Policy Research Working Paper 10284 Firms’ Digitalization during the COVID-19 Pandemic A Tale of Two Stories Edgar Avalos Xavier Cirera Marcio Cruz Leonardo Iacovone Denis Medvedev Gaurav Nayyar Santiago Reyes Ortega Finance, Competitiveness and Innovation Global Practice January 2023 Policy Research Working Paper 10284 Abstract The COVID-19 pandemic accelerated the digital transfor- lagged in their response to the pandemic, increasing the mation of businesses. Using a unique global panel dataset, gap with those that were more digitally ready. Moreover, this paper documents the patterns of digital adoption although the share of online sales across firms for all size during the pandemic across firms in 57 (mostly develop- groups increased, there is a growing concentration of online ing) countries. The data show the tale of two stories. On sales among top firms. The paper discusses some of the one hand, the pandemic drove firms to increase the use factors associated with this increase in the digital divide of digital platforms and invest in digital solutions. On and find that changes in digitalization remain even after the other hand, there is evidence that the digital divide mobility restrictions have eased. The analysis suggests that increased. There remain substantial gaps between small and the pandemic has accelerated digitalization, but some firms large firms as well as across sectors, particularly for new disproportionately benefited from the digital transforma- investments in digital solutions. Firms that did not use tion, potentially increasing the digital divide. any digital platform or channel before the pandemic, also This paper is a product of the Finance, Competitiveness and Innovation Global Practice. 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 xcirera@worldbank.org or marciocruz@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 Firms’ Digitalization during the COVID-19 Pandemic: A Tale of Two Stories∗ Edgar Avalos Xavier Cirera** Marcio Cruz** Leonardo Iacovone Denis Medvedev Gaurav Nayyar Santiago Reyes Ortega JEL classification: D22, L20, L25, O10 Keywords: Coronavirus (COVID-19), Firm-level innovation, Digitalization, Digital divide, Innovation and Technology Policy ∗ The authors gratefully acknowledge comments and feedback on the survey questionnaires as well as on earlier drafts of the paper. Authors thank the Enterprise Analysis Unit of the Development Economics Global Indicators Department of the World Bank Group for making the Enterprise Survey data available. This was a background paper to the "Bridging the Technological Divide: Technology Adoption by Firms in Developing Countries" report. The views expressed in this article are solely those of the authors and do not necessarily reflect those of the World Bank or its Board. ** Correspondent authors: Marcio Cruz (marciocruz@ifc.org); Xavier Cirera (xcirera@worldbank.org). 1. Introduction The shock associated with the COVID-19 pandemic has been large and persistent. As documented in Apedo-Amah et al. (2020) and Cirera et al. (2021), the private sector has experienced a large, persistent negative impact on sales across all regions of the world. Most firms had still not recovered the levels of sales in 2019 more than a year after the beginning of the pandemic (Cirera et al., 2021) and some sectors are experiencing structural changes in the way products or services are produced and demanded. These changes and the associated reallocation of resources have already been identified in the data from the United States (Barrero et al., 2021). A critical question for policy is how these changes will affect firm growth and aggregate produc- tivity. Are surviving firms coming out stronger from the pandemic? The answer to this question depends on many different channels that affect productivity, which in many cases are in opposing directions. For example, Harris and Moffat (2021) implemented a survey of UK firms and found that, on the one hand, 18% of firms have stopped and 45% reduced doing R&D but, on the other hand, 40% of firms have increased ICT investments. Andrews et al. (2021) find a positive reallocation effect for three OECD countries, mainly driven by technology readiness to face the shock and the additional investments in technology. This paper explores firms’ digitalization trends during the COVID-19 pandemic using a novel firm-level data set of 57 countries across three periods of time since the start of the pandemic. The paper identifies the digitalization effects on firms’ resilience, differences in adoption, and changes in market share distribution. There are at least two forces associated with the pandemic, in opposite directions, driving the decision of businesses to invest in and expand the use of digital technologies. On the one hand, the reduction of economic activity and the increased uncertainty may reduce the overall willingness to invest. On the other hand, because a key source of the pandemic restrictions is associated with a lack of mobility of consumers and workers, digital technologies have the potential to minimize disruptions by facilitating home-based work and online sales 1 . So far, the evidence from different sources suggests that the increased demand for the use of digital technologies is prevailing. High- frequency indicators suggest a significant jump in the demand for online sales. For example, Google trend indices for the expression “online shopping” in different languages suggest that interest in the use of digital technology reached a historical peak at earlier stages of the pandemic (Figure 1). 1 For example Apedo-Amah et al. (2020) shows how digital sales channels helped firms coping with the first wave of the pandemic, and Kim (2020) shows how consumers preferences changed with the pandemic, allowing firms that digitalized to access a larger market 2 120 100 80 Google Trend Index 60 40 20 0 English Spanish French Note: It shows the trend for the expression “online shopping” in English, Spanish or Portuguese, and French. Figure 1: Google trend for the expression “online shopping” This wave of digitalization comes, however, with an important risk; the deepening of the digital divide. Literature has shows that significant differences in the use of digital technologies across firms existed pre-pandemic (Cirera et al. (2022)). These can be explained by several factors such as access to infrastructure and finance or capabilities to successfully implement these technologies. While the shock on digital demand for businesses can offer an opportunity for smaller firms and those less exposed to digital technologies to catch up, it can also lead to an increasing gap, if more capable firms are disproportionately able to respond faster to this shock. For example, in retail and wholesale, the shock may lead to an increasing market share towards well-established e-commerce platforms. The objective of this paper is to describe the use of digital technologies during the pandemic and the heterogeneity in its use for 57 (mostly developing) countries, where the evidence is very scarce. Specifically, the paper analyzes the gap in digital responses to COVID-19 across businesses. It relies on three waves of the World Bank’s Business Pulse Surveys (BPS) and COVID-19 Follow- up Enterprise Surveys (WBES), which provide comprehensive information on the impact of the COVID-19 pandemic on businesses worldwide, including digital adjustments made by firms in response to the shock.2 The paper aim to address three key questions: i) What is the share of firms that started or increased the use of digital technologies in response to COVID-19 and for what business purpose? ii) How has the digital response varied across sector, size, and digital readiness of the firms? and, iii) Has the inequality in the digital sales markets increased over this period? This analysis is complemented by testing if these patterns on digitization are associated with firm 2 See Apedo-Amah et al. (2020) for more details on the BPS and www.enterprisesurveys.org for more details on the WBES. 3 capabilities or access to finance conditions pre-COVID-19. Results show that almost half of the firms increased the use of digital platforms in response to the pandemic, and about a quarter invested in new digital solutions. However, the response was not equal. Larger and more digitally ready (pre-pandemic) firms increased and invested more in digital solutions, increasing the digital divide. Inequality and concentration measures in online sales show that even if more firms are using digital channels, the gap remains due to a disproportionate increase in digital markets concentration by some firms. The remainder of this paper is structured as follows. Section 2 describes the data and method- ology used in the analysis. Section 3 provides some general descriptive statistics on the share of firms using digital technology in response to COVID-19 and analyze the main purpose for which the technology is being used for. Section 4 compares the gap in the use of and investment in digital technologies along three key dimensions: country, sector, and size of the firms. Section 5 analyzes the inequality across firms regarding online and offline sales, and remote and non-remote workers. Section 6 explores four potential factors associated with our findings: access to finance, regulations, infrastructure, and firm capabilities. The last section concludes. 2. Data and methodology The World Bank, through the Business Pulse Surveys (BPS) and the Enterprise Surveys (WBES), has collected data on more than 100,000 firms in more than 65 countries since the start of the COVID- 19 pandemic. This paper uses a harmonized dataset of both surveys, and in particular, the module on the digital response that firms’ managers implemented to counteract the negative impact brought by mobility restrictions and a change in consumption patterns (see Watanabe and Omori (2020)). In most of the paper, unless otherwise noted, the analysis restricts the sample to firms with 5 or more employees, which was a threshold used in most countries for which data is available.3 Three main indicators regarding the digital response by firms were collected. First, whether the firm has started or increased the use of digital platforms in response to the pandemic. Second, if the firm invested in the implementation or use of digital solutions; and third, the share of sales done through online channels. Throughout the paper, analyses explore how firms differed in their digital response towards the pandemic, and the possible implications of such responses. For the analysis, the paper will focus on the 57 countries 4 with data on digital changes. In total, there is information on 65,130 firms, where 25,255 are panel firms. This feature of the data allows to analyze how digital changes happened since the start of the pandemic5 . The first wave of surveys 3 The WBES uses the threshold of 5+ employees. 4 Countries include: Argentina, Bangladesh, Benin, Brazil (Sao Paulo), Bulgaria, Cambodia, Chile, Croatia, Cyprus, Czechia, the Arab Republic of Egypt, El Salvador, Estonia, Georgia, Greece, Guatemala, Honduras, Hungary, India, Indonesia, Italy, Jordan, Kenya, Kosovo, Kyrgyzstan, Latvia, Lesotho, Lithuania, Madagascar, Malaysia, Malta, Moldova, Mongolia, Morocco, Mozambique, Nepal, Nicaragua, Pakistan, Paraguay, Poland, Portugal, Romania, Senegal, Sierra Leone, the Slovak Republic, Slovenia, Somalia, South Africa, Sri Lanka, Tajikistan, Tanzania, Tunisia, Türkiye, Uzbekistan, Vietnam, Zambia, and Zimbabwe) 5 Not all countries implemented the three rounds of surveys. Within each country, some firms were observed during 4 occurred between May and October 2020, and two rounds of follow-up surveys were implemented. Wave 2 was implemented between November 2020 and April 2021; and wave 3 was implemented between May 2021 and November 20216 . The analysis is restricted to small, medium and large firms, unless otherwise specified, and excludes firms that were closed at the time of the survey, or those in the health or education sector. Throughout the paper, three main approaches are used to analyze the digital divide in response to COVID-19 across firms. First, the paper compare the digital response across groups of firms based on basic observable characteristics (e.g., country, sector, size). This exercise mostly relies on estimating and comparing the likelihood that firms in these respective groups have increased the use of or invested in digital technology in response to COVID-19. The second approach analyzes measures of inequality with focus on online sales, to assess if there was an increase in the concentration towards firms with the highest digital market share at the start of the pandemic. A third approach is decomposing the explanatory power of different set of variables (size, sector, management quality, among others) to explain changes in digitalization. These approaches are complementary and allows to investigate if the digital divide across businesses has changed across different dimensions. To implement the first approach a probit model is used to compare the likelihood of the digital response across groups of firms. Estimations control for pre-pandemic firm size, sub-sector, country fixed effects, and the constructed measure of the severity of the crisis faced by firms at the time of the survey. More specifically, the estimations follow this equation: Yit = α + βXit + γSectorst + θCountryct + δShockict + it , (1) where Yit refers to the outcome of interest of firm i in time t. The main two outcomes are dummy variables indicating whether the firm increased the use of digital platforms, and whether the firm invested in digital solutions. Time t is split into two periods, broadly defined as the first and last rounds in which the survey was conducted. For all specifications the vector of firm characteristics X includes dummies for the size of the firm based on number of workers (small; medium; and large). Estimations also control for 12 sectors and country fixed effects. To control for variation in the severity of the shock within the period the survey is implemented, Shockict includes the severity of the crisis measure using Google mobility trends around transit stations.7 All variables are interacted with a time measure t to allow for changes in the coefficients between the first and the latest rounds. Analysis estimate the margins at the means for the variable of interest (e.g., size, sector, or country), for which results are compared across time, keeping the distribution of size, sector, and country constant across waves to ensure that the results are not driven by differences in composition. To the first round but did not respond to the follow-up survey, and some of them are replacement firms collected in follow-up surveys. The methodology section in Cirera et al. (2021) describes the robustness checks implemented and the handling of the different sample of firms. 6 For more details on the surveys and aggregate results see Cirera et al. (2021) 7 Authors construct an indicator of the severity of the crisis that is a weighted average of 30-day periods since the start of the pandemic until the date of the survey. Specifically, the 30-day period average just before the survey has a weight of 1, the average from day 31-60 has a weight of 1/2, the average from day 61 to 90 has a weight of 1/3, and so on until the start of the pandemic. This approach is aligned with Apedo-Amah et al. (2020) and Cirera et al. (2021), which analyze the overall results from the 1st and 2nd waves of BPS and the WBES follow up. 5 control for variations in the sample size by countries, estimations are weighted by using the inverse of the number of observations in each country, that is, the sum of weights of each country has the same weight in our summary statistics.8 For the second approach, the paper compares measures of inequality based on the distribution of total online sales across firms. First, the average monthly sales pre-pandemic is multiplied by the change in sales observed in each wave of the BPS as a reference to estimate the monthly sales during COVID-19.9 Then, a multiplication of the estimated sales for each firm by the share of online sales is calculated, which allows to compute the value of online and off-line sales. The paper then uses these variables to estimate the relative mean deviation (RMD) for the share of online sales. This mean deviation measure is calculated as follows: N n=1 |Yi − µ| RM D = (2) N This analysis is complemented by computing the Gini coefficient for online and offline sales and compare them across time. A third approach used in the paper is a Shapley decomposition. To calculate the Shapley values for each group of variables in a regression (i.e. size, sector, management quality), n! regressions are estimated by eliminating each component at once and then taking the average of the contributions of the component. Shapley values represent the proportion of the R-squared explained by each component. It is important to note that these values do not point out the direction of the effect, but rather identify which grouping of variables contributes the most at explaining differences in each of the outcomes of interest. 3. An unprecedented wave of digitalization The stringency of the measures adopted to reduce the spread of the pandemic created a large disruption to the way firms had to operate. Restrictions restrained workers’ mobility, introduced bottlenecks in supply chains and made it difficult for firms to reach consumers in person. These constraints, in turn, induced changes in production and demand patterns to which firms had to rapidly respond if they wanted to continue to operate. Indeed, Apedo-Amah et al. (2020) found evidence of accelerated digitalization among firms as an immediate response to the pandemic with BPS data collected between April and August 2020. Using a subsequent wave of these data, Cirera et al. (2021) document that this pattern of digitization persisted between September 2020 and February 2021 in medium-term patterns (for more details on digital adoption see Comin et al. 8 In some countries, sampling weights are available in order to produce nationally representative results at the country level, but for cross-country comparison purposes, these weights are not included. 9 This exercise is based on strong assumptions of no seasonality, which means that sales per month in 2019 were constant so total sales can be recovered using the percentage change in sales in each wave. Because there is no information on differences in sales across firms in levels, one way to think about this exercise is that we weight the changes in sales observed in the BPS by the participation in the total sales of the economy pre-COVID 19. 6 (2022)). Specifically, firms responded by accelerating the use of existing digital technologies and investing in new digital solutions. This expansion in digitalization allowed firms to better reach consumers (or reach new ones) and organize production more flexibly and efficiently. Figure 2 shows that nearly 45% of businesses had increased the use of digital platforms in the latest available data, compared to 32% in the first wave. Furthermore, 28% of firms had invested in new digital solutions, a 8 percentage points increase from wave 1. Although the intensity of these investments is not measured in these survey data, the number of firms taking actions to accelerate digitization has been unprecedented. 60 45 44 40 Share of firms 28 23 20 0 Increase in digitalization Increased use of digital platforms Have online sales Invested in digital solutions Have remote workers Note: Results from the latest round of available data. Figure 2: Fraction of businesses by responses to the shock. In addition, the increase in digitalization has been steady. For countries with all three rounds of survey data, the percentage of firms increasing the use of digital technologies has gone up from 28 percent to 41 percent, and 47 percent in the latest round (Figure 3). Moreover, the increase in digitalization has been widespread across the world. Even after controlling for the magnitude of restrictions on mobility, there was little difference in the uptake of digital technologies between countries at different levels of income. In fact, the advent of digitalization in response to the pandemic was lower in many high-income countries across Europe, compared with low- and middle-income countries across Asia and Africa. For example, the likelihood of firms starting or increasing their use of digital technologies in Hungary, Portugal, Italy, Slovenia, and the Slovak Republic was less than 20% during the first wave of the survey and did not exceed 30% during the second wave of the survey. In contrast, the corresponding share increased from around 50% to more than 80% in Kenya, from around 30% to more than 60% in Tunisia, and from around 30% to almost 60% in El Salvador between the two survey rounds. Nevertheless, pre-pandemic gaps in digitalization still remain. More firms in high income countries use some digital technologies but have not increased 7 their use after the pandemic. (a) By BPS round (b) By income groups Prob. of starting or increasing use of digital 60 100 47 80 Cumulative predicted probability 41 40 60 28 40 20 20 0 0 Developing countries High income 1 2 3 Survey Round Started Increased Use but didn't increase Note: (a) Result from 24 countries with three rounds of data. Aggregate indicator might differ from figure 2 which includes all available data. (b) BPS countries includes all 4 categories. For WBES data, previous rounds are used to identify firms that used digital platforms but not increased use. Figure 3: Likelihood of starting or increasing the use of digital technology 4. The digital divide Even before COVID-19 hit there was growing evidence of a digital divide in firms (see Bach et al. (2013)). Results show differences in digital adoption across firm sector, size or manager characteristics that over time have persisted. Different channels such as access to finance, access to human capital, or differences in management practices have been hypothesized as mechanisms that might prevent this gap from closing. When lock-downs and customers’ behavior changed due to the pandemic, many of those channels were exacerbated. A report from the European Commission shows that about a third of jobs in the EU could be done remotely, and that the COVID-19 shock expanded remote work to younger workers in "blue-collar" occupations. Nevertheless, its expansion and persistence depends on improving firm digital capabilities. Vo et al. (2022) show that firms with higher digital infrastructure before COVID-19 were able to perform better during the crisis. Furthermore, the digital divide can be higher in developing countries where a large percentage of firms face access restrictions to infrastructure, skills, and networks (Ndulu et al. (2022)). This sections analyzes the heterogeneity in the uptake of digital technologies across sectors, firm-size categories, and business function purpose, in response to the COVID-19 pandemic. 4.1 Differences across sectors The increase in digitalization after the pandemic shock amplified existing differences by sectors. The firms with the highest likelihood of starting or increasing the use of digital technologies, with more than 60%, were firms in knowledge intensive sectors (ICT and financial services). This is not entirely surprising given that these services are the most amenable to home-based work (Dingel 8 and Neiman, 2020). In contrast, the likelihood of firms starting or increasing the use of digital technologies, at around 35%, was the lowest in agriculture, mining, manufacturing, construction and transportation - sectors least amenable to home-based work. There is a set of sectors where the uptake of digital technologies lies somewhere in between these two extremes. These sectors probably are the ones where COVID-19 accelerated their digitalization processes. For example, the likelihood of firms starting or increasing the use of digital technologies was between 47% and 51% in accommodation, food services, and wholesale and retail trade. These sectors are among those most intensive in face-to-face interactions with consumers (Avdiu and Nayyar, 2020). And while the share of jobs that can be done from home are low because in-person delivery remains necessary, digital platforms have enabled these service providers to replace in- person transactions with online transactions. These patterns across sectors are analogous when looking at firms’ investment in digital tech- nologies. The likelihood of such investment ranged from around 43% in ICT and financial services, and around 30% in all other sectors. This positive result shows that even firms in sectors that tend to have lower investments in digital technologies have found reasons to make such investments since the start of COVID-19. 60 60 51 47 43 44 Predicted probability 40 38 35 31 31 28 27 27 20 0 Knowledge Retail Accomm & Other Manufacturing Agriculture intensive food serv. services services Increasing use of digital platforms Investing in digital solutions Figure 4: Predicted probability of digital response to COVID-19 by sector The digital divide within sectors is increasing, especially in less technology intensive sectors. Figure 5 shows that for all sectors, firms that were more digital ready pre-pandemic were also more likely to increase the use of digital technologies relative to those that were lagging behind before the pandemic. In almost all cases, the more digital advanced firms pre-pandemic were two or three times more likely to increase the use of digital platforms than those less prepared. These differences were particularly relevant for agriculture where differences in scale and access to infrastructure and skills play a key role in amplifying the digital divide10 . 10 For example, see some discussions about the digital divide in agriculture Warren et al. (2002) for the United Kingdom and in Moon et al. (2012) for the Republic of Korea. 9 Probability of increasing use of digital tech. 80 71 70 68 66 63 64 60 40 37 23 24 20 20 21 16 0 Agriculture Manufacturing Retail Other Accomm & Knowledge services food serv. intensive services Low digital readiness firms (pre-pandemic) High digital readiness firms (pre-pandemic) Figure 5: Predicted probability of digital response to COVID-19 by digital readiness (pre-pandemic) and sector 4.2 Differences across firm size The increase in the use of digital technologies during the COVID-19 pandemic is correlated with firm size. Large firms performed better than small firms with respect to the uptake of digital solutions. The likelihood of starting or increasing the use of digital technologies ranged from 34% for micro firms11 to 43% for small firms, 50% for medium-sized firms, and 52% for large firms (Figure 6). These differences across firm size groups are starker when looking at investments in technology adoption.The likelihood of investing in digital solutions ranged from 20% for micro firms to 49% for large firms. This is not surprising as investments in new digital solutions often entail high fixed costs. The World Bank (2022) using more granular data for Malaysian firms identifies that the digital divide across firm size is not only present in the extensive margin. Medium firms invested 3 times more as a share of their sales than small firms, and large firms invested five times more. As will be shown in the next subsection, not only the size but also the focus of such investments has been different. 11 Only for this analysis micro firms are included for the subset of countries with available data 10 60 52 50 49 43 Predicted probability 40 38 34 29 20 20 0 Micro (0-4) Small (5-19) Medium Large (100+) (20-99) Increasing use of digital platforms Investing in digital solutions Figure 6: Predicted probability of digital response to COVID-19 by size Differences in the use of digital platforms across size groups, before and after the pandemic, are also persistent. These differences both in terms of use of digital technologies, as well as in investment decisions, show an increase in the digital divide across firms’ size. There are high levels of heterogeneity in firms’ digital response to the shock within firm size groups. Firms with no or low use of digital technologies before the pandemic are the ones with the slowest response to the shock, creating a digital divide also within firm size. Moreover, differences across firm size can be explained to a larger extent due to differences in responses by the laggards (Figure 7). Probability of increasing use of digital tech. 80 68 68 69 60 60 40 33 26 27 20 16 0 Micro (0-4) Small (5-19) Medium Large (100+) (20-99) Low digital readiness firms (pre-pandemic) High digital readiness firms (pre-pandemic) Figure 7: Predicted probability of digital response to COVID-19 by digital readiness (pre-pandemic) and size 11 4.3 Differences across business functions An additional element to assess whether there is uneven or incomplete digitalization is to understand what are the business functions where firms invest in digital solutions. The first wave of BPS collected information on the business function associated with investments in digital solutions in response to COVID-19. Figure 8 shows a large concentration of digital investments in functions related to customer relations - sales, marketing and payments. Specifically, 64% of firms increased digitalization in marketing, 55% in sales, and 37% in payments. Overall, among firms that have adopted digital technologies in response to the pandemic, 56% used it for internal (management and production) and external (sales, marketing, etc.) business functions. Another 34% used it only for external purposes, while around 10% used it only for internal purposes. This responds to some of the most pressing challenges imposed by the restrictions associated with the pandemic, since it provides flexibility to connect with consumers. But it also suggests that the digital transformation may not be as complete as expected, and a large number of firms are still not applying digital technologies to management and production functions. Note: Results obtained from Wave 1 which asked for a disaggregation by business functions Figure 8: Predicted probability of type of digital technologies used. Conditional on using digital technologies The same World Bank report using data from Malaysian firms find that "while 71 percent of surveyed SMEs used social media for product communication and marketing purposes, only 44 percent actually engaged in e-commerce activities. Even when they did engage in such activities, most payments were still transacted in cash or through a separate banking transaction, rather than through an integrated payment gateway to enable a seamless online transaction" (World Bank (2022)). Furthermore the report shows that the focus of investments in digital platforms has not been equal across firm size. While smaller firms are investing in online sales channels and social media for marketing, investments in more advanced technologies such as CRM, SRM, or ERP have been minimal. 12 4.4 Differences across firm capabilities (pre-pandemic) An important factor associated with the digital divide in response to the pandemic is the gap in digital readiness and overall firm capabilities pre-pandemic. Figure 9 shows the distribution of the information about increasing use and investment in digital technologies based on three groups of firms classified by the levels of digitalization pre-pandemic.12 The figure shows significant persistence in digitalization, with a ratio close to 3 in likelihood of using and investing in digital technologies between ex-ante highly digitalized companies compared to those that had prior low digitalization. These results are aligned to previous findings combining the BPS data with very gran- ular information on pre-existent digital technologies, from the Firm-level Adoption of Technology for firms in Brazil, Senegal, and Vietnam, as described by (Comin et al., 2022). 80 68 60 Predicted probability 52 45 40 28 29 20 16 0 Low Medium High Digital readiness score (pre-pandemic) Increasing use of digital tech. Investing in digital solutions Figure 9: Predicted probability of digital response to COVID-19 by digital readiness (pre-pandemic) A potential explanation for this persistence in digitalization might be associated with the capabil- ities for digital upgrading and awareness. When comparing similar firms in terms of size and sector but differences in management practices, Figure 10 shows that among firms that did not increase the use of digital technologies in response to the pandemic, those adopting better managerial practices pre-pandemic were less likely to do report that they did not need digital upgrading.13 . The role 12 Groups are created using the number of digital technologies used by the firms before the pandemic from the following list: digital platforms for sales or digital payments; use of online chat, social media or big data for marketing and product development; software for customer or supply chain relationship management; and enterprise resource planning for operations or business administration. Low are those firms with 0 or 1 technology, medium those with 2, and high those with 3 or 4. 13 Groups are created using the number of good management practices implemented by firms before the pandemic from the following list: compared sales to targets at least once per month, advertised the business, or made a promotion based solely on performance and ability factors. Low group are those firms with 0 practices, medium those with 1, and high those with 2 or 3. 13 of management quality in digitalization has been documented before. For example, Grimes et al. (2012) show the correlation between management quality and broadband access, and Ifinedo (2011) identifies management commitment as a key driver for the adoption of e-business technologies in firms. While the pandemic shock provided a large monetary incentive for firms to digitalize, management barriers are still present and will limit any further progress especially in smaller firms where low management is more likely to be present. % of firms that do not see the need of upgrading digitally 60 50 42 40 33 20 0 Low Medium High Management score (pre-pandemic) Figure 10: Probability of reporting no need for digital upgrade (conditional on not increasing use of digital technology) 5. Inequality in online sales The previous section provides a comprehensive perspective of the increasing digital divide in response to the pandemic, but how large is this gap? Despite small firms catching up on using digital technology in response to the pandemic, the size of the gap has increased. Even if a larger share of firms are relying on digital technologies for sales, a small number of firms disproportionately benefit from online sales. This is particularly relevant for sectors such as retail, where the "winner takes all" mechanism may prevail due to strong network effects (i.e., the larger the number of consumers and suppliers, the larger the value of the platform). To investigate this issue, this section focuses on the digital divide regarding online sales. Figure 11 shows a significant heterogeneity in the share of firms having online sales across sectors and firm size groups. Firms in the hospitality sectors are more likely to have online sales, as well as firms in knowledge intensive services and in retail. Interestingly, firms in sectors that do not rely so much on face-to-face interactions with customers, such as manufacturing or agriculture, still have a large share of firms (between 34% and 42%) using online sales. In terms of firms’ size, larger firms are more likely to have online sales channels compared to micro or small firms. Nevertheless, 40% of micro firms have established those channels, which highlights the relevance that digital platforms are having across all types of firms. 14 60 60 54 51 51 50 48 % of firms that have online sales % of firms that have online sales 42 42 44 40 40 40 34 20 20 0 0 Hospitality Knowledge Retail Manufacturing Other Agriculture intensive services Micro (0-4) Small (5-19) Medium Large (100+) services (20-99) Figure 11: Predicted probability of having online sales 5.1 The intensive margin of digital sales To explore the size of the gap, this section starts by examining how dissimilar is the use of technologies across firms and how that has changed with the pandemic. Figure 12 and Figure 13 show a bright story about how the pandemic has pushed more firms into the digital sphere. The total value of online sales as a share of 2019 sales has increased over time, as captured by the different rounds of the BPS survey. This comes from more firms establishing online sales channels, as well as online sales having a more prominent role in total sales. At all percentiles of intensity of online sales in wave 1, the value of online sales has increased. For example, while the median firm did not generate any online sale during wave 1, during the latest wave, it generated 7.9% of its 2019 sales via online channels; the 75th percentile went from 2.3% to 28.8%, and the 90th percentile from 19.9% to 64%. 15 200 Value of online sales / Monthly sales 2019 150 100 50 0 0 20 40 60 80 100 Percentile of intensity of online sales Wave 1 Wave 2 Wave 3 (May,2020-Oct,2020) (Nov,2020-Apr,2021) (May,2021-Nov,2021) Figure 12: Distribution of share of online sales across firms and waves Figure 13 also shows how, in relative terms, the differences in the share of online sales across firms in different sectors have closed. The relative mean deviation measures how disperse is the share of online sales is within each sector. An indicator closer to 1 shows high variance (i.e. some firms with high participation in online markets and many with low or no participation), and an indicator closer to 0 shows that all firms have the same share of online sales. The figure shows that even in sectors such as agriculture or manufacturing where at the start of the pandemic there was a large variance in terms of participation of online channels, the latest round of data shows a decrease in the relative mean. The largest drop comes from firms in the hospitality sectors. This result is consistent with the magnitude of the shock for these firms and how their recovery has relied on digital channels. 16 1.00 0.80 0.73 Relative mean deviation 0.70 0.71 0.71 0.70 0.66 0.68 0.61 0.60 0.60 0.55 0.55 0.52 0.40 0.20 0.00 Agriculture Manufacturing Other Hospitality Retail Knowledge services intensive services First round Latest round Note: Relative mean deviation range: [0,1), where values closer to 1 indicate a higher dispersion in the group. Figure 13: Inequality and levels of online sales by sector 5.2 A deeper look into online sales. The case of Vietnam To further measure the magnitude of the digital divide between firms, this section compares the online and offline Gini coefficients for sales for the overall sample and the retail sector in Vietnam.14 The retail sector was among those where online sales were more relevant in general, and in the context of the pandemic. Figure 14 and Figure 15 show the other tale of the digitalization story during the pandemic. The figures show the Lorenz curve that plots the accumulated concentration in online and offline total sales. The results suggest three main facts: First, online markets are dominated by few firms. The top 5th percentile represents more than 80% of the total share of online sales. Second, that inequality is 10 percent higher than what is observed in offline channels. And third, even if more firms entered the digital markets after the pandemic, this inequality has remained stable or even increased during the peak of the pandemic. These facts are observed both for the overall sample and for a restricted sample of retail firms highlighting the importance of having a dominant position in the online markets, and how the shock has made it more relevant. 14 The choice of Vietnam was motivated by the fact that the data has a good coverage for the 2019 sales. 17 All Retail 1 1 Cumulative sum of online sales .8 .8 Cumulative sum of online sales .6 .6 .4 .4 .2 .2 0 0 0 20 40 60 80 100 0 20 40 60 80 100 Population percentage Population percentage Jun, 2020 (Gini = .971) Oct, 2020 (Gini = .983) Feb, 2021 (Gini = .973) Jun, 2020 (Gini = .926) Oct, 2020 (Gini = .946) Feb, 2021 (Gini = .938) Figure 14: Lorenz curve for online sales by wave All Retail 1 1 .8 .8 Cumulative sum of in-person sales Cumulative sum of in-person sales .6 .6 .4 .4 .2 .2 0 0 0 20 40 60 80 100 0 20 40 60 80 100 Population percentage Population percentage Jun, 2020 (Gini = .879) Oct, 2020 (Gini = .891) Feb, 2021 (Gini = .879) Jun, 2020 (Gini = .827) Oct, 2020 (Gini = .883) Feb, 2021 (Gini = .848) Figure 15: Lorenz curve for offline sales by wave Another way of measuring inequality is showing the ratio between the median firm and the 90th percentile firm. Figure 16 shows this measure for the sales per worker by channel - offline vs online sales. As mentioned above, inequality is larger in the online sales market. For every dollar that the median firm sells online, the 90th percentile firm sells 24.4. This differs vastly from offline channels where one dollar for the median firm corresponds to 10.5 in the 90th percentile. This is 2.3 times lower than in online markets. Overall, this brings a new perspective and challenge for competition policy. 18 25 24.4 Gap in sales per channel per worker 20 P90 vs Median 15 10.6 10 5 0 Offline Online Conditional on having each channel Figure 16: Ratio of sales value by channel P90 relative to median 6. What drives these results? Several factors may explain firms’ decisions to increase digital investments as a response to the pandemic, and the heterogeneous response that explains an increasing digital divide. This section tries to understand more formally what are the firm characteristics pre-pandemic that may explain these firm digital choices. For this exercise, The analysis is restricted to the sample to countries for which pre-pandemic information from WBES was available. Specifically, this analysis estimates Equation 1 using some variables extracted from the baseline WBES survey,15 and check the correlation for two potential channels: i) The technological capabilities of firms before the shock; ii) Access to finance. Table 1 and Table 2 summarize the main results. The first hypothesis is that firms with higher levels of technology capabilities pre-pandemic were able to adjust faster to digital adoption. Comin et al. (2022) combine data from the Firm-level Adoption of Technology survey with the BPS for Brazil, Senegal, and Vietnam and show that firms using a higher level of technology sophistication pre-pandemic were significantly more likely to increase the use to digital technology, which also contributed to a better sales performance at the initial stage of the shock. Using a different sample and simpler proxies for technological capabilities pre-pandemic - such as having own website and spending on R&D - the data also shows that these characteristics are significantly and positively associated with the likelihood that firms start or increase the use of digital technologies, increase the share of remote workers, and the share of online sales after COVID-19. Thus, technological capabilities pre-pandemic are also associated with larger digital adoption during the pandemic, which suggests persistence and an increasing risk of the digital divide moving forward. 15 The surveys was implemented over the 2018-2019 period. 19 Using digital technologies Having remote workers (1) (2) (3) (4) (5) (6) Having own website 0.094∗∗∗ 0.094∗∗∗ 0.088∗∗∗ 0.028∗∗ 0.028∗∗ 0.024∗ (0.013) (0.013) (0.013) (0.012) (0.012) (0.012) Spending on RD 0.027 0.030 0.032 0.068∗∗∗ 0.069∗∗∗ 0.062∗∗∗ (0.020) (0.020) (0.020) (0.019) (0.019) (0.019) Banking account 0.025∗∗ 0.027∗∗ 0.030∗∗ -0.006 -0.005 -0.005 (0.013) (0.013) (0.013) (0.012) (0.012) (0.012) Exporter 0.023 0.024 0.024 0.032∗∗ 0.032∗∗ 0.032∗∗ (0.014) (0.014) (0.014) (0.013) (0.014) (0.013) Productivity quantile (1-100) -0.000 -0.000 -0.000 0.001∗∗∗ 0.001∗∗∗ 0.001∗∗∗ (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Electrical outages (region) -0.019 -0.018 (0.014) (0.013) Finance constraint (region) -0.001 0.001 (0.001) (0.001) Age 5-9 years -0.055 -0.055 -0.059 -0.022 -0.024 -0.015 (0.043) (0.043) (0.045) (0.034) (0.035) (0.036) Age 10+ years -0.081∗∗ -0.080∗ -0.077∗ -0.030 -0.032 -0.022 (0.041) (0.041) (0.043) (0.031) (0.031) (0.032) Medium (20-99) 0.034∗∗ 0.035∗∗ 0.032∗∗ 0.034∗∗∗ 0.035∗∗∗ 0.039∗∗∗ (0.015) (0.015) (0.014) (0.013) (0.013) (0.013) Large (100+) 0.041∗∗ 0.040∗∗ 0.038∗∗ 0.144∗∗∗ 0.145∗∗∗ 0.148∗∗∗ (0.018) (0.018) (0.017) (0.018) (0.018) (0.017) Const and utilities -0.029 -0.030 -0.018 0.043∗ 0.041∗ 0.043∗ (0.022) (0.022) (0.021) (0.023) (0.023) (0.023) Retail and wholesale 0.171∗∗∗ 0.170∗∗∗ 0.166∗∗∗ -0.001 -0.002 -0.005 (0.016) (0.016) (0.016) (0.015) (0.015) (0.014) Transp and storage 0.078∗∗ 0.076∗∗ 0.078∗∗ 0.139∗∗∗ 0.137∗∗∗ 0.135∗∗∗ (0.038) (0.038) (0.037) (0.039) (0.039) (0.039) Accomm 0.085∗ 0.086∗ 0.121∗∗ 0.085∗∗ 0.084∗∗ 0.071∗ (0.044) (0.044) (0.047) (0.039) (0.039) (0.038) Food prep serv 0.143∗∗∗ 0.143∗∗∗ 0.140∗∗∗ 0.002 -0.001 -0.007 (0.039) (0.039) (0.036) (0.032) (0.031) (0.033) ICT 0.192∗∗∗ 0.192∗∗∗ 0.191∗∗∗ 0.237∗∗∗ 0.235∗∗∗ 0.230∗∗∗ (0.043) (0.043) (0.044) (0.041) (0.041) (0.042) Other serv 0.027 0.029 0.017 0.043 0.044 0.035 (0.036) (0.036) (0.035) (0.041) (0.042) (0.041) Constant -0.093 -0.091 -0.006 -0.161∗∗ -0.160∗∗ -0.102∗∗ (0.084) (0.084) (0.053) (0.075) (0.075) (0.048) Observations 8842 8797 8842 8711 8668 8711 R2 0.117 0.117 0.143 0.190 0.190 0.212 Note. Results from linear probability model. Clustered standard errors at the country-region-size-sector in parentheses. Specifications (1)-(2) and (4)-(5) include country dummies. Specifications (3) and (6) include country-region dummies. All specifications control for the severity of the crisis based on Google mobility (See section 2). ∗ p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01. ‘Electrical outage (region)’ refers to the share of firms experiencing electrical outages during the previous fiscal year in the same sub-national region. ‘Finance constraint (region)’ refers to the share of firms identifying access to finance as a major constraint in the same sub-national region. ‘Banking account’ refers to having checking or saving account. Table 1: Factors associated with increasing use of digital and remote work 20 Having online sales Share of online sales (1) (2) (3) (4) (5) (6) Having own website 0.064∗∗∗ 0.064∗∗∗ 0.060∗∗∗ 1.557∗∗ 1.552∗∗ 1.382∗∗ (0.016) (0.017) (0.017) (0.625) (0.629) (0.630) Spending on RD -0.001 0.001 0.001 0.151 0.238 0.262 (0.021) (0.021) (0.022) (0.801) (0.804) (0.837) Banking account 0.022 0.022 0.026∗ 1.170∗ 1.187∗ 1.240∗∗ (0.015) (0.015) (0.016) (0.603) (0.615) (0.620) Exporter -0.010 -0.008 -0.005 1.103∗ 1.134∗ 1.260∗∗ (0.017) (0.017) (0.017) (0.642) (0.639) (0.634) Productivity quantile (1-100) -0.000 -0.000 -0.000 -0.004 -0.003 -0.005 (0.000) (0.000) (0.000) (0.009) (0.009) (0.009) Electrical outage (region) -0.010 -0.594 (0.016) (0.632) Finance constraint (region) -0.001 -0.025 (0.001) (0.023) Age 5-9 years -0.001 -0.003 -0.010 0.830 0.784 0.550 (0.064) (0.065) (0.067) (3.370) (3.414) (3.468) Age 10+ years -0.011 -0.014 -0.014 -0.980 -1.042 -1.014 (0.060) (0.061) (0.062) (3.247) (3.290) (3.346) Medium (20-99) 0.021 0.021 0.022 -1.632∗∗∗ -1.610∗∗∗ -1.585∗∗∗ (0.017) (0.017) (0.017) (0.590) (0.589) (0.576) Large (100+) 0.039∗ 0.039∗ 0.037∗ -1.210∗ -1.187∗ -1.327∗ (0.021) (0.021) (0.021) (0.701) (0.702) (0.695) Const and utilities -0.062∗∗ -0.064∗∗ -0.058∗∗ -0.688 -0.749 -0.652 (0.025) (0.025) (0.025) (0.793) (0.794) (0.758) Retail and wholesale 0.116∗∗∗ 0.116∗∗∗ 0.115∗∗∗ 2.872∗∗∗ 2.836∗∗∗ 2.764∗∗∗ (0.023) (0.023) (0.023) (0.673) (0.675) (0.647) Transp and storage 0.014 0.011 0.008 2.961∗ 2.894∗ 2.834∗ (0.042) (0.043) (0.041) (1.638) (1.652) (1.550) Accomm 0.195∗∗∗ 0.196∗∗∗ 0.223∗∗∗ 7.402∗∗∗ 7.453∗∗∗ 8.717∗∗∗ (0.050) (0.050) (0.051) (2.687) (2.699) (2.676) Food prep serv 0.165∗∗∗ 0.165∗∗∗ 0.163∗∗∗ 5.783∗∗∗ 5.755∗∗∗ 5.504∗∗∗ (0.045) (0.045) (0.043) (1.906) (1.899) (1.889) ICT 0.083∗∗ 0.083∗ 0.078∗ 6.807∗∗∗ 6.828∗∗∗ 6.721∗∗∗ (0.042) (0.042) (0.042) (2.166) (2.181) (2.168) Other serv -0.030 -0.029 -0.035 -0.096 -0.066 -0.371 (0.036) (0.036) (0.036) (1.215) (1.214) (1.214) Constant -0.231∗∗ -0.227∗∗ 0.524∗∗∗ -5.022 -4.688 23.159∗∗∗ (0.096) (0.096) (0.108) (4.210) (4.220) (4.836) Observations 6324 6295 6324 6324 6295 6324 R2 0.082 0.082 0.101 0.099 0.100 0.115 Note. Results from linear probability model. Clustered standard errors at the country-region-size-sector in parentheses. Specifications (1)-(2) and (4)-(5) include country dummies. Specifications (3) and (6) include country-region dummies. All specifications control for the severity of the crisis based on Google mobility (See section 2). ∗ p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01. ‘Electrical outage (region)’ refers to the share of firms expe- riencing electrical outages during the previous fiscal year in the same sub-national region. ‘Finance constraint (region)’ refers to the share of firms identifying access to finance as a major constraint in the same sub-national region. ‘Banking account’ refers to having checking or saving account. Table 2: Factors associated with increasing in online sales A second hypothesis is related to access to finance and the fact that many firms were financially constrained to invest in digital technologies to respond to the pandemic. While the data set has 21 limited variables to capture this information, the analysis suggests that "having a checking or savings account" is also positively associated with increasing the use of digital technology and having online sales. Using similar data from WBES, Amin and Viganola (2021) show that firms with better access to finance were significantly less likely to experience a decline in sales during COVID-19.16 An important element is that even after controlling for proxies that aim to capture potential drivers of adoption, results still suggest a robust gap between firm size groups in the participation in digital markets. For investment in digital solutions, size seems the most robust observable characteristic for which significant differences are observed. In terms of sectors, ICT, retail and wholesale, and food services are among those with a more robust response towards a significant increase along different dimensions of digital adoption. Retail and food services are typically characterized by having a large share of SMEs and hiring a large number of workers. Even after controlling for several factors, there is low explanatory power about the decision of starting or increasing the use of digital technologies. To investigate this issue, figure 17 presents a Shapley decomposition for differences of the estimations described in columns (3) and (6) of Table 2. This exercise decomposes these differences by factors attributable to specific sources. The results show that differences in observable characteristics of the firm and differences across firms do not explain a large portion of the observable differences in digitalization indicators. Furthermore, differences across countries explain only 7% of the differences in digitalization and about 90% remains unexplained. This result highlights the relevance of having a more in-depth analysis of how and which technologies are adopted by firms. For example Cirera, Comin, and Cruz (2022) show that technology adoption heterogeneity is not only large across similar firms but also within firms; across business functions inside the firm. Taking into account this heterogeneity is a key step to design better support mechanisms for firms’ digitalization. 16 The authors used as a proxy for access to finance the average of firms within sector/size/region that apply for a loan and the loan amount was approved in full or did not apply for a loan because it had enough internal funds. 22 89.9 88.5 80 Relative Shapley Value (%) R2 explained by group 60 40 20 7.2 7.7 .9 .8 2 3.1 0 Diff. across Diff. across Diff. across Unexplained firms firms countries performance characteristics Having online sales Share of online sales Firm decisions include having a website, spend in R&D, having a bank account, size, productivity and being an exporter. Firm characteristics include subsector and age groups Figure 17: Shapley-Shorrocks decomposition of firms’ digitalization decisions 7. Conclusion and policy implications The unprecedented shock caused by the COVID-19 pandemic has led to an acceleration of the digital transformation worldwide. This acceleration of digitization, however, has been very uneven across firms. the evidence presented here suggests an increase in the digital divide along several dimensions: across countries, sectors, firms size and business functions. The paper identifies several factors that could be behind this uneven acceleration of the digital transformation. Among these factors, the results highlight the role of firm capabilities in digital uptake and the persistence in digitization - those more digital ready pre-pandemic are also those that are more likely to invest in new digital tools. These results have important implications for policy. First, the worldwide increase in the use of digital technologies shows that some of the perceptions or reluctance among entrepreneurs to digitalize their businesses may have changed, and digital upgrading programs may be able to overcome some of the traditional lack of take up. Second, targeting, which has been always a challenge in this type of program, is ever more important given the increase in the digital divide. Increasing the participation of smaller firms in the use of digital markets is critical. But also ensuring that firms fully digitalize their business, and not only those functions more related to dealing with customers - such as marketing and online sales - that were more pressing responses to the pandemic restrictions, but also other areas of the firm such as production or supply chain management. Quite likely, these are the areas where more information is needed among smaller firms on what technologies are available, and how to successfully adopt them. The fact that firms capabilities continue to be a critical factor in explaining persistence in the adoption of digital technologies also suggest the need to work with entrepreneurs on supporting the accumulation of this knowledge 23 and organizational capabilities as part of the process of digital adoption. References M. Amin and D. Viganola. Does better access to finance help firms deal with the covid-19 pandemic? evidence from firm-level survey data. Policy Research Working Paper n. 9697, 2021. D. Andrews, A. Charlton, and A. Moore. COVID-19, productivity and reallocation: Timely evi- dence from three OECD countries. OECD Economics Department Working Papers 1676, OECD Publishing, July 2021. URL https://ideas.repec.org/p/oec/ecoaaa/1676-en.html. M. C. Apedo-Amah, B. Avdiu, M. C. Xavier Cirera, E. Davies, A. Grover, L. Iacovone, U. Kilinc, D. Medvedev, F. O. Maduko, S. Poupakis, J. Torres, and T. T. Tran. Businesses through the Covid-19 Shock: Firm-Level Evidence from Around the World. Policy Research Working Paper 9434, World Bank, October 2020. B. Avdiu and G. Nayyar. When face-to-face interactions become an occupational hazard: Jobs in the time of covid-19. Economics Letters, 197:109648, 2020. M. P. Bach, J. Zoroja, and V. B. Vukšić. Determinants of firms’ digital divide: A review of recent research. Procedia Technology, 9:120–128, 2013. W. Bank. Digitalizing smes to boost competitiveness, 2022. J. M. Barrero, N. Bloom, S. J. Davis, and B. H. Meyer. COVID-19 Is a Persistent Reallocation Shock. AEA Papers and Proceedings, 111:287–291, May 2021. doi: 10.1257/pandp.20211110. URL https: //ideas.repec.org/a/aea/apandp/v111y2021p287-91.html. X. Cirera, M. Cruz, A. Grover, L. Iacovone, D. Medvedev, M. Pereira-Lopez, and S. Reyes. Firm recovery during covid-19: Six stylized facts. Policy Research Working Paper 9810, World Bank, October 2021. X. Cirera, D. Comin, and M. Cruz. Bridging the Technological Divide: Technology Adoption by Firms in Developing Countries. World Bank Publications, 2022. D. A. Comin, M. Cruz, X. Cirera, K. M. Lee, and J. Torres. Technology and resilience. Working Paper 29644, National Bureau of Economic Research, January 2022. URL http://www.nber.org/papers/ w29644. J. I. Dingel and B. Neiman. How many jobs can be done at home? Journal of Public Economics, 189:104235, 2020. ISSN 0047-2727. doi: https://doi.org/10.1016/j.jpubeco.2020.104235. URL https://www.sciencedirect.com/science/article/pii/S0047272720300992. A. Grimes, C. Ren, and P. Stevens. The need for speed: impacts of internet connectivity on firm productivity. Journal of Productivity Analysis, 37(2):187–201, 2012. 24 R. Harris and J. Moffat. The impact of the covid-19 pandemic on the level and distribution of intangibles investment in the uk. Applied Economics Letters, 0(0):1–5, 2021. doi: 10.1080/13504851. 2021.1954591. URL https://doi.org/10.1080/13504851.2021.1954591. P. Ifinedo. Internet/e-business technologies acceptance in canada’s smes: an exploratory investigation. Internet Research, 21(3):255–281, 2011. R. Y. Kim. The impact of covid-19 on consumers: Preparing for digital sales. IEEE Engineering Management Review, 48(3):212–218, 2020. J. Moon, M. D. Hossain, H. G. Kang, and J. Shin. An analysis of agricultural informatization in korea: the government’s role in bridging the digital gap. Information Development, 28(2):102–116, 2012. B. Ndulu, N. X. Ngwenya, and M. Setlhalogile. The digital divide in south africa: Insights from the covid-19 experience and beyond. In The Future of the South African Political Economy Post-COVID 19, pages 273–295. Springer, 2022. L. Vo, T.-H. L. Le, and D. Park. Digital divide decoded: Can e-commerce and remote workforce enhance enterprise resilience in the covid-19 era? Asian Development Bank Economics Working Paper Series, (667), 2022. M. F. Warren et al. Adoption of ict in agricultural management in the united kingdom: the intra-rural digital divide. ZEMEDELSKA EKONOMIKA-PRAHA-, 48(1):1–8, 2002. T. Watanabe and Y. Omori. Online consumption during the COVID-19 crisis: Evidence from japan. Covid Economics: Vetted and Real-Time Papers, 32, 06 2020. 25 A. Appendix Country Wave 1 Wave 2 Balanced Panel Country Wave 1 Wave 2 Balanced Panel BEN∗ - 252 - MDG 341 427 198 BGR† 1154 1022 727 MNG∗ 271 222 211 BRA 474 310 309 MWI - 671 - CYP∗ 147 153 121 MYS 1094 1055 1045 CZE∗ 373 370 314 NIC∗ 170 177 136 EST∗ 257 274 231 PAK 694 267 267 GEO∗ 488 468 436 POL† 1924 1471 1312 GHA 1155 929 928 PRT∗ 683 709 603 GRC∗ 507 524 478 ROU† 1207 935 882 GTM∗ 183 177 143 SEN 449 444 331 HND∗ 148 143 115 SLE 120 126 42 HRV∗ 328 317 293 SLV∗ 355 374 293 HUN∗ 583 601 520 SVK∗ 298 278 255 IDN 645 521 521 SVN∗ 242 238 171 ITA∗ 407 409 346 TUN 1455 850 846 KEN 1019 889 876 TUR 831 1056 208 KHM 401 327 320 TZA 496 450 277 LTU∗ 198 197 153 VNM 461 449 448 LVA∗ 232 241 174 ZAF 1351 364 359 MAR∗ 741 673 559 ZMB∗ 505 503 472 MDA∗ 264 258 231 ∗ WBES countries. † WBES and BPS observations. Table 3: List of countries used in the analysis 26 Note: Excluded wave 1 of MDG and TUR due to replacement observations. Figure 18: Likelihood of starting or increasing the use of digital technology by country 27