles of concrete, cement and plaster 5% 47.5778 Manufacture of tobacco products 10% 49.09625 Wholesale of textiles, clothing and footwear 25% 51.65359 Sale of motor vehicle parts and accessories Mean 54.59439 Retail sale via stalls and markets of food, beverages and tobacco products 50% 55.08822 Short term accommodation activities 75% 57.95426 Retail sale of audio and video equipment in specialized stores 90% 59.86957 Retail sale of food in specialized stores 95% 60.74328 Retail sale in non-specialized stores with food, beverages or tobacco predominating 99% 63.10362 Other retail sale of new goods in specialized stores Largest 70.12779 Retail sale of clothing, footwear and leather articles in specialized stores Note: The Face-to-Face (F2F) score used in our analysis is taken from Avdiu and Nayyar (2020), which measures the importance and intensity of F2F interactions with other people for more than 900 occupations based on 4-digit International Standard of Industrial Classification (ISIC) codes. Table A.3: Monthly Income Prediction Model Variable Monthly Income Coefficient S.E Mdate 31.99*** (1.464) Age 25.43*** (0.872) Male 1,833*** (23.92) Education Level Less than elementary 600.3*** (39.37) Elemetary 836.9*** (38.00) Secondary 1,229*** (42.06) High school 1,583*** (51.52) College/University 2,057*** (76.08) More than university 5,694*** (460.8) Industry 1.Manu-Construction -418.5*** (74.44) 2.Wholesale-Retail -529.6*** (82.38) 3.Transportation -103.6 (96.38) 4.Accomm-Food-Entertainment -1,128*** (84.65) 5.Other -911.1*** (82.62) Occupation Level 2.Professional -4,429*** (444.9) 3.Technicians and Associate Professionals -3,977*** (436.1) 4.Clerical Support Workers -3,309*** (446.9) 5.Service and Sales Workers -5,987*** (428.0) 6.Skilled Agricultural, Forestry and Fishery Workers -7,181*** (471.4) 7.Craft and related trades workers -6,962*** (428.6) 8.Plant and machine operators, and assemblers -6,361*** (430.8) 9.Elementary occupations -7,829*** (428.5) Constant -12,650*** (1,123) Observations 321,015 R-squared 0.050 Note: This table presents our linear monthly income prediction model using the data from January 2018-January 2020 (pre-COVID) period. Standard errors are presented in parentheses. *** p<0.01, ** p<0.05, * p<0.1. 25 Table A.4: Event-study estimates on household business suspensions and closures by industry and month (HBTC) Monthly Suspension Filings Monthly Closure Filings Manufacturing Wholesale Transportation Hotel-Food- Other Manufacturing Wholesale Transportation Hotel-Food- Other Construction Retail Entertainment Construction Retail Entertainment (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) [Panel A] Aggregate effect Post Covid -0.00772 3.499*** 0.132*** 1.119*** 2.582*** 0.0221*** 0.367*** 0.0206*** 0.0977*** 0.0403*** (0.0274) (0.254) (0.0255) (0.0547) (0.0865) (0.00270) (0.0388) (0.00327) (0.0107) (0.00524) [Panel B] Month-by-month effect Feb x Post Covid -0.808*** -1.535*** -0.219*** -0.213*** 0.832*** -0.0202*** -0.140*** -0.0193*** -0.0367*** -0.0302*** (0.0485) (0.229) (0.0239) (0.0257) (0.0720) (0.00177) (0.0126) (0.00224) (0.00521) (0.00408) Mar x Post Covid -0.234*** 1.274*** -0.00987 0.292*** 1.337*** 0.0414*** 0.280*** 0.0279*** 0.0669*** 0.0324*** (0.0251) (0.293) (0.0344) (0.0437) (0.0807) (0.0115) (0.0270) (0.00726) (0.00839) (0.00656) Apr x Post Covid 1.140*** 15.66*** 0.836*** 4.351*** 7.375*** 0.0171*** 0.240*** 0.0292*** 0.0903*** 0.0201** (0.0868) (0.819) (0.0678) (0.171) (0.249) (0.00535) (0.0252) (0.0106) (0.0102) (0.00819) May x Post Covid 0.438*** 5.169*** 0.349*** 1.806*** 3.620*** 0.0271*** 1.128*** 0.0226*** 0.298*** 0.112*** (0.0660) (0.357) (0.0460) (0.0950) (0.139) (0.00609) (0.252) (0.00783) (0.0556) (0.0215) Jun x Post Covid -0.315*** 1.434*** 0.0884*** 0.661*** 1.614*** 0.0188** 0.304*** 0.0322*** 0.0599*** 0.0221* (0.0413) (0.247) (0.0325) (0.0627) (0.0868) (0.00798) (0.0345) (0.00834) (0.0175) (0.0116) Jul x Post Covid -0.192*** 0.691*** -0.0694*** 0.174*** 1.293*** 0.0622*** 0.556*** 0.0364*** 0.155*** 0.0834*** (0.0321) (0.183) (0.0242) (0.0335) (0.0673) (0.00742) (0.0416) (0.00818) (0.0203) (0.0120) Aug x Post Covid -0.0834** 1.802*** -0.0478 0.763*** 2.006*** 0.00818* 0.202*** 0.0149** 0.0506** 0.0417*** (0.0418) (0.295) (0.0351) (0.0814) (0.117) (0.00489) (0.0372) (0.00669) (0.0250) (0.0147) Mean (pre-COVID) 0.90 3.73 0.57 0.68 0.70 0.03 0.27 0.04 0.09 0.07 26 Observations 355,080 355,080 355,080 355,080 355,080 371,635 371,635 371,635 371,635 371,635 Observation level Commune Commune Commune Commune Commune Commune Commune Commune Commune Commune Month FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Commune FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Notes: This table presents the event-study estimates of the heterogeneous impacts of COVID-19 on household business across sectors. Panel A shows coefficients of aggregate impacts, estimated using Equation 1. Panel B shows coefficients for each month after January 2020, estimated using Equation 2. Data on household business suspension and closure filings are obtained from the Household Business Tax Census. Standard errors are clustered at the commune level and are presented in parentheses. *** p<0.01, ** p<0.05, * p<0.1. Table A.5: Event-study estimates on household business suspensions and closures by revenue and month (HBTC) Monthly Suspension Filings Monthly Closure Filings <=25th 25th-50th 50th-75th >75th <=25th 25th-50th 50th-75th >75th Percentile Percentile Percentile Percentile Percentile Percentile Percentile Percentile Revenue Revenue Revenue Revenue Revenue Revenue Revenue Revenue (1) (2) (3) (4) (5) (6) (7) (8) [Panel A] Aggregate effect Post Covid 2.195*** 1.838*** 1.516*** 1.775*** 0.0551*** 0.0459*** 0.0380*** 0.409*** (0.0883) (0.0792) (0.0844) (0.211) (0.00493) (0.00415) (0.00369) (0.0438) [Panel B] Month-by-month effect Feb x Post Covid -0.692*** -0.404*** -0.447*** -0.400*** -0.0653*** -0.0562*** -0.0483*** -0.0768*** (0.112) (0.0991) (0.0900) (0.104) (0.00464) (0.00510) (0.00485) (0.0111) Mar x Post Covid 0.712*** 0.625*** 0.549*** 0.773*** 0.0949*** 0.0926*** 0.0749*** 0.186*** (0.119) (0.0967) (0.0943) (0.209) (0.0107) (0.00825) (0.00811) (0.0343) Apr x Post Covid 8.651*** 7.397*** 6.390*** 6.921*** 0.135*** 0.0925*** 0.0991*** 0.0699*** (0.281) (0.269) (0.288) (0.649) (0.0185) (0.0110) (0.0122) (0.0199) May x Post Covid 3.572*** 2.886*** 2.308*** 2.615*** 0.0415*** 0.0306*** 0.00248 1.514*** (0.145) (0.128) (0.138) (0.299) (0.0107) (0.00952) (0.00674) (0.281) Jun x Post Covid 1.321*** 0.975*** 0.611*** 0.575*** 0.0301*** 0.0157* 0.00296 0.388*** (0.0978) (0.0871) (0.0882) (0.200) (0.0105) (0.00906) (0.00701) (0.0512) Jul x Post Covid 0.580*** 0.412*** 0.312*** 0.593*** 0.127*** 0.117*** 0.112*** 0.538*** (0.0678) (0.0556) (0.0540) (0.163) (0.0111) (0.00999) (0.00958) (0.0543) Aug x Post Covid 1.222*** 0.978*** 0.893*** 1.346*** 0.0223** 0.0296*** 0.0233*** 0.243*** 27 (0.123) (0.0972) (0.105) (0.239) (0.0105) (0.00791) (0.00735) (0.0651) Mean (pre-COVID) 2.08 1.63 1.45 1.42 0.11 0.09 0.08 0.20 Observations 355,080 355,080 355,080 355,080 371,635 371,635 371,635 371,635 Observation level Commune Commune Commune Commune Commune Commune Commune Commune Month FE Yes Yes Yes Yes Yes Yes Yes Yes Commune FE Yes Yes Yes Yes Yes Yes Yes Yes Notes: This table presents the event-study estimates of the heterogeneous impacts of COVID-19 on household business by the size of revenue. Panel A shows coefficients of aggregate impacts, estimated using Equation 1. Panel B shows coefficients for each month after January 2020, estimated using Equation 2. Data on household business suspension and closure filings are obtained from the Household Business Tax Census. Standard errors are clustered at the commune level and are presented in parentheses. *** p<0.01, ** p<0.05, * p<0.1. Table A.6: Event-study estimates on household business workload and income by industry and month (LFS) Weekly Work Hours Monthly Income (residuals) Manufacturing Wholesale Transportation Hotel-Food- Other Manufacturing Wholesale Transportation Hotel-Food- Other Construction Retail Entertainment Construction Retail Entertainment (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) [Panel A] Aggregate effect Post Covid -1.354*** -3.630*** -6.402*** -4.634*** -4.709*** -737.9*** -947.8*** -1,676*** -1,125*** -838.2*** (0.173) (0.199) (0.376) (0.295) (0.333) (39.43) (84.48) (100.3) (97.10) (94.35) [Panel B] Month-by-month effect Feb x Post Covid -4.374*** -3.481*** -3.191*** -2.092*** -4.946*** -452.8*** -494.7*** -512.5** -87.94 -822.9*** (0.567) (0.479) (0.841) (0.601) (0.946) (111.8) (147.8) (233.0) (190.0) (272.6) Mar x Post Covid 4.270*** -0.862** -1.118 -1.250** -0.999 -273.3*** -685.2*** -627.6*** -258.1 -97.43 (0.428) (0.430) (0.708) (0.543) (0.732) (100.6) (129.0) (238.8) (171.0) (334.1) Apr x Post Covid -6.078*** -14.45*** -25.65*** -18.86*** -16.00*** -1,106*** -1,118*** -2,528*** -1,809*** -1,252*** (0.415) (0.569) (1.086) (0.847) (0.896) (77.88) (142.0) (219.0) (145.3) (195.5) May x Post Covid -2.919*** -3.447*** -7.551*** -5.460*** -5.856*** -1,368*** -1,764*** -2,849*** -2,292*** -2,355*** (0.411) (0.423) (0.869) (0.651) (0.771) (111.1) (179.3) (236.1) (212.7) (220.4) Jun x Post Covid -0.283 -0.710* -1.072* -1.218** -2.122*** -559.0*** -790.5*** -1,407*** -807.4*** -772.2*** (0.332) (0.381) (0.645) (0.536) (0.679) (89.29) (127.5) (202.1) (131.0) (208.2) Jul x Post Covid -0.472* -0.879** -1.218* 0.340 -1.129* -678.1*** -797.6*** -1,750*** -1,224*** -325.3* (0.272) (0.356) (0.648) (0.488) (0.625) (75.97) (179.3) (225.0) (252.5) (168.6) Aug x Post Covid -0.683* -2.146*** -2.474*** -2.695*** -2.745*** -521.8*** -928.0*** -1,428*** -1,034*** -429.5* (0.398) (0.396) (0.726) (0.756) (0.725) (92.39) (303.3) (203.5) (224.2) (242.9) Sept x Post Covid 0.373 -1.174*** -1.637** -2.343*** -2.097*** -649.1*** -883.6*** -1,759*** -1,018*** -748.8*** 28 (0.390) (0.381) (0.759) (0.667) (0.775) (87.77) (125.1) (200.0) (136.1) (183.5) Mean (pre-COVID) 47.41 47.88 48.05 46.40 44.32 6217.53 6589.74 7830.67 5765.21 5997.87 Observations 145,771 127,061 26,336 62,976 45,943 145,771 127,061 26,336 62,976 45,943 Observation level Individual Individual Individual Individual Individual Individual Individual Individual Individual Individual Individual controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Month FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Commune FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Notes: This table presents the event-study estimates of the heterogeneous impacts of COVID-19 on household business across sectors. Panel A shows coefficients of aggregate impacts, estimated using Equation 1. Panel B shows coefficients for each month after January 2020, estimated using Equation 2. Data on individual work hours and income are obtained from the Vietnamese Labor Force Surveys. Monthly income (in thousand VND) is residualized from pre-COVID linear trend and observable individual factors including age, gender, educational level, and occupation group. Standard errors are clustered at the commune level and are presented in parentheses. *** p<0.01, ** p<0.05, * p<0.1. Table A.7: Difference-in-difference estimates with F2F index – robustness check with 2018 HBTC baseline Monthly Suspensions Monthly Closures Weekly Work Hours Monthly Income (residuals) (1) (2) (3) (4) Panel A: Aggregate effects Post Covid x F2F Index 2.993*** 0.327*** -0.732*** -30.16 (0.689) (0.0674) (0.179) (42.70) Panel B: Month-by-month effects Feb x F2F Index 1.891** -0.0578*** -0.554 8.839 (0.751) (0.0214) (0.509) (102.9) Mar x F2F Index 2.004*** 0.166*** -0.478* -81.05 (0.654) (0.0502) (0.290) (81.26) Apr x F2F Index 10.36*** 0.133** -1.965*** 40.11 (1.519) (0.0532) (0.386) (71.63) May x F2F Index 3.078*** 1.715*** -0.945** -110.8 (0.847) (0.435) (0.370) (95.82) Jun x F2F Index 1.097 0.180*** -0.580** 3.212 (0.669) (0.0484) (0.277) (68.73) Jul x F2F Index 0.756 0.177** -0.434* -11.11 (0.533) (0.0758) (0.251) (67.82) Aug x F2F Index 1.765** -0.0363 -0.0915 -36.46 (0.740) (0.0657) (0.376) (89.88) Sep x F2F Index - - -0.750** -90.62 - - (0.325) (71.01) Observations 322,960 333,355 401,591 401,164 Observation level Commune Commune Individual Individual Month FE Yes Yes Yes Yes Commune FE Yes Yes Yes Yes Individual controls - - Yes Yes Notes: This table presents robustness results following Table 4. Panel A presents the difference-in-difference estimates of the aggregate impacts of COVID-19 on household business outcome variables. The coefficients are estimated using Equation 3. Panel B presents the estimated impacts for each month after January 2020. The coefficients are estimated using Equation 4. Data on household business suspension and closure filings are obtained from the Household Business Tax Census (HBTC). Data on individual work hours and income are obtained from the Vietnamese Labor Force Surveys. Data on F2F index is derived from Avdiu and Nayyar (2020), constructed using the 2018 HBTC industrial composition in each commune. Monthly income is residualized from pre-COVID linear trend and observable individual factors including age, gender, educational level, and occupation level. Standard errors are clustered at the commune level and are presented in parentheses: *** p<0.01, ** p<0.05, * p<0.1. 29