Policy Research Working Paper 10633 Disruptive Technologies and Finance An Analysis of Digital Startups in Africa Marcio Cruz Mariana Pereira-Lopez Edgar Salgado International Finance Corporation A verified reproducibility package for this paper is December 2023 available at http://reproducibility.worldbank.org, click here for direct access. Policy Research Working Paper 10633 Abstract This paper investigates the relationship between disruptive offerings compared to other regions, except for mobile technologies and access to finance for digital tech firms in payments. Second, incorporating these technologies is asso- Africa. Through textual analysis of data from Crunchbase ciated with more funding, but this link is weaker in Africa and Pitchbook, the study explores how firms across differ- than in other regions. These results hold when excluding ent age cohorts incorporate disruptive technologies into mobile payments and addressing potential endogeneity their offerings in e-commerce, fintech, and information using instrumental variables. Third, firms that do incorpo- technology services. The findings reveal three key insights rate disruptive technologies tend to secure funding earlier, for African digital tech startups. First, African startups are with lower initial amounts, but are more likely to succeed less likely to incorporate disruptive technologies into their in terms of exit or valuation growth than their peers. This paper is a product of the International Finance Corporation. It is part of a larger effort by the World Bank Group 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 marciocruz@ifc.org. A verified reproducibility package for this paper is available at http://reproducibility.worldbank. org, click here for direct access. RESEA CY LI R CH PO TRANSPARENT ANALYSIS S W R R E O KI P NG PA 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 Disruptive Technologies and Finance: An Analysis of Digital Startups in Africa* Marcio Cruz† , Mariana Pereira-Lopez ‡, Edgar Salgado§ Originally published in the Policy Research Working Paper Series on December 2023. This version is updated on April 2024. To obtain the originally published version, please email prwp@worldbank.org. JEL classifications: D22, O31, O33 Keywords: digital technologies, disruption, financing, startups, entrepreneurship * We thank Susan Lund, Paolo Mauro, Denis Medvedev, Samuel Edet, Santiago Reyes, Florian Molders, ¨ Georges Houngbonon, Zineb Benkirane, Matthew Saal, Ami Dalal, Verena Wiedemann, Davide Stru- sani, Stephan Dreyhaupt, and participants of the workshop at IFC for their valuable comments. Lucien Ahoubange provided extraordinary research support. This paper was prepared as a background study for the IFC’s ”Digital Opportunities in African Businesses” project. This paper is a product of the IFC Economic Research Department, in partnership with the Government of Japan. † International Finance Corporation, email: marciocruz@ifc.org (Corresponding author) ‡ International Finance Corporation, email: mpereiralopez@ifc.org § International Finance Corporation, email: esalgadochavez@ifc.org 1 Introduction Digital technologies have been seen as a potential driver of leapfrogging in Africa. The African digital start-up ecosystem, though still nascent, is among the fastest-growing in the world. It is the only region where the median deal size did not decline during 2022 (CB Insights, 2023). Although framed as a footloose sector, digital business providers are widely perceived as an economic activity with the potential and opportunity to bring growth to deprived or lagged regions and enabling technological catching up (Friederici et al., 2017, 2020). The remarkable speed of mobile payment adoption by individuals and firms in Africa are practical examples of the potential implications of disruptive technolo- gies. This process has been driven by leading digital tech platforms and firms, such as M-pesa and Wave, by incorporating innovative and disruptive technological solutions into their offerings, allowing them to disrupt markets through better and cheaper solu- tions for customers. However, apart from a few well-known successful cases, there is little evidence available on how digital startups in Africa have been incorporating disruptive technologies into their offerings, a key condition to play an enabling role for digital dis- ruptiveness in the continent, or on the role of finance in this process. This paper analyzes the extent to which digital tech firms in Africa incorporate disrup- tive technologies and how this relates to access to finance along the firm life cycle. First, it compares the level of disruptive technologies incorporated by digital businesses in Africa with the Latin American market,1 and a global frontier from which a substantial propor- tion of digital technologies have emerged, constructed using business data for cities such as Seattle, Palo Alto, Tokyo, and London. Second, the paper documents the variation in the incorporation of disruptive technologies into their offerings along the life cycle of digital tech firms in Africa. Following the literature on firm dynamics and entrepreneur- ship, which suggests that young firms (startups) are more innovative (Decker et al., 2014), the paper compares the differences in the incorporation of disruptive technologies along the life cycle of tech startups. Third, the paper evaluates to what extent incorporating these disruptive technologies into the firms’ products and services is associated with the amount of funding received. It is expected that these high-potential firms incorporat- ing disruptive technologies should exhibit a higher probability of attracting funding, but there could be different market failures affecting this process. To address these questions, we use the technologies identified as disruptive by Bloom et al. (2021), based on patents for the U.S., and the frequency with which these technolo- gies are mentioned in earning calls. Using textual analysis, we identify those firms in Crunchbase and Pitchbook2 that include any of the bigrams associated with these tech- nologies, either in their short or in their long description. We focus on three main digital sectors: Fintech, E-commerce, and Information Technology (IT),3 which were selected 1 Latin America represents a good comparison group because it comprises emerging markets and, similar to Africa, it has significantly increased its tech adoption and entrepreneurship in recent years, and these dynamics are relatively inward-related, in sharp contrast with other emerging regions like Asia, where the dynamics tend to be trade and outward-related. 2 Crunchbase (2023) data obtained from www.crunchbase.com and PitchBook (2023) data from https://pitchbook.com/. 3 It refers to the provision of digital services, such as software and data. 2 given their relevance in the high-growth pattern of startups in Africa. Comparing the disruptive component among African digital tech firms with the fron- tier (advanced) markets, defined as cities where important technological hubs are located (Palo Alto, Seattle, London, and Tokyo), we find that firms in Africa are 25% less likely to incorporate disruptive technologies than the average digital tech firm on the frontier. However, the recent wave of adoption of mobile payments in the region underestimates this gap. Excluding mobile payments and Wi-Fi technologies, African firms become 36% less likely to incorporate disruptive technologies than the frontier. Along the life cycle, we find that African startups are more likely to incorporate dis- ruptive technologies than older digital tech firms. This pattern is similarly observed among their counterparts in Latin America and the frontier. However, not only the whole distribution in Africa is below the frontier markets, which is expected due to different levels of development, but the African curve of disruptiveness along firm age is slightly flatter, indicating that the difference between startups and older firms is smaller than their counterparts in LAC and the frontier. This is particularly relevant for the Fintech sector, where Kenya and Nigeria, which are the leading countries in the region, show more dis- ruption coming from older firms. Moreover, young startups in Africa are less likely to incorporate disruptive technologies than their counterparts in Latin America. This pat- tern becomes clearer when excluding mobile payments and Wi-Fi technologies. Our results show that in frontier cities, startups that incorporate any disruptive tech- nologies receive 263% more funding than non-disruptive ones. This funding premium shrinks to 100% in LAC and to 40% in Africa. Excluding mobile payments and Wi-Fi technologies from the disruptive group, the results suggest a similar pattern. Additional evidence suggests that being disruptive is associated with earlier financing but with lower tickets and higher growth in valuation and future funding deal size. We used an instrumental variable (IV) to test if access to funding is driven by the in- corporation of disruptive technologies into African firms’ offerings. This strategy aims to address potential concerns of reverse causality (firms incorporate further disruptive technologies due to more access to finance) and endogeneity driven by omitted variables biases (e.g., unobserved factors may drive both access to finance and disruptive technolo- gies). For this purpose, we instrument the incorporation of disruptive technology into the services provided with the share of firms in a similar sector incorporating these technolo- gies in the frontier markets and the relative supply of skilled human capital in the local labor market, controlling for country fixed effects. Our IV results suggest that firms in- corporating disruptive technologies into their offerings are significantly more likely to be funded and receive more funding. The results are robust to incorporating all technologies into the estimations. Our analysis contributes to a wider literature studying the characteristics of digital entrepreneurship. Most previous studies have focused on explaining the growth of en- trepreneurship in Africa without further characterizing the technological content or the disruptiveness potential of these ventures (Friederici et al., 2020). In this sense, the main contribution of this study is the measurement and systematic analysis of the extent to which these disruptive technologies are incorporated into African entrepreneurship com- pared to other markets. The second contribution is that we leverage and analyze a large amount of commercial data for Africa, a region that has persistently lacked high-quality 3 datasets, to go beyond previous small-sample qualitative studies and understand the al- location of funding to digital entrepreneurship along the life cycle of firms. The paper is structured as follows. Section 2 explains the definition and characteristics of disruptive technologies that are used throughout the paper. Section 3 explains the data sources and the methodologies used to identify disruptive technologies and the empiri- cal strategy. Section 4 analyzes the main results of the paper, including the life cycle of disruptive technologies and the differences in securing funding, and section 5 concludes. 2 What makes firms disruptive, and where is the frontier? Disruptive technologies are those with the potential to create new markets and eventu- ally overtake leading companies in mainstream markets (Bower and Christensen, 1995). The technological architecture embedded in these technologies starts to become a part of established technologies, and thus disruptive technologies show steep increases in perfor- mance. As explained by Bower and Christensen (1995) and Christensen (1997), even large established companies that are leaders in the market can be substantially affected by fail- ing to implement these technologies, reducing their market share or, in some cases, even exiting the market. An example of disruptive technologies is artificial intelligence, which was started by creating a new and very specialized market. Furthermore, it has gradually been incorporated into other digital solutions offered to consumers and businesses. A firm becomes disruptive with innovation or the adoption of new technologies that have the potential to alter the way markets behave. The disruption produced by these new technologies can propagate via input or output markets. For instance, firms with digital financial technologies, usually known as Fintech, use digital technologies to im- prove their interaction with clients through the use of mobile apps, ultimately expanding coverage because the interaction simplifies. This reduces the cost of market outreach, and the firm eventually expands in the output market. At the same time, the massive col- lection of data through the mobile app allows the firm to analyze the characteristics of its clients to provide services that are even more tailored to its customers, furthering the impact on the output market. But to analyze these data, the firm needs highly qualified data scientists, which ultimately influences the labor (input) market. Not all cutting-edge technologies are disruptive technologies. Some products or ser- vices with high technological content are too specialized to change markets. On the other hand, there are products for which the technology is not necessarily new, but the potential for creating markets and overtaking existing technologies makes them disruptive. Take, for example, the case of the personal computer and Apple; as explained by Christensen (1997), the technology was not necessarily new but was disruptive because it could bring the existing computer technology to millions of households and businesses. Something similar was observed with the Ford Model T in the U.S., where the car technology was not new, but it was produced in a way that allowed the car to transition from a luxury product to a mass-market good (Eli et al., 2023). Given the specific characteristics of disruptive technologies, it has been challenging to establish a rigorous classification of disruptive technologies. Bloom et al. (2021) use an innovative and systematic approach. They identify 29 disruptive technologies based on 4 information from millions of U.S. patents and validate this list by analyzing the frequency at which these technologies were mentioned in hundreds of thousands of earnings confer- ence calls. Therefore, this strategy considers both the technology and the potential market components of disruptive technologies. As their exercise is based in the U.S., we consider these technologies a good proxy for the type of technologies that startups and innovative entrepreneurs might decide to incorporate into their products or services. 3 Data and methodology 3.1 Data The main data source used in this paper is Crunchbase (www.crunchbase.com). Cre- ated in 2007, the use of Crunchbase as a primary source of business information and for academic research has gradually increased, with over 55 million users currently. As explained by Dalle et al. (2017), Crunchbase is updated on a daily basis, and the infor- mation is obtained from large investor networks, a community of contributors, global investment firms that submit their whole portfolio, as well as entrepreneurs and inde- pendent investors that provide their own information. All these data are validated and cross-linked with funders, firm staff, and management profiles. This self-reported infor- mation is complemented through Artificial Intelligence and Machine Learning algorithms that web scrape to enrich these profiles. According to Dalle et al. (2017), comparing Crunchbase with the OECD Entrepreneur- ship Financing Database, the coverage of Crunchbase appears to be very comprehensive for young firms, especially in the case of the U.S. As this exercise was conducted in 2017, and given the exponential growth of this platform, we can expect that the coverage is even better now. Given the constraints on data availability in Africa, the increasing use of Crunchbase globally, the emergence of other platforms that concentrate information on Africa and the methodology (web scraping and self-reporting) that Crunchbase uses to gather and validate their information, the coverage for Africa appears to be relatively good for these types of exercises. For this study, we use data from 32,530 digital businesses collected by Crunchbase as of February 2023. While this paper focuses on Africa, we use information from other regions as a benchmark. To define the location, we use information on the firm’s head- quarters. We classify 7,785 digital businesses as located in Africa, 12,816 in Latin America (South America and Mexico), and 11,929 in any of the frontier cities (London, Palo Alto, Seattle, and Tokyo). These frontier cities are selected based on the fact that they are loca- tions characterized or known as technological hubs.4 At the sectoral level, we focus on three sectors: Fintech, E-Commerce, and IT. The selection of these three sectors is directly related to the distribution of recent investments 4 Wedefine alternative options for the frontier based on the top Crunchbase ranking and top revenue of the dataset, but we only use them for robustness checks. The main reason for not taking these measures into consideration in the core analysis is that these measures are a combination of technology and funding and, therefore, are already highly correlated with funding. We want a benchmark more related to technology than funding, as this later variable is one of the outcomes or firm characteristics we analyze. 5 in startups in Africa and their importance in startups’ growth in Africa. Fintech is the fastest growing startup industry in Africa. According to Disrupt Africa (2022), Fintech is the power behind the region’s startup revolution. Only in 2022, it raised 43% of the total funding. The second sector in terms of funding was E-commerce and retail tech, with 17% of the region’s total. When we analyze the total number of companies in the African Entrepreneurial Ecosystem, almost 20% of the companies are in Fintech, while 13% are E-commerce firms (Briter Bridges, 2022). The rationale for selecting the IT sector has to do with its characteristics, as it includes technologies like software, IT infrastructure and services, and data storage and management, which are associated with core activities of the digital sector. For the purpose of this paper, we define startups as digital businesses belonging to these sectors (Fintech, E-Commerce, or IT), younger than ten years of age. The age defini- tion is arbitrary, but follows the convention observed in many regulatory acts aiming to promote innovative and high-growth startups. For example, Tunisia’s Start-Up Act, ap- proved by the parliament in 2018, defines up to eight years as the threshold. This law has been used as a reference by many other African countries (e.g., Senegal). This threshold is used as a reference when comparing the level of disruptiveness between digital startups and older firms.5 One limitation of the Crunchbase data is that it has only a cross-section. Although it includes very detailed information about the companies, such as the year when the firm was founded, similar firms, the amounts and different rounds of funding, and character- istics of the founders and investors, among others; it is a cross-section and takes the most recent values or attributes of the firm. This does not allow us to analyze the evolution of these firms over time. Therefore, any reference to the life cycle in this paper refers to the comparison of firms across different age groups. We complement the analysis with data from Pitchbook (https://pitchbook.com/), an- other VC/PE private repository that provides detailed characteristics and financial and financing history information on more than 3.6 million companies worldwide. Pitch- book focuses on the entire life-cycle of investments and collects information mostly on investment-ready firms. Figure 1 shows the spatial distribution of the African digital firms analyzed in this paper. As shown, South Africa, Nigeria, and the Arab Republic of Egypt stand out as digital hubs, where most of the firms in our sample are located. Appendix Figure B1 shows the spatial concentration of African digital firms in the main city of each country. In most countries, a large share of the firms is located in the capital city. From the so- called Big Four countries (Egypt, Nigeria, Kenya, and South Africa), where most of the digital firms operate, it is only in South Africa where the main city concentrates less than 50% of all digital firms operating in the market. 5 We also test the robustness of our results with a threshold of five years of age for startups. 6 Figure 1: Geographical location of digital firms Note: Own calculations using Crunchbase data. www.crunchbase.com 3.2 Methodology 3.2.1 Identifying disruptive firms To identify which firms can be considered disruptive, we rely on the list of 29 technologies identified by Bloom et al. (2021).6 As previously mentioned, this list results from the tex- tual analysis of millions of U.S. patents and hundreds of thousands of earnings conference calls. Therefore, it has the advantage of including both characteristics of disruptive tech- nologies: the cutting-edge technological content (from patents) and the potential market (from earnings conference calls). Another advantage of this list is that it is exogenous to the region of interest (Africa) and the benchmark emerging economies’ region (LAC). We use the bigrams associated with each of the 29 technologies identified by Bloom et al. (2021) as disruptive and search for them in the description of the products and ser- 6 See Table A1 in the Appendix for the list of technologies and bigrams. 7 vices of the firm. In practice, we run the search within the short and long descriptions, as well as the industry description and keywords listed by both Crunchbase and Pitchbook. An example of how these exercises work is included in Figure B3 of the Appendix. Im- portantly, when we capture a disruptive technology into the offering description of the firm, we are not only identifying the adoption of these technologies, but that they are also part of their core services or products, which enable their customers to benefit from these technologies and, therefore, facilitate diffusion across the economy. Then, we count the number of disruptive technologies a firm has incorporated into its offering. If a firm has many bigrams included in the classification by Bloom et al. (2021), but all of them are associated with the same disruptive technology, the firm obtains a value of one. A firm for which none of these bigrams can be found in its description receives a value of zero in this score. Figure 2 ranks the most frequent technologies among the three regions we study. Not all technologies are found in the textual analysis. “Lithium Battery”, “Software Defined Radio,” and “Wireless Charging” are non-existent in our sample, while “Drug Conju- gates” is part of the technologies found in frontier countries but not in Africa nor LAC. This mostly responds to the sectoral scope of our analysis as we are only focusing on E-commerce, Fintech, and IT. “Cloud Computing” is the most prevalent disruptive tech- nology in the sample, with a higher incidence among Latin American countries. Within frontier countries, the second most frequent is “Machine Learning/AI” but less frequent in Africa and LAC. “Mobile Payments” is the second most frequent technology in Africa, well above the prevalence observed in the frontier countries and in LAC. This technology is associated with the recent Fintech boom in the region. Therefore, to test the robustness of our results, we perform some analyses excluding mobile payments-related technolo- gies as well as Wi-Fi, given that due to the wide adoption of these two technologies among digital companies, they might be taken by the companies as given and, therefore, might not be included in their descriptions. Figure B2 in the appendix shows the average probability of incorporating a disrup- tive technology across regions. Panel (a) shows the average probability that a firm uses disruptive technologies (extensive margin) without controlling for any firm-level char- acteristics, while panel (b) shows the average number of disruptive technologies condi- tional on reporting at least one (intensive margin). The green bars report both margins after considering any of the 29 technologies discussed above, while the orange bars use only 27 technologies after excluding “Mobile Payments” and “Wi-Fi” from the analysis to account for the Fintech boom in Africa. As shown in the Figure, there are substantial dif- ferences against the frontier, but mostly against those firms at the Top of the Crunchbase distribution on the extensive margin. Once firms have implemented at least one technol- ogy, they tend to be slightly above one in the number of disruptive technologies with no substantial difference. However, Africa and LAC are slightly below the frontier. 8 Figure 2: The Distribution of 29 Technologies among Digital Firms Note: Own calculations using Crunchbase data. www.crunchbase.com 3.2.2 Measuring the distance to the frontier and the life cycle To explore the distance to the frontier, we start by estimating gaps in both extensive and intensive margins of the African and LAC digital firms with respect to the reference cities. The estimation is: ykri = α + β kr Rkri + β Xi + ε kri (1) where ykri is either a dummy variable to identify whether the firm i in industry k (E- Commerce, Fintech or IT) and region r (Africa or LAC) uses at least one disruptive tech- nology (extensive margin), or the number of disruptive technologies among disruptive firms (intensive margin). R is a set of dummies for either Africa or LAC, while leav- ing cities in the frontier as the reference category. β is the coefficient that measures the gap after controlling for the IHS transformation of the age of the firm,7 the number of overlapping sectors in the country where the firm operates,8 and a set of nine labor-size dummies.9 To estimate the number of disruptive technologies along the age distribution, we run Poisson regressions: 7 We use the Inverse Hyperbolic Sine (IHS) transformation as suggested by Bellemare and Wichman (2020). 8 Up to two overlapping sectors if the firm i operates in the three sectors simultaneously within the country. 9 1-10 employees, 11-50, 51-100, 101-250, 251-500, 501-1000, 1001-5000, 5001-10000, and 10000+. 9 Tkri = α + γkr ln agekri + β Xi + ε kri (2) where Tkri is the number of disruptive technologies of firm i in region r and sector k. γkr is the coefficient associated with the natural log of firm’s age in years. We allow Tkri to be zero for non-disruptive firms, and Xi contains the same set of controls described above (except for age). 3.2.3 Funding and disruptive technologies Finally, to investigate whether there are differences in the allocation of funding, we esti- mate the following regression: Fi = α + θ1 Ti + θ R ∗ Ri ∗ Ti + β Xi + γR Ri + ε i (3) where Fi is the outcome variable of funding, which could be either a dummy variable indicating that the firm received funding, or the IHS transformation of the amount of funding in U.S. Dollars received (in total or during the last funding round), as reported in February 2023. Xi is the same set of controls outlined above plus a variable that includes the number of similar companies to also control for competition, and γR are region effects (Africa and Latin America).10 An Instrumental Variable Approach A crucial feature of the estimates regarding funding and disruptive technologies is that the direction of causality is not clear. It could be the case that firms incorporating dis- ruptive technologies receive more funding or, on the contrary, that having received more funding allows them to incorporate these technologies into their offering. To identify the effect of disruptive technologies on access to funding, in this section, we propose an Instrumental Variables (IV) approach. The proposed instrument combines country and firm-specific characteristics to create a vector of exposure to the adoption of disruptive technologies. The main assumption behind the instrument is that firms from sectors that are more intensive in disruptive technologies (defined exogenously by what is observed in the frontier) and that are located in countries with better skills should in- corporate more disruptive technologies into their offering. The interaction of these two variables should only affect firm-level financing through the capability of the firm to in- corporate disruptive technologies. The first part of the instrument uses information from countries in the frontier to cal- culate the share of disruptive firms by industry vertical and impose this on African firms based on their own industry verticals. This strategy based on sectoral intensity follows the idea behind Rajan and Zingales (1998) and has been used recently in the Information Technology (IT) use literature (Bloom et al., 2012; Fernandes et al., 2019; Iacovone et al., 2023). Using only information from the frontier is plausibly exogenous since this excludes country-specific barriers that would prevent the adoption of disruptive technologies in in- dustry verticals that otherwise would be technology-demanding. Crunchbase classifies 10 In all our estimations, standard errors are clustered at the level of the country. 10 approximately 600 industry verticals, and for each, we calculate the share of disruptive firms listed in these verticals. Then, we match these industry vertical shares to all African firms in the sample. Firms could be linked to one or many verticals. The strategy exploits the fact that most firms are associated with more than one vertical and thus the final ex- posure of a firm is the combination of all the shares depending on the number of verticals the firm has. The idea is that the exposure to the adoption of disruptive technologies will depend on the industry vertical mix of every firm. The second component of the instrument, the country characteristic that predicts the adoption of technologies, is the degree of skill mismatch in the country. Bloom et al. (2021) show that the emergence of disruptive technologies is more likely in geographies with a high concentration of skilled labor. Porzio (2023) finds that countries where skill mis- match in the labor force is high, settle in a low equilibrium of low technology adoption. We use skill mismatch data produced by ILO prior to 2010.11 Mismatch occurs when the level of education of the person in employment does not correspond to the level of educa- tion required to perform their job. ILO uses a normative approach to identify mismatched workers. It is based on the educational requirements set out in the International Standard Classification of Occupations (ISCO) for each one-digit ISCO occupational group, and on the level of education of each person in employment. Three categories are possible; the first, are matched individuals whose education level corresponds to the education re- quirements described in the occupation; the second category is over-educated workers, while the third corresponds to workers with an education below the one required by the occupation. These three groups are computed for employees and self-employed workers. Our instrument uses the share of matched or over-educated workers among employed individuals. The idea is that countries where workers are adequately or over-skilled are more likely to have the conditions for disruptive technology adoption. Being as this vari- able is defined at the country level, it should be exogenous to the firm. The instrument, ultimately, is the interaction of these two components: K ∑k= i 1 θk IVi = Mc (4) Ki where Mc is the mismatch share in country c, θk is the share of disruptive firms from the frontier in industry vertical k, Ki is the number of industry verticals in firm i. We calculate this IV for both the full set of disruptive technologies and the set that excludes mobile payments and Wi-Fi. We focus on the sample of firms in the Africa region. The first stage is defined as: Ti = α + β IVi + β Xi + ε kri (5) where Ti is the number of disruptive technologies, IVi is the instrument defined in equation 4, and Xi are the same controls included in equation 3. 11 International Labour Organization (2023), https://ilostat.ilo.org/resources/concepts-and- definitions/description-education-and-mismatch-indicators/ 11 The second stage, similar to equation 3, is defined as: ˆi + β Xi + ε i Fi = λ + θ1 T (6) ˆi is the predicted value from the first stage (equation 5). where T 4 Results 4.1 To what extent are digital entrepreneurs in Africa incorporating dis- ruptive technologies in their offering? Figure 3 displays the estimated gap in disruptiveness using equation (1) by region and industry. Panel (a) includes all technologies, while panel (b) excludes mobile payments and Wi-Fi. The left-hand side figures describe the extensive margin (if the firm incor- porates any disruptive technology), while the right-hand side figures show the intensive margin (the number of disruptive technology incorporated by the firm, conditional on incorporating disruptive technologies). All figures are normalized to the frontier (zero on the horizontal axis). Overall, when considering all technologies, we observe a large difference across industries, with Fintech being closer to the frontier on the extensive margin. The comparison between panels (a) and (b) shows that the inclusion of “mobile payments” and “Wi-Fi” among the group of disruptive technologies leads to significant differences for African countries, particularly for Fintech. This reflects the recent trends in mobile payments in the region.12 On the extensive margin, panel (a.1) shows that African firms are, on average, six percentage points less likely to incorporate disruptive technologies. As, in the frontier, 24% of the digital firms report incorporating at least one disruptive technology into their offering; this means that relative to that, African digital firms are 25% less disruptive than the frontier firms in the extensive margin. LAC firms are also less disruptive relative to the frontier by almost seven percentage points, representing a gap of 28%. Excluding mobile payment and WiFi is particularly relevant for Africa as, in the lower panel (panel b.1), the gap against the frontier in Africa represents 36%, while for LAC, it is 32%. This difference is mainly driven by the Fintech sector, which moves from being not statistically different to the frontier (panel a.1) to being about ten percentage points less likely to incorporate other disruptive technologies. On the intensive margin, analyzing the gap in the number of disruptive technologies conditional on being disruptive, differences against the frontier are observed for all sec- tors in Africa, except e-commerce (panel a.2). The distance to the frontier on the intensive margin for Africa is larger and more heterogeneous across industries when excluding mo- bile payments and Wi-Fi (panel b.2). While African firms operating in E-Commerce show a smaller gap against the frontier, Fintech and IT are even less disruptive compared to the frontier. 12 Given the rapid diffusion of mobile payments in Africa and potential concerns raised by practitioners that some firms under-emphasize the provision of Wi-Fi, we provide results excluding both technologies as robustness. 12 Figure 3: Gap with respect to the Disruptive Frontier (a) Including all disruptive technologies (b) Excluding Mobile Payments and Wi-Fi Note: Frontier includes the pooled number of start-ups in London, Palo Alto, Seattle, and Tokyo. Africa includes all African countries, while LAC includes South America plus Mexico. Panel (a) shows the extensive and intensive margins by industry, including all technologies. Panel (b) excludes mobile payments and Wi-Fi To better understand what each sector comprises, figures B5-B7 in the Appendix show the difference in the most cited 25 words describing each sector between African and frontier firms. For instance, Figure B5 suggests that e-commerce firms in Africa describe themselves more with words linked to “logistics” than e-commerce firms in the frontier. 13 Figure B7 shows the same analysis for IT firms: while African firms in IT seem to de- scribe themselves more with words like “information technology”, they less frequently use “saas” or “artificial intelligence” in their description than IT firms in the frontier. Taking all these results together, both the extensive and intensive margins, they point to a general picture of a gap in disruptiveness that, however, disguises important differ- ences by region and industry. 4.2 What is the life cycle of disruptiveness in Africa? Haltiwanger et al. (2013) suggest that young firms (i.e., start-ups) play a predominant role in job creation, and therefore, their role in innovation and productivity is also of importance. This is also aligned with the ”Silicon Valley” narrative that promotes new firms and young entrepreneurs as positive attributes that lead to transformational and high-growth ventures (Friederici et al., 2020). Figure 4 shows the average number of disruptive technologies by age estimated using equation (2). These averages are net of size (labor) to allow comparing firms of similar characteristics. The pattern in the frontier firms indicates that the youngest use about twice the number of disruptive technologies than the group of 30-year-old firms. The same pattern emerges in African and Latin American firms, and for both of them, the life cycle of disruptiveness is below the frontier. Interestingly, young LAC firms are more disruptive than African firms, but this rank reverts for firms seven years old and beyond when we include all disruptive technolo- gies (panel A). We take this as an early indication of further heterogeneity along the life cycle by region and that there is an important degree of action in the right tail of the age distribution in Africa, which could be driven by the characteristics of large sectors, like Fintech, and countries that account for an important proportion of the activity (the big four). This hypothesis is further reinforced by the results shown in panel (b) of Figure 4 where, excluding mobile payments, the differences between firms in LAC and Africa in the number of disruptive technologies incorporated are almost negligible for firms older than 15 years. Figure 5 looks at the life-cycle patterns by industry. There is a great sectoral hetero- geneity as, while the IT sector shows that Africa is below LAC over the whole life-cycle, in E-commerce, younger firms in LAC tend to be more disruptive, but Africa catches up as firms age. The patterns of these two sectors do not show significant differences if we exclude mobile payments and Wi-Fi from the definition of disruptive technologies. Firms in Fintech, on the other hand, exhibit a life-cycle profile that significantly dif- fers from the other two sectors. In general, the degree of disruptiveness seems relatively constant along the life cycle in the three regions but slightly less so for LAC. Indeed, the estimated coefficient for age is not statistically significant in any of the regions within Fin- tech. In contrast with E-commerce and IT, excluding mobile payments and Wi-Fi makes a big difference in this sector. As panel (a) of Figure 5 shows, when these technologies are considered in the definition of disruptiveness, Africa is very close to the frontier and even slightly above for young firms, though older firms tend to lag slightly behind. Excluding these technologies significantly increases the distance to the frontier for Africa but less so 14 for LAC, with both regions consistently below the frontier along their whole life cycle. Still, the flatter pattern remains. At this point, it is clear that the Fintech industry faces a different context than the rest. This could be the reflection of higher entry costs, stricter regulations, or sectoral cross- overs (i.e., before becoming Fintech, firms may need to consolidate in adjacent industries). This will be further explored in the following sections. Figure 6 takes a deeper look into the age profiles by country in Africa. As previously mentioned, South Africa, Nigeria, Egypt, and Kenya are the four largest markets in digi- tal entrepreneurship (the so-called Big Four), and, therefore, these four countries explain the shape of the age profile in the entire Africa. However, some interesting differences in patterns emerge. Older Fintech firms in Kenya seem to be more disruptive than young firms in this sector. South Africa and, to a lower extent, Nigeria show a similar pattern in Fintech, while this is the case for Nigerian IT firms (panel b). Egypt is the only country where excluding mobile payments changes the pattern completely. When mobile pay- ments are considered (panel a), older firms tend to incorporate more of the disruptive technologies, while the exclusion of mobile payments leads to the opposite pattern for this country. This, once again, highlights the critical role that mobile payments have had in the leading African countries. In the case of Fintech in Kenya, for example, this different pattern can be easily ex- plained using, for example, the high-growth cases or success stories. As explained by Chitavi et al. (2021), there are some key factors associated with these cases in the Fintech sector in Kenya. First of all, there is a preference from consumers for bundled services, as they reduce the information, search, and implementation costs. Secondly, financing en- tails trust, and though young firms can offer agility and technological architecture, more established and, therefore, older firms can transfer trust from previous related brands. Take, for example, the case of M-Pesa, which, as explained by Chitavi et al. (2021), started in 2007 as a new Fintech product that built upon the trust that Safaricom, the telecom- munications giant, already had. On the other hand, the team leader of M-Pesa launched M-Kopa in 2012 (which is among our disruptive companies) and took advantage of the infrastructure and connectivity to M-Pesa. All these results point to the existence of country and sector differences that produce deviations from what we observe in the frontier. 15 Figure 4: Age Profile of Disruptive Technologies (a) Including all disruptive technologies (b) Excluding Mobile Payments and Wi-Fi Note: Frontier includes the pooled number of start-ups in London, Palo Alto, Seattle, and Tokyo. Africa includes all African countries, while LAC includes South America plus Mexico. 16 Figure 5: Age Profile of Disruptive Technologies by Industry (a) Including all disruptive technologies (b) Excluding Mobile Payments and Wi-Fi Note: Frontier includes the pooled number of start-ups in London, Palo Alto, Seattle, and Tokyo. Africa includes all African countries, while LAC includes South America plus Mexico. 17 Figure 6: Age Profile of Disruptive Technologies by Industry and Country in Africa (a) Incuding all disruptive technologies (b) Excluding Mobile Payments and Wi-Fi Note: Frontier includes the pooled number of start-ups in London, Palo Alto, Seattle, and Tokyo. Africa includes all African countries, while LAC includes South America plus Mexico. Panel a shows the gap in the extensive margin by industry. Panel b shows the gap in the intensive margin 18 4.3 The Allocation of funding How does the capital market allocate funding to disruptive firms? Being disruptive en- tails risks but also the promise of future revenue. Since the definition of disruptiveness adopted by Bloom et al. (2021) implies that these disruptive technologies are being in- creasingly discussed in earnings conference calls, it is possible to expect a funding pre- mium or return for disruptive firms. We explore this with equation (3) and show the results in Tables 1, 2, and 3. In the frontier countries, firms incorporating disruptive technologies into their offers are 16 percentage points more likely to receive funding at some point in their life cycle. Table 1 shows that these results contrast with digital tech firms in LAC, where this prob- ability is only seven percentage points higher, and Africa, where it is even lower, three percentage points larger than firms with no disruptive technologies. The magnitudes are slightly lower when we only consider the last round of funding (column 2), but the dif- ferences between the frontier, LAC, and Africa have the same magnitudes. The results do not change much when we exclude mobile payments and Wi-Fi from the disruptive technologies (columns 3 and 4).13 Table 2 shows the elasticities of funding (measured as the IHS of funding value) on being disruptive (incorporating at least one disruptive technology into the firm’s offer- ing), considering both the extensive and the intensive margins of funding. The refer- ence category consists of firms with no disruptive technologies. Panel (a) focuses on all firms. Being disruptive in the frontier has a funding premium of 263% compared to non- disruptive. This means that while non-disruptive firms in the frontier receive funding of about US$ 6 million, disruptive firms will get funded by US$ 21.8 million. Being disrup- tive in Africa also attracts funding, but to a lesser degree: the disruptiveness premium is 40%, while in LAC it is 100%. Once again, as shown in column (2), the magnitudes are lower when we consider the last funding, but the conclusions are the same. Even consid- ering the importance of mobile payments, the elasticities do not change much when we exclude these technologies along with WiFi from the classification of disruptiveness. Panel (b) turns the attention to startups, here classified as not older than ten years. Among start-ups in the frontier, being disruptive has a premium of 217%, as indicated in column (1). This coefficient is qualitatively smaller than the overall coefficient from column (1) in panel (a), suggesting that, at least in the frontier, the attraction of funding among disruptive firms occurs more among older firms. Panel (c) confirms this by show- ing that the disruptiveness premium among firms older than ten years is 294%. In the case of Africa, younger firms have a premium to disruptiveness of 28%, while older firms have a premium of more than twice (60%). Once again, this suggests that older firms ex- hibit a higher correlation between disruptive technologies and funding. In sharp contrast with Africa, the premium in LAC is almost the same between startups and older firms and, as previously explained, larger than the one observed in Africa. These results, in terms of the lower return to disruptiveness observed in Africa, are not 13 Appendix Figure B4 explores some of the technologies individually and finds that the positive associa- tion is statistically dominated by a few technologies: although most associations are positive, only a few are statistically significant. This, however, could be the result of a lack of statistical power given the constraint to every specific technology. 19 associated with the lower amount of funding (in absolute terms) that these countries re- ceive. Initially, the lower level of funding, which is indeed observed, is already accounted for by the regional terms and the intercept. Secondly, when we estimate split regressions for each region, we still observe that disruptive firms have a higher probability and re- ceive a higher amount of funding than non-disruptive firms in the same region, even in specifications where we control for country effects. The lower return to disruptiveness observed in these regions has, therefore, to do with other factors. One of them could be related to a potentially higher perception of risk compared to frontier countries, where there are lower information failures and risks can be more thoroughly evaluated. When we restrict our sample to firms obtaining funding, that is, the intensive margin of funding, in Table 3, the differences between disruptive and non-disruptive firms in the frontier and Africa become very small and not significant. This points to a critical role of the extensive margin of funding. Once firms obtain financing, there are no important differences in the value of funding according to disruptiveness. However, in the case of LAC, the premium, conditional on receiving funding, becomes negative, particularly for startups. A potential explanation for this result is that these firms associated with disrup- tive technologies might be perceived as relatively riskier and, therefore, obtain relatively lower tickets (around 27% lower).14 Table 1: Disruptiveness and Probability of Receiving Funding Any Technology Excluding Mobile Payments and Wi-Fi Ever Funded Last Funded Ever Funded Last Funded (1) (2) (3) (4) Disruptive 0.164*** 0.135*** 0.179*** 0.149*** (0.016) (0.011) (0.017) (0.013) Disruptive × Africa -0.134*** -0.103*** -0.152*** -0.114*** (0.018) (0.013) (0.020) (0.016) Disruptive × LAC -0.093*** -0.072*** -0.101*** -0.079*** (0.024) (0.016) (0.030) (0.021) R2 0.214 0.185 0.215 0.186 Observations 25464 25464 25464 25464 Countries 33 33 33 33 Clustered errors in parentheses. *** 10%, ** 5% and * 1%. All regressions include as controls age (in IHS transformation), dummies for 9 categories of employment size, and the number of similar firms (in IHS transformation), dummies for the number of overlapping sectors where the firm operates (one or two additional sectors). The dependent variable is a dummy variable that takes a value of one if the firm received funding. Standard errors clustered at the country level. 14 Tables A2 and A3 in the appendix re-estimates the tables 2 and 3 using a different definition of start-up: firm no older than 5 years. Results suggest again that most of the financial gain among disruptive firms accrues for older firms. 20 Table 2: Disruptiveness and Elasticity of Funding Any Technology Excluding Mobile Payments and Wi-Fi Ever Funded Last Funded Ever Funded Last Funded (1) (2) (3) (4) Panel a. All Firms Disruptive 2.634*** 2.090*** 2.847*** 2.297*** (0.284) (0.195) (0.257) (0.178) Disruptive × Africa -2.239*** -1.683*** -2.505*** -1.878*** (0.300) (0.204) (0.294) (0.210) Disruptive × LAC -1.635*** -1.245*** -1.776*** -1.374*** (0.387) (0.249) (0.456) (0.300) R2 0.241 0.208 0.242 0.209 Observations 25464 25464 25464 25464 Countries 33 33 33 33 Panel b. Start-ups (10 years old or less) Disruptive 2.172*** 1.795*** 2.417*** 2.029*** (0.107) (0.080) (0.074) (0.100) Disruptive × Africa -1.891*** -1.451*** -2.150*** -1.609*** (0.173) (0.126) (0.144) (0.145) Disruptive × LAC -1.297*** -1.024*** -1.407*** -1.114*** (0.225) (0.173) (0.364) (0.260) R 2 0.229 0.193 0.230 0.195 Observations 15634 15634 15634 15634 Countries 33 33 33 33 Panel c. Older firms (11 years old or more) Disruptive 2.944*** 2.161*** 3.095*** 2.328*** (0.397) (0.135) (0.411) (0.151) Disruptive × Africa -2.339*** -1.616*** -2.526*** -1.840*** (0.489) (0.241) (0.485) (0.246) Disruptive × LAC -2.137*** -1.461*** -2.294*** -1.643*** (0.504) (0.243) (0.533) (0.284) R 2 0.162 0.128 0.163 0.129 Observations 9830 9830 9830 9830 Countries 33 33 33 33 Clustered errors in parentheses. *** 10%, ** 5% and * 1%. All regressions include as controls age (in IHS transformation), dummies for 9 categories of employment size, and the number of similar firms (in IHS transformation), dummies for the number of overlapping sectors where the firm operates (one or two additional sectors). The dependent variable is the IHS transformation of the amount of funding in US Dol- lars received in the last funding round. Standard errors clustered at the country level. 21 Table 3: Disruptiveness and elasticity of funding conditional on receiving funding Any Technology Excluding Mobile Payments and Wi-Fi Ever Funded Last Funded Ever Funded Last Funded (1) (2) (3) (4) Panel a. All Firms Disruptive 0.036 0.024 0.016 0.043 (0.127) (0.119) (0.117) (0.134) Disruptive × Africa -0.023 -0.077 -0.077 -0.190 (0.217) (0.194) (0.275) (0.290) Disruptive × LAC -0.306** -0.344** -0.352** -0.387** (0.131) (0.128) (0.129) (0.150) R2 0.524 0.476 0.524 0.521 Observations 6519 5791 6519 5791 Countries 33 32 33 32 Panel b. Start-ups (10 years old or less) Disruptive 0.035 0.063 0.025 0.054 (0.130) (0.133) (0.120) (0.136) Disruptive × Africa -0.035 -0.086 -0.059 -0.175 (0.200) (0.199) (0.254) (0.261) Disruptive × LAC -0.324** -0.374** -0.374*** -0.398*** (0.143) (0.143) (0.126) (0.144) R 2 0.531 0.485 0.531 0.527 Observations 5520 4907 5520 4907 Countries 33 32 33 32 Panel c. Older firms (11 years old or more) Disruptive 0.036 -0.138 -0.009 0.010 (0.198) (0.181) (0.206) (0.223) Disruptive × Africa 0.184 -0.058 -0.270 -0.317 (0.663) (0.681) (0.602) (0.666) Disruptive × LAC -0.134 -0.177 -0.150 -0.259 (0.301) (0.215) (0.302) (0.298) R2 0.458 0.428 0.458 0.461 Observations 999 884 999 884 Countries 24 22 24 22 Clustered errors in parentheses. *** 10%, ** 5% and * 1%. All regressions include as controls age (in IHS transformation), dummies for 9 categories of employment size, number of similar firms (in IHS transformation), dummies for the number of overlapping sectors where the firm operates (one or two additional sectors). The dependent variable is the IHS transformation of the amount of funding in US Dollars received in the last funding round. Standard errors clustered at the country level. 22 4.3.1 Instrumental variable approach Table A4 in the Appendix shows the first stage results of our specification using the instru- ment described in subsection 3.2.3. The coefficient associated with the IV is statistically significant at the 1% level, and the F-stat in every specification is above the rule of thumb of ten. We implement the IV estimation on the sample of African firms born after 2010 to ensure that the IV represents characteristics at baseline. This results in a drop in the number of observations mainly due to a lack of ILO information on the skill mismatch for African countries. Table 4 presents the results15 for the probability of receiving funding, the IHS trans- formation of the amount received, and the IHS transformation of the amount received conditional on being funded among African firms.16 Regarding the probability of receiving funding, it is worth recalling that the overall coefficient for Africa derived from column 1 in table 1 is three percentage points (0.164 - 0.134). Column 1 from panel (a) in Table 4 suggests a higher coefficient (7 percentage points). Results are similar for the probability of receiving recent funding (last funded) or if we use the alternative definition of disruptive technologies that excludes mobile payments and Wi-Fi. Similarly, the estimated elasticity of funding in panel (b) of Table 4 is 1.05 p.p., which is higher than the elasticity estimated in table 2, 0.40 (2.634-2.239). The conditional elasticity shown in panel (c) becomes statistically significant for the disruptive measure that con- siders all technologies (columns 1 and 2). Disruptive firms with access to finance receive more funding than non-disruptive firms with access to finance. Overall, the IV results confirm the findings of Tables 1 to 3 but tend to be larger in magnitude. Similar to the case of Iacovone et al. (2023), the larger magnitudes obtained in the IV specifications could be related to heterogeneity of treatment, in this case, in the returns to disruptiveness.17 The results are robust to including country fixed effects and, as shown in Appendix Table A5, the results are qualitatively similar.18 15 The number of countries and observation reduces significantly for this IV regression due to the fact that we are restricting the sample to firms that were in operation before 2010 and to the data availability of mismatch indicators from ILO. Yet, the original OLS results for this subsample are similar to those reported in Tables 1 and 2. 16 Results for LAC are reported in Appendix tables A6 and A7. 17 Card (2001) presents a discussion of the cases in which OLS estimators are lower than IV in the context of the returns to education. 18 We also used a measure of mismatch with only the matched skills –i.e., excluding over-skilled workers, and results (available upon request) were similar. 23 Table 4: IV Estimates of the Effect of Disruptiveness on the Probability and Elasticity of Funding Any Technology Excluding Mobile Payments and Wi-Fi Ever Funded Last Funded Ever Funded Last Funded (1) (2) (3) (4) Panel a. Probability of Being Funded Disruptive 0.070** 0.076*** 0.061* 0.066** (0.030) (0.024) (0.035) (0.029) R2 0.144 0.123 0.145 0.125 Observations 2428 2428 2428 2428 Countries 9 9 9 9 Panel b. Elasticity of Funding Disruptive 1.049** 1.091*** 0.922** 0.965*** (0.412) (0.301) (0.454) (0.337) R2 0.163 0.142 0.164 0.144 Observations 2428 2428 2428 2428 Countries 9 9 9 9 Panel c. Elasticity of Funding Conditional on Being Funded Disruptive 0.422* 0.343 0.278 0.216 (0.231) (0.288) (0.320) (0.493) R2 0.437 0.436 0.439 0.436 Observations 548 481 548 481 Countries 9 9 9 9 Clustered errors in parentheses. *** 10%, ** 5% and * 1%. All regressions include as controls age (in IHS transformation), dummies for 9 categories of employment size, number of similar firms (in IHS transformation), dum- mies for the number of overlapping sectors where the firm operates (one or two additional sectors). The dependent variable is either a dummy variable indicating if the firm received funding (panel a), or the IHS transformation of the amount of funding in US Dollars received in the last funding round (panels b and c). Standard errors clustered at the country level. 4.4 Disruptiveness and performance To analyze the correlation between the incorporation of disruptive technologies in firms’ products and services offering and firm-level performance, we rely on Pitchbook data, which includes more detailed information in terms of valuation and growth. Due to data restrictions, in this case, we are only able to compare Africa with LAC. In Table 5, we estimate similar models to the ones shown in previous tables but using some performance-related indicators as dependent variables. As shown in panel (a) of the table, disruptive firms in Africa receive funding 0.66 years earlier than non-disruptive 24 firms. Still, the value of funding is 15% lower, while the first valuation of the firm is 12% lower. Once again, this could potentially be explained by some risk considerations. On the other hand, these firms exhibit 4.7 percentage points higher probability of success and an average 46% larger valuation growth over the life cycle of the firm. Furthermore, they tend to receive larger deals later on. The results do not change much if we only con- sider those firms that were born after the technology; that is, those firms that potentially were created with these disruptive technologies embedded in their offering rather than adopted them afterward.19 The results for LAC do not differ significantly against Africa, except for the case of the probability of success, which, though still larger than non-disruptive firms, is lower than the one observed for disruptive firms in Africa. Still, when we consider only those firms that started operations after the technology was created, we see that the ticket value, though still lower for disruptive firms, is not as low as the one observed in Africa. Table 5: Disruptiveness and Performance Years Until First First Success Value Value Annual Deal First Deal Deal Value Valuation Probability Growth Growth Growth (1) (2) (3) (4) (5) (6) (7) Panel a. All Firms Disruptive -0.655*** -0.153*** -0.116*** 4.662*** 0.455** 0.053 0.039* (0.190) (0.031) (0.030) (1.035) (0.168) (0.041) (0.023) Disruptive × LAC 0.304 0.071 0.052 -2.990* -0.142 -0.116** -0.004 (0.275) (0.054) (0.049) (1.760) (0.167) (0.056) (0.032) R2 0.286 0.172 0.088 0.235 0.231 0.106 0.152 Observations 2065 2327 2327 2327 365 344 1851 Countries 51 53 53 53 34 34 47 Panel b. All Firms born after technology Disruptive -0.917*** -0.172*** -0.129*** 5.517*** 0.477*** 0.055 0.049*** (0.277) (0.042) (0.031) (0.795) (0.169) (0.044) (0.018) Disruptive × LAC 0.160 0.089 0.047 -3.287** -0.118 -0.118* 0.003 (0.346) (0.058) (0.053) (1.482) (0.164) (0.061) (0.027) R2 0.266 0.170 0.089 0.239 0.234 0.105 0.158 Observations 2023 2273 2273 2273 354 334 1813 Countries 51 53 53 53 34 34 47 Clustered errors in parentheses. *** 10%, ** 5% and * 1%. All regressions include as controls age (in IHS transformation), dummies for 9 categories of employment size, number of similar firms (in IHS transformation, dummies for the number of overlapping sectors where the firm operates (one or two additional sectors). Standard errors clustered at the country level. 19 ’Success probability’ is based on the PitchBook VC Exit Predictor model, which leverages machine learning and our vast database of information about VC-backed companies, financing rounds, and investors to assess a startup’s prospect of a successful exit. ’Value growth’ is defined as the ratio between the last and the first valuation. 25 5 Conclusion This paper analyzed the level of disruptiveness of digital businesses in Africa along their life cycle and its association with funding. Using data from Crunchbase and through tex- tual analysis, we identified which firms incorporate into their products and services tech- nologies considered disruptive according to Bloom et al. (2021) classification, based on earning conference calls and patents information. To produce meaningful comparisons, we collected data on Africa, Latin America, and a group of cities at the technological frontier like Seattle, Palo Alto, London, and Tokyo. Our results indicate that African digital firms incorporate a lower level of disruptive technologies into their offering than the frontier. Still, there is a great deal of sectoral heterogeneity, especially when we take into account mobile payments, particularly in the case of Fintech, where Africa is significantly closer to the frontier, particularly on the extensive margin (probability of incorporating at least one of these technologies). The region is significantly below the frontier on the intensive margin (number of different technologies incorporated) in all the subsectors analyzed. Although startups in Africa tend to be more disruptive than older firms, the level is much lower than the frontier. Still, we observe a great level of heterogeneity in the life- cycle of disruptiveness by sector, with Fintech showing a different pattern, particularly for Kenya, South Africa, and Nigeria, which exhibit more disruption from older firms in this sector. These results are consistent with different characteristics of the sector, where the level of trust, technology infrastructure, and the possibility of bundling different prod- ucts, along with potential regulatory frictions, can be critical factors. We also find that, although disruptiveness positively correlates with funding, this re- lationship is weaker compared to the frontier. In addition, we find that even though being disruptive helps attract more funding compared to not being disruptive, this is pre- eminently occurring among older firms and mostly on the extensive margin of funding. These results are consistent with frictions affecting both the life cycle and the allocation of financing in the African market. Additional results indicate that being disruptive is associated with earlier financing but with lower tickets and higher growth in valuation and future funding deal size. Once again, these results point to some market imperfec- tions (e.g., information failures) related to the assessment of risk among these potentially disruptive and high-growth firms, which limits their funding. Our analyses can inform the intervention of multilateral development institutions in terms of their role in creating markets, contributing to de-risking some of these invest- ments, and, therefore, promoting growth in underdeveloped markets. 26 References B ELLEMARE , M. F. AND C. J. W ICHMAN (2020): “Elasticities and the Inverse Hyperbolic Sine Transformation,” Oxford Bulletin of Economics and Statistics, 82, 50–61. B LOOM , N., T. A. H ASSAN , A. K ALYANI , J. L ERNER , AND A. TAHOUN (2021): “The dif- fusion of disruptive technologies,” Tech. rep., National Bureau of Economic Research. B LOOM , N., R. S ADUN , AND J. VAN R EENEN (2012): “Americans Do IT Better: US Multi- nationals and the Productivity Miracle,” The American Economic Review, 102, 167–201. B OWER , J. L. AND C. M. C HRISTENSEN (1995): “Disruptive Technologies: Catching the Wave,” Harvard Business Review, 43. B RITER B RIDGES (2022): “Briter Intelligence,” Accessed on February, 2023. C ARD , D. (2001): “Estimating the return to schooling: Progress on some persistent econo- metric problems,” Econometrica, 69, 1127–1160. CB I NSIGHTS (2023): “State of venture 2023 report,” . C HITAVI , M., L. C OHEN , AND S. H AGIST (2021): “Kenya Is Becoming a Global Hub of FinTech Innovation,” Harvard Business Review, 21. C HRISTENSEN , C. M. (1997): The innovator’s dilemma: when new technologies cause great firms to fail, Harvard Business Review Press. C RUNCHBASE (2023): “Crunchbase,” https://www.crunchbase.com. D ALLE , J.-M., M. D EN B ESTEN , AND C. M ENON (2017): “Using Crunchbase for eco- nomic and managerial research,” . D ECKER , R., J. H ALTIWANGER , R. J ARMIN , AND J. M IRANDA (2014): “The role of en- trepreneurship in US job creation and economic dynamism,” Journal of Economic Per- spectives, 28, 3–24. D ISRUPT A FRICA (2022): “African Tech Startups Funding Report,” Tech. rep., Disrupt Africa. E LI , S., J. K. H AUSMAN , AND P. R HODE (2023): “The Model T,” Tech. rep., National Bureau of Economic Research. F ERNANDES , A. M., A. M ATTOO , H. N GUYEN , AND M. S CHIFFBAUER (2019): “The in- ternet and Chinese exports in the pre-ali baba era,” Journal of Development Economics, 138, 57–76. F RIEDERICI , N., S. O JANPER A ¨ , AND M. G RAHAM (2017): “The impact of connectivity in Africa: Grand visions and the mirage of inclusive digital development,” The Electronic Journal of Information Systems in Developing Countries, 79, 1–20. 27 F RIEDERICI , N., M. WAHOME , AND M. G RAHAM (2020): Digital entrepreneurship in Africa: How a continent is escaping Silicon Valley’s long shadow, The MIT Press. H ALTIWANGER , J., R. S. J ARMIN , AND J. M IRANDA (2013): “Who creates jobs? Small versus large versus young,” Review of Economics and Statistics, 95, 347–361. ´ I ACOVONE , L., M. P EREIRA -L OPEZ , AND M. S CHIFFBAUER (2023): “Competition makes it better: Evidence on when firms use it more effectively,” Research Policy, 52, 104786. I NTERNATIONAL L ABOUR O RGANIZATION (2023): “Education and Mismatch Indicators database (EMI),” ilostat.ilo.org/data. P ITCH B OOK (2023): “PitchBook,” https://pitchbook.com/. P ORZIO , T. (2023): “Cross-Country Differences in the Optimal Allocation of Talent and Technology,” Tech. rep., Mimeo. R AJAN , R. AND L. Z INGALES (1998): “Financial Dependence and Growth,” The American Economic Review, 88, 559–86. 28 A Appendix tables Table A1: Bigrams associated with the 29 disruptive technologies identified by Bloom et al. (2021) Technology Keywords 3D printing 3d printer; 3d printing; additive manufacturing; 3d printed Autonomous cars Self-driving car; robot car; autonomous vehicles; autonomous car; autonomous cars; automated driving; driverless car; autonomous driving; autonomous vehicle; driverless truck Bispecific monoclonal antibody bispecific monoclonal; the bispecific; bispecific antibody Cloud computing paas; cloud infrastructure; distributed cloud; cloud provider; cloud offerings; cloud service; cloud applications; community cloud; private cloud; public cloud; cloud deployments; cloud environments; cloud management; cloud services; cloud security; enterprise class; iaas; hybrid cloud; cloud platform; cloud providers; cloud hosting; personal cloud; enterprise network; cloud computing; cloud based; saas; cloud storage; enterprise applications; cloud solution; enterprise cloud; cloud solutions; cloud deployment Computer vision pose estimation; motion estimation; visual servoing; facial recognition; gesture recognition; computer vision; image recognition; sensor fusion; object recognition Drug conjugates kinase inhibitor; drug conjugate; antibody drug; drug conjugates Electronic gaming social game; video games; social games; video game; game content; electronic gaming; gaming products Millimeter wave; millimeter wave Fingerprint sensor fingerprint sensor; fingerprint scanner Fracking fracking; fraccing; hydrofracking; hydrofracturing; hydraulic fracturing GPS gps systems; global positioning; navigation devices Hybrid vehicle/Electric car hybrid vehicle; electric vehicle; electric motorcycle; vehicle charging; hybrid electric; plugin hybrids; electric buses; electrical vehicles; electric car; electric vehicles Lane departure warning lane departure; departure warning Lithium battery ion battery; lithium ion battery; lithium ion batteries; lithium batteries; ion batteries; lithium polymer; lithium ion; lithium battery Machine Learning/AI neural network; deep learning; language processing; machine learning; machine intelligence; natural language; artificial intelligence; AI technology; supervised learning; learning algorithms; unsupervised learning; reinforcement learning; AI machine Mobile payment mobile transfer; mobile commerce; mobile payment; mobile wallet; mobile money OLED display; oled 29 Online streaming streaming content; music streaming; interactive tv; live stream; digital video; video conferencing; online streaming; online video; mobile video; streaming services; streaming media; live video; video ondemand; live streaming; video ad; internet radio; video streaming; streaming video RFID frequency identification; keyless entry; rfid tags; rfid Search Engine search engine; search engines Smart devices mobile devices; tablet computers; wearable devices; tablet pcs; smartphone tablet; android phones; media devices; smart phones; smart devices; smart tvs; smart speaker; smart watch; smart car; smart phone; iphone ipad; portable media; smart tablets; connected devices; smartphones tablets; android smartphones; phones tablets; android devices; smart refrigerator; smartcar; smartphone; smart tv; smart band Social networking user generated; user generated content; social platforms; networking sites; social channels; social media; social networking; social networks; social network Software defined radio defined radio Solar Power solar wafer; rooftop solar; solar modules; solar cells; crystalline silicon; silicon solar; solar panel; solar power; solar wafers; solar energy; solar applications; solar module; solar cell; solar pv; solar grade; solar panels; photovoltaic; solar thermal Stent graft stent graft Touch screen touch controller; touch panel; capacitive touch; touchscreen; touch screens; touch sensor Virtual reality virtual reality; augmented reality; mixed reality; extended reality Wi-Fi wifi hotspots; wifi network; wifi; broadband connectivity; wireless networks Wireless charging wireless charging; inductive charging Table A2: Disruptiveness and Elasticity of Funding: alternative definition of startup Any Technology Excluding Mobile Payments and Wi-Fi Ever Funded Last Funded Ever Funded Last Funded (1) (2) (3) (4) Panel a. All Firms Disruptive 2.634*** 2.090*** 2.847*** 2.297*** (0.284) (0.195) (0.257) (0.178) Disruptive × Africa -2.239*** -1.683*** -2.505*** -1.878*** (0.300) (0.204) (0.294) (0.210) Disruptive × LAC -1.635*** -1.245*** -1.776*** -1.374*** (0.387) (0.249) (0.456) (0.300) R2 0.241 0.208 0.242 0.209 Observations 25464 25464 25464 25464 Countries 33 33 33 33 Panel b. Start-ups (5 years old or less) Disruptive 1.477*** 1.229*** 1.703*** 1.486*** (0.048) (0.047) (0.154) (0.156) Disruptive × Africa -1.476*** -1.113*** -1.558*** -1.172*** (0.153) (0.139) (0.268) (0.243) Disruptive × LAC -0.838** -0.693** -0.848 -0.752 (0.336) (0.294) (0.570) (0.497) R2 0.233 0.202 0.235 0.203 Observations 8683 8683 8683 8683 Countries 33 33 33 33 Panel c. Older firms (6 years old or more) Disruptive 3.072*** 2.402*** 3.277*** 2.576*** (0.298) (0.204) (0.307) (0.228) Disruptive × Africa -2.486*** -1.844*** -2.811*** -2.078*** (0.353) (0.262) (0.354) (0.281) Disruptive × LAC -2.148*** -1.582*** -2.329*** -1.720*** (0.407) (0.266) (0.443) (0.296) R 2 0.223 0.177 0.224 0.178 Observations 16781 16781 16781 16781 Countries 33 33 33 33 Clustered errors in parentheses. *** 10%, ** 5% and * 1%. All regressions include as controls age (in IHS transformation), dummies for 9 categories of employment size, and the number of similar firms (in IHS transformation), dummies for the number of overlapping sectors where the firm operates (one or two additional sectors). The dependent variable is the IHS transformation of the amount of funding in US Dol- lars received in the last funding round. Standard errors clustered at the country level. 30 Table A3: Disruptiveness and elasticity of funding conditional on receiving funding: al- ternative definition of startup Any Technology Excluding Mobile Payments and Wi-Fi Ever Funded Last Funded Ever Funded Last Funded (1) (2) (3) (4) Panel a. All Firms Disruptive 0.036 0.024 0.016 0.043 (0.127) (0.119) (0.117) (0.134) Disruptive × Africa -0.023 -0.077 -0.077 -0.190 (0.217) (0.194) (0.275) (0.290) Disruptive × LAC -0.306** -0.344** -0.352** -0.387** (0.131) (0.128) (0.129) (0.150) R2 0.524 0.476 0.524 0.521 Observations 6519 5791 6519 5791 Countries 33 32 33 32 Panel b. Start-ups (5 years old or less) Disruptive -0.081 -0.002 -0.120** -0.075 (0.060) (0.077) (0.055) (0.077) Disruptive × Africa -0.004 -0.135 -0.019 -0.139 (0.213) (0.175) (0.264) (0.258) Disruptive × LAC -0.345*** -0.481*** -0.388*** -0.426*** (0.101) (0.097) (0.061) (0.079) R2 0.500 0.462 0.501 0.496 Observations 3461 3125 3461 3125 Countries 32 32 32 32 Panel c. Older firms (6 years old or more) Disruptive 0.137 0.062 0.137 0.146 (0.163) (0.154) (0.145) (0.173) Disruptive × Africa 0.103 0.099 -0.064 -0.158 (0.300) (0.292) (0.348) (0.439) Disruptive × LAC -0.168 -0.134 -0.224 -0.252 (0.168) (0.159) (0.173) (0.201) R2 0.543 0.501 0.543 0.543 Observations 3058 2666 3058 2666 Countries 31 28 31 28 Clustered errors in parentheses. *** 10%, ** 5% and * 1%. All regressions include as controls age (in IHS transformation), dummies for 9 categories of employment size, number of similar firms (in IHS transformation), dummies for the number of overlapping sectors where the firm operates (one or two additional sectors). The de- pendent variable is the IHS transformation of the amount of funding in US Dollars received in the last funding round. Standard errors clustered at the country level. 31 Table A4: First Stage: Pr of Being Disruptive Any Technology Excluding Mobile Payments and Wi-Fi (1) (2) (3) (4) IV 4.143*** 4.143*** (0.115) (0.115) IV-non mobile 3.239*** 3.239*** (0.131) (0.131) R2 0.360 0.360 0.225 0.225 Observations 2620 2620 2620 2620 F-Value 60.87 90.23 31.44 45.88 Country FE No Yes No Yes Clustered errors in parentheses. *** 10%, ** 5% and * 1%. All regressions include as controls age (in IHS transforma- tion), dummies for 9 categories of employment size, num- ber of similar firms (in IHS transformation, dummies for the number of overlapping sectors where the firm operates (one or two additional sectors). The dependent variable is a dummy variable that takes a value of one if the firm re- ceived funding. Standard errors clustered at the country level. 32 Table A5: IV Estimates of the Effect of Disruptiveness on the Probability and Elasticity of Funding in Africa: Country Fixed Effects Any Technology Excluding Mobile Payments and Wi-Fi (1) (2) (3) (4) Ever Funded Last Funded Ever Funded Last Funded Panel a. Probability of Being Funded Disruptive 0.062* 0.064** 0.049 0.051* (0.034) (0.027) (0.039) (0.028) R2 0.153 0.131 0.154 0.132 Observations 2428 2428 2428 2428 Countries 9 9 9 9 Panel b. Elasticity of Funding Disruptive 0.894* 0.920** 0.721 0.749** (0.514) (0.372) (0.568) (0.347) R2 0.171 0.148 0.172 0.150 Observations 2428 2428 2428 2428 Countries 9 9 9 9 Panel c. Elasticity of Funding Conditional on Being Funded Disruptive 0.429*** 0.493*** 0.318** 0.511* (0.143) (0.129) (0.132) (0.282) R2 0.473 0.478 0.476 0.479 Observations 548 481 548 481 Countries 9 9 9 9 Clustered errors in parentheses. *** 10%, ** 5% and * 1%. All regressions include as controls age (in IHS transformation), dummies for 9 categories of employment size, number of similar firms (in IHS transformation), dummies for the number of overlapping sectors where the firm operates (one or two additional sectors). The dependent variable is the IHS transformation of the amount of funding in US Dollars received in the last funding round. Standard errors clustered at the country level. 33 Table A6: IV Estimates of the Effect of Disruptiveness on the Probability and Elasticity of Funding: LAC Any Technology Excluding Mobile Payments and Wi-Fi (1) (2) (3) (4) Ever Funded Last Funded Ever Funded Last Funded Panel a. Probability of Being Funded Disruptive 0.124*** 0.113*** 0.125*** 0.114*** (0.016) (0.014) (0.015) (0.013) R2 0.102 0.089 0.105 0.092 Observations 5209 5209 5209 5209 Countries 9 9 9 9 Panel b. Elasticity of Funding Disruptive 1.603*** 1.402*** 1.559*** 1.363*** (0.250) (0.224) (0.229) (0.221) R2 0.141 0.120 0.143 0.123 Observations 5209 5209 5209 5209 Countries 9 9 9 9 Panel c. Elasticity of Funding Conditional on Being Funded Disruptive -0.463*** -0.569*** -0.679*** -0.839*** (0.109) (0.116) (0.175) (0.171) R2 0.573 0.529 0.568 0.519 Observations 1386 1225 1386 1225 Countries 9 9 9 9 Clustered errors in parentheses. *** 10%, ** 5% and * 1%. All regressions include as controls age (in IHS transformation), dummies for 9 categories of employment size, number of similar firms (in IHS transformation), dummies for the number of overlapping sectors where the firm operates (one or two additional sectors). The dependent variable is the IHS transformation of the amount of funding in US Dollars received in the last funding round. Standard errors clustered at the country level. 34 Table A7: IV Estimates of the Effect of Disruptiveness on the Probability and Elasticity of Funding: including country fixed effects, LAC Any Technology Excluding Mobile Payments and Wi-Fi (1) (2) (3) (4) Ever Funded Last Funded Ever Funded Last Funded Panel a. Probability of Being Funded Disruptive 0.128*** 0.117*** 0.129*** 0.119*** (0.016) (0.014) (0.016) (0.015) R2 0.106 0.093 0.108 0.096 Observations 5209 5209 5209 5209 Countries 9 9 9 9 Panel b. Elasticity of Funding Disruptive 1.685*** 1.488*** 1.667*** 1.472*** (0.237) (0.226) (0.230) (0.244) R2 0.144 0.124 0.146 0.126 Observations 5209 5209 5209 5209 Countries 9 9 9 9 Panel c. Elasticity of Funding Conditional on Being Funded Disruptive -0.415*** -0.526*** -0.629*** -0.789*** (0.096) (0.088) (0.167) (0.136) R2 0.577 0.533 0.573 0.525 Observations 1386 1225 1386 1225 Countries 9 9 9 9 Clustered errors in parentheses. *** 10%, ** 5% and * 1%. All regressions include as controls age (in IHS transformation), dummies for 9 categories of employment size, number of similar firms (in IHS transformation), dummies for the number of overlapping sectors where the firm operates (one or two additional sectors). The dependent variable is the IHS transformation of the amount of funding in US Dollars received in the last funding round. Standard errors clustered at the country level. 35 B Appendix figures Figure B1: Concentration of digital firms in main cities 36 Note: Own calculations using Crunchbase data. www.crunchbase.com Figure B2: Disruptiveness in the Extensive and Intensive Margins 37 Note: Frontier includes the pooled number of start-ups in London, Palo Alto, Seattle and Tokyo. Africa includes 20 African countries with information in Crunchbase, while LAC includes South America plus Mexico. “Top 1000” are the first 1000 firms in the Crunchbase ranking for these three industries. “Top 1000 Revenue” uses estimated revenue to rank firms starting from the largest. Panel a shows the gap in the extensive margin by industry. Panel b shows the gap in the intensive margin Figure B3: Example identification of disruptive technologies 38 Figure B3 continued: Example identification of disruptive technologies 39 Figure B4: Heterogeneity in the Probability of Being Ever Funded This table shows the differential effect of being disruptive on the probability of being ever funded by each individual technology. Due to power limitations, some had to be grouped in “Rest”. 40 Figure B5: Appearance Gap of Keywords (Africa - Frontier): E-Commerce 41 Figure B6: Appearance Gap of Keywords (Africa - Frontier): Fintech 42 Figure B7: Appearance Gap of Keywords (Africa - Frontier): IT 43