Policy Research Working Paper 10579 Do Shocks Perpetuate Disparities within and across Informal Firms? Evidence from the COVID-19 Pandemic in South Asia Arti Grover Mariana Pereira-López International Finance Corporation & South Asia Region October 2023 Policy Research Working Paper 10579 Abstract Using three rounds of data from the Business Pulse Survey In particular, necessity firms experience a larger drop in in South Asia, this paper studies the differential effects of sales relative to the parasitic type of informal firms. To add the COVID-19 shock on informal firms. It also captures to this, the adjustment response (for example, the use of heterogeneity within informal firms based on the degree digital platforms) of informal firms is smaller, which perpet- and motivation of informality. The findings suggest that uates the gap between formal and informal firms. Within the severity of the impact of the COVID-19 shock and informal firms, the parasitic type typically have a smaller the recovery speed are strongly associated with the degree adjustment response. These findings have implications for of informality. Firms’ external attributes, such as size, sector, policies to support the private sector in the presence of age, and gender of the owner, do not explain the depth of informality, including considerations pertaining to target- the impact. Internal characteristics such as poor manage- ing, modality of support, and the instruments required for ment capabilities and education of the manager and owners designing more impactful programs during shocks. are strong predictors of vulnerability among informal firms. This paper is a product of the International Finance Corporation and the Office of the Chief Economist, South Asia 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 agrover1@ifc.org and mpereiralopez@ifc.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 Do Shocks Perpetuate Disparities within and across Informal Firms? Evidence from the COVID-19 Pandemic in South Asia∗ Arti Grover† & Mariana Pereira-López‡ JEL classification: D22, L20, L25, O17 Keywords: COVID-19, Crisis, Firms, Informality, Recovery, Digital ∗ Acknowledgments: The authors are grateful to the South Asia Regional Trade Facilitation Program (SARTFP) for funding support and Denis Medvedev for his strategic guidance during the inception of the project. The authors thank several members of the World Bank teams for their efforts in data collection, including Amila Indeewari Dahanayake, Besart Avdiu, Hosna Ferdous Sumi, Mauricio Alejandro Pinzon Latorre, Mohammad Sulaiman Akbari Ojashwi Samser JB Rana, Peter Mousley, Rafay Khan, Ruchita Mangh- nani and Subika Farazi. The team is grateful to Franklin Maduko, Sarah Hebous, and Reyes Aterido for their careful work in data harmonization. The team thanks Maurizio Bussolo and Siddarth Sharma for their helpful feedback. † World Bank email: agrover1@ifc.org ‡ World Bank email: mpereiralopez@ifc.org 1. Introduction The COVID-19 crisis represented an unprecedented shock for the private sector, affecting businesses through many channels at an uncontrolled speed and spanning a duration of more than two years. In the process, it unveiled the underlying vulnerabilities of firms, sectors, and countries. These types of crises pose pressing questions not only about the magnitudes and responses to economic shocks in general, but also about the heterogeneity of their effects. For instance, evidence from the COVID-19 shock and recovery indicates that small firms were disproportionately affected (Apedo-Amah et al., 2020; Cirera et al., 2021). Given that most informal firms tend to be small, it is crucial to understand not only the effects of a shock on informal firms, but also if those effects are explained by external characteristics such as firm size or by other internal attributes of informality. A priori, given the characteristics of informal firms, the magnitude and direction of the impact of crises and, in particular, of the COVID-19 shock are not entirely clear. On the one hand, initially, informal firms are expected to exhibit a worse performance due to their higher vulnerability, including the lack of access to finance from formal sources and the difficulty of public support reaching these firms. Specific to the COVID-19 crisis, these firms are concentrated in sectors requiring more face-to-face interactions and were, therefore, more exposed to the lockdown restrictions. On the other hand, informal firms can avoid certain taxes and regulatory burden and have more flexibility in terms of employment (including lower hiring and firing costs), location, resource management, and market competition (Loayza, 2018). Therefore, informal firms and jobs can more easily recover given their lower organization capital and entry costs (Alfaro et al., 2020). Using several rounds of Business Pulse Survey (BPS) data of the World Bank for six countries in South Asia, this paper quantifies the heterogeneity in the impact of the COVID-19 crisis and the recovery of firms by their informality level. Our results suggest that: First, there is a huge overlap in the productivity distribution across formal and informal firms, suggesting that the reasons for staying informal at the lower end of the distribution might be very different than at the higher end. This also points to the possibility that the shock may not affect all informal firms uniformly. Second, in line with the existing literature, our BPS data confirms that compared with formal firms, informal firms are smaller, younger, and in sectors with low entry costs. They have lower productivity, with consistently lower levels of management and technological capabilities. Even the better-managed informal firms do not seem to grow in size, especially 2 the parasitic ones and thereby perpetuate the misallocation of resources. Third, the severity of the impact of the COVID-19 shock and the recovery speed are associated with the degree of informality. Moreover, firms’ external attributes such as size, sector, age, and gender of the owner do not explain the depth of the impact. Characteristics of informal firms that are intrinsic, such as poor management capabilities and education of the manager and owners are strong predictors of poor outcomes among informal firms. Fourth, results, including split regressions, fully interacted models, Shapley decomposition, and Oaxaca-Blinder decomposition, suggest that informality by itself is a critical factor in explaining the impact of the shock on firms that goes beyond the firms’ choice of sector, their size and age. Informal firms driven by necessity witness a larger drop in sales during the shock period compared to the parasitic type of informal firms. Fifth, the adjustment response of informal firms to crisis is lower even after accounting for their lower initial capabilities. In particular, the adoption of digital technology that was a key adjustment response to lockdown restrictions and partially helped firms mitigate the negative demand shock of the crisis was lower in informal firms and particularly the parasite firms. Encouragingly, female-led informal firms were more likely to start the use of digital technologies and sell online but less likely to invest in such technologies. Sixth, Access to public support by informal firms is significantly lower. We also suspect mistargeting in support, given that the probability of accessing support is uncorrelated with initial productivity and positively associated with the level of change in sales. Moreover, the instruments used do not address the underlying factors that make informal firms vulnerable, nor do they use the crisis as an opportunity to provide incentives for firms to formalize. Although a large informal sector is a distinctive characteristic of many developing countries (Perry and Maloney, 2007; La Porta and Shleifer, 2014; Ulyssea, 2020), countries in South Asia exhibit even higher rates of informality (Bussolo et al., 2020) such that more than 90% of the businesses in South Asia are informal. Thus, South Asia provides a unique setting for this type of analysis. In addition, South Asia was also among the regions that were disproportionately affected by the COVID-19 shock (Brucal et al., 2021). The combination of these two factors reinforces the importance of analyzing the role of informality in explaining firm-level outcomes. Our study makes four main contributions to the literature. First, our analysis takes ad- vantage of several rounds of novel firm-level data specifically collected to analyze the 3 COVID-19 shock. We rely on BPS for Afghanistan, Bangladesh, India, Nepal, Pakistan, and Sri Lanka, which includes information that spans from the onset of the crisis and the implementation of lockdown measures to the recovery phase, where mostly all firms were fully open.1 Second, we construct a composite measure of informality that accounts for its multidimensionality through the continuous index based on a Tetrachoric Principal Component Analysis (PCA). This approach allows us to measure the degree of informality on a continuum. Third, we distinguish between low-productivity informal firms born out of “necessity” versus those that are comparable to formal firms but they still choose to remain informal to evade costs associated with formality (“parasite” firms).2 Fourth, we separate the role of external attributes and internal characteristics in correlating the effect of the crisis on informal firms. The remainder of the paper is structured as follows: Section 2 summarizes the literature on the definition and measurement of the informal sector focused on firm informality. In Section 3, we explain the methodological aspects of the analysis, including the unique char- acteristics of the Business Pulse Survey dataset, the structure of the informality composite index that we propose, and our analysis of measures that differentiate across different views of informality such as necessity, De Soto, and “parasite” firms (See Ulyssea, 2018, 2020). We also highlight the identification strategy. Section 4 presents descriptive analyses of the variables of interest in the sample. Section 5 presents the main results of our analysis of the heterogeneous effects of the crisis according to informality, while Section 6 concludes. 2. Related Literature 2.1 Definition and Measurement of Informality Informality may include definitions from the perspective of the workers, firms, and gov- ernment. These may be inter-related but have different implications. On the workers’ side, a job that does not provide standard labor protection falls under informality (Perry and Maloney, 2007), while on the firm’s side, informality ranges from firms that do not exist in the records of the authorities (entirely under the radar) to much larger firms complying with many regulations, hiring formal workers, but failing to declare some of their revenues and workers to partially avoid some tax payments (La Porta and Shleifer, 2014) which makes them informal from the perspective of government. A wide array of combinations with different degrees of informality exist in between adhering to all regulatory requirements of 1 These data are obtained over three rounds of data collection, except for the case of Afghanistan and Sri Lanka, where political risks prevented us from implementing a third round of surveys. 2 See Ulyssea (2018, 2020). 4 being fully formal, which makes it a multidimensional concept (Perry and Maloney, 2007; Ulyssea, 2018, 2020). The literature uses both macro and micro methods to measure informality.3 Our work mea- sures informality from the firm’s side and relies on microeconomic approaches that produce disaggregate indicators at the sectoral, regional, and enterprise levels. Given the obvious problems of self-reporting by firms about their non-compliance, proxies (e.g., self-employed and microenterprises) are widely used to measure informality, though with the caveat that very small firms can still comply with the law and be formal (Restrepo Echavarria, 2015). Firm-level surveys that include direct questions about the various forms of informality, such as the one used in our work, can be considered a lower bound of the size of informality (Putnin , š and Sauka, 2015), given that some firms might misreport (to avoid being on the radar of the government).4 However, in most cases, researchers rely on only one aspect of informality ignoring the multidimensional nature of the issue (Restrepo Echavarria, 2015; Perry and Maloney, 2007). 2.2 Stylized Facts on Informality The informality literature has identified a set of empirical regularities about informality (non-necessarily causal) across developing countries. First, the share of informality in developing countries is large and persistent Ulyssea (2020), accounting for 70%-80% of the labor force and about 20% of GDP (Loayza, 2018). Second, informal firms tend to be a drag on growth not only because these firms have lower productivity (Perry and Maloney, 2007; La Porta and Shleifer, 2014; La Porta et al., 2008) but also because they use valuable resources leading to misallocation and aggregate productivity losses. Third, informal firms tend to employ low-skilled or low-educated managers and workers (La Porta and Shleifer, 2014; De Mel et al., 2010). Fourth, informal firms lag behind in innovation, investment in new capital, and technology adoption, perhaps due to smaller size and lack of complementary factors (intangibles) such as human capital and managerial capabilities. 2.3 Informality Types or Views: Exit and Exclusion In exploring the root cause of informality in Latin America, Arias et al. (2010) propose two distinct but complementary lenses: informality driven by ‘exclusion’ from state benefits vis- 3 See Medina and Schneider (2019) for an example of macroeconomic-focused methods, Medina and Schneider (2018) for a survey of macro methods and Elgin et al. (2021) for a comprehensive database using both model-based and survey-based measures of informality that covers more than 160 economies for the period 1990-2018. 4 Some surveys use indirect questions about the sector in cases where firms might have strong incentives to provide incorrect responses (Putnin, š and Sauka, 2015; Gërxhani, 2004; Reilly et al., 2019). 5 a-vis that driven by voluntary ‘exit’ decisions resulting from private cost-benefit calculations that lead firms to opt out of formal institutions. The former views informality as resulting from either a “necessity” that creates a class of subsistence low-productive firms that would otherwise exit the market to be profitable enough to survive (La Porta and Shleifer, 2014). Alternatively, such informal firms may have untapped potential but excluded from the formal market due to burdensome regulations (De Soto, 1989). By comparison, the latter perceives informal firms as “parasites” that have productivity comparable to their formal counterparts but still take advantage of the lower costs of informality to remain more profitable (Levy Algazi et al., 2018). Recent works by Ulyssea (2018, 2020) allow for these views of informality to coexist and explain the heterogeneity and complexity of the informality problem in developing countries. 2.4 Crisis and Informality Several studies have analyzed the role of the informal sector in the context of crises (e.g., Elgin et al., 2021, 2022) and specifically if it can serve as a safety net in the face of employment losses during shocks. Studies suggest that informality may indeed be a safety net (Loayza and Rigolini, 2011; Colombo et al., 2019; Fernández and Meza, 2015) although it may depend inversely on the size of informality. To the best of our knowledge, so far the evidence on the impact of the pandemic on the informal economy has been limited to macro measures (Ohnsorge and Yu, 2022) or on informal employment or workers (Alfaro et al., 2020, ILO, 2020). The closest work is related to firm-level evidence on the impact of the COVID-19 shock on small firms (Bartik et al., 2020; Humphries et al., 2020; Apedo-Amah et al., 2020; Cirera et al., 2021; Brucal et al., 2021). 3. Data and Methodology 3.1 Data To analyze the role of informality in the context of the COVID-19 crisis, we use data from the World Bank BPS. This survey has collected data for thirty-three months for over 150,000 firms in 87 countries. The BPS survey instrument includes questions that have focused on the different shocks that firms suffered in the aftermath of COVID (change in sales, uncertainty, expectations, supply shocks, financial fragility), the adjustment mechanisms they had to implement (employment adjustments, digitalization, and innovation efforts), as well as the amount and characteristics of the public support received.5 5 It also contains detailed information about firm-level baseline characteristics such as firm size, sector, age, gender of the owner, indebtedness, access to credit, and pre-COVID sales, among other indicators. For details on the first three waves, see Apedo-Amah et al. (2020) and Cirera et al. (2021). 6 In the case of South Asia, the BPS includes data for six countries: Afghanistan, Bangladesh, India, Nepal, Pakistan, and Sri Lanka, over three waves of data collection. The first one was conducted at the onset of the COVID-19 crisis (between May and July 2020), a follow-up round was collected between May and August 2021, and the third and final round was collected between May and July 2022. The last round was not implemented in Sri Lanka and Afghanistan due to fragile political situations in the two countries (see Figure A1 for the timing of the surveys and Table B1 for the observation count in each round). 3.2 Measuring Informality Starting from the second round, the BPS in South Asia also captures the multidimensional nature of informality by including six questions relating to business license, registration under company law, contributions to employee provident funds, value-added tax (VAT) registration, the firm holding a bank account, and the separation of this account for business use.6 3.2.1 Informality Composite Index To measure the multidimensionality problem of informality, we follow an approach similar to Medvedev and Oviedo (2016) in constructing an informality index that relies on a combination of informality indicators. We take the first component of a Tetrachoric Principal Component Analysis (PCA), given that our informality variables are binary, and build three scores to capture distinct aspects of informality. Such an approach helps reduce the multidimensionality of informality into one index and allows us to build a continuous quantitative measure. The first index, PCA 5, uses all five measures captured in the BPS, except for the bank accounts mix, as this variable is conditional on having a bank account. We build two sub-indices, for robustness checks. PCA 1 restricts to operational indicators that measure the extensive margin of informality (Ulyssea, 2018, 2020) accounting for aspects relating to business license, VAT registration, registration under company law.7 PCA 2 focuses on intensive and direct measures of informality (Ulyssea, 2018, 2020) pertaining to a firm’s contributions to social insurance schemes and holding a bank account, which may likely suffer less from misreporting issues. Table B2 shows the tetrachoric correlations, eigenvalues, and factor loadings for the main index and the two subindices. Of the total variance of the main Tetrachoric PCA composite 6 As we cannot track transitions into and out of informality, we are implicitly assuming informality is time-invariant in our analysis. 7 These dimensions are, initially, more sensitive when asked directly and, therefore, more subject to misreporting as they hold a veil of illegality. 7 measure (PCA 5), 57.52% is explained by the extensive/operational measures of informality (no license, no registration, no VAT registration number). The remaining 42.48% is explained by the two intensive proxies of informality (contributions to employees’ social security and bank account). 3.2.2 Types of Informal Firms: Necessity, Parasite, and De Soto Informal firms can either be born out of necessity with a survival motive; are parasitic such that they choose to remain informal to evade costs; or are high-potential but per the De Soto view remain informal due to burdensome regulations. The type of informality is correlated with the initial productivity of firms when they enter the market and is contingent on the human capital of the entrepreneurs (Ulyssea, 2018). We use the pre-pandemic productivity measure to analyze the distributions across formal and informal firms. To this end, we take the following steps: First, we regress productivity on firm attributes such as age and other initial characteristics, including size, sector, whether the firm is female-led, and country fixed effects: log (sales/worker)it0 = β0 + β1 ageit0 + β2 age2 it0 + β3 f emalei + γs + ϕc + νj + uit (1) Where γs are sector fixed effects; ϕc are country fixed effects; and νj are size effects, where j =small, medium, and large. Second, using the residual of this regression, that is, unexplained labor productivity, we calculate the 25th and the 75th percentiles of the distribution of this conditional productivity by country and sector for fully formal firms - on both extensive and intensive margins - as measured by PCA(5). In the last step, we map the distribution of informal firms to one of the three groups: below 25th percentile; between 25th and 75th percentile, and; above 75th percentile. We expect that informal firms below the 25th percentile threshold should be more similar to necessity or survival firms as they have very low productivity compared with formal firms of similar characteristics. For the other two types of informal firms, the distinction is more difficult with our dataset. For example, in contrast with Ulyssea (2018), we do not have the pre-entry productivity signal but labor productivity before the pandemic, implying that the firm has already internalized the cost-benefits trade-offs and decided to remain informal. Given that the costs relative to productivity benefits are lower for parasite firms (compared to De Soto), we expect firms between the bottom and top quartile of the productivity distribution would 8 fit the De Soto view, where firms have some potential to grow, yet due to burdensome regulations, they are not able to join the formal sector. Firms above the 75th percentile threshold could be classified as "parasite" firms as, conditional on their characteristics, they have a level of productivity comparable to or even higher than the top quartile of formal firms but they still remain in the informal sector.8 Under this framework, we can classify all firms in our sample into four categories: (1) Informal-necessity, (2) Informal-De Soto view, (3) Informal-parasite, and (4) Fully formal. Figure 1, shows the kernel density of this conditional productivity for the four types of firms in our sample, using PCA(5) as the informality index. As expected, the informal firm’s productivity distribution has a heavy left tail compared to formal firms that have a thicker right tail. Yet, there is a huge overlap in the productivity distribution across these firms, suggesting that the reasons for staying informal at the lower end of the distribution might be very different than at the higher end.9 This also points to the possibility that the shock may not affect all informal firms uniformly. Figure 1: Distribution of conditional productivity across types of informality .4 .3 Kernel density .2 .1 0 -5 0 5 10 IHS transformation labor productivity conditional on size, sector, age, female ownership, and country Fully formal Necessity De Soto Parasite Informal (all types) Source: Authors’ calculations using data from the World Bank Business Pulse Survey (BPS). 8 Arguably, our classification hinges on the arbitrary quartile definition, but, in general, firms classified under these three categories are at least similar to the informality view they intend to represent. We also tested a stricter threshold with necessity firms below the minimum of the conditional productivity, and the results (available upon request) do not change much, but standard errors increase significantly due to the smaller number of observations in the necessity category of informality. 9 By construction, we observe more overlap between the distribution of De Soto type of firms and the distribution of formal firms, while the density for necessity firms is shifted to the left. The classification is constructed at the country-sector level, and this figure pools all countries and sectors together. 9 3.3 Estimation Methods It is widely known that informal firms are small, more likely present in certain services sectors with a higher level of participation of women in these sectors, among others. There- fore, we correlate the adverse effect of the pandemic (or the adjustments response of firms) with informality by itself once we control for these firm-level characteristics. We estimate the model shown in equation 2: yit = β0 + β1 Inf ormality + βj Xit + uit (2) The main controls included in Xit are indicator variables to identify microenterprises (less than nine employees), age, sectors, and a female-leadership. We also control for country and time-fixed effects, the severity of the lockdown (using the Google mobility indicator), and a dummy variable for whether the firm is a panel firm. The latter controls for any recall bias in retrospective variables and differences in new firms’ characteristics. Our main outcome variable is the change in sales since the start of the pandemic which measures the size of the shock.10 We implement a Shorrocks-Shapley decomposition for the change in sales to analyze the relative impact of informality against other firm-level characteristics.11 To understand the heterogeneity in the impact of the shock and the adjustment responses among informal firms, we estimate the model in equation 3: yit = β0 + β1 Inf ormalityi + βj Xit + γk Inf ormality i ∗ Zit + uit (3) Zit is each of the different groups of variables included in vector Xit taken separately. Finally, we estimate split regressions, to allow for different coefficients for different types of firms (necessity, De Soto, parasite, and fully formal).12 10 In addition to the change in sales, we also consider other outcome variables such as the firm’s adjustment response to the pandemic through technology adoption and innovation in products and services. 11 Such decomposition has been widely used to understand inequality and poverty measures (See Shorrocks and Shorrocks, 1999; Shorrocks, 2013; Sastre and Trannoy, 2002). 12 To analyze the statistical significance of the differences between coefficients, we also estimate a fully- interacted model of the types of informality categorical variable, which is equivalent to the split regressions setting. 10 4. Descriptive Statistics The descriptive statistics shown in Table B5 suggest that 42% of the firms are microen- terprises and that the service sector comprises 43% of the observations. The hospitality subsector, which was one of the hardest hit sectors in the pandemic, accounts for almost one-third of the firms in the service sector. Among the firms, 40% are young (less than five years), 13% are maturing (between five and fourteen years), and 47% are what we call established (with fifteen years or more in the market). Only about one-fifth of the firms in our sample are female-led. The primary outcome of interest, the change in sales compared to the value before the pandemic, has been previously studied in (Apedo-Amah et al., 2020; Cirera et al., 2021). The three rounds of data show that, on average, and without controlling for any other factor, firms exhibit a decrease of 29% percentage points against their pre-pandemic level. This value, of course, includes periods of recovery. We also analyze recovery for the last two rounds, measured as ∆Change in salest,t−1 . Though recovery is, on average, around 11 percentage points, the median is still zero, indicating that half of the firms have not recovered and that the dispersion of recovery is very high with the standard deviation being around four times the mean, which underscores the need to analyze sources of heterogeneity in recovery after shocks. In addition to the firm outcomes affected by the shock, we also consider variables related to the adjustment response by firms (e.g., use of digital technologies) and access to public support. About 50% of the firms started or increased their use of digital technologies, and 61% report having online sales. Only 22% of firms report having received support from the government, which varied widely across the three rounds.13 5. Results 5.1 Informal firms tend to be smaller, in sectors with low entry costs (retail and services) and also younger Informality is highly correlated with small businesses (see, for example, La Porta and Shleifer, 2014; Ulyssea, 2018, 2020). The BPS data for South Asia shows that the share of microenterprises (fewer than nine employees) in the formal sector, after controlling for firm characteristics such as sector, country, time, and age, is 38%. This contrasts sharply 13 During the first round, on average, only 5% of the firms received support, while for the second round, support increased to around 20% of the firms. 11 with the share in informal sector and, in particular, for parasite firms, which have a share of 80% (Figure 2, panel a).14 A quarter of formal firms are in manufacturing (25%), while informal firms have a participation of more than 83% in the services and retail sectors (Figure 2, panel b). Informal firms also tend to be relatively younger than formal firms perhaps because of faster churning in the informal sector (Figure 2, panel c), while nearly half of the firms in the formal sector are established (with more than 15 years in operations). This result is consistent with the recent findings in Vietnam (Mccaig and Pavcnik, 2021) where entry and exit are relatively higher in the informal sector compared to that observed in the formal sector. Given that the attributes of our sample from BPS data are in line with documented evidence about informal firms we are confident that our data are suitable for analysis on informality. Figure 2: Characteristics of firms by type of informality (a) Size (b) Sector 100 100 Manufacturing Share of microenterprises (less than 9 employees) 90 90 Retail 80 Other services 80 78 80 80 72 73 73 Share of firms in each sector) 69 70 70 60 60 50 50 40 38 40 30 30 25 20 20 17 17 10 12 10 10 10 3 3 0 0 Necessity De Soto Parasite Fully formal Necessity De Soto Parasite Fully formal (c) Age 100 Young 90 Maturing Established 80 Share of firms in each sector) 70 60 50 47 41 40 37 36 38 34 32 34 30 30 27 25 19 20 10 0 Necessity De Soto Parasite Fully formal Notes: Controlling by wave, country, severity of the crises and other firm-level characteristics, except the one depicted in each panel. Source: Authors’ calculations using data from the World Bank Business Pulse Survey (BPS). 14 This is consistent with what we would expect for this type of firm, as these are relatively productive firms that tend to stay under the radar to take advantage of the lower costs of informality. The results do not change much if we restrict our sample only to panel firms (results available upon request). 12 5.2 Informal firms have lower productivity, with consistently lower levels of manage- ment and technological capabilities across all types of informal firms Consistent with the finding in the literature that informal firms are highly inefficient (La Porta and Shleifer, 2014), the BPS data for South Asia shows that informal firms are largely present in the first two terciles of pre-pandemic labor productivity (Figure 3, panel a)15 and that there is a clear negative relationship between the degree of informality and productivity (Figure 3, panel b).16 Figure 3: Informality according to normalized polychoric PCA 5 vs. labor productivity (a) Proportion of formal and informal by tercile of labor (b) Labor productivity vs. Informality productivity .4 Formal .4 Informal Labor productivity standardized at country level .3 .2 Proportion of firms .2 0 .1 -.2 0 -.4 1 2 3 -.2 0 .2 .4 .6 .8 Terciles of initial labor productivity PCA Informality Notes: Panel (a) classifies as informal firms those with a value above the median in our continuous measure of informality (PCA 5) Panel (b) controls for size, sector, age, country, and severity of the crisis. Source: Authors’ calculations using data from the World Bank Business Pulse Survey (BPS). These results hold consistently for other predictors of firms’ total factor productivity, such as managerial capabilities (see Bloom et al., 2019, 2022; Grover and Torre, 2019) and tech- nological readiness.17 Management and digital readiness are negatively correlated with the degree of informality (Figure 4, panels a and b), and especially so for parasite firms. While fully formal firms apply, on average, 54% of structured management practices, parasite firms implement around one-third (4, panel c). The results for digital readiness are very similar (Figure 4, panel d) in line with the management literature confirming that managerial 15 Informal firms are defined as those with above median value of PCA 5. 16 These results are robust to using our other measures of informality, as shown in Figures A2 and A3. 17 The management score is computed using three dichotomous variables that indicate whether the firm monitors targets at least once a month, advertises the business at least once every six months, and promotes staff promotions based solely on performance, in a way comparable to (for example, Bloom et al., 2019). While we present the results using PCA 5 as a measure of informality, they are robust to using other alternative measures (PCA 1 and PCA 2, as shown in Figures A4 and A5 of the Online Appendix. Technological readiness is computed based on the pre-pandemic period status of firms’ use of digital payments solutions, social media, or big data analytics for marketing and Enterprise Resource Planning (ERP) for business administration and operations. 13 capabilities are complementary factors for other investments, such as the adoption and use of Information Technologies (IT) (see, for example Bloom et al., 2012, 2016).18 Figure 4: Informality according to normalized polychoric PCA 5 vs. Management practices and digital readiness (a) Management vs. Informality index (b) Digital readiness vs. Informality .7 2 1.8 .6 Normalized Management score Initial Digital Readiness Score 1.6 .5 1.4 .4 1.2 1 .3 -.2 0 .2 .4 .6 .8 -.2 0 .2 .4 .6 .8 PCA Informality PCA Informality (c) Management score by informality type (d) Digital readiness by informality type 0.7 1.5 1.4 1.34 0.6 1.3 0.54 1.2 Mean digital readiness score 1.1 Mean management score 0.5 1.0 0.40 0.88 0.9 0.4 0.37 0.78 0.33 0.8 0.69 0.7 0.3 0.6 0.5 0.2 0.4 0.3 0.1 0.2 0.1 0.0 0.0 Necessity De Soto Parasite Fully formal Necessity De Soto Parasite Fully formal Notes: Both panels control for size, sector, age, and country. Source: Authors’ calculations using data from the World Bank Business Pulse Survey (BPS). In line with the extant literature on management practices (see Bloom et al., 2019, 2022; Grover and Torre, 2019), we also observe a positive relationship between management and firm size (Figure 5, panel a). That is, the selection mechanism through which more capable firms survive and grow is at work; however, informal firms exhibit a weaker size- management relationship indicating that there is some degree of misallocation perpetuated due to informality, particularly by parasite firms where this relationship is the weakest. A similar pattern is also observed with respect to digital readiness (Figure 5, panel b). This suggests that the faster churning in the informal sector is not necessarily linked to productivity improvements, as also evidenced in Vietnam (Mccaig and Pavcnik, 2021). 18 In terms of educational distribution, fully formal firms exhibit a much higher proportion of managers with a college degree compared to informal firms (Figure A6). 14 Figure 5: Size vs. firm-level capabilities (a) Size vs. management (b) Digital readiness vs. management 3 3 ln(number of employees) ln(number of employees) 2.5 2.5 2 2 1.5 1.5 -.5 0 .5 1 1.5 -1 0 1 2 3 4 Normalized management score Initial Digital readiness score Necessity De Soto Parasite Fully formal Necessity De Soto Parasite Fully formal Notes: Both panels control for sector, age, and country. Source: Authors’ calculations using data from the World Bank Business Pulse Survey (BPS). 5.3 The severity of the impact of the COVID-19 shock and the recovery speed is nega- tively associated with the degree of informality Using drop in sales across the three waves of data collection as our main outcome variable, we document that sales decreased significantly more for informal firms (an average 45% drop in sales during the first wave compared to 36% for formal firms, controlling for firm- level attributes) (Figure 6, panel a) and remain significantly below their pre-pandemic sales (17% below). These results are comparable across the various types of informal firms, with the necessity firms being hit slightly harder. Nevertheless, there is a tight negative correlation between the degree of informality and the change in sales, pooling the three waves together (Figure 6, panel b). Similar results are obtained with respect to recovery across waves (∆Change in salest,t−1 ), that is, as the degree of informality increases, recovery is slower (Figure A7). The external attributes of firms that are associated with informality are also the ones that have been negatively associated with firm performance during the pandemic. For example, micro and small firms witnessed a larger drop their sales and a slower recovery (Cirera et al., 2021). Given that when data are limited such attributes are used as proxy for informality, (e.g., Perry and Maloney (2007) justifies using firm size to indicate informality because "most microenterprises are informal"), we test if the poor performance of firms during crisis is explained by factors in addition to these attributes. To this end, we implement the following approaches: Weighted cumulative distribution function: Specifically for firm size, the differences in cumula- tive distribution function of change in sales between formal and informal firms persist even 15 Figure 6: Change in sales by informality (a) Change in sales vs. 2019 according to the type of (b) Change in sales vs. informality informality 0 -20 -10 -8 -25 -20 -17 -17 -20 Change in sales -30 -30 -40 -36 -38 -35 -45 -44 -50 -47 -46 -46 -49 -60 -40 Fully formal Necessity De Soto Parasite -.2 0 .2 .4 .6 .8 Round 1 Round 2 Round 3 Informality PCA score Notes: Controls for size, sector, age, country, gender of the owner, and severity of the crisis. Source: Authors’ calculations using data from the World Bank Business Pulse Survey (BPS). when weighted by size (Figure A8), implying that there is still a role for informality as an additional factor explaining firm performance during crisis. Regression framework : Results from Table 1, suggest that even after external firm attributes such as size, sector, age, and gender of the owner, fully informal firms (a value of one on the informality index) suffered ten percentage points higher decline in sales relative to formal firms (column 1). If we further condition by initial labor productivity (column 2), the magnitude does not change much. This indicates that even though an important part of the performance of firms can indeed be explained by their characteristics, as the coefficient on informality reduces significantly from the unconditional (-36%) to the conditional specification, there is still a role for informality by itself.19 Within the types of informal firms, those driven by necessity have worse performance, an average four-percentage points difference against parasite firms (columns 3 and 4). Recovery exhibits a similar pattern, with fully informal firms showing a much worse performance, even after controlling for initial productivity (columns 5-8). To understand the role of each firm attribute, we extend our regression framework to also include interactions of these traits with informality, one at a time. The results from our split regressions, presented in Table B7, suggest that female-led informal firms perform relatively worse.20 19 The results for the unconditional specification are not shown here but are available upon request. 20 In the fully-interacted specifications, the coefficients of female-led are negative for necessity and De Soto firms, while for parasite firms, the coefficient associated with female-led is positive but not statistically significant (Table B7). 16 Shapley decomposition: Decomposing change in sales to attributable factors, such decomposi- tions show that informality accounted for 33% of the variation in the change in sales in the first round and 56% in the second round (Figure 7). As economies began to recover by the third round, the role of informality in explaining the drop in sales decreased but was still larger than firm size, sector and other firm observables. Oaxaca-Blinder decomposition: Conditioning on the performance of firms, our results from Oaxaca-Blinder decomposition suggest that though the endowments (firm attributes) play an important role, there is still a non-negligible effect of being informal and of the interaction between informality and these attributes (Table 3). 17 Table 1: Change in monthly sales relative to 2019 and recovery Change in sales Recovery (1) (2) (3) (4) (5) (6) (7) (8) Conditional Cont. initial Conditional Cont. initial Conditional Cont. initial Conditional Cont. initial productivity productivity productivity productivity Informality -10.08** -10.41** -5.317** -6.058** (1.702) (2.029) (2.349) (2.924) Necessity -11.17** -9.289** -6.397** -7.103** (1.347) (1.432) (1.580) (1.703) De Soto -8.770** -7.929** -4.855** -5.198** (1.151) (1.231) (1.319) (1.429) Parasite -8.428** -5.553** -6.577** -2.056 (1.208) (1.691) (1.343) (1.960) Fully formal Baseline Baseline Baselne Baselne log(Initial labor productivity) 0.430** 0.324** 0.0223 -0.100 (0.110) (0.114) (0.110) (0.111) Lag change in sales -0.724** -0.754** -0.724** -0.756** (0.0130) (0.0152) (0.0127) (0.0150) 18 Panel controls Yes Yes Yes Yes Yes Yes Yes Yes Other controls Yes Yes Yes Yes Yes Yes Yes Yes Initial productivity No Yes No Yes No Yes No Yes Lag change in sales No No No No Yes Yes Yes Yes Observations 19569 15190 19984 15365 10203 8100 10417 8191 Notes: All continuous variables are winsorized at the 1% and 99%. Recovery is calculated as the difference in the changes in sales across rounds of data collection. Robust standard errors in parentheses * p<0.10, ** p<0.05, ** p<0.01 Source: Authors’ calculations using data from the World Bank Business Pulse Survey (BPS). Table 2: Change in monthly sales relative to 2019 and recovery-Heterogeneity Change in sales Recovery (1) (2) (3) (4) (5) (6) (7) (8) Size Age Female-led Subsector Size Age Female-led Subsector Informality -18.20** -11.70** -8.740** -13.49** -9.919** -9.339** -4.412* -8.810** (2.637) (2.145) (1.760) (2.316) (3.263) (2.873) (2.452) (3.083) Micro (less than 9 employees) × Informality 11.65** 6.759* (2.832) (3.479) Young × Informality 8.130** 7.172 (3.917) (5.274) Established × Informality 0.239 6.652** (2.345) (3.090) Female-led × Informality -7.718** -5.479 (2.917) (4.074) Const and utilities × Informality 12.36** 2.728 (6.116) (8.221) Retail and wholesale × Informality 14.87** 12.49** (3.075) (3.931) Transp and storage × Informality 0.457 -0.923 (6.366) (9.204) Accomm × Informality -7.319 -6.129 19 (6.396) (8.865) Food prep serv × Informality 5.216 6.367 (4.751) (5.479) ICT × Informality 12.02 3.710 (8.789) (12.69) Financial serv × Informality 1.373 14.24 (12.11) (28.55) Education × Informality -12.45 -19.75 (12.56) (23.37) Health × Informality -16.60* -21.10 (9.896) (14.19) Other serv × Informality 1.567 5.053 (3.242) (4.438) Panel controls Yes Yes Yes Yes Yes Yes Yes Yes Other controls Yes Yes Yes Yes Yes Yes Yes Yes Initial productivity No No No No No No No No Lag change in sales No No No No Yes Yes Yes Yes Observations 19569 19569 19569 19569 10203 10203 10203 10203 Notes: Each regression also includes a control for the variable interacted with the informality measure. All continuous variables are winsorized at the 1% and 99%. Recovery is calculated as the difference in the changes in sales across rounds of data collection. Robust standard errors in parentheses * p<0.10, ** p<0.05, ** p<0.01 Source: Authors’ calculations using data from the World Bank Business Pulse Survey (BPS). Figure 7: Change in sales: Relative Shapley decomposition 59.7 58.6 60 Relative Shapley Value (%) - Change in sales R2 explained by group 40 32.8 32.9 30 21.6 20 17.2 13.3 12.8 7.6 2.6 2.4 1 .9 1.4 1.5 1.3 2.2 .2 0 .1 0 Round 1 Round 2 Round 3 Informality Size Age Sector Female-led Severity of the crisis (weighted Google mobility index) Initial prod. Notes: The regressions also control for the country, but only the relative contribu- tion of firm-level characteristics, including informality, are presented here. Source: Authors’ calculations using data from the World Bank Business Pulse Survey (BPS). Table 3: Oaxaca Blinder decomposition Dependent variable: Change in sales (1) (2) Conditional Cont. initial productivity Differential Prediction formal -32.80*** -31.27*** (0.404) (0.446) Prediction informal -44.81*** -44.94*** (0.600) (0.747) Difference 12.01*** 13.67*** (0.724) (0.870) Decomposition Endowments 7.625*** 9.297*** (0.641) (0.762) Coefficients 2.535*** 2.499*** (0.772) (0.898) Interaction 1.845*** 1.875** (0.681) (0.780) Observations 13,364 10,066 Notes: All continuous variables are winsorized at the 1% and 99%. Robust standard errors in parentheses * p<0.10, ** p<0.05, ** p<0.01 Source: Authors’ calculations using data from the World Bank Business Pulse Survey (BPS). 5.4 The adjustment response of informal firms to the crisis is lower even after account- ing for their lower initial capabilities Although physical distancing imposed during the pandemic accelerated the adoption of digital technologies, it also disproportionately benefited the larger firms and widened the digital divide with smaller firms (Cirera et al., 2021; Avalos et al., 2023). Given the negative 20 correlation between firm size and informality, it is not surprising that the probability of starting or increasing the use of digital technologies is significantly higher for fully formal firms (on average, ten percentage points higher (Figure 8 panel a), even after controlling for firm-level characteristics and initial productivity. Similar results are observed for other digital outcomes such as the probability of having online sales, investing in digital technologies, and the share of employees working remotely (Figure 8 panels b-d). Parasite firms consistently lag behind in their adjustment response relative to the other two types of informal firms. Figure 8: Digitalization (a) Started or increased use of digital (b) Has online sales 60.0 60.0 54.82 Probability of increasing or starting using digital 50.0 50.0 47.77 48.48 46.73 Probability of having online sales 42.37 40.0 40.0 34.46 35.83 34.32 30.0 30.0 20.0 20.0 10.0 10.0 0.0 0.0 Necessity De Soto Parasite Fully formal Necessity De Soto Parasite Fully formal (c) Has invested in digital technologies (d) Share of employees working remotely 40.0 14.0 12.0 10.97 Probability of investing in digital 30.0 28.13 Share of remote workers 10.0 9.71 8.72 7.94 8.0 20.0 18.25 17.52 15.15 6.0 10.0 4.0 2.0 0.0 0.0 Necessity De Soto Parasite Fully formal Necessity De Soto Parasite Fully formal Notes: Controls for size, sector, age, country, gender of the owner, and severity of the crisis. Source: Authors’ calculations using data from the World Bank Business Pulse Survey (BPS). The descriptive finding above is also confirmed in a regression setting using PCA 5, as the continuous measure for informality (Table 4). Conditional on external attributes (column 1) and initial digital readiness (column 2), the probability of starting to use or increasing the use of digital is substantially lower for informal firms. Yet, there are no significant differences when analyzing heterogeneity by different views of informality (columns 3 and 4) in the use of digital technologies but for other measures of adjustment response (e.g., online sales and investment in digital technologies), parasite firms perform worse. 21 Table 4: Digitalization (1) (2) (3) (4) (5) (6) (7) (8) Conditional Cont. initial Conditional Cont. initial Conditional Cont. initial Conditional Cont. initial digital digital digital digital Panel A Started using or increased digital Has online sales Informality -0.292** -0.215** -0.242** -0.166** (0.0194) (0.0226) (0.0251) (0.0323) Necessity -0.111** -0.0667** -0.0657** -0.0267 (0.0145) (0.0154) (0.0163) (0.0182) De Soto -0.0979** -0.0592** -0.0591** -0.0255* (0.0124) (0.0130) (0.0136) (0.0154) Parasite -0.112** -0.0602** -0.116** -0.0792** (0.0125) (0.0132) (0.0149) (0.0169) Fully formal Baseline Baseline Baseline Baseline Initial digital readiness score 0.0981** 0.102** 0.0600** 0.0617** (0.00408) (0.00407) (0.00515) (0.00514) Observations 14037 11870 14324 12138 14124 11969 14296 12122 Panel B Invested in digital technologies Share of employees working remotely (%) 22 Informality -0.240** -0.230** -3.819** -4.523** (0.0203) (0.0256) (1.235) (1.490) Necessity -0.0875** -0.0744** -2.247** -2.471** (0.0134) (0.0146) (0.980) (1.077) De Soto -0.0950** -0.0799** -3.027** -3.960** (0.0114) (0.0124) (0.822) (0.912) Parasite -0.121** -0.0863** -1.253 -0.938 (0.0119) (0.0130) (0.843) (0.937) Fully formal Baseline Baseline Baseline Baseline Initial digital readiness score 0.0571** 0.0591** 0.0806 0.166 (0.00397) (0.00394) (0.276) (0.277) Observations 14214 12047 14484 12298 19033 16869 19432 17249 Panel controls Yes Yes Yes Yes Yes Yes Yes Yes Other controls Yes Yes Yes Yes Yes Yes Yes Yes Initial productivity No Yes No Yes No Yes No Yes Change in sales No No No No Yes Yes Yes Yes Notes: All continuous variables are winsorized at the 1% and 99%. Robust standard errors in parentheses * p<0.10, ** p<0.05, ** p<0.01 Source: Authors’ calculations using data from the World Bank Business Pulse Survey (BPS). Table 5: Digitalization heterogeneity (1) (2) (3) (4) (5) (6) (7) (8) Size Age Female-led Subsector Size Age Female-led Subsector Panel A Started using or increased digital Has online sales Informality -0.335** -0.254** -0.306** -0.358** -0.251** -0.228** -0.259** -0.258** (0.0323) (0.0259) (0.0206) (0.0306) (0.0400) (0.0339) (0.0261) (0.0380) Micro (less than 9 employees) × Informality 0.0658* 0.0134 (0.0360) (0.0437) Young × Informality 0.0244 0.0524 (0.0451) (0.0536) Established × Informality -0.110** -0.0662* (0.0324) (0.0400) Female-led × Informality 0.0870** 0.102** (0.0406) (0.0480) Retail × Informality 0.110** -0.0471 (0.0433) (0.0588) Other Services × Informality 0.101** 0.0431 (0.0356) (0.0428) Observations 14037 14037 14037 14037 14124 14124 14124 14124 Panel B 23 Invested in digital technologies Share of employees working remotely (%) Informality -0.283** -0.238** -0.231** -0.268** -5.071** -3.655** -4.251** -1.373 (0.0329) (0.0270) (0.0211) (0.0317) (1.978) (1.507) (1.244) (1.739) Micro (less than 9 employees) × Informality 0.0656* 1.797 (0.0361) (2.001) Young × Informality 0.0388 -1.772 (0.0449) (2.729) Established × Informality -0.0267 0.390 (0.0328) (1.569) Female-led × Informality -0.0691* 2.495 (0.0413) (2.198) Retail × Informality 0.0434 -2.716 (0.0447) (1.834) Other Services × Informality 0.0408 -4.529** (0.0356) (1.819) Observations 14214 14214 14214 14214 19033 19033 19033 19033 Panel controls Yes Yes Yes Yes Yes Yes Yes Yes Other controls Yes Yes Yes Yes Yes Yes Yes Yes Initial productivity No Yes No Yes No Yes No Yes Change in sales No No No No Yes Yes Yes Yes Notes: All continuous variables are winsorized at the 1% and 99%. Robust standard errors in parentheses * p<0.10, ** p<0.05, ** p<0.01 Source: Authors’ calculations using data from the World Bank Business Pulse Survey (BPS). Split regressions that interact firm attributes with informality (Table 5) consistently suggest that female-led informal businesses have a higher probability of starting or increasing their use of digital and selling online but are less likely to invest in digital technologies. These results are broadly observed in fully-interacted models as well (See Table B8). 5.5 Access to public support by informal firms is limited and not necessarily targeted to alleviate the most pressing internal constraint The COVID-19 pandemic prompted an unprecedented amount of government support where decisions regarding the scope, targeting, policy instruments had to be mobilized in a very short span of time. As the crisis unfolded, the support provided to firms increased, but with important country-level heterogeneity (Cirera et al., 2021). On average, and across the three rounds of the survey, about 22% of the firms in South Asia received public support (Table B5). Informal firms have nearly 8 percentage points lower probability of receiving public support than that observed for formal firms (Table 6, column 1), with all types of informal firms being equally hurt (column 3). The probability of accessing support appears to be uncorrelated with initial productivity but positively associated with the level of change in sales (Table 6, column 2).21 Both these factors are suggestive of mistargeting in the allocation of support.22 5.6 Robustness To test the robustness of our results, we conduct a large set of additional exercises. Restricted sample: Given that India has a much larger sample compared to the other countries in our sample, we re-estimated our specifications, excluding India. With this change, the magnitude of the coefficients is within the one percentage point range of the base specification, and our conclusions remain robust.23 Balanced panel: We test whether the unbalanced panel structure could be driving our results in a number of ways. First, we restrict our sample to firms that appear three times in our dataset (balanced panel), which reduces our sample size significantly. Second, we drop firms that appear only once in our sample. Finally, we analyze the most restrictive case in 21 These results are robust to alternative specifications such as the inclusion of panel firms only and so on. 22 Within the set of instruments used for supporting informal firms, the probability of receiving technical support is lower implying that the capabilities of informal firms are not addressed as part of the public programs even when these capabilities have been critical in coping with COVID-19 (Grover and Karplus, 2020). 23 The results are not shown here but are available upon request. 24 which we ignore the data collected from previous waves (pseudo-panel data) as it could entail some recall bias (we are asking for information about the first round in subsequent rounds). Though there is some small variation in terms of the magnitudes of the coefficients (between one and two percentage points in the case of change in sales), the qualitative interpretation and the statistical significance of the results do not change. Table 6: Probability of receiving support Dependent variable=1 if received support Conditional Cont. initial Conditional Cont. initial productivity productivity Informality -0.0977** -0.125** (0.0176) (0.0219) Necessity -0.0462** -0.0622** (0.0133) (0.0149) De Soto -0.0370** -0.0450** (0.0110) (0.0123) Parasite -0.0314** 0.00729 (0.0113) (0.0163) % change in sales during past 30 days 0.000586** 0.000697** 0.000574** 0.000675** (0.0000937) (0.000110) (0.0000925) (0.000109) Intial productivity -0.000408 -0.00198* (0.00104) (0.00109) Panel controls Yes Yes Yes Yes Other controls Yes Yes Yes Yes Observations 14078 11086 14368 11227 Notes: All continuous variables are winsorized at the 1% and 99%. Recovery is calculated as the difference in the changes in sales across rounds of data collection. Robust standard errors in parentheses * p<0.10, ** p<0.05, ** p<0.01 Source: Authors’ calculations using data from the World Bank Business Pulse Survey (BPS). Initial productivity thresholds for classifying informal firms: A third exercise related to our classification of firms according to the views of informality is related to the threshold in terms of the level of pre-pandemic productivity of formal firms (quartiles). We tested with alternative thresholds, for example, top and bottom 30% or top and bottom 20%. In all cases, our results remain robust, with the magnitude of the coefficients being within the one percentage point range of the base model. Alternative definition of informality: A final robustness check is related to using a sharper definition of informality following Ulyssea (2018) and Ulyssea (2020). To do this, we define an extensive margin as those firms that do not have a registration and do not pay taxes. We use an intensive margin definition identifying among those that are formal on the extensive margin, which ones do not provide social security for their workers or do not use a bank account. Our results using such dichotomous variables (instead of the ones from continuous measures or types of informality) show very similar magnitudes. 25 6. Conclusions Using three waves of Business Pulse Survey data collected in South Asia during the COVID- 19 crisis, we disentangle the impact of the shock on informal firms, their adjustment response, and the reasons underlying their fragility. We construct a measure of informality that captures its multidimensionality. In our analysis, we also incorporate the various views characterizing informality by comparing the pre-pandemic productivity distribution of these firms – necessity, De Soto, and parasite firms – with that of formal firms. These features allow us to test the heterogeneity in impact, recovery, and response to the pandemic by the degree of informality (our composite index) and the type or motivation for informality. Before delving into analyzing the impact of the crisis, we compare the stylized facts on informality using our survey data. In line with the existing literature, our data confirm that informal firms, as defined by our composite index, tend to be smaller in size, relatively younger, and more prevalent in sectors with low entry costs. Additionally, informal firms have lower labor productivity, possibly driven by poor management capabilities and digital readiness. Such lack of capabilities potentially makes informal firms more vulnerable to shocks even after accounting for external attributes. That is, informality by itself leads to worse outcomes and, in particular, for firms driven by necessity that remain subsistence. These factors also limit the adjustment response to shocks. For instance, the gap between formal and informal in terms of digital adoption is not only salient prior to the pandemic but also widening. Given our finding that size cannot proxy for informality due to the multi-dimensional nature of the problem, we propose considering the following aspects in supporting the private sector in the context of informality: Targeting – Who to support: In general, sector or firm selection can be distortionary; however, during shocks the efficiency of instruments can be most effective for the industry or group of firms that are most affected. Based on our analysis, we suggest that considering the severity and type of informality is critical in assessing the modality and instrument for support. For instance, although necessity firms are particularly vulnerable, support for these firms should consider approaches similar to those used for helping households (e.g., cash transfers). By comparison, the De Soto and Parasite firms lag behind in their adjustment response to the pandemic and may be eligible for support on digitization. The selection of sector or firm groups should be carefully evaluated after supply-chain analysis and focus on strengthening the weak linkages, where the failure of a few key firms in a value chain can have big repercussions for the rest of the economy. 26 Modality – How to support: First, crowding in the private sector is crucial to help firms during the crisis and especially when governments face tighter fiscal space. For instance, several Development Financial Institutions (DFIs) supported firms during the COVID-19 pandemic, including large and mid-sized ones. Support to large and mid-sized firms can be made conditional on passing on some benefits to their key suppliers who remain informal (mainly due to regulatory burdens). Second, scaling up access to finance for informal businesses could be critical, given the fragility and credit crunch during the crisis. Governments or Multilateral Development Institutions may be able to avoid direct targeting by ramping up credit through financial institutions. However, concessional finance to financial institutions by DFIs can be made conditional on supporting certain types of informal businesses. Instruments – What to support: Our results suggest that relative to the formal sector, informal firms have lower levels of management capabilities and lag behind in digital readiness, which makes them even more vulnerable during the shock. Such crises are an opportunity to reduce this capabilities gap. There are three possible market failures that hinder the adoption of technology among informal firms: (i) information gaps: firms do not have information on the possible technology/adjustment models (ii) capabilities failure: firms do not have the required skills to adopt technology or adjust business models, and (iii) credit market failure: financial institutions may not lend for upgrading programs as their value may not be known or intangible. Although informal firms face such gaps in capabilities, the support reaching them barely addresses these issues.24 24 Some examples of support in this intervention include skills training and providing access to inexpensive smartphones with a short-term subscription. Improving the capabilities of informal firms can enhance their resilience to shocks and potentially encourage formalization in the long run. 27 References L. Alfaro, O. Becerra, and M. Eslava. EMES and COVID-19: Shutting Down in a World of Informal and Tiny Firms. NBER Working Paper Series, 2020. URL http://www.nber.org/papers/w27360. M. C. Apedo-Amah, B. Avdiu, X. Cirera, M. Cruz, E. Davies, A. Grover, L. Iacovone, U. Kilinc, D. Medvedev, F. O. Maduko, S. Poupakis, J. Torres, and T. T. Tran. Unmasking the Impact of COVID- 19 on Businesses. Policy Research Working Paper; No. 9434., 10 2020. doi: 10.1596/1813-9450-9434. URL https://openknowledge.worldbank.org/handle/10986/34626. J. Arias, O. Azuara, P. Bernal, J. J. Heckman, and C. Villarreal. Policies to promote growth and economic efficiency in mexico. Technical report, National Bureau of Economic Research, 2010. E. Avalos, X. Cirera, M. Cruz, L. Iacovone, D. Medvedev, G. Nayyar, and S. R. Ortega. Firms’ Digitalization during the COVID-19 Pandemic. Policy Research Working Paper; No. 10284, 2023. A. W. Bartik, M. Bertrand, Z. Cullen, E. L. Glaeser, M. Luca, and C. Stanton. The impact of COVID-19 on small business outcomes and expectations. Proceedings of the National Academy of Sciences, 117 (30):17656–17666, 2020. N. Bloom, R. Sadun, and J. Van Reenen. Americans do IT better: US multinationals and the productivity miracle. American Economic Review, 102(1):167–201, 2012. N. Bloom, R. Sadun, and J. Van Reenen. Management as a technology? Technical report, Centre for Economic Performance Discussion Paper 1433, 2016. N. Bloom, E. Brynjolfsson, L. Foster, R. Jarmin, M. Patnaik, I. Saporta-Eksten, and J. V. Reenen. What drives differences in management practices? aeaweb.org, 109(5):1648–1683, 2019. doi: 10.1257/aer.20170491. URL https://www.aeaweb.org/articles?id=10.1257/aer.20170491. N. Bloom, L. Iacovone, M. Pereira-Lopez, J. V. Reenen, M. I. Mexico, and W. Bank. Management and misallocation in Mexico. NBER Working Paper Series, 2022. URL http://www.nber.org/papers/ w29717. A. Brucal, A. Grover, and S. R. Ortega. Damaged by the disaster. Policy Research Working Paper; No. 9604., 3 2021. doi: 10.1596/1813-9450-9604. URL https://openknowledge.worldbank.org/handle/ 10986/35357. M. Bussolo, S. Sharma, and H. Timmer. COVID-19 has worsened the woes of South Asia’s informal sector, 2020. URL https://blogs.worldbank.org/endpovertyinsouthasia/ covid-19-has-worsened-woes-south-asias-informal-sector. X. Cirera, M. Cruz, A. Grover, L. Iacovone, D. Medvedev, M. Pereira-Lopez, and S. Reyes. Firm Recovery during COVID-19. 10 2021. doi: 10.1596/1813-9450-9810. URL https://openknowledge. worldbank.org/handle/10986/36428. 28 E. Colombo, L. Menna, and P. Tirelli. Informality and the labor market effects of financial crises. World Development, 119:1–22, 2019. S. De Mel, D. McKenzie, and C. Woodruff. Who are the microenterprise owners? evidence from sri lanka on tokman versus de soto. In International differences in entrepreneurship, pages 63–87. University of Chicago Press, 2010. H. De Soto. The other path. Harper & Row New York, 1989. C. Elgin, M. A. Kose, F. Ohnsorge, and S. Yu. Understanding informality. 2021. C. Elgin, C. C. Williams, G. Oz-Yalaman, and A. Yalaman. Fiscal stimulus packages to covid-19: The role of informality. Journal of International Development, 34(4):861–879, 2022. A. Fernández and F. Meza. Informal employment and business cycles in emerging economies: The case of Mexico. Review of Economic Dynamics, 18(2):381–405, 2015. K. Gërxhani. The informal sector in developed and less developed countries: A literature survey. Public Choice, 120(3-4):267–300, 9 2004. ISSN 00485829. doi: 10.1023/B:PUCH.0000044287.88147. 5E. Google LLC. Google COVID-19 community mobility reports. URL https://www.google.com/ covid19/mobility. A. Grover and V. J. Karplus. The energy-management nexus in firms which practices matter, how much and for whom? Policy Research Working Paper No. 9397. World Bank, Washington, DC., 2020. A. Grover and I. Torre. Management Capabilities and Performance of Firms in the Russian Federation, Policy Research Working Paper No. 8996. World Bank, Washington, DC, 2019. T. Hale, N. Angrist, R. Goldszmidt, B. Kira, A. Petherick, T. Phillips, S. Webster, E. Cameron-Blake, L. Hallas, S. Majumdar, and H. Tatlow. A global panel database of pandemic policies (Oxford COVID-19 Government Response Tracker). Nature Human Behaviour, 5(4):529–538, 4 2021. ISSN 23973374. doi: 10.1038/S41562-021-01079-8. J. E. Humphries, C. Neilson, and G. Ulyssea. The evolving impacts of COVID-19 on small businesses since the CARES Act. 2020. R. La Porta and A. Shleifer. Informality and development. Journal of Economic Perspectives, 28(3): 109–126, 6 2014. doi: 10.1257/JEP.28.3.109. R. La Porta, A. Shleifer, C. Jones, P. Klenow, G. Mankiw, J. Rauch, J. Stein, and W. Nordhaus. The Unofficial Economy and Economic Development. 12 2008. doi: 10.3386/W14520. URL https://www.nber.org/papers/w14520. 29 S. Levy Algazi et al. Under-rewarded efforts: The elusive quest for prosperity in mexico. IDB Publications (Books), 2018. N. Loayza. Informality: why is it so widespread and how can it be reduced? World Bank Research and Policy Briefs, (133110), 2018. URL https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3360124. N. V. Loayza and J. Rigolini. Informal employment: safety net or growth engine? World Development, 39(9):1503–1515, 2011. B. Mccaig and N. Pavcnik. Entry and Exit of Informal Firms and Development. IMF Economic Review, 69:540–575, 2021. doi: 10.1057/s41308-021-00142-8. URL https://doi.org/10.1057/ s41308-021-00142-8. L. Medina and F. Schneider. Shedding light on the shadow economy: A global database and the interaction with the official one. Available at SSRN 3502028, 2019. L. Medina and M. F. Schneider. Shadow economies around the world: what did we learn over the last 20 years? International Monetary Fund, 2018. D. Medvedev and A. M. Oviedo. Informality and profitability: Evidence from a new firm survey in ecuador. The Journal of Development Studies, 52(3):412–427, 2016. F. Ohnsorge and S. Yu. The long shadow of informality: Challenges and policies. World Bank Publications, 2022. G. Perry and W. F. Maloney. Informality: Exit and exclusion. World Bank Publications, 2007. , š and A. Sauka. Measuring the shadow economy using company managers. Journal of T. J. Putnin Comparative Economics, 43(2):471–490, 5 2015. ISSN 0147-5967. doi: 10.1016/J.JCE.2014.04.001. B. Reilly, G. K. Panoeconomicus, and u. 2018. Shadow Economy-is an Enterprise Survey a Preferable Approach? panoeconomicus.org, 66:589–610, 2019. doi: 10.2298/PAN161108022R. URL http: //panoeconomicus.org/index.php/jorunal/article/view/429. P. Restrepo Echavarria. Measuring the Underground Economy is Difficult | St. Louis Fed, 1 2015. URL https://www.stlouisfed.org/publications/regional-economist/january-2015/ underground-economy. M. Sastre and A. Trannoy. Shapley inequality decomposition by factor components: Some methodological issues. Journal of Economics, 77(S1):51–89, 12 2002. ISSN 0931-8658. doi: 10.1007/BF03052500. A. F. Shorrocks. Decomposition procedures for distributional analysis: A unified framework based on the Shapley value. Journal of Economic Inequality, 11(1):99–126, 3 2013. ISSN 15738701. doi: 10.1007/S10888-011-9214-Z. 30 A. F. Shorrocks and A. F. Shorrocks. Decomposition Procedures for Distributional Analysis: A Unified Framework Based on the Shapley Value. 1999. URL http://citeseerx.ist.psu.edu/viewdoc/ summary?doi=10.1.1.471.2940. G. Ulyssea. Firms, Informality, and Development: Theory and Evidence from Brazil. American Economic Review, 108(8):2015–47, 8 2018. ISSN 0002-8282. doi: 10.1257/AER.20141745. G. Ulyssea. Informality: Causes and Consequences for Development. 2020. doi: 10.1146/ annurev-economics. URL https://doi.org/10.1146/annurev-economics-. 31 A. Online Appendix Figures Figure A1: Timing of the Business Pulse Survey SAR countries (a) Afghanistan (b) Bangladesh AFG BGD Weighted average in Gooble Mobility trends Weighted average in Gooble Mobility trends 300 100 80 80 200 60 lockdown measures lockdown measures Stringency of Stringency of (baseline=0) (baseline=0) 60 40 100 50 40 0 20 0 20 -50 -100 0 -100 0 01/01/20 07/01/20 01/01/21 07/01/21 01/01/22 07/01/22 01/01/20 07/01/20 01/01/21 07/01/21 01/01/22 07/01/22 Round 1 Round 2 Round 1 Round 2 Round 3 Stringency of Stringency of Weighted average in lockdown Weighted average in lockdown Google Mobility trends Google Mobility trends restrictions restrictions Weighted average is constructed using 30 day-periods of Google mobility around transit stations. Weighted average is constructed using 30 day-periods of Google mobility around transit stations. Weight of past 30 days is equal to 1, weight of period 2 (days 31-60) is equal to 1/2, so on and so forth. Weight of past 30 days is equal to 1, weight of period 2 (days 31-60) is equal to 1/2, so on and so forth. (c) India (d) Nepal IND NPL Weighted average in Gooble Mobility trends Weighted average in Gooble Mobility trends 100 300 100 80 80 200 lockdown measures lockdown measures 150 Stringency of Stringency of (baseline=0) (baseline=0) 60 60 100 100 50 40 40 0 0 20 20 -50 -100 0 -100 0 01/01/20 07/01/20 01/01/21 07/01/21 01/01/22 07/01/22 01/01/20 07/01/20 01/01/21 07/01/21 01/01/22 07/01/22 Round 1 Round 2 Round 3 Round 1 Round 2 Round 3 Stringency of Stringency of Weighted average in lockdown Weighted average in lockdown Google Mobility trends restrictions Google Mobility trends restrictions Weighted average is constructed using 30 day-periods of Google mobility around transit stations. Weighted average is constructed using 30 day-periods of Google mobility around transit stations. Weight of past 30 days is equal to 1, weight of period 2 (days 31-60) is equal to 1/2, so on and so forth. Weight of past 30 days is equal to 1, weight of period 2 (days 31-60) is equal to 1/2, so on and so forth. (e) Pakistan (f) Sri Lanka PAK LKA Weighted average in Gooble Mobility trends Weighted average in Gooble Mobility trends 100 100 80 80 200 lockdown measures lockdown measures Stringency of Stringency of (baseline=0) (baseline=0) 60 60 100 50 40 40 0 0 -50 20 20 -100 -100 0 -150 0 01/01/20 07/01/20 01/01/21 07/01/21 01/01/22 07/01/22 01/01/20 07/01/20 01/01/21 07/01/21 01/01/22 07/01/22 Round 1 Round 2 Round 3 Round 1 Round 2 Weighted average in Stringency of Weighted average in Stringency of Google Mobility trends lockdown Google Mobility trends lockdown restrictions restrictions Weighted average is constructed using 30 day-periods of Google mobility around transit stations. Weighted average is constructed using 30 day-periods of Google mobility around transit stations. Weight of past 30 days is equal to 1, weight of period 2 (days 31-60) is equal to 1/2, so on and so forth. Weight of past 30 days is equal to 1, weight of period 2 (days 31-60) is equal to 1/2, so on and so forth. Notes: The figure shows the timing of the survey along with the trends in terms of the stringency index of lockdown restrictions from the Oxford Government Response Tracker, where a higher score means stricter policies (See Hale et al., 2021), and the Mobility Index based on Google mobility reports around transit stations (Google LLC). 32 Figure A2: Informality according to normalized polychoric PCA 1 vs. labor productivity (a) Proportion of formal and informal by tercile of labor productivity (b) Labor productivity vs. Informality .4 .4 Labor productivity standardized at country level .3 .2 Proportion of firms .2 0 .1 -.2 0 1 2 3 Terciles of initial labor productivity -.4 Formal Informal -.5 0 .5 1 PCA 2 Notes: Panel (b) controls for size, sector, age, country, and severity of the crisis. Figure A3: Informality according to normalized polychoric PCA 2 vs. labor productivity (a) Proportion of formal and informal by tercile of labor productivity (b) Labor productivity vs. Informality .4 .4 Labor productivity standardized at country level .3 Proportion of firms .2 .2 0 .1 0 -.2 1 2 3 Terciles of initial labor productivity Formal Informal -.2 0 .2 .4 .6 .8 PCA 1 Notes: Panel (b) controls for size, sector, age, country, and severity of the crisis. Figure A4: Informality according to normalized polychoric PCA 1 vs.capabilities (a) Management vs. Informality (b) Digital readiness vs. Informality 2 2 1.8 1.8 Digital readiness score Management score 1.6 1.6 1.4 1.4 1.2 1.2 -.5 0 .5 1 -.5 0 .5 1 PCA 2 PCA 2 Notes: controls for size, sector, age, country, and severity of the crisis. 33 Figure A5: Informality according to normalized polychoric PCA 2 vs.capabilities (a) Management vs. Informality (b) Digital readiness vs. Informality 2 2 1.5 Digital readiness score Management score 1.5 1 1 .5 -.5 0 .5 1 -.5 0 .5 1 PCA 1 PCA 1 Notes: controls for size, sector, age, country, and severity of the crisis. Figure A6: Educational level by informality type (a) All firms (b) By gender Male-led Female-led 1 Less than high-school 0.34 0.40 0.35 0.85 1 1 High-school 0.34 0.42 0.39 0.86 0.35 0.32 0.26 0.85 College or more .8 .8 .8 0.23 0.20 0.22 0.27 .6 0.20 0.21 0.23 .6 .6 Share of firms Share of firms 0.21 0.23 0.50 0.46 .4 0.43 0.46 0.45 .4 .4 0.40 0.41 0.36 0.35 .2 .2 .2 0.11 0.11 0.10 0 0.04 0.05 Necessity De Soto Parasite Fully formal 0 Necessity De Soto 0.03 Parasite Fully formal 0 Necessity De Soto Parasite Fully formal Notes: Unconditional distribution. Figure A7: Recovery change in sales by informality (a) Recovery according to the type of informality (b) Recovery vs. informality 35 33 25 30 29 27 27 25 20 Recovery change in sales 20 Recovery change in sales 15 15 10 5 3 0 10 -5 -2 -3 -3 -10 5 -15 -20 Fully formal Necessity De Soto Parasite 0 -.2 0 .2 .4 .6 .8 Round 2 Round 3 Informality PCA score Notes: Controls for size, sector, age, country, gender of the owner, and severity of the crisis. Source: Authors’ calculations using data from the World Bank Business Pulse Survey (BPS). 34 Figure A8: CDF % change in sales during past 30 days Unweighted Weighted by size 1 1 .8 .8 Cumulative Probability Cumulative Probability .6 .6 c.d.f. of Formal .4 .4 c.d.f. of Informal .2 .2 0 0 -100 -50 0 50 -100 -50 0 50 % change in sales during past 30 days % change in sales during past 30 days Unweighted Weighted by size 1 1 .8 .8 Cumulative Probability Cumulative Probability c.d.f. of Necessity .6 .6 c.d.f. of De Soto .4 .4 c.d.f. of Parasite c.d.f. of Fully formal .2 .2 0 0 -100 -50 0 50 -100 -50 0 50 % change in sales during past 30 days % change in sales during past 30 days Figure A9: Change in sales: Relative Shapley decomposition-without controlling for initial productivity 60.3 60 Relative Shapley Value (%) - Change in sales 52.7 40.7 R2 explained by group 40 35.8 34.8 19.1 19.4 20 15.2 7.9 3.7 3.7 1.6 2.4 .5 .1 .9 .6 .9 0 Round 1 Round 2 Round 3 Informality Size Age Sector Female-led Severity of the crisis (weighted Google mobility index) Notes: The regressions also control for the country, but only the relative contribu- tion of firm-level characteristics, including informality, are presented here. Source: Authors’ calculations using data from the World Bank Business Pulse Survey (BPS). 35 B. Online Appendix Tables Table B1: Countries included in the sample in each round Country Round 1 Round 2 Round 3 Total obs. Afghanistan 353 548 0 901 Bangladesh 823 923 994 2,740 India 681 3,071 3,084 6,836 Sri Lanka 481 1,008 0 1,489 Nepal 508 1,253 1,218 2,979 Pakistan 1,443 1,629 1,488 4,560 Total 4,289 8,432 6,784 19,505 *The agricultural sector is excluded from the sample as it wasn’t part of the sample frame in all countries. Source: Authors’ calculations using data from the World Bank Business Pulse Survey (BPS). B.1 Tetrachoric PCA calculations Table B2: Informality Tetrachoric correlation matrices No business Not registered No contributions No VAT No bank license social security registration account PCA 5 No business license 1 Not registered 0.68013512 1 No contributions social security 0.58932312 0.58252556 1 No VAT registration 0.68132262 0.75081026 0.69021685 1 No bank account 0.43967786 0.29887387 0.69624241 0.50824859 1 PCA 1 No business license 1 Not registered .68417095 1 No VAT registration .68450466 .75205983 1 PCA 2 No contributions social security 1 No bank account .69703171 1 Table B3: Informality tetrachoric Eigenvalues k Eigenvalues Proportion explained Cum. explained PCA 5 1 3.386497 0.677299 0.677299 2 0.815763 0.163153 0.840452 3 0.354599 0.07092 0.911372 4 0.240837 0.048167 0.959539 5 0.202304 0.040461 1 PCA 1 1 2.414315 0.804772 0.804772 2 0.337746 0.112582 0.917353 3 0.247940 0.082647 1.000000 PCA 2 1 1.697032 0.848516 0.848516 2 0.302968 0.151484 1.000000 Table B4: Informality tetrachoric: First component loadings (squared values add to one) Variable Comp1 PCA 5 Comp1 PCA 1 Comp1 PCA 2 No business license 0.4508 0.5647 Not registered 0.4434 0.5835 No contributions social security 0.4704 0.7071 No VAT registration 0.485 0.5836 No bank account 0.3791 0.7071 36 Table B5: Descriptive Statistics Variable mean sd p10 p50 p90 N Firm characteristics Micro (9 employees or less) 0.423 0.494 0.000 0.000 1.000 20,160 Manufacturing 0.385 0.487 0.000 0.000 1.000 19,826 Retail 0.188 0.391 0.000 0.000 1.000 19,826 Services 0.427 0.495 0.000 0.000 1.000 19,826 Young 0.401 0.490 0.000 0.000 1.000 20,092 Maturing 0.127 0.333 0.000 0.000 1.000 20,092 Established 0.472 0.499 0.000 0.000 1.000 20,092 Female-led 0.211 0.408 0.000 0.000 1.000 20,160 Outcomes and variables of interest Change in sales vs. 2019 (%) -28.534 39.603 -90.000 -25.000 20.000 20,160 Recovery (∆Change in sales) 11.237 41.971 -35.000 0.000 65.000 10,510 =1 if uses digital 0.496 0.500 0.000 0.000 1.000 14,602 =1 if has online sales 0.612 0.487 0.000 1.000 1.000 14,468 =1 if has repackaged products 0.392 0.488 0.000 0.000 1.000 15,174 Share of remote workers (%) 9.806 24.478 0.000 0.000 40.000 19,601 Access to public support 0.218 0.413 0.000 0.000 1.000 8,122 Source: Authors’ calculations using data from the World Bank Business Pulse Survey (BPS). Table B6: Firm size vs. initial capabilities (1) (2) Independent variables Dependent variable= ln(employees) Management Digital capabilities Firm capabilities measure 0.390*** 0.0477* (0.0729) (0.0247) Firm capabilities*De Soto -0.0548 0.0230 (0.0923) (0.0322) Firm capabilities*Parasite -0.174* 0.0600* (0.101) (0.0346) Firm capabilities*Fully formal 0.167* 0.210*** (0.0915) (0.0348) N 17418 17418 Firm-level attributes Yes Yes Robust standard errors in parentheses * p<0.10, ** p<0.05, *** p<0.01 The model used includes interactions between the categories of informality and all other variables Source: Authors’ calculations using data from the World Bank Business Pulse Survey (BPS). 37 Table B7: Change in monthly sales relative to 2019 and recovery-Heterogeneity Change in sales Recovery (1) (2) (3) (4) (5) (6) (7) (8) Necessity De Soto Parasite Fully formal Necessity De Soto Parasite Fully formal Micro (less than 9 employees) -2.235 -5.791** -6.966** -4.196** -2.887 -5.585** -5.593** -2.808** (1.970) (1.420) (1.745) (1.426) (2.533) (1.883) (1.937) (1.344) Maturing Baseline Baseline Baseline Baseline Baseline Baseline Baseline Baseline Young 0.923 4.295** 2.131 1.106 0.576 5.248** -1.945 2.510 (2.578) (1.965) (1.747) (3.030) (3.363) (2.569) (2.082) (3.085) Established -2.765 -0.916 -0.841 -0.708 -0.660 1.718 0.557 -1.412* (1.978) (1.330) (1.466) (0.867) (2.569) (1.822) (1.745) (0.805) Female-led -2.015 -1.468 2.723 4.293** 0.949 -0.484 1.907 2.311** (2.440) (1.693) (1.683) (1.067) (3.139) (2.242) (2.113) (1.074) Manuf Baseline Baseline Baseline Baseline Baseline Baseline Baseline Baseline Const and utilities 0.843 -0.744 3.983 -5.432* -3.252 -2.278 -0.453 -4.121 (4.352) (3.671) (4.208) (2.928) (5.447) (5.346) (4.338) (2.958) Retail and wholesale 4.738** 3.820** 5.856** -2.391** 2.528 4.543* 7.537** -3.194** (2.404) (1.845) (2.057) (1.148) (2.926) (2.524) (2.377) (1.039) Transp and storage -3.653 -5.090 -4.564 -4.609 -1.600 -0.370 3.970 -2.939 38 (4.513) (5.245) (4.142) (3.012) (6.410) (6.208) (5.147) (3.238) Accomm -7.072 -14.88** -11.52** -9.584** 0.292 -13.87** -3.334 -5.705** (5.816) (2.825) (4.187) (1.588) (8.571) (3.712) (5.159) (1.376) Food prep serv 0.747 -4.911** -2.226 -4.309* 1.113 -4.318 1.463 -5.158** (3.625) (2.302) (2.494) (2.328) (4.620) (3.040) (2.723) (1.907) ICT 2.301 9.694* 14.16** 3.429 4.821 8.045 10.31 3.944** (7.696) (5.607) (5.804) (2.535) (8.664) (7.747) (7.263) (2.010) Financial serv -7.804 0.0764 15.90** 4.936** 3.780 14.03* 9.380 4.662** (10.23) (6.308) (7.531) (1.488) (18.36) (8.146) (16.19) (1.365) Education 8.550 -4.355 3.732 4.801 14.89 4.758 4.282 7.807 (9.672) (6.035) (4.895) (6.811) (14.34) (16.19) (8.651) (6.637) Health 21.83** 4.690 11.53** 18.17** -4.343 -2.075 9.912 6.484* (10.55) (5.824) (4.294) (3.596) (21.13) (8.123) (6.296) (3.316) Other serv -3.102 -3.781* -0.124 -1.218 -4.654 0.909 0.811 -2.861** (3.139) (2.067) (2.108) (1.534) (4.506) (2.808) (2.602) (1.421) Panel controls Yes Yes Yes Yes Yes Yes Yes Yes Other controls Yes Yes Yes Yes Yes Yes Yes Yes Initial productivity No No No No No No No No Lag change in sales No No No No Yes Yes Yes Yes Observations 2257 4231 4761 8735 1041 1931 2180 5265 Notes: All continuous variables are winsorized at the 1% and 99%. Recovery is calculated as the difference in the changes in sales across rounds of data collection. Robust standard errors in parentheses * p<0.10, ** p<0.05, ** p<0.01 Source: Authors’ calculations using data from the World Bank Business Pulse Survey (BPS). Table B8: Digitalization: split regressions by type of informality (1) (2) (3) (4) (5) (6) (7) (8) Necessity De Soto Parasite Fully formal Necessity De Soto Parasite Fully formal Panel A Started using or increased digital Has online sales Micro (less than 9 employees) -0.120** -0.113** -0.157** -0.140** -0.0703** -0.0640** -0.108** -0.0692** (0.0237) (0.0185) (0.0226) (0.0173) (0.0264) (0.0224) (0.0274) (0.0175) Maturing Baseline Baseline Baseline Baseline Baseline Baseline Baseline Baseline Young 0.0276 -0.000609 0.0217 0.0458* 0.0265 0.00107 0.0647** -0.0547 (0.0301) (0.0233) (0.0197) (0.0260) (0.0353) (0.0287) (0.0254) (0.0334) Established -0.0241 -0.0465** -0.0207 0.0192 -0.0671** -0.0321 -0.0198 -0.0173 (0.0232) (0.0172) (0.0177) (0.0117) (0.0277) (0.0207) (0.0219) (0.0111) Female-led 0.0784** 0.0451* 0.0764** 0.0291** 0.0809** 0.0293 0.0727** 0.000212 (0.0315) (0.0254) (0.0242) (0.0142) (0.0365) (0.0287) (0.0284) (0.0121) Manuf Baseline Baseline Baseline Baseline Baseline Baseline Baseline Baseline Const and utilities 0.129** -0.0326 0.0653 -0.0374 0.118* -0.0164 0.0226 -0.0582* (0.0621) (0.0373) (0.0428) (0.0357) (0.0709) (0.0483) (0.0528) (0.0334) Retail and wholesale 0.0217 0.0872** 0.0242 -0.0193 0.0222 0.0230 -0.0104 0.0302** (0.0275) (0.0264) (0.0238) (0.0147) (0.0343) (0.0287) (0.0305) (0.0142) 39 Transp and storage -0.0351 0.0250 -0.0395 -0.0729** 0.0152 0.0951 0.110 0.0322 (0.0621) (0.0459) (0.0455) (0.0358) (0.103) (0.0781) (0.0766) (0.0352) Accomm 0.0777 0.105** 0.0568 0.0286 0.0934 0.0679 -0.0349 0.0330 (0.0778) (0.0443) (0.0555) (0.0208) (0.0827) (0.0459) (0.0535) (0.0201) Food prep serv 0.0364 0.0463* -0.0314 0.0434 0.0258 0.00648 -0.0154 0.0512** (0.0383) (0.0276) (0.0259) (0.0268) (0.0433) (0.0327) (0.0342) (0.0255) ICT 0.382** 0.410** 0.255** 0.131** 0.121 0.176** -0.0155 0.0969** (0.0958) (0.0710) (0.0645) (0.0275) (0.105) (0.0849) (0.0627) (0.0285) Financial serv 0.273** 0.338** 0.207** 0.0992** 0.0719 0.0671 0.158 0.0721** (0.120) (0.117) (0.0981) (0.0213) (0.138) (0.120) (0.119) (0.0198) Education 0.332** 0.385** 0.286** 0.0830 0.127 0.124 0.0278 0.131** (0.106) (0.0868) (0.0814) (0.0620) (0.0884) (0.0900) (0.0972) (0.0499) Health 0.113 0.00669 0.0657 -0.0209 0.0926 -0.148** -0.0232 0.0730 (0.120) (0.0633) (0.0565) (0.0701) (0.134) (0.0602) (0.0670) (0.0646) Other serv 0.113** 0.0904** 0.0867** 0.0402** 0.0706* 0.0262 0.0663** 0.0395** (0.0379) (0.0254) (0.0256) (0.0193) (0.0417) (0.0303) (0.0312) (0.0184) % change in sales during past 30 days 0.000525* 0.000520** 0.000906** 0.000701** 0.000773** 0.000948** 0.00134** 0.00111** (0.000281) (0.000198) (0.000195) (0.000183) (0.000309) (0.000240) (0.000233) (0.000186) Observations 1739 3254 3323 6008 1226 2490 2368 8212 Panel controls Yes Yes Yes Yes Yes Yes Yes Yes Other controls Yes Yes Yes Yes Yes Yes Yes Yes Initial digital readiness No No No No No No No No Notes: All continuous variables are winsorized at the 1% and 99%. Robust standard errors in parentheses * p<0.10, ** p<0.05, ** p<0.01 Source: Authors’ calculations using data from the World Bank Business Pulse Survey (BPS). Table B8 continued: Digitalization: split regressions by type of informality (1) (2) (3) (4) (5) (6) (7) (8) Necessity De Soto Parasite Fully formal Necessity De Soto Parasite Fully formal Panel B Invested in digital technologies Share of employees working remotely (%) Micro (less than 9 employees) -0.0249 -0.0813** -0.0917** -0.105** -1.042 -2.915** 0.0592 -1.788* (0.0170) (0.0145) (0.0189) (0.0179) (1.460) (1.065) (1.224) (0.984) Maturing Baseline Baseline Baseline Baseline Baseline Baseline Baseline Baseline Young 0.00779 0.0164 0.00866 0.00151 -3.299* -0.566 -1.235 0.0185 (0.0221) (0.0195) (0.0160) (0.0393) (1.783) (1.276) (1.068) (1.960) Established -0.00598 -0.0375** -0.0239* -0.0132 -0.895 -0.647 -0.735 -0.818 (0.0176) (0.0128) (0.0138) (0.0132) (1.428) (0.962) (0.971) (0.634) Female-led 0.0447* 0.0277 0.0505** 0.0517** 5.225** 3.812** 3.492** 2.270** (0.0248) (0.0202) (0.0202) (0.0157) (1.908) (1.475) (1.345) (0.732) Manuf Baseline Baseline Baseline Baseline Baseline Baseline Baseline Baseline Const and utilities 0.0544 0.0000631 0.00783 0.0258 -2.129 1.689 -0.842 0.816 (0.0444) (0.0326) (0.0281) (0.0443) (2.716) (2.211) (2.281) (2.515) Retail and wholesale 0.0341 0.0282 0.0279 -0.00322 -5.923** -0.0783 -1.443 -0.795 (0.0210) (0.0189) (0.0188) (0.0169) (1.656) (1.141) (1.293) (0.747) 40 Transp and storage -0.0537* 0.0225 0.0495 0.00838 -3.631 8.138* 1.150 0.924 (0.0285) (0.0369) (0.0485) (0.0434) (3.720) (4.212) (2.766) (2.098) Accomm 0.0428 0.0299 0.0103 0.0293 -7.466** -2.674 -6.838** 0.228 (0.0554) (0.0268) (0.0442) (0.0230) (2.880) (1.828) (2.390) (1.055) Food prep serv 0.0447 -0.0180 -0.00573 0.0461 -3.192 2.093 0.102 -0.346 (0.0274) (0.0176) (0.0199) (0.0311) (2.956) (1.652) (1.396) (1.336) ICT 0.275** 0.165** 0.201** 0.141** 10.77* 18.11** 8.960** 18.62** (0.0836) (0.0660) (0.0633) (0.0331) (6.513) (5.906) (3.914) (2.467) Financial serv 0.214** 0.115 0.123* 0.108** 16.03** 3.367 -4.182* 3.723** (0.0807) (0.0933) (0.0729) (0.0252) (7.515) (5.408) (2.448) (1.046) Education 0.208** 0.150** 0.260** 0.168* 12.78* 10.78** 2.286 11.21 (0.101) (0.0766) (0.0810) (0.0912) (7.419) (5.356) (4.774) (7.137) Health 0.0681 0.0878 0.0125 0.0356 -0.704 5.878 -6.309** -0.751 (0.0935) (0.0622) (0.0381) (0.0837) (8.378) (4.544) (1.765) (3.814) Other serv 0.0606** 0.0587** 0.0666** 0.0446** -2.987 3.952** -1.924 0.302 (0.0279) (0.0199) (0.0208) (0.0225) (2.377) (1.560) (1.252) (1.091) % change in sales during past 30 days 0.000783** 0.000778** 0.000672** 0.00219** 0.00570 -0.0187* 0.000772 0.00580 (0.000193) (0.000158) (0.000159) (0.000225) (0.0188) (0.0109) (0.0123) (0.0123) Observations 1738 3143 3246 6357 2197 4144 4464 8627 Panel controls Yes Yes Yes Yes Yes Yes Yes Yes Other controls Yes Yes Yes Yes Yes Yes Yes Yes Initial digital readiness No No No No No No No No Notes: All continuous variables are winsorized at the 1% and 99%. Robust standard errors in parentheses * p<0.10, ** p<0.05, ** p<0.01 Source: Authors’ calculations using data from the World Bank Business Pulse Survey (BPS). Table B9: Probability of receiving support by instrument Conditional Cont. initial Conditional Cont. initial Conditional Cont. initial productivity productivity productivity Dependent variable: Technical support Grants New credits Compared to no support Informal-normalized PCA informality measures -0.0108** -0.0129* 0.00489 -0.000991 -0.0803** -0.102** (0.00518) (0.00691) (0.0110) (0.0137) (0.0114) (0.0143) Observations 10831 8373 11292 8697 12093 9436 Conditional on having support Informal-normalized PCA informality measures 0.00949 0.0196 0.245** 0.239** -0.269** -0.308** (0.0284) (0.0354) (0.0506) (0.0577) (0.0624) (0.0729) Observations 2537 2174 2562 2194 2584 2209 Panel controls Yes Yes Yes Yes Yes Yes Other controls Yes Yes Yes Yes Yes Yes Change in sales Yes Yes Yes Yes Yes Yes Initial productivity No Yes No Yes No Yes Observations 10831 8373 11292 8697 12093 9436 Notes: All continuous variables are winsorized at the 1% and 99%. Recovery is calculated as the difference in the changes in sales across rounds of data collection. Robust standard errors in parentheses * p<0.10, ** p<0.05, ** p<0.01 Source: Authors’ calculations using data from the World Bank Business Pulse Survey (BPS). 41 C. Questionnaire-Informality Informality module Is this establishment registered for a business license? Cov6a6 1= Yes 2= No INSTRUCTION: Only ask in countries where sampling -9=Don’t know (spontaneous) frame may include informal firms Does this establishment have a tax registration number? Cov6e10 1= Yes 2= No INSTRUCTION: Only ask in countries where sampling -9=Don’t know (spontaneous) frame may include informal firms Is this Establishment registered under company law as a Cov6e7 1= Yes sole proprietorship, partnership or limited liability 2= No company? -9=Don’t know (spontaneous) INSTRUCTION: Only ask in countries where sampling frame may include informal firms Is this establishment’s business bank account separate Cov6e8 1= Yes, Establishment has a separate from any personal accounts of the owners? bank account 2= No, Establishment has a mixed business/personal account 3= No, Establishment or the owner don’t have a bank account Does this Establishment contribute to the employees’ Cov6e9 1= Yes provident fund or government employee insurance 2= No scheme for its employees? -9=Don’t know (spontaneous) INSTRUCTION: Only ask in countries where sampling frame may include informal firms 42