Policy Research Working Paper 10641 Government Support and Firm Performance during COVID-19 Miriam Bruhn Asli Demirguc-Kunt Dorothe Singer Development Economics A verified reproducibility package for this paper is Development Research Group available at http://reproducibility.worldbank.org, December 2023 click here for direct access. Policy Research Working Paper 10641 Abstract This paper assesses the medium-run effects of government that received support in Round 1 performed better in terms support to firms during the COVID-19 crisis and whether of Round 3 sales, but only if they did not have continued the effectiveness of this support varied with its timing. Using support. Firms that also received support in Round 2 had data from three rounds of the World Bank’s Enterprise Sur- similar Round 3 sales as those that received no support veys COVID-19 Follow-up Surveys carried out between and were more likely to decrease employment. Firms that May 2020 and April 2022, it relates government support received government support only in Round 2 experienced in Round 1 (received in the first half of 2020) and Round no boost in Round 3 performance. The findings suggest that 2 (received during the second half of 2020 or early 2021) government support should be provided promptly, but it with firm performance in Round 3 (generally mid-2021). should also be phased out quickly. Controlling for a host of background characteristics, firms This paper is a product of the Development Research Group, Development Economics. 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 dsinger@worldbank.org. A verified reproducibility package for this paper is available at http://reproducibility. worldbank.org, click here for direct access. 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 Government Support and Firm Performance during COVID-19 * Miriam Bruhn,a Asli Demirguc-Kunt,b Dorothe Singerc Keywords: Government support, COVID-19, productivity, firms, Europe and Central Asia (ECA) JEL codes: D22, D24, H81, O47 * Nikkie Pacheco provided excellent research assistance. a mbruhn@worldbank.org, World Bank. b ademirguckunt@gmail.com, Non-Resident Fellow, Center for Global Development. c dsinger@worldbank.org, World Bank. 1. Introduction Governments around the world provided unprecedented support to firms when the COVID-19 pandemic started. For example, the World Bank’s SME-Support Measure Dashboard tracked 1,500 government measures to support small and medium-size enterprises (SMEs) in 132 countries. 1 According to policy makers, these initial support measures aimed to prevent mass insolvency of viable firms and related knock-on effects for the financial sector, to preserve jobs and firm-specific intangible capital, and to reduce the friction costs of firms temporarily exiting the market (World Bank 2021). Evidence from several countries suggests that government support measures were successful in helping firms weather the crisis and preserving jobs in the short run (see Chen et al. 2022, on China; Guerrero- Amezaga et al. 2022, on Latin America, and Turkson et al. 2021, on Italy), although at a high cost (Autor et al. 2022, on the US). However, longer-run effects of COVID-19 support to firms are uncertain (Staples and Krumel 2023), in part since these measures may simply have delayed insolvency for some firms (Dörr, Licht, and Murmann 2022). We contribute to the literature by assessing the medium-run effects of government support to firms during the COVID-19 crisis. We use data from three rounds of the World Bank’s Enterprise Surveys (ES) COVID-19 Follow-up Surveys, allowing us to look at firm performance in mid-2021, over a year after support measures were first enacted. Another contribution of our paper is that we test whether the effectiveness of government support varied with its timing. The data allow us to examine support received at the onset of the crisis (during the first half of 2020, as reported in Round 1 of the ES COVID-19 Follow-up Surveys), as well as support received later (during the second half of 2020 or early 2021, as reported in Round 2 of the surveys). While initial support may be effective, if it continues over an extended period, it can also distort markets and firms’ incentives (Pop and Amador 2020; World Bank 2021). Thus, the effects of initial support may differ from the effects of subsequent support. We analyze data for 15 emerging markets and developing countries in Europe and Central Asia (ECA), to focus on a homogeneous environment. First, we document that 42 percent of firms received government support by Round 1 of the survey, and 23 percent received support between Rounds 1 and 2, giving a cumulative 50 percent of firms receiving support. The support measure reaching by far the largest share of firms was wage subsidies, with 8 out of 10 firms that received any type of support receiving such subsides. The second most wide-reaching measure was cash transfers (less than 4 out of 10 firms), followed by fiscal relief, payment deferrals, and new credit (less than 2 out of 10 firms each). Then, we ask how receiving government support by Round 1 and Round 2 of the ES COVID-19 Follow-up Surveys affected firm performance in Round 3 of the survey. Here, we examine the relationship between Round 3 performance and support in Rounds 1 and 2 to minimize reverse causality. To reduce omitted variable bias, we control for a host of firm characteristics that may have affected both the probability of receiving support and firm performance. We find that firms that received government support in Round 1 performed better in terms of Round 3 sales compared to firms that did not receive any support, but only if they did not have continued 1 https://dataviz.worldbank.org/authoring/SME-COVID19/Overview 2 support in the following period. Interestingly, firms that continued to receive government support in the second round performed no better or even worse than those that did not receive any support. Also, firms that received government support only in the second round experienced no performance boost compared to those that received no support. These results hold up through various robustness checks, including controlling for firm performance in Rounds 1 and 2. Our findings suggest that, for effective policy, government support should be provided to firms quickly without delay, but it should also be phased out quickly. The paper is organized as follows. Section 2 presents the data and summary statistics. Section 3 describes the empirical strategy. Section 4 discusses the main results. Section 5 includes several robustness checks, and Section 6 concludes. 2. Data and Summary Statistics We use data from ES COVID-19 Follow-up Surveys to capture government support received by firms during the COVID-19 pandemic and to measure firm performance. 2 The World Bank has collected one to three rounds of these surveys in 45 countries starting in May 2020. 3 To explore how government support was repeated over time, we use data for countries with three survey rounds as of May 9, 2022. We conduct our analysis on a sample of 15 emerging markets and developing countries in ECA to focus on a homogeneous environment (see Table A.1 in the Appendix). 4 The three rounds of surveys were completed between May 2020 and April 2022 for this sample of 15 countries. 5 The sampling frame for the ES COVID-19 Follow-up Surveys includes all firms that replied to the latest pre-COVID-19 ES, making it possible to link performance during the COVID-19 pandemic back to firm characteristics collected through the ES. The ES cover formal (registered) firms with five or more employees and are designed to be nationally representative for the manufacturing and service sectors. For the countries in our sample, the most recent ES was conducted in 2019 or early 2020. 6 The ES COVID-19 Follow-up Surveys include questions on government support received during the pandemic. They also ask firms about changes in performance during the crisis, as well as expectations for the next months. Table A.2 provides the definitions of the variables included in our analysis. Tables 1 and 2 present the summary statistics of the share of firms that received government support and when they did so. Table 3 presents the summary statistics on firm characteristics and performance. 2 More information on the ES COVID-19 Follow-up Surveys is available at https://www.enterprisesurveys.org/en/covid-19. 3 Four rounds of data were collected in Jordan. 4 At least three rounds of survey data are also available for 3 countries in the Middle East and North Africa, 2 countries in Sub-Saharan Africa, and 4 countries in high-income Europe and Central Asia. 5 Armenia and North Macedonia have later Rounds 2 and Rounds 3 compared to the other 13 countries in our sample. Our results in this paper are robust for a sub-sample of 13 countries which drops Armenia and North Macedonia from the original sample. 6 For most countries, the surveys were completed before March 2020. For Romania, 92 percent of the interviews were completed before March 2020. Our results are similar when we exclude the 8 percent of observations from Romania that were collected in or after March 2020. 3 In all tables and analysis, we use sampling weights provided in the ES COVID-19 Follow-up Surveys to correct for unequal probability of selection as well as ineligibility for all reported statistics. We also rescale the sampling weights to give equal weight to all countries. 2.1 How many firms received government support and when? Our sample includes 3,647 firms across 15 countries that reported whether they received government support in Rounds 1, 2, and 3 and reported changes in performance in Round 3. Table 1 reports the share of firms that received government support by survey round. About 4 in 10 firms reported receiving support in Round 1; in Round 2 and Round 3 the share of firms reported doing so drops to about a quarter. Cumulatively, by Round 3 just over half of firms (55 percent) reported having received support. The support measure reaching by far the largest share of firms was wage subsidies, with 44 percent of firms – or 8 out of 10 firms that received any type of support – receiving such subsides. The second most wide-reaching measure was cash transfers (22 percent), followed by fiscal relief and payment deferrals (12 percent each). New credit was the least commonly reported support measure (8 percent). To minimize reverse causality in examining the relationship between firm performance in Round 3 and government support, we will focus on support in Rounds 1 and 2 in our analysis below. The relative ordering of the most common support measures is similar. Table 2 also reports the share of firms that received government support but does so by the share of firms that received government support in each combination of rounds. Considering all three rounds (Panel A), firms could receive up to 3 rounds of support in 8 possible combinations: from receiving support in each of the three rounds to none. Most firms which received any government support did so only in Round 1: 20 percent of firms or about one-third of firms receiving support. This was followed by firms receiving support in all three rounds (9 percent), and Rounds 1 and 2 and Rounds 1 and 3 (7 percent each). The combinations of receiving support in Round 2 only, Round 3 only, or Rounds 2 and 3 was reported by fewer than 5 percent of firms each. About 45 percent of firms received no support at all. The ordering of combinations is proportional when considering only Rounds 1 and 2 (Panel B). 2.2 What are the firm characteristics and changes in performance during COVID-19? Table 3 Panel A presents the summary statistics of firm background characteristics from the pre-COVID- 19 ES. To measure productivity, we use labor productivity, defined as (log of) sales over total number of permanent full-time employees. 7 Labor productivity is preferred to total factor productivity due to limitations in measuring non-labor inputs in the ES (World Bank 2021a). The average firm in our sample had a log of labor productivity of 10.59. At the same time, the average firm had 28 employees (2.62 in log units) and was 20 years old (2.88 in log units). Only 20 percent of the firms had a female top manager. The sample includes few firms with at least 10 percent state or foreign ownership (less than 1 and 7 percent, respectively). Less than half the firms reported that they had a loan (42 percent) or innovated a product or process during 2017 to 2019 (43 percent), but 65 percent owned a website. Most firms reported that their main market was national (48 percent) or local (41 percent) versus 7 In calculating labor productivity, the outliers are eliminated by first log-transforming total annual sales and the number of permanent full-time employees, and then trimming at plus and minus three standard deviations from the mean, as described in World Bank (2021a). 4 international (12 percent). Slightly over half the firms are in the service sector (53 percent), followed by manufacturing (27 percent) and retail (20 percent). The ES COVID-19 Follow-up Surveys ask firms about changes in performance during the crisis, as well as expectations for the next months. Table 3 Panel B presents the summary statistics of firm performance by Round 3. Firms reported that their sales remained an average of 2 percent lower compared to the same month a year earlier. Round 3 surveys were typically conducted in mid-2021 (see Table A.1), meaning that firms reported their change in sales compared to mid-2020, a time when firms typically experienced steep declines in sales compared to pre-COVID-19 (Apedo-Amah and others 2020; Bruhn, Demirguc-Kunt, and Singer, forthcoming). Table A.1 also shows drops in sales reported in Round 1 and Round 2, which are all relative to the same month one year earlier. Taken together, the numbers suggest that, in mid-2021, sales of the average firm were still about 21 percent lower than before COVID-19. Unlike sales, employment outcomes in all rounds are measured compared to December 2019. The Round 3 data in Table 2 Panel B shows that the number of permanent full-time workers remained 15 percent lower compared to pre-COVID-19, with almost half of firms (47 percent) having decreased their number of workers. About 21 percent reported that they anticipated falling into arrears on outstanding liabilities in the next six months. 2.3 How are government support and firm characteristics correlated? Table 4 presents the correlation coefficients between firms that received government support and firm characteristics. Overall, not many firm characteristics are consistently or significantly correlated with having received government support across different rounds of government support. While many firm characteristics are significantly correlated with having received government support in Round 1 or 2, many of these correlations turn insignificant when we unbundle government support into Round 1 versus Round 2 support. Large firms, and those with foreign ownership are more likely to have received support in both rounds, possibly because these firms have stronger political connections. Also, having a credit or loan, and having a website are characteristics that are positively correlated with receiving government support in both rounds. This may be because firms with an existing financial relationship and online presence may be better placed to apply for and receive government support electronically where this option is available. 8 8 We also find inconsistent or insignificant associations between government support and firm characteristics when we consider a multivariate OLS regression controlling for the same firm characteristics and country fixed effects. These results differ from our earlier findings reported in Demirguc-Kunt, Bruhn, and Singer (forthcoming) where we show in a different sample of government support data—corresponding to Round 1 in 11 countries and Round 2 in 13 countries—that more productive firms were less likely to receive any COVID-19 government support. 5 3. Empirical Strategy We examine the relationship between government support and firm performance by estimating the following regression specification with OLS 9: 3 (1) = β0 + + ℎ0 + + + ε where Firm performance of firm i in sector j and country k is captured by one of four variables: percentage change in sales; percentage change in employment; decreased employment; and anticipate falling into arrears. All firm performance variables are measured in Round 3 of the ES COVID-19 Follow- up Surveys. Government Support is a vector of one or more indicator variables equal to one if the firm received any type of government support and equal to zero otherwise. 10 We first use an indicator for having received government support in Round 1 or 2 of the ES COVID-19 Follow-up Surveys. Here, the coefficient represents the difference in Round 3 performance between firms that received support in Round 1 or 2 and firms that did not receive support in either of these rounds. Next, we aim to unpack the role of government support in Round 1 versus Round 2. In this regression, we include three mutually exclusive indictors for government support: (i) receiving support only in Round 1, (ii) receiving repeated support in Rounds 1 and 2, and (iii) receiving support only in Round 2. This regression thus compares Round 3 performance of firms in these three categories to firms that did not receive any support in Rounds 1 and 2. In identifying the causal effect of government support on firm performance, we face two main issues. The first is reverse causality. That is, government support may be determined by firm performance, for example if worse performing firms are more likely to receive support. We mitigate reverse causality by looking at firm performance one round after receiving government support, so that support is not directly influenced by firm performance. That is, we use Round 3 firm performance and government support in Round 1 and 2. The second threat to identification is omitted variable bias. That is, other factors may determine both government support and firm performance. We mitigate omitted variable bias by controlling for firm background characteristics, as well as sector and country fixed effects (S and C, respectively). Firm Characteristics is a vector of firm-level background variables measured before the COVID-19 crisis, as described in Section 2 and Table 3 (at time t = 0). These variables may determine firm performance for the following reasons. More productive, larger, and older firms may weather a crisis better because they are more established and have more financial and organizational resources to do so. These firms may also have strong 9 We chose a linear probability model since it is simpler and tends to yield similar marginal effects as a nonlinear model (Angrist and Pischke 2009). Our results are similar if we use a probit model to estimate the regressions with binary outcome variables. 10 One caveat in this analysis is that we do not know how much government assistance firms receive, but the amount of assistance has been shown to correlate with firm outcomes (Dvoulety and others 2021). 6 political connections. More innovative firms can be more likely to adapt to a crisis. State- or foreign- owned firms may be more resilient to crises through access to more resources, connections, or know- how. At the same time, state-owned firms may be less adaptable. Firms with pre-crisis credit may have easier access to finance during a crisis, allowing them to mitigate its impacts. With respect to gender, Liu, Wei, and Xu (2021) document that women-led businesses were more likely to close, close for longer, and anticipate falling into arrears than men-led businesses during the COVID-19 pandemic. We control for owning a website since Wagner (2021) shows that firms with a website were more likely to survive during the pandemic, potentially because the online presence allowed firms to interact with customers even when in-person contact was restricted. Finally, the location of the firm’s main market (local, national, or international) may matter for performance during the pandemic since movement restrictions may have impacted firms with different main markets differently. To further address omitted variable bias, we also run a robustness check controlling for management quality. This variable is not available for the full sample, which is why we include it only in the robustness check. Similarly, we run a robustness check controlling for firm performance in Rounds 1 and 2. Performance in Rounds 1 and 2 may have influenced the probability to getting support in Rounds 1 and 2 and could also be related to Round 3 performance due to persistence. 4. Results Tables 5 and 6 show the results from estimating equation 1 to examine the relationship between receiving government assistance and firm performance. Our main finding emerging from this analysis is that firms that received government support in Round 1 performed better than firms that received no support, in terms of sales, but that subsequent support had no or a negative impact on firm performance. The initial results in Table 5 show that firm performance is not correlated with having received any government support in Round 1 or 2. However, when we unbundle government support and control for Round 1 and 2 support separately in Table 6, firms that received support in Round 1 reported increased sales by Round 3 relative to one year ago (i.e., typically mid-2020). This result is driven by firms that received support in Round 1 only, doing so increased sales by 8 percentage points, compared to firms that received no support. Subsequent support in Round 2 – either as Round 1 and 2 support or Round 2 only – had no statistically significant effect on sales. In terms of employment outcomes, firms that received support only in Round 1 reported increased employment by 5 percentage points relative to December 2019, although this coefficient is not statistically significant. To the contrary, firms that also received support in Round 2 were more likely to decrease their employment relative to December 2019. Tables 5 and 6 also show that government support is not correlated with a firm’s likelihood of anticipating falling into arrears on outstanding liabilities in the next 6 months. What matters are a firm’s productivity and age: more productive firms and older firms are less likely to anticipate falling into arrears. Overall, our results suggest that the initial support provided to firms during the COVID-19 crisis helped firms, but there is no evidence that subsequent support did so. This pattern implies that if any support is provided, it should be phased out quickly. 7 5. Robustness Checks 5.1 What is the effect of government support on exit and non-response? Our analysis in Section 4 uses the sample of firms for whom government support is observed in all three rounds of the ES COVID-19 Follow-up Surveys, thus dropping any firms that closed between rounds or that did not respond to the later survey rounds for other reasons. This section asks whether using this sample of responding firms could bias our results, which may be the case if there is a relationship between government support and exit or non-response. We start by pointing out that confirmed exit rates are very low in our sample. Table A.3 shows exit rates for the sample of 4,279 firms that reported government support in both Rounds 1 and 2 (i.e., firms that had not closed by Round 2). Only 1 percent of these firms were confirmed to have exited by Round 3. The exit rate is higher, 9 percent, if we include firms that are assumed to have closed since they could not be contacted in Round 3. We also lose firms in our balanced sample due to interview refusal in Round 3 or not reporting government support in Round 3. In total, 19 percent of firms drop out of the sample of 4,279 for any of the reasons mentioned above (confirmed exit, assumed exit, interview refusal, and not reporting government support), giving 3,647 firms in our balanced sample. Tables A.4 and A.5 follow the same format as Tables 5 and 6, to examine whether receiving government support in Rounds 1 and 2 is associated with subsequent firm exit or non-response to Round 3. We find no relationship between government support and confirmed firm exit. The only variable that shows a statistically significant correlation with confirmed exit is firm age, where older firms were less likely to exit than younger firms (Table A.4 Column 1). When also including assumed exit as the outcome, firms with a female top manager are more likely to exit and state-owned firms are less likely to exit (Table A.4 Column 2). Column 3 of Tables A.4 and A.5 use an outcome variable that is equal to one if the firm was in Round 1 but is not in our balanced sample for any reason (confirmed exit, assumed exit, or non-response). Firms that received any support in Round 1 or 2 are less likely to drop out of the sample (Table A.4). However, this relationship does not persist when we break down government support by round (Table A.5). Since government support does not consistently predict exit and non-response in our main regressions, we conclude that the analysis in Section 4, which uses a balanced panel to look at the effect of support on firm performance, likely does not suffer from non-response bias. 5.2 Controlling for the competition environment Bruhn, Demirguc-Kunt, and Singer (forthcoming) show that countries’ competition environment affected the reallocation of economic activity across firms during the COVID-19 crisis. Countries with a strong competition environment experienced more reallocation from less productive to more productive firms than countries with a weak competition environment. Here, we test whether the competition environment also influenced the relationship between government support and firm performance. 8 Following Bruhn, Demirguc-Kunt, and Singer (forthcoming), Table A.6 uses the 2020 Bertelsmann Stiftung Transformation Index (BTI) as a measure of competition. As in the earlier paper, we interact this measure with labor productivity (the main effect is subsumed in the country fixed effect). Importantly, we also add interaction terms of the BTI with our measures of government support. These interaction terms are not statistically significant in Table A.6. Thus, we do not find evidence that the relationship between government support and firm performance depends on the competition environment. With respect to reallocation from less productive to more productive firms, the results in Table A.6 suggest that the effect of competition persists in the medium run. The interaction terms between the BTI and labor productivity are of a similar magnitude as in Bruhn, Demirguc-Kunt, and Singer (forthcoming), although the coefficient for “Anticipate falling into arrears” in Table A.6 is not statistically significant at conventional levels. The exception is that the coefficient for “Percentage change in sales” is much smaller than in Bruhn, Demirguc-Kunt, and Singer (forthcoming), which may be because the reference period here is different. In the short run, change in sales was relative to 2019, while change in sales here is relative to the previous year. That is, competition influenced the initial reallocation in sales from low to high productivity firms but did not determine additional reallocation. It is worth nothing that this additional reallocation was also much smaller in magnitude, with the average percentage change in sales being -1.98 in the medium run (Table A.1), compared to -23.98 in Bruhn, Demirguc-Kunt, and Singer (forthcoming). 5.3 Using country-sector fixed effects The analysis in Section 4 controls for country and sector fixed effects. Here, we do a robustness check with country-sector fixed effects since governments may have directed support to those sectors that were hit harder in their country. Table A.7 replicates the results in Tables 5 and 6 but replacing country and sector fixed effects with country-sector fixed effects. The results are similar, showing that our findings are robust to controlling for country-sector fixed effects. 5.4 Controlling for management quality Although the analysis in Table 5 and 6 controls for many firm characteristics, the findings could be subject to omitted variable bias. Most notably, an unobserved measure of firm quality could drive both the probability of receiving government support and subsequent firm performance. For example, firms with better managers may be more likely to apply for government support and may also perform better during a crisis. The ES include a measure of management quality, based on 11 questions that are aggregated following Grover and Karplus (2021) and Bloom et al. (2019). Unfortunately, the management quality measure is only available for firms with 20 or more employees. Table A.8 Panel A1 replicates the results from Table 5 in this smaller sample. As in Table 5, we find no statistically significant relationship between receiving government support in Round 1 or 2 and firm performance in Round 3. We then control for management quality in Panel A2 of Table A.8, which leaves the coefficients on Round 1 or 2 support basically unchanged. 9 Panel B1 in Table A.8 replicates Table 6 in the smaller sample for which management quality is available. Most findings are similar, except that receiving only Round 1 support is not positively related with change in sales in the sample of firms with 20 or more employees, suggesting that Round 1 support mostly benefitted smaller firms. Controlling for management quality in Panel B2 of Table A.8 changes the coefficients on government support very little. Overall, the fact that accounting for management quality does not meaningfully change the results in Table A.8 alleviates concerns about omitted variable bias. 5.5 Controlling for firm performance in Rounds 1 and 2 Table A.9 accounts for the fact the government support in Round 1 or 2 may be related to contemporaneous firm performance. For example, firms that saw larger declines in sales in Round 1 may have been more likely to receive support in Round 1. This pattern could bias our results if there is persistence in firm performance over time. Columns 1 through 4 in Table A.9 replicate the results from Tables 5 and 6, controlling for four additional variables: the percentage changes in sales and employment reported in Rounds 1 and 2. Columns 5 through 8 take a slightly different approach, controlling instead for the Round 1 and 2 values of the outcome variable in the respective column. The results in Table A.9 are consistent with our earlier findings. The effect of government support on sales in Round 3 is stronger, but still driven by Round 1 support only. There is weak evidence that receiving only Round 2 support increased employment in Round 3, but this effect is not robust. Receiving repeated support in Round 1 and 2 does not lead to better performance in Round 3. 5.6 Types of government support So far, we have analyzed the relationship between any type of government support and Round 3 firm performance. In this subsection, we ask if this relationship varies with the type of government support received. We focus on the two most prominent types of support: wage subsidies and cash transfers, with the caveat that wage subsidies were more than twice as frequent as cash transfers. Table 1 illustrates that 40 percent of firms received wage subsidies by Round 2, with only 17 percent of firms receiving cash transfers. Table A.10 looks at the relationship between receiving wage subsidies/cash transfers in Rounds 1 and 2 and firm performance in Round 3. Wage subsidies in Round 1 are associated with better sales performance in Round 3. The correlation between cash transfers and Round 3 sales is of the same magnitude as the one for wage subsidies, but it is not statistically significant, perhaps since fewer firms received cash transfers, lowering statistical power for detecting their effects. We can thus not draw any strong conclusions regarding the relative effects of wage subsidies and cash transfers. Table A.10 also shows that continued support in Rounds 1 and 2 is associated with worse employment outcomes in Round 3. Support in Round 2 is not correlated with Round 3 performance. These findings are consistent with our results from Section 4, supporting the argument that any support should be given promptly and phased out quickly. 10 6. Conclusion We use data from three rounds of ES COVID-19 Follow-up Surveys for 15 emerging markets and developing economies in ECA to assess the effect of government support during the COVID-19 crisis on firm performance by mid-2021, over a year after support measures were first enacted. We first document that 42 percent of firms received government support by Round 1 of the survey and 23 percent received support between Rounds 1 and 2, for a cumulative 50 percent of firms that received government support. We find that the effect of government support on firm performance varies with the timing of the support. Firms that received support in Round 1 performed better in terms of Round 3 sales, but only if they did not have continued support in Round 2. Firms that received support in Round 2 had a similar sales performance in Round 3 and were more likely to decrease employment than those that received no support. Overall, these results suggest that during economic crises such as the COVID-19 pandemic, government support should be provided to firms quickly, but that it should also be phased out quickly. Doing so ensures that support measures help firms weather the crisis but do not delay insolvencies for firms that may no longer be competitive. This conclusion is consistent with the notion that fiscal stimulus should be timely and temporary (Elmendorf and Furman 2008). In practice, responses to previous crises have often been implemented too slowly and were not rolled back fully, leading to increasingly higher government spending (Taylor and Castillo O'Sullivan 2015). By mid-2021 the COVID-19 crisis was still ongoing. An avenue for future research is to investigate the effects of government support measures during the COVID-19 crisis on long-run firm performance once such data becomes available. To help inform the government response to future crises, researchers could also try to examine the relative merit of different firm support measures such as wage subsidies compared to cash subsidies. 11 References Angrist, J., & Pischke, J.-S. 2009. “Mostly harmless econometrics: An empiricist’s companion.” Princeton University Press. https://doi.org/10.2307/j.ctvcm4j72 Apedo-Amah, M. C., Avdiu, B., Cirera, X., Cruz, M., Davies, E., Grover, A., Iacovone, L., Kilin, U., Medvedev, D., Okechukwu Maduko, F., Poupakis, S., Torres, J., & Tran, T. T. 2020. “Unmasking the impact of COVID-19 on businesses: Firm level evidence from across the world.” Policy Research Working Paper 9434, World Bank, Washington, DC. https://openknowledge.worldbank.org/handle/10986/34626. Autor, D., Cho, D., Crane, L.D., Goldar, M., Lutz, B., Montes, J., Peterman, W.B., Ratner, D., Villar D., & Yildirmaz, A. 2022. "The $800 Billion Paycheck Protection Program: Where Did the Money Go and Why Did It Go There?" Journal of Economic Perspectives, 36(2): 55-80. https://doi.org/ 10.1257/jep.36.2.55 Bertelsmann Stiftung Transformation Index (BTI). 2020. “BTI 2020 Codebook for Country Assessments.” https://bti-project.org/en/methodology#Index%20aggregation. Bloom, N., Brynjolfsson E., Foster L., Jarmin R., Patnaik M., Saporta-Eksten I., & Van Reenen, R. 2019. “What drives differences in management practices?” American Economic Review, 109(5): 1648– 1683. https://doi.org/ 10.1257/aer.20170491 Bruhn, M., Demirguc-Kunt, A., & Singer, D. Forthcoming. “Competition and firm recovery post-COVID- 19.” Small Business Economics. https://doi.org/10.1007/s11187-023-00750-w Chen, J., Cheng, Z., Gong, R.K., & Li, J. 2022. “Riding out the COVID-19 storm: How government policies affect SMEs in China” China Economic Review, Volume 75, October 2022, 101831. https://doi.org/10.1016/j.chieco.2022.101831. Dörr, J., Licht, G., & Murmann, S. 2022. “Small firms and the COVID-19 insolvency gap.” Small Business Economics, 58, 887–917. https://doi.org/10.1007/s11187-021-00514-4 Dvoulety, O., Srhoj, S., & Pantea, S. 2021. “Public SME grants and firm performance in European Union: A systematic review of empirical evidence.” Small Business Economics, 57(1): 243–263. https://doi.org/10.1007/s11187-019-00306-x Elmendorf, D.W., & Furman, J. 2008. “If, When, How: A Primer on Fiscal Stimulus.” The Hamilton Project Strategy Paper. The Brookings Institution. https://www.brookings.edu/articles/if-when-how-a- primer-on-fiscal-stimulus/ Guerrero-Amezaga, M. E., Humphries, J. E., Neilson, C. A., Shimberg, N., & Ulyssea, G. 2022. “Small firms and the pandemic: Evidence from Latin America.” Journal of Development Economics, 155, [102775]. https://doi.org/10.1016/j.jdeveco.2021.102775 Grover, A., & Karplus, V. 2021. “Coping with COVID-19: Does Management Make Firms More Resilient?” Policy Research Working Paper 9514, World Bank, Washington, DC. https://doi.org/10.1596/1813- 9450-9514 Liu, Yu., Wei, S., & Jian, Xu. 2021. “COVID-19 and women-led businesses around the world.” Finance Research Letters, 43. https://doi.org/10.1016/j.frl.2021.102012 Staples, A.J., & Krumel, T.P. 2023. “The Paycheck Protection Program and small business performance: Evidence from craft breweries.” Small Business Economics, 61: 931–956. https://doi.org/10.1007/s11187-022-00717-3 12 Turkson, D., Addai, N.B., Chowdhury, F. et al. 2021. “Government policies and firm performance in the COVID-19 pandemic era: a sectoral analysis.” SN Bus Econ 1, 168. https://doi.org/10.1007/s43546- 021-00170-6 Taylor, J., & Castillo O'Sullivan, A. 2015. “Timely, Targeted, and Temporary?' an Analysis of Government Expansions Over the Past Century.” MERCATUS RESEARCH, Available at SSRN: http://dx.doi.org/10.2139/ssrn.3211609 Pop, G., & and Amador, A.. 2020. “To Aid and How to Aid: Policy Options to Preserve Markets.” European State Aid Law Quarterly 19 (2): 127–36, https://doi.org/10.21552/estal/2020/2/4. Wagner, J. 2021. “With a little help from my website, firm survival and web presence in times of COVID- 19: Evidence from 10 European countries.” Working Paper Series in Economics No. 399, University of Lüneburg, Lower Saxony, Germany, http://hdl.handle.net/10419/234588. World Bank. 2021. Supporting Firms in Restructuring and Recovery. EFI Insight: Finance, Competitiveness and Innovation, Washington, DC. World Bank. 2021a. “Firm Level productivity estimates.” Methodological Note. World Bank, Washington, DC. Available at: www.enterprisesurveys.org. 13 Table 1: Government Support by Survey Round Variable Round 1 Round 2 Round 3 Obs Mean Obs Mean Obs Mean Per round Received government support 3,647 0.42 3,647 0.23 3,647 0.25 Cash transfer 3,635 0.13 3,644 0.05 3,645 0.08 Payment deferrals 3,639 0.06 3,645 0.03 3,645 0.04 New Credit 3,638 0.05 3,645 0.02 3,646 0.02 Fiscal relief 3,634 0.06 3,644 0.04 3,646 0.05 Wage subsidies 3,643 0.32 3,643 0.17 3,645 0.19 Cumulative Received government support 3,647 0.50 3,647 0.55 Cash transfer 3,633 0.17 3,637 0.22 Payment deferrals 3,636 0.09 3,635 0.12 New Credit 3,636 0.07 3,635 0.08 Fiscal relief 3,631 0.09 3,635 0.12 Wage subsidies 3,638 0.40 3,640 0.44 Source: Enterprise Surveys COVID-19 Follow-up Surveys. Notes: Includes the 3,647 firms that reported whether they received government support in Rounds 1, 2, and 3. Some firms did not specify the type of support, which is why the number of observations varies. Table 2: Number of Government Support Rounds Received No. of Rounds Rounds Obs Mean Considering all 3 Rounds 3 All 3 Rounds 3,647 0.09 2 Rounds 1 and 2 3,647 0.07 2 Rounds 1 and 3 3,647 0.07 2 Rounds 2 and 3 3,647 0.05 1 Round 1 only 3,647 0.20 1 Round 2 only 3,647 0.03 1 Round 3 only 3,647 0.05 0 No Rounds 3,647 0.45 Considering Rounds 1 and 2 only 2 Rounds 1 and 2 3,647 0.15 1 Round 1 only 3,647 0.27 1 Round 2 only 3,647 0.08 0 No Rounds 3,647 0.50 Source: Enterprise Surveys COVID-19 Follow-up Surveys. Note: Includes the 3,647 firms that reported whether they received government support in Rounds 1, 2, and 3. 14 Table 3: Firm Characteristics and Performance Summary Statistics Variable Obs Mean SD p10 p90 Min Max a. Firm characteristics (2019) Log(labor productivity) 3,647 10.59 1.20 8.98 12.01 3.95 16.45 Log(number of employees) 3,647 2.62 0.97 1.61 3.93 0.00 7.16 Log(firm age) 3,647 2.88 0.55 2.08 3.40 0.69 5.03 Top manager female 3,647 0.20 Innovated during 2017-19 3,647 0.43 State ownership (10%) 3,647 0.00 Foreign ownership (10%) 3,647 0.07 Has line of credit or loan 3,647 0.42 Owns a website 3,647 0.65 Main market Local 3,647 0.41 National 3,647 0.48 International 3,647 0.12 Sector Manufacturing 3,647 0.27 Retail 3,647 0.20 Other services 3,647 0.53 b. Firm performance (Round 3) Percentage change in monthly sales relative to one year earlier 3,507 -1.99 33.91 -40.00 25.00 -100.00 400.00 Percentage change in number of permanent full-time workers since December 2019 3,546 -14.60 51.38 -80.00 28.57 -200.00 200.00 Reduced number of permanent full-time workers since December 2019 3,546 46.96 Anticipate falling into arrears on outstanding liabilities in the next 6 months 3,247 20.85 Sources: Enterprise Surveys and Enterprise Surveys COVID-19 Follow-up Surveys. Note: In our sample, the percentage of firms who have state ownership (10%) is 0.17 percent. 15 Table 4: Correlations between government support and firm characteristics Government support Round 1 or 2 Round 1 only Rounds 1 and 2 Round 2 only Log(labor productivity) 0.015 0.028* 0.030* -0.060*** Log(number of employees) 0.050*** 0.005 0.064*** -0.002 Log(firm age) -0.003 0.032* -0.013 -0.039** Top manager female 0.001 -0.004 -0.010 0.021 Innovated during 2017-19 0.060*** 0.031* 0.025 0.025 State ownership (10%) -0.022 -0.006 -0.008 -0.020 Foreign ownership (10%) 0.028* -0.011 0.053*** -0.001 Has line of credit or loan 0.056*** 0.019 0.038** 0.019 Owns a website 0.086*** 0.039** 0.079*** -0.013 Local market dummy -0.076*** -0.072*** -0.019 0.005 National market dummy 0.029* 0.041** -0.001 -0.013 International market dummy 0.056*** 0.035** 0.025 0.011 Manufacturing sector 0.032* 0.042** -0.008 0.001 Retail sector -0.059*** -0.035** -0.022 -0.021 Other services sector 0.015 -0.016 0.027* 0.017 Sources: Enterprise Surveys COVID-19 Follow-up Surveys and Enterprise Surveys. Notes: Correlations matrix based on 3,647 observations. *** p < 0.01, ** p < 0.05, * p < 0.1. 16 Table 5: Combined Effect of Round 1 or 2 Support on Round 3 Performance Anticipate Percentage Percentage Decreased falling change in change in employment into sales employment arrears Received any support in 4.315 1.013 1.476 2.669 Round 1 or 2 (3.739) (2.516) (2.357) (3.412) Log(labor productivity) 1.547* 1.688 -2.883 -2.730*** (0.768) (2.653) (2.180) (0.825) Log(number of 1.362 3.499 2.157* -1.242 employees) (1.022) (2.068) (1.053) (1.017) Log(firm age) -2.071 0.668 -0.146 -8.339*** (2.188) (2.694) (2.095) (2.004) Top manager female -0.219 2.601 1.627 -1.188 dummy (3.123) (3.955) (2.649) (1.889) Innovated during 2017-19 -0.839 1.464 -0.458 2.908 (1.704) (5.135) (4.828) (2.412) State ownership (10%) -10.590 -10.220 -9.962 -7.598 (8.382) (24.472) (17.058) (10.223) Foreign ownership (10%) -1.441 8.682 -5.473 -5.805 (4.724) (13.390) (8.920) (5.457) Has line of credit or loan 1.634 2.041 1.431 -0.096 (2.157) (2.591) (1.921) (2.586) Owns a website 2.244 -1.305 2.697 0.858 (2.000) (4.110) (3.838) (2.932) National market dummy 4.945* -0.180 -0.521 2.157 (2.540) (5.171) (4.276) (3.454) International market 3.991 -6.101 1.691 -2.143 dummy (3.942) (7.897) (4.999) (4.544) Constant -22.878* -40.982 67.940** 74.194*** (12.041) (29.101) (23.510) (9.539) R 2 0.049 0.101 0.060 0.229 Number of observations 3,507 3,546 3,546 3,247 Sources: Enterprise Surveys COVID-19 Follow-up Surveys and Enterprise Surveys. Notes: All regressions are ordinary least squares and include sector and country fixed effects. Standard errors are clustered at the country level. *** p < 0.01, ** p < 0.05, * p < 0.1. 17 Table 6: Separate Effects of Round 1 and 2 Support on Round 3 Performance Percentage Anticipate Percentage Decreased change in falling into change in sales employment employment arrears Received in Round 1 only dummy 8.415** 4.972 -1.915 3.757 (3.821) (3.445) (3.197) (5.190) Received in Rounds 1 and 2 dummy 0.655 -4.233 8.670** 3.017 (4.475) (3.947) (3.605) (3.009) Received in Round 2 only dummy -0.849 -1.776 -0.192 -0.599 (7.075) (4.657) (4.313) (4.826) R2 0.055 0.104 0.065 0.230 Number of observations 3,507 3,546 3,546 3,247 Sources: Enterprise Surveys COVID-19 Follow-up Surveys and Enterprise Surveys. Notes: All regressions are ordinary least squares and include sector and country fixed effects, as well as the control variables listed in Table 5. Standard errors are clustered at the country level. *** p < 0.01, ** p < 0.05, * p < 0.1. 18 Appendix Table A.1: Completion Dates of Survey Fieldwork for the World Bank Enterprise Surveys (with average percentage change in sales relative to one year earlier) Round 1 Round 2 Round 3 Country ES Baseline Percentage Percentage Percentage ES COVID-19 ES COVID-19 ES COVID-19 change in change in change in Survey date Survey date Survey date sales sales sales Armenia December 2020 April 2021 -23.14 November 2021 3.31 April 2022 0.39 Bulgaria March 2020 September 2020 -21.56 December 2020 -20.80 May 2021 -6.85 Croatia November 2019 September 2020 -16.78 January 2021 -17.09 June 2021 5.39 Czechia March 2020 October 2020 -15.27 February 2021 -22.27 June 2021 -5.80 Estonia January 2020 October 2020 -10.94 February 2021 -10.91 August 2021 8.45 Georgia January 2020 June 2020 -45.97 November 2020 -27.81 October 2021 -3.99 Hungary March 2020 September 2020 -15.34 February 2021 -14.67 June 2021 0.98 Latvia January 2020 November 2020 -8.65 February 2021 -20.03 August 2021 -7.01 Lithuania January 2020 October 2020 -14.00 February 2021 -28.89 August 2021 -9.31 Moldova November 2019 May 2020 -58.91 November 2020 -28.43 June 2021 -10.27 North Macedonia October 2019 November 2020 -26.91 June 2021 -2.48 January 2022 -5.49 Poland December 2019 August 2020 -14.52 December 2020 -14.84 June 2021 -5.00 Romania June 2020 September 2020 -17.32 December 2020 -10.67 June 2021 6.25 Slovak Republic March 2020 October 2020 -13.33 February 2021 -24.70 June 2021 -4.54 Slovenia November 2019 August 2020 -11.59 December 2020 -12.08 June 2021 7.10 -20.95 -16.82 -1.98 Note: Dates shown are for the last completed survey for each round in a country. For Romania, 92% of the ES Baseline interviews were completed before March 2020. 19 Table A.2: Description of Main and Control Variables Variable Description Equals 1 if firm received any national or local government assistance Received government support provided in response to the COVID-19 outbreak and 0 otherwise. Equals 1 if the government support received by firm was a cash transfer Cash transfer and 0 otherwise. Equals 1 if the government support received by firm was a payment Payment deferrals deferral and 0 otherwise. Equals 1 if the government support received by firm was new credit and New Credit 0 otherwise. Equals 1 if the government support received by firm was fiscal relief and Fiscal relief 0 otherwise. Equals 1 if the government support received by firm was wage subsidies Wage subsidies and 0 otherwise. Received any support in Round 1 or 2 Equals 1 if firm received any government support during Round 1 or 2 Received any support in Round 1 Equals 1 if firm received any government support during Round 1 Received any support in Round 2 Equals 1 if firm received any government support during Round 2 Received in all 3 Rounds dummy Equals 1 if firm received any government support during all 3 Rounds Received in Rounds 1 and 2 dummy Equals 1 if firm received any government support during Rounds 1 and 2 only Received in Rounds 1 and 3 dummy Equals 1 if firm received any government support during Rounds 1 and 3 only Received in Rounds 2 and 3 dummy Equals 1 if firm received any government support during Rounds 2 and 3 only Received in Round 1 only dummy Equals 1 if firm received any government support during Round 1 only 20 Received in Round 2 only dummy Equals 1 if firm received any government support during Round 2 only Received in Round 3 only dummy Equals 1 if firm received any government support during Round 3 only No Rounds dummy (did not receive any support at all) Equals 1 if firm did not receive any government support at all Log of annual sales divided by the number of full-time permanent Log(labor productivity) employees (in USD 2009) Log(number of employees) Log of total number of full-time employees Log(firm age) Log of number of years that the firm has been operating. Top manager female Equals 1 if firm's top manager is female and 0 otherwise. Equals 1 if firm innovated a product or process in the last 3 years and 0 Innovated during 2017-19 otherwise. State ownership (10%) Equals 1 if firm is at least 10% state owned and 0 otherwise. Foreign ownership (10%) Equals 1 if firm is at least 10% foreign owned and 0 otherwise. Has line of credit or loan Equals 1 if firm has line of credit or bank loans and 0 otherwise Owns a website Equals 1 if firm has its own website and 0 otherwise. Equals 1 if firm's main product is sold mostly in same municipality where Local market firm is located and 0 otherwise. Equals 1 if firm's main product is sold mostly across the country where National market the firm is located and 0 otherwise. Equals 1 if firm's main product is sold mostly internationally and 0 International market otherwise. Manufacturing Equals 1 if firm is in the manufacturing sector and 0 otherwise. Retail Equals 1 if firm is in the retail sector and 0 otherwise. Equals 1 if firm is in the selected services sector, excluding retail, and 0 Other services otherwise. 21 Percentage change in monthly sales Average percentage change in monthly sales compared to the same relative to one year earlier month one year earlier Average percentage change in permanent full-time workers since Percentage change in number of December 2019. The formula is: ((a1-a0)/[(a1+a0)/2])*100; where a1 = permanent full-time workers since permanent full-time workers, end of last completed month & a0 = December 2019 permanent full-time workers, end of December 2019 Reduced number of permanent full- Equals 100 if firm reduced number of permanent full-time workers since time workers since December 2019 December 2019 and 0 otherwise. Anticipate falling into arrears on Equals 100 if firm anticipates falling into arrears on outstanding liabilities outstanding liabilities in the next 6 in the next 6 months and 0 otherwise. months Table A.3: Exit Rates between Rounds 2 and 3 Exit measure Obs Mean Confirmed exit 4279 0.01 Assumed exit 4279 0.09 Did not answer Round 3 4279 0.19 Source: Enterprise Surveys COVID-19 Follow-up Surveys. 22 Table A.4: Combined Effect of Round 1 or 2 Support on Round 3 Exit Probability of Probability of being Probability of being Confirmed Confirmed or dropping out of exit Permanently Assumed sample and not Closed by Round Permanently Closed making it to 3647 3 by Round 3 baseline sample Received any support in Round 1 -0.004 -0.016 -0.026* or 2 (0.005) (0.021) (0.012) Log(labor productivity) -0.006 0.006 0.014** (0.005) (0.005) (0.006) Log(number of employees) -0.001 -0.002 -0.010 (0.002) (0.008) (0.013) Log(firm age) -0.008* -0.004 -0.027 (0.004) (0.011) (0.019) Top manager female dummy 0.005 0.040* 0.040 (0.004) (0.022) (0.027) Innovated during 2017-19 0.004 -0.011 -0.008 (0.006) (0.012) (0.023) State ownership (10%) -0.000 -0.046** -0.127** (0.005) (0.020) (0.055) Foreign ownership (10%) -0.003 -0.001 0.095 (0.002) (0.051) (0.061) Has line of credit or loan 0.001 0.019 0.011 (0.003) (0.016) (0.023) Owns a website -0.002 0.012 0.034* (0.005) (0.013) (0.018) National market dummy -0.002 -0.011 0.003 (0.004) (0.011) (0.014) International market dummy -0.004 0.034 0.055 (0.007) (0.036) (0.040) Constant 0.109 0.029 0.093 (0.066) (0.080) (0.075) R2 0.035 0.146 0.134 Number of observations 4,279 4,279 4,279 Sources: Enterprise Surveys COVID-19 Follow-up Surveys and Enterprise Surveys. Notes: All regressions are ordinary least squares and include sector and country fixed effects. Standard errors are clustered at the country level. *** p < 0.01, ** p < 0.05, * p < 0.1. 23 Table A.5: Separate Effects of Round 1 and 2 Support on Round 3 Exit Probability of Probability of Probability of dropping out of being confirmed being confirmed exit sample and or assumed permanently not making it to permanently closed by Round 3 3647 baseline closed by Round 3 sample Received in Round 1 only dummy -0.003 -0.020 -0.023 (0.005) (0.021) (0.016) Received in Rounds 1 and 2 dummy -0.007 -0.002 -0.008 (0.008) (0.020) (0.030) Received in Round 2 only dummy -0.001 -0.025 -0.061 (0.004) (0.046) (0.046) R2 0.035 0.146 0.135 Number of observations 4,279 4,279 4,279 Sources: Enterprise Surveys COVID-19 Follow-up Surveys and Enterprise Surveys Notes: All regressions are ordinary least squares and include sector and country fixed effects, as well as the control variables listed in Table A.4 Standard errors are clustered at the country level. *** p < 0.01, ** p < 0.05, * p < 0.1. 24 Table A.6: Effects of Round 1 and 2 Support on Round 3 Performance (with BTI Interaction Variables) Percentage Percentage Anticipate Decreased change in change in falling into employment sales employment arrears Panel A: Combined effect of Round 1 or 2 support Received any support in Round 1 or 2 4.316 1.303 1.278 2.899 (3.708) (2.716) (2.135) (3.863) Received any support in Round 1 or 2 * BTI -0.290 -0.757 -0.982 -1.035 (2.278) (2.907) (1.761) (3.065) Log(labor productivity) 1.479** 2.340 -3.799* -2.872*** (0.683) (2.404) (1.873) (0.767) Log(labor productivity)*BTI -0.188 1.938 -2.662** -0.881 market organization (0.702) (1.187) (1.007) (0.687) R2 0.049 0.103 0.065 0.230 Number of observations 3,507 3,546 3,546 3,247 Panel B: Separate effects of Round 1 and 2 support Received in Round 1 only dummy 8.686** 5.478 -2.105 4.227 (3.793) (3.778) (3.071) (5.183) Received in Rounds 1 and 2 dummy 0.984 -2.906 7.803** 4.493 (4.209) (3.767) (3.184) (3.187) Received in Round 2 only dummy -0.469 -1.538 -0.433 -0.404 (7.043) (4.585) (4.250) (4.566) Received in Round 1 only dummy * BTI -1.000 -0.680 -1.615 -1.445 (1.907) (3.757) (2.408) (3.679) Received in Rounds 1 and 2 dummy * BTI -0.911 -2.880 0.456 -3.113 (3.671) (2.804) (2.136) (2.895) Received in Round 2 only dummy * BTI 2.876 2.747 -3.061 0.887 (5.175) (3.521) (2.850) (3.530) Log(labor productivity) 1.586** 2.416 -3.741* -2.724*** (0.629) (2.363) (1.899) (0.775) Log(labor productivity)*BTI -0.184 1.944 -2.638** -0.905 market organization (0.734) (1.211) (0.999) (0.710) R2 0.057 0.107 0.070 0.231 Number of observations 3,507 3,546 3,546 3,247 Sources: Enterprise Surveys COVID-19 Follow-up Surveys and Enterprise Surveys; Bertelsmann Stiftung Transformation Index (BTI) 2020. Notes: All regressions are ordinary least squares and include sector and country fixed effects, as well as the controls listed in Table A.4. BTI market organization is based on responses to the question: "To what level have the fundamentals of market-based competition developed?" BTI market organization is centered on its mean to facilitate interpretation of the coefficients. Standard errors are clustered at the country level. *** p < 0.01, ** p < 0.05, * p < 0.1. 25 Table A.7: Effects of Round 1 and 2 Support on Round 3 Performance (Using Country-Sector Fixed Effects) Percentage Percentage Anticipate Decreased change in change in falling into employment sales employment arrears Panel A: Combined effect of Round 1 or 2 Support Received any support in Round 1 or 2 4.374 1.233 2.035 2.554 (3.601) (2.530) (2.514) (3.279) R2 0.074 0.133 0.082 0.242 Number of observations 3,507 3,546 3,546 3,247 Panel B: Separate effects of Round 1 and 2 Support Received in Round 1 only dummy 8.011* 4.668 -0.937 3.577 (3.776) (3.360) (3.186) (5.097) Received in Rounds 1 and 2 dummy 0.562 -4.249 8.626** 3.262 (4.659) (3.975) (3.855) (2.934) Received in Round 2 only dummy 0.465 0.173 0.084 -1.134 (6.316) (4.443) (4.894) (4.650) R2 0.080 0.136 0.085 0.243 Number of observations 3,507 3,546 3,546 3,247 Sources: Enterprise Surveys COVID-19 Follow-up Surveys and Enterprise Surveys. Notes: All regressions are ordinary least squares and include country-sector fixed effects, as well as the controls listed in Table A.4. Standard errors are clustered at the country level. *** p < 0.01, ** p < 0.05, * p < 0.1. 26 Table A.8: Effects of Round 1 and 2 Support on Round 3 Performance (Controlling for Management Quality) Percentage Percentage Anticipate Decreased change in change in falling into employment sales employment arrears Panel A1: Combined effect of Round 1 or 2 support (comparison sample, not controlling for management quality) Received any support in Round 1 or 2 -0.875 0.297 6.877 3.209 (2.728) (4.847) (4.610) (4.694) R2 0.106 0.107 0.086 0.255 Number of observations 1,813 1,821 1,821 1,714 Panel A2: Combined effect of Round 1 or 2 support (controlling for management quality) Received any support in Round 1 or 2 -0.805 0.284 7.034 3.059 (2.727) (4.796) (4.522) (4.559) Management quality -4.275 0.597 -7.181 5.272 (5.604) (9.526) (12.336) (5.673) R2 0.107 0.107 0.086 0.256 Number of observations 1,813 1,821 1,821 1,714 Panel B1: Separate effects of Round 1 and 2 support (Comparison sample, not controlling for management quality) Received in Round 1 only dummy -0.455 3.859 6.407 3.535 (2.298) (3.498) (5.476) (5.059) Received in Rounds 1 and 2 dummy 3.354 -4.641 12.123 8.679* (3.802) (7.816) (6.990) (4.818) Received in Round 2 only dummy -6.756 0.264 0.963 -4.856 (4.889) (7.516) (4.930) (5.965) R2 0.112 0.111 0.088 0.261 Number of observations 1,813 1,821 1,821 1,714 Panel B2: Separate effects of Round 1 and 2 support (Controlling for management quality) Received in Round 1 only dummy -0.378 3.846 6.524 3.360 (2.303) (3.403) (5.377) (4.942) Received in Rounds 1 and 2 dummy 3.463 -4.660 12.300* 8.529* (3.814) (7.861) (6.950) (4.679) Received in Round 2 only dummy -6.727 0.241 1.171 -4.910 (4.884) (7.426) (4.928) (5.901) Management quality -4.541 0.781 -7.186 4.934 (5.561) (9.494) (12.417) (5.558) R2 0.113 0.111 0.089 0.262 Number of observations 1,813 1,821 1,821 1,714 Sources: Enterprise Surveys COVID-19 Follow-up Surveys and Enterprise Surveys. Notes: All regressions are ordinary least squares and include sector and country fixed effects, as well as the controls listed in Table A.4. Management quality is only available for firms with 20 or more employees. Panels A1 and B1 show our main results (from Tables 5 and 6) in this smaller sample. Standard errors are clustered at the country level. *** p < 0.01, ** p < 0.05, * p < 0.1. 27 Table A.9: Effects of Round 1 and 2 Support on Round 3 Performance (Controlling for Round 1 and 2 Performance) Controlling for Sales and Employment Controlling for Outcome Variables Percentage Percentage Anticipate Percentage Percentage Anticipate Decreased Decreased change in change in falling into change in change in falling into employment employment sales employment arrears sales employment arrears Panel A: Combined effect of Round 1 or 2 support Received support in Round 1 or 2 7.522** 3.454 -3.148 -0.161 7.413** 2.099 0.935 0.332 (2.801) (3.308) (3.790) (3.126) (2.908) (3.088) (2.818) (2.790) R2 0.103 0.302 0.195 0.239 0.098 0.321 0.307 0.333 Number of observations 3,208 3,273 3,273 3,005 3,295 3,474 3,489 2,868 Panel B: Separate effects of Round 1 and 2 support Received in Round 1 only dummy 9.550** 3.862 -3.188 1.906 9.661** 2.853 1.689 1.158 (3.260) (3.211) (3.176) (4.549) (3.280) (3.372) (2.953) (4.108) Received in Rounds 1 and 2 dummy 6.130 -0.694 2.417 0.061 5.740 -1.791 4.826 0.550 (3.919) (5.306) (4.779) (3.317) (4.030) (4.026) (3.956) (2.585) Received in Round 2 only dummy 3.899 8.418* -11.261 -6.129 3.563 5.932 -7.130* -2.270 (5.603) (4.727) (7.606) (3.677) (5.742) (4.859) (3.990) (3.723) R2 0.105 0.304 0.199 0.241 0.100 0.322 0.310 0.334 Number of observations 3,208 3,273 3,273 3,005 3,295 3,474 3,489 2,868 Sources: Enterprise Surveys COVID-19 Follow-up Surveys and Enterprise Surveys. Notes: All regressions are ordinary least squares and include sector and country fixed effects, as well as the controls listed in Table A.4. Columns 1 through 4 include four additional controls: Percentage change in sales in Round 1, percentage change in sales in Round 2, percentage change in employment in Round 1, and percentage change in employment in Round 2. Instead, Columns 5 includes percentage change in sales in Round 1 and percentage change in sales in Round 2, while Columns 6 includes percentage change in employment in Round 1 and percentage change in employment in Round 2; Column 7 includes decreased employment in Round 1 and decreased employment in Round 2; and Column 8 includes anticipate falling into arrears in Round 1 and anticipate falling into arrears in Round 2. Standard errors are clustered at the country level. *** p < 0.01, ** p < 0.05, * p < 0.1. 28 Table A.10: Effects of Round 1 and 2 Support on Round 3 Performance (Wage Subsidies and Cash Transfers) Wage Subsidies Cash Transfers Percentage Percentage Anticipate Percentage Percentage Anticipate Decreased Decreased change in change in falling into change in change in falling into employment employment sales employment arrears sales employment arrears Panel A: Combined effect of Round 1 or 2 support Received support in Round 1 or 2 3.986 1.922 2.552 3.043 6.441 2.603 -0.242 6.182 (2.755) (3.874) (2.809) (3.099) (4.062) (3.552) (3.094) (6.060) R2 0.048 0.101 0.061 0.228 0.049 0.101 0.061 0.228 Number of observations 3,500 3,538 3,538 3,241 3,495 3,533 3,533 3,241 Panel B: Separate effects of Round 1 and 2 support Received in Round 1 only dummy 7.526** 2.334 1.353 3.646 8.803 5.651 -4.382 8.984 (2.596) (3.075) (2.741) (4.830) (5.950) (5.653) (3.838) (9.526) Received in Rounds 1 and 2 dummy 0.573 -3.826 11.988* 2.326 -5.336 -17.262 13.165 -4.497 (4.910) (6.042) (5.612) (3.214) (8.283) (10.063) (8.215) (5.713) Received in Round 2 only dummy -0.822 4.111 -0.045 1.416 7.912 1.813 4.092 4.023 (5.810) (8.623) (4.642) (4.217) (5.546) (3.349) (5.681) (5.435) R2 0.053 0.102 0.064 0.228 0.053 0.104 0.063 0.229 Number of observations 3,501 3,539 3,539 3,241 3,494 3,532 3,532 3,241 Sources: Enterprise Surveys COVID-19 Follow-up Surveys and Enterprise Surveys. Notes: All regressions are ordinary least squares and include sector and country fixed effects, as well as the controls listed in Table A.4. Standard errors are clustered at the country level. *** p < 0.01, ** p < 0.05, * p < 0.1. 29