## INTRODUCTION: SMALL BUSINESS FINANCING NEEDS AN INNOVATIVE APPROACH ## Source details **Canonical URL:** [INTRODUCTION: SMALL BUSINESS FINANCING NEEDS AN INNOVATIVE APPROACH](https://www.imf.org/-/media/files/publications/gfs-notes/2021/english/gsnea2021002.pdf) ## Other formats - [Markdown version](/-/media/files/publications/gfs-notes/2021/english/gsnea2021002.pdf.md) - [Structured JSON version](/-/media/files/publications/gfs-notes/2021/english/gsnea2021002.pdf.json) --- ### Key challenges and COVID-19 effects - Main barriers to bank financing: high cost, physical distance, and lack of proper documentation (financial history and soft information), producing information asymmetry. - International Finance Corporation estimate: "40 percent of small businesses had an unmet financing need of US$5.2 trillion every year." - COVID-19 amplified vulnerabilities: - Small-business–concentrated sectors (notably services) were highly exposed to lockdowns. - Small businesses typically have limited cash buffers. - Pandemic increased perceived credit risk and accentuated information asymmetries. - Digital financial services offer potential to accelerate and enhance financial inclusion amid social distancing. ### Policy responses during COVID-19 - Measures to facilitate access to financing included: - Extending credit lines. - Allowing delays of loan repayments and moratoriums on principal for affected small businesses. - Providing guarantees for uncollateralized small business loans. - Reduction of fees and taxes and direct fiscal subsidies and grants. - Use of digital technology by member countries: - Digital platforms connecting small businesses with free services, training, and vendors (examples cited: Greece, Italy). - Grants for digital consulting services (example cited: Germany). - Credit facilities to buy or lease digital equipment or services (examples cited: Malaysia, Spain). ### Advantages and evidence from digital banks (MYbank and others) - Digital banks can leverage network effects, e-commerce platforms, and social networks; technology firms already active in lending: Mercado Credito (Argentina), Paytm (India), Amazon Lending (United States). - China examples: Ant Group, Baidu, Tencent, and JD.com making inroads; some obtained banking licenses. - MYbank characteristics and outcomes: - Borrower profile: - About "80 percent" of MYbank borrowers have less than 10 employees. - More than "70 percent" had difficulties accessing loans previously from any financial institutions. - Loan characteristics and performance: - Loans are generally smaller, shorter in duration, and used primarily for operational purposes. - By the end of 2020, MYbank had served more than "35 million MSEs", providing them with "¥31,000 (US$4,500)" in loans on average. - MYbank’s average nonperforming loan (NPL) ratio remained below "2 percent" even after an uptick post-COVID-19 onset. - Operating cost of MYbank loans: "¥2.3 or US$0.4 per loan" (MYbank 2018). - Operational strengths during COVID-19: - Gathering electronic data trails to track business activities and assess business continuity. - Providing remote lending quickly using proprietary transaction data and traditional credit/tax histories (example: XWBank approved "330,000 loan applications a day with only 270 lending officers"). - Leveraging data analytics and machine learning to predict cash-flow needs and default rates by customer group (location, sector, activity, business relationship). - Rapid product development (example: WeBank launched new products in "10 days"). - MYbank dynamic pre-approved credit lines and the "3-1-0 model": registration/application within "3 minutes", money transferred to an Alipay account within "1 second", and "0 human intervention". ### MYbank customer vulnerability and financing needs during the pandemic - Customer composition: - Many offline merchants are self-employed urban or rural businesses: about "30 percent" restaurants, "20 percent" local convenience stores, "10 percent" fresh food sellers, remainder local retailers and service providers. - Loan size comparisons: - Average loan amount needed by these micro firms: "¥25,000 (about US$3,500)". - Average commercial loan size from traditional banks: "¥1 million (US$150,000)". - Survey findings (February/March 2020; 20,165 valid questionnaires): - More than half of surveyed MSEs in China expected a "50 percent or larger year-over-year decline" in sales revenue in 2020:Q1 due to the pandemic; declines were especially severe for service-sector and smaller firms. - More than two-thirds of MSEs expected financing gaps of between "¥10,000 (US$1,500) and ¥500,000 (US$70,000)". - More than two-thirds expected loans with durations of between "six months and a year". - Sample regional distribution: Eastern China "38.2 percent", Southern China "17.8 percent", Western China "17.1 percent", Central China "16.7 percent", Northern China "10.2 percent". - Firm size in 2019: "72.9 percent" had sales revenue of less than "¥1 million", and "76.8 percent" employed fewer than "10 employees". ### Empirical methodology to assess economic impact of digital lending - Objective: investigate correlation and causal link between digital bank lending at pandemic onset (February 2020) and firm performance during the pandemic, focusing on MSEs that typically lacked access to traditional banks. - Data universe: micro-level data from MYbank customer pool; samples not restricted to firms with certain credit line levels. - Two-step propensity score matching (PSM) approach: - Step 1: Estimate probability of borrowing using cross-sectional probit regressions on 2019 data. Dependent variable: whether a firm had borrowed from MYbank in 2019:Q4. Independent variables include sector, region, average monthly sales and loan balance, and standard deviations of these variables during 2019:Q1–Q3. Sales and loan balance are in absolute Yuan. Sector classification uses Chinese Merchant Category Code; region uses provinces. - Step 2: Use propensity scores to match borrowers to nonborrowers via k nearest neighbor (k-NN). Treatment group: randomly selected "20,000 MSEs" that borrowed in February 2020; control group: randomly selected "20,000 MSEs" that did not borrow in February 2020. The mean difference of sales growth between groups measures the average treatment effect. - Endogeneity and identification: - Proxy variable: firms’ pre-shock loan balance as of 2019:Q4 (zero or positive) used to separate treatment/control and to support causal interpretation since it precedes the pandemic. - Verification: firms with positive 2019:Q4 loan balances had higher actual borrowing from MYbank during the pandemic, holding other firm characteristics constant. - Robustness checks: - Subsample analysis for offline service sector firms produced similar findings. - Gender and regional differences between treatment and control groups were examined and found similar to whole-sample results. - Notable model detail: the probit specification referenced Φ−1(Yi)=β0+β1sectori+β2regioni+β3FirmChari where Y is dummy for borrowing in 2019:Q4 and FirmChar includes past monthly average sales, loan balance, and standard deviations (2019:Q1–Q3). ### Empirical results: impact of digital lending on MSE sales and activity - Sample: random sample of "40,000 MSEs" (treatment and control of "20,000" each). - Main findings: - MSEs that borrowed in February 2020 were more active on average than nonborrowers. - Borrowers registered positive year-over-year sales growth rates while nonborrowers had much lower and even negative year-over-year sales growth rates. - Within February 2020 borrowers, firms that continued to borrow in March and April showed significantly higher sales growth. - Offline service-sector firms that received loans at pandemic onset significantly outperformed those that did not, despite similar past performance in 2019. - Sales growth of the treatment group still declined from "38 percent" in March 2019 to "10 percent" in March 2020, indicating they were not immune to the shock but fared relatively better. - Estimated causal impact (sales response per loan amount): - Every ¥1,000 lent led to an increase in the firm’s sales revenue by ¥1,170 in April 2020. - Every ¥1,000 lent led to an increase in the firm’s sales revenue by ¥1,340 in May 2020. - Every ¥1,000 lent led to an increase in the firm’s sales revenue by ¥1,130 in June 2020. - Interpretation: digital banking’s ability to leverage digital data and platforms for remote lending played a positive role in supporting small businesses amid the pandemic and complemented traditional bank credit. ### MSE business continuity (Feb–Jun 2020) and heterogeneity - Definitions and patterns: - Non-activity defined as having low (less than 20 percent of average monthly sales in 2019) or zero monthly sales. - In February 2020 the share of active MSEs with positive monthly sales declined sharply and bounced back starting in March. - Gender heterogeneity: - Random draws: 10,000 female-owned MSEs and 10,000 male-owned MSEs with similar distributions of city tiers and 2019 sales. - Year-over-year growth rates of digital lending to female-owned MSEs were positive between January and May 2020 and were 15 to 25 percentage points higher than that of male-owned MSEs. - Relationship slightly reversed in June 2020, but the share of active female-owned MSEs remained higher than that of male-owned MSEs. - Geographic heterogeneity: - Random draws: 10,000 MSEs from higher-tier cities (first- and second-tier) and 10,000 from lower-tier cities (third-, fourth-, and fifth-tier), ensuring similar distributions of genders and 2019 sales. - Lending growth rates to MSEs in less developed regions outpaced those in developed regions during the pandemic, though this relationship reversed starting in May 2020. - In both region groups, the share of active MSEs with positive sales declined sharply in February and bounced back afterward. - Even when lending growth rates to less developed regions became lower than to developed regions in May and June 2020, the share of active MSEs in less developed regions was higher than that in more developed regions. ### Limitations, risks, and robustness - Limitations acknowledged: - Endogeneity concerns: firms may borrow because of better sales prospects; pandemic could systematically affect borrowing decisions and sales. - Identification relies on pre-COVID borrowing (2019:Q4) and zero loan balances in Jan 2020 to mitigate endogeneity. - Regression results available upon request; confidence bands are not available, so results are indicative. - Analysis is based on five-month data during the lockdown; long-term impact remains to be seen. - Overall support from digital banks to the real economy is still limited given their small size in the whole financial system. - Risks related to digital lending: - Need to strengthen regulation and ensure financial stability, including data privacy and security. - Balance benefits from financial innovation with guarding against potential risks (Jeffrey 2021). - Regulatory and supervisory frameworks should consider mitigated risks (machine learning and big data reducing information asymmetry) and increased risks (cyber security and data security). ### Policy implications and recommendations - Role of digital banks: - Digital banks’ ability to remotely assess creditworthiness and lend highlights the importance of digital lending to facilitate business continuity for small businesses during pandemics and economic downturns. - Digital banks (e.g., MYbank, WeBank, XW Bank) appear to have reduced the financing gap and supported small business continuity during the COVID-19 lockdown. - Digital lending enabled higher sales growth and improved access to finance for segments less served by traditional banks (female-owned MSEs and MSEs in less developed regions). - Regulatory and structural policies: - Consider empowering digital banks to support small businesses and achieve more inclusive growth, particularly for vulnerable groups during pandemics and recessions. - Policies to enhance open banking—allowing data to move securely between traditional banks and digital banks in a standardized way—could further support small business financing and financial inclusion. - Countries could strengthen digital banking services based on specific circumstances; making information easier to share could boost supply of and demand for credit, increasing SME lending without increasing system risks (Bank of England 2020). - Policymakers could focus on digitally enabled short-term liquidity products without collateral, with higher interest rates for appropriate risk compensation and dynamic credit line adjustments—recognizing limitations of big data and machine learning where data are limited. - Supervision and future research: - Digital banks carry their own risks which need to be carefully understood, monitored, and managed; these topics warrant future research beyond the scope of this paper. *Source: INTRODUCTION: SMALL BUSINESS FINANCING NEEDS AN INNOVATIVE APPROACH (gsnea2021002)* ### INTRODUCTION: SMALL BUSINESS FINANCING NEEDS AN INNOVATIVE APPROACH ### INTRODUCTION: SMALL BUSINESS FINANCING NEEDS AN INNOVATIVE APPROACH ### Key challenges for small businesses and financing needs - Main barriers to bank financing: high cost, physical distance, and lack of proper documentation (financial history and soft information), producing information asymmetry. - International Finance Corporation estimate: "40 percent of small businesses had an unmet financing need of US$5.2 trillion every year." - COVID-19 amplified vulnerabilities: small-business–concentrated sectors (notably services) were highly exposed to lockdowns; small businesses typically have limited cash buffers. - Pandemic increased perceived credit risk and accentuated information asymmetries; digital financial services offer potential to accelerate and enhance financial inclusion amid social distancing. ### Policy responses during COVID-19 (measures noted in the source) - Measures to facilitate access to financing included: - Extending credit lines. - Allowing delays of loan repayments and moratoriums on principal for affected small businesses. - Providing guarantees for uncollateralized small business loans. - Reduction of fees and taxes and direct fiscal subsidies and grants. - Use of digital technology by member countries: - Digital platforms connecting small businesses with free services, training, and vendors (examples cited: Greece, Italy). - Grants for digital consulting services (example cited: Germany). - Credit facilities to buy or lease digital equipment or services (examples cited: Malaysia, Spain). ### Advantages of digital banks for small business financing - Digital banks can leverage network effects, e-commerce platforms, and social networks; technology firms already active in lending: Mercado Credito (Argentina), Paytm (India), Amazon Lending (United States). - China examples: Ant Group, Baidu, Tencent, and JD.com making inroads; some obtained banking licenses. - Characteristics and outcomes from MYbank experience: - Borrower profile: - About "80 percent" of MYbank borrowers have less than 10 employees. - More than "70 percent" had difficulties accessing loans previously from any financial institutions. - Loan characteristics: - Loans are generally smaller, shorter in duration, and used primarily for operational purposes. - By the end of 2020, MYbank had served more than "35 million MSEs", providing them with "¥31,000 (US$4,500)" in loans on average. - MYbank’s average nonperforming loan (NPL) ratio remained below "2 percent" even after an uptick post-COVID-19 onset. - Operating cost of MYbank loans: "¥2.3 or US$0.4 per loan" (MYbank 2018). - Digital-bank operational strengths during COVID-19: - Gathering electronic data trails to track business activities and assess business continuity. - Providing remote lending quickly using proprietary transaction data and traditional credit/tax histories (example: XWBank approved "330,000 loan applications a day with only 270 lending officers"). - Leveraging data analytics and machine learning to predict cash-flow needs and default rates by customer group (location, sector, activity, business relationship). - Rapid product development (example: WeBank launched new products in "10 days"). ### MYbank customer vulnerability and financing needs during the pandemic - Customer composition (survey basis and sectoral exposure): - Many offline merchants are self-employed urban or rural businesses: about "30 percent" restaurants, "20 percent" local convenience stores, "10 percent" fresh food sellers, remainder local retailers and service providers. - Loan size comparisons: - Average loan amount needed by these micro firms: "¥25,000 (about US$3,500)". - Average commercial loan size from traditional banks: "¥1 million (US$150,000)". - Survey findings (February/March 2020; 20,165 valid questionnaires): - More than half of surveyed MSEs in China expected a "50 percent or larger year-over-year decline" in sales revenue in 2020:Q1 due to the pandemic; declines were especially severe for service-sector and smaller firms. - More than two-thirds of MSEs expected financing gaps of between "¥10,000 (US$1,500) and ¥500,000 (US$70,000)". - More than two-thirds expected loans with durations of between "six months and a year". - Sample regional distribution: Eastern China "38.2 percent", Southern China "17.8 percent", Western China "17.1 percent", Central China "16.7 percent", Northern China "10.2 percent". - Firm size in 2019: "72.9 percent" had sales revenue of less than "¥1 million", and "76.8 percent" employed fewer than "10 employees". ### Empirical methodology to assess economic impact of digital lending - Objective: investigate correlation and causal link between digital bank lending at pandemic onset (February 2020) and firm performance during the pandemic, focusing on MSEs that typically lacked access to traditional banks. - Data universe: micro-level data from MYbank customer pool; samples not restricted to firms with certain credit line levels. - Two-step propensity score matching (PSM) approach to address selection bias: - Step 1: Estimate probability of borrowing using cross-sectional probit regressions on 2019 data. Dependent variable: whether a firm had borrowed from MYbank in 2019:Q4. Independent variables include sector, region, average monthly sales and loan balance, and standard deviations of these variables during 2019:Q1–Q3. Sales and loan balance are in absolute Yuan. Sector classification uses Chinese Merchant Category Code; region uses provinces. - Step 2: Use propensity scores to match borrowers to nonborrowers via k nearest neighbor (k-NN). Treatment group: randomly selected "20,000 MSEs" that borrowed in February 2020; control group: randomly selected "20,000 MSEs" that did not borrow in February 2020. The mean difference of sales growth between groups measures the average treatment effect. - Addressing endogeneity: - Proxy variable: firms’ pre-shock loan balance as of 2019:Q4 (zero or positive) used to separate treatment/control and to support causal interpretation since it precedes the pandemic. - Verifying assumption: firms with positive 2019:Q4 loan balances had higher actual borrowing from MYbank during the pandemic, holding other firm characteristics constant. - Robustness checks: - Subsample analysis for offline service sector firms (heavily hit) produced similar findings. - Gender and regional differences between treatment and control groups were examined and found similar to whole-sample results. - Notable methodological details: - The probit model specification referenced: Φ−1(Yi)=β0+β1sectori+β2regioni+β3FirmChari where Y is dummy for borrowing in 2019:Q4 and FirmChar includes past monthly average sales, loan balance, and standard deviations (2019:Q1–Q3). - MYbank dynamic pre-approved credit lines and the "3-1-0 model": registration/application within "3 minutes", money transferred to an Alipay account within "1 second", and "0 human intervention". ### Empirical results: economic impact of digital lending - Main sample: random sample of "40,000 MSEs" (treatment and control of "20,000" each). - Findings: - MSEs that borrowed in February 2020 were more active on average than nonborrowers. - Borrowers registered positive year-over-year sales growth rates while nonborrowers had much lower and even negative year-over-year sales growth rates. - Within February 2020 borrowers, firms that continued to borrow in March and April showed significantly higher sales growth. - Offline service-sector firms that received loans at pandemic onset significantly outperformed those that did not, despite similar past performance in 2019. - Sales growth of the treatment group still declined from "38 percent" in March 2019 to "10 percent" in March 2020, indicating they were not immune to the shock but fared relatively better. - Interpretation: - Results support the hypothesis that digital banking’s ability to leverage digital data and platforms for remote lending played a positive role in supporting small businesses amid the pandemic and that digital banking credit complemented traditional bank credit. *Source: INTRODUCTION: SMALL BUSINESS FINANCING NEEDS AN INNOVATIVE APPROACH (gsnea2021002)* ### 1. MSE Business Continuity ### 1. MSE Business Continuity ### MSE activity and business continuity during Feb–Jun 2020 - Panel definitions: - Non-activity defined as having low (less than 20 percent of average monthly sales in 2019) or zero monthly sales. - Non-activity does not necessarily lead to defaults. - Observations (Feb–Jun 2020): - Share of non-active MSEs plotted between February and June 2020 (Panels refer to figures in source). - In February 2020 the share of active MSEs with positive monthly sales declined sharply and bounced back starting in March (noted across gender and regional comparisons). ### Causal analysis of digital bank lending on MSE sales - Sample and identification: - A random sample of another 40,000 MSEs was used to investigate a possible causal relationship between digital bank lending and higher sales growth. - Treatment group constructed by selecting firms that borrowed in 2019:Q4; control group were similar MSEs that did not borrow in 2019:Q4. - Propensity score matching (PSM) used to ensure sector, region, prior sales, and borrowing before 2019:Q4 were similar between groups. - Additional requirement: firms in both groups had a zero loan balance in Jan 2020 to ensure similar credit demand immediately before the pandemic. - Identification assumption: borrowing in 2019:Q4 proxies higher likelihood to borrow during the pandemic but is uncorrelated with the exogenous impact from the pandemic. - Estimation and results: - Causal effect calculated as: (difference of the average sales revenue between the two groups) divided by (difference of the average amount of new loans between the two groups). - Estimated impact of MYbank lending on MSE sales: - Every ¥1,000 lent led to an increase in the firm’s sales revenue by ¥1,170 in April 2020. - Every ¥1,000 lent led to an increase in the firm’s sales revenue by ¥1,340 in May 2020. - Every ¥1,000 lent led to an increase in the firm’s sales revenue by ¥1,130 in June 2020. - These results support that access to credit from the digital bank led to higher sales during the pandemic. ### Digital lending inclusivity: gender and geography - Gender differences: - Random draws: 10,000 female-owned MSEs and 10,000 male-owned MSEs ensuring similar distributions of city tiers and of sales in 2019. - Year-over-year growth rates of digital lending to female-owned MSEs were positive between January and May 2020 and were 15 to 25 percentage points higher than that of male-owned MSEs. - Relationship slightly reversed in June 2020, but the share of active female-owned MSEs remained higher than that of male-owned MSEs, partially reflecting better business continuity for female-owned MSEs. - Geographic differences: - Random draws: 10,000 MSEs from higher-tier cities (first- and second-tier) and 10,000 from lower-tier cities (third-, fourth-, and fifth-tier), ensuring similar distributions of genders and sales in 2019. - Lending growth rates to MSEs in less developed regions (third-, fourth-, and fifth-tier cities) outpaced those in developed regions (first- and second-tier cities) during the pandemic, though this relationship reversed starting in May 2020. - In both region groups, the share of active MSEs with positive sales declined sharply in February and bounced back afterward. - Even when lending growth rates to less developed regions became lower than to developed regions in May and June 2020, the share of active MSEs in less developed regions was higher than that in more developed regions, partially reflecting better business continuity in less developed regions. ### Limitations and robustness notes - Endogeneity concerns acknowledged: firms may borrow because of better sales prospects; pandemic could systematically affect borrowing decisions and sales. - Identification strategy relied on pre-COVID borrowing (2019:Q4) and zero loan balances in Jan 2020 to mitigate endogeneity. - Regression results are available upon request; as confidence bands are not available, results should be seen as indicative. - Analysis is based on five-month data during the lockdown; long-term impact remains to be seen. - The overall support from digital banks to the real economy is still limited given their small size in the whole financial system. ### Policy implications and recommendations - Role of digital banks: - Digital banks’ ability to remotely assess creditworthiness and lend highlights the importance of digital lending to facilitate business continuity for small businesses during pandemics and economic downturns. - Digital banks (e.g., MYbank, WeBank, XW Bank) appear to have reduced the financing gap and supported small business continuity during the COVID-19 lockdown. - Digital lending enabled higher sales growth and improved access to finance for segments less served by traditional banks (female-owned MSEs and MSEs in less developed regions). - Regulatory and structural policies: - Consider empowering digital banks to support small businesses and achieve more inclusive growth, particularly for vulnerable groups during pandemics and recessions. - Policies to enhance open banking—allowing data to move securely between traditional banks and digital banks in a standardized way—could further support small business financing and financial inclusion. - Countries could strengthen digital banking services based on specific circumstances; making information easier to share could boost supply of and demand for credit, increasing SME lending without increasing system risks (Bank of England 2020). - Policymakers could focus on digitally enabled short-term liquidity products without collateral, with higher interest rates for appropriate risk compensation and dynamic credit line adjustments—recognizing limitations of big data and machine learning where data are limited. - Regulation and supervision: - Need to strengthen regulation and ensure financial stability, including data privacy and security. - Balance benefits from financial innovation with guarding against potential risks (Jeffrey 2021). - Regulatory and supervisory frameworks should consider both mitigated risks (machine learning and big data reducing information asymmetry) and increased risks (cyber security and data security) associated with wide use of data in small business financing. - Digital banks carry their own risks which need to be carefully understood, monitored, and managed; these topics are beyond the scope of this paper and warrant future research. *IMF | Monetary and Capital Markets* --- _Source: https://www.imf.org/-/media/files/publications/gfs-notes/2021/english/gsnea2021002.pdf_