## wpiea2021090-print-pdf

## Source details

**Canonical URL:** [wpiea2021090-print-pdf](https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021090-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2021/english/wpiea2021090-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2021/english/wpiea2021090-print-pdf.pdf.json)

---

### Introduction and key contribution
- Objective: construct a digital financial inclusion index covering 52 emerging markets and developing economics (EMDEs) for 2014 and 2017, and combine it with a traditional financial inclusion index to form a comprehensive aggregate index.
- Focus: payments aspect of financial inclusion enabled by digital financial services (DFSs); indices exclusively focus on the payments aspects of financial inclusion.
- Data sources: World Bank Global Findex Database, IMF Financial Access Survey (FAS), International Telecommunication Union (ITU), and GSMA Mobile Money Dataset.
- Principal findings:
  - Digital financial inclusion increased between 2014 and 2017 across all countries in the sample, even where traditional financial inclusion stalled or declined.
  - Most countries saw increases in both access and usage dimensions.
  - Regional leaders on digital inclusion: Africa and Asia and the Pacific.
- Limitations:
  - Sample limited to 52 EMDEs and two years (2014 and 2017).
  - Databases do not differentiate DFS providers nor the range of services a user has access to.
  - Lack of granular and long time-series data on digital financial inclusion indicators.

### Definitions (Box 1)
- Fintech: technology-enabled innovation in financial services that could result in new business models, applications, processes or products with an associated material effect on the provision of financial services.
- Financial inclusion: “access to and use of formal financial services,” capturing transactions, savings, credit, and insurance for individuals and firms.
- Digital financial inclusion / fintech-enabled financial inclusion: digital access to and usage of formal financial services, such as through mobile phone (both smart and non-smart phones) and computers (to access the internet). Includes services provided by fintech companies and financial institutions.
- Digital payment: payment executed digitally; includes payments using mobile phones, computer and/or internet; does not include card payments.
- Mobile banking: use of an application on a mobile device to access and execute banking services.
- Mobile money: pay-as-you-go digital medium of exchange and store of value using mobile money accounts, facilitated by mobile money agents; electronically stored monetary value available to a user to conduct transactions through a mobile device, issued upon receipt of funds, accepted as a means of payment by persons other than the issuer, and redeemable for cash.

### Methodology and index construction
- Indices constructed:
  - Stage 1: Access and usage sub-indices.
  - Stage 2: Traditional financial inclusion index and digital financial inclusion index.
  - Stage 3: Comprehensive aggregate financial inclusion index (combines traditional and digital indices).
- Statistical method: three-stage principal component analysis (PCA) to determine weights on indicators and mitigate bias from correlated variables.
- Normalization and comparability:
  - Each index is constructed and normalized separately based on data for both 2014 and 2017.
  - Levels can be compared over time for each index but not directly compared across indices.
  - Index scores are normalized between 0 and 1 using a global min-max procedure.
- Rationale for payments focus:
  - Payments often serve as the entry point to broader financial services.
  - Mobile money payment services play a greater role in low-income and lower middle-income countries.
  - Alternative finance credit and insurance remain small (total outstanding alternative finance is less than 0.1 percent of GDP in 2017 for most countries in sample, except China).
  - Value of mobile money transactions examples: around 20 percent of GDP in Bangladesh and Senegal to over 140 percent of GDP in Zimbabwe in 2018.

### Indicators and conceptual mapping
- Access indicators:
  - Traditional: ATMs per 100,000 adults; commercial bank branches per 100,000 adults (IMF FAS).
  - Digital: mobile subscription per 100 people (ITU); % of population with access to internet (ITU); number of registered mobile money agents per 100,000 adults (IMF FAS; GSMA; staff estimates).
- Usage indicators:
  - Traditional: % adults with an account at a financial institution; % who save at a financial institution; % with debit cards; % who received wages through a financial institution account; % who use a financial institution account for utility payments (World Bank Global Findex).
  - Digital: % adults with a mobile account; % who use internet to pay; % who use mobile phone to receive wages; % who use mobile phone to make utility payments (World Bank Global Findex).
- Weighting overview (from Appendix Table II.2):
  - Access to bank infrastructure: 0.25
  - Access to digital infrastructure: 0.125
  - Usage (traditional): 0.25
  - Usage (digital): 0.125
  - Note: number of registered mobile money agents per 100,000 adults listed under access with staff estimate.

### Data construction and coverage notes
- Missing FAS data on ATM per 100,000 adults and bank branches per 100,000 adults are interpolated using proxies (ATM per 10,000 km2 and branches per 10,000 km2); if proxies unavailable, filled with general past trend.
- FAS includes annual mobile money transactions and volume but not for enough countries; these variables were excluded to retain sample size.
- A novel dataset on number of mobile money agents was constructed from multiple sources; regional GSMA and IMF FAS data supplemented with provider websites, IFC reports, news articles, and staff estimates.
- Mobile-related variables receive the same weights as traditional bank-related variables in final index because weights derive from PCA; potential bias where DFSs have smaller presence.

### Findings — stylized facts and usage patterns
- ATM per population increased notably over the last decade especially in middle income countries; LICs have ATMs per population at less than 1/10th of the global average.
- Bank branches per population remained stable overall and declined on average in high-income countries.
- Share of adults with financial institution accounts roughly doubled in lower middle-income and low-income countries in 2011–17.
- Improvements in account holding for LICs are pronounced in broader account holdings, likely reflecting mobile money spread.
- More active measures (saved or borrowed from a financial institution) saw more muted improvement.
- Mobile subscription increased sharply in LICs in the mid-2000s; mobile phones are main vehicle to access internet in lower income countries.
- Use of mobile phone for sending/receiving domestic remittances roughly doubled in lower middle income countries and LICs between 2014 and 2017.
- About half the population received or sent domestic remittances using mobile phones in LICs in 2017.
- Share of population using internet to pay bills or make purchases increased across countries and is higher in higher per capita income countries.
- Number of mobile money operators increased to over 250 in 2018.
- Number of active mobile money accounts almost tripled between 2013 and 2017 in lower-middle and low income countries.
- Over 2.9 million active mobile money agents operate in 90 countries; cash-in/cash-out services account for 55 percent of total value of mobile money transactions in 2017.
- Number of mobile money agent outlets is roughly the same as the number of ATMs globally.
- Value of mobile money transactions in 2018 examples range from around 20 percent of GDP in Bangladesh and Senegal to over 140 percent of GDP in Zimbabwe.

### Findings — Traditional Financial Inclusion Index
- Regions with high traditional inclusion: Asia and the Pacific; Latin America and the Caribbean; Emerging Europe.
- Traditional financial inclusion is highly associated with GDP per capita.
- High-ranking countries in 2017: Mongolia, China, Thailand, Brazil, Turkey, Namibia.
- Majority of bottom-quartile countries are in Africa.
- Traditional index remained broadly unchanged between 2014 and 2017 for most countries (slow-moving access indicators).
- Changes in traditional access index concentrated around zero compared to usage index.
- Some mid- to high-level traditional inclusion countries in Asia and Latin America saw relatively large improvements.
- Low-level traditional inclusion countries (primarily in Africa) saw limited improvements.
- Eight countries experienced a decline in traditional financial inclusion (example: Nigeria and South Africa).

### Findings — Digital Financial Inclusion Index and progress (2014–2017)
- Most of the 52 EMDEs saw an increase in the digital financial inclusion index between 2014 and 2017.
- Improvement particularly large on average in African countries.
- Highest gainers include Ghana, Benin, and Senegal.
- Digital inclusion improved in all countries except Panama.
- Both access and usage dimensions increased for most countries; in a small number improvements were driven primarily by usage.
- Regions — sample composition and patterns:
  - Africa (19)
  - Asia and the Pacific (12)
  - Latin America and the Caribbean (13)
  - Middle East and Central Asia (6)
  - Emerging Europe (2)
- Africa featured prominently among the top quartile in 2017, led by Ghana, Kenya and Senegal.
- Within-region differences notable (e.g., Ghana versus Nigeria; Bangladesh versus Myanmar).

### Digital versus traditional patterns and substitution/complementarity
- Two clusters identified:
  - High digital / low-to-medium traditional: mostly African countries (e.g., Ghana, Kenya, Senegal, Uganda, Rwanda) — fintech filling gaps left by financial institutions.
  - High traditional / medium digital: includes Brazil, Romania, Panama and Guatemala — well-developed banking sector penetration.
- Exceptions where both digital and traditional are high: China and Malaysia.
- Cases of substitution (digital replacing traditional) more common in lower per capita income countries; complementarity more common in higher per capita income countries.
- Eight countries with increased digital index and decreased traditional index: Botswana; El Salvador; Mexico; Nigeria; Romania; Rwanda; South Africa; Zimbabwe.
  - In all but two cases (Zimbabwe and Romania) the decline in traditional inclusion driven more by fall in demand (usage) than by access (supply).
  - Pattern could reflect substitution by technology-related financial services and/or banks shifting toward technology-based delivery instead of physical presence.

### Comprehensive Financial Inclusion Index — methodology and implications
- Comprehensive index incorporates digital indicators alongside traditional measures for a more realistic estimate of cross-country differences.
- Data used: Global Findex usage for mobile money and online financial services; IMF FAS data on access to mobile money accounts; supplementary data on mobile money agents.
- Three-stage PCA:
  1. Compute access and usage sub-indices (first-stage PCA).
  2. Combine sub-indices into traditional and digital indices (second-stage PCA); mobile money agents added in second-stage for digital index.
  3. Form weighted combination of traditional and digital indices to produce comprehensive measure (third-stage PCA).
- Incorporating digital indicators significantly improves ranking of countries with high digital but low traditional inclusion and lowers rankings for countries with strong bank infrastructure but low fintech adoption.
- Example ranking changes:
  - Kenya, Botswana and Jordan ranked similarly on traditional inclusion; in the comprehensive index Kenya ranks in top group, Botswana remains similar, Jordan around bottom 1/3.
  - Uganda ranks among top quartile and Togo in bottom quartile in comprehensive measure, while both were in bottom quartile in traditional inclusion.

### Comprehensive Financial Inclusion Index — country rankings (selected)
- Comprehensive Financial Inclusion Index ranking (country — ranking):
  - Mongolia 1
  - China 2
  - Kenya 3
  - Malaysia 4
  - Ghana 5
  - Namibia 6
  - Turkey 7
  - Thailand 8
  - Chile 9
  - Brazil 10
  - South Africa 11
  - Uganda 12
  - Rwanda 13
  - Senegal 14
  - Dominican Republic 15
  - Indonesia 16
  - Romania 17
  - Armenia 18
  - Zimbabwe 19
  - Sri Lanka 20
  - Bangladesh 21
  - Argentina 22
  - Panama 23
  - Botswana 24
  - Gabon 25
  - Cote d'Ivoire 26
  - Colombia 27
  - Benin 28
  - Peru 29
  - Guatemala 30
  - Zambia 31
  - Bolivia 32
  - Mexico 33
  - Tunisia 34
  - India 35
  - Honduras 36
  - Jordan 37
  - El Salvador 38
  - Philippines 39
  - Togo 40
  - Vietnam 41
  - Pakistan 42
  - Nicaragua 43
  - Cambodia 44
  - Nigeria 45
  - Cameroon 46
  - Mauritania 47
  - Congo, Democratic Republic of 48
  - Congo, Republic of 49
  - Myanmar 50
  - Madagascar 51
  - Afghanistan 52

### Summary statistics and PCA details (selected exact figures)
- Appendix Table II.3 (observations = 104 for all listed variables):
  - ATM per 100,000 population: Mean 32.17; Standard Deviation 29.49; Range 109.31.
  - Bank branches per 100,000 population: Mean 11.47; Standard Deviation 7.84; Range 32.66.
  - Account at a F.I. (%): Mean 40.57; Standard Deviation 21.25; Range 75.56.
  - Saving at a F.I. (%): Mean 15.31; Standard Deviation 9.34; Range 35.52.
  - Mobile subscription per 100 ppl.: Mean 105.16; Standard Deviation 31.03; Range 130.48.
  - Internet (%): Mean 33.00; Standard Deviation 18.76; Range 66.26.
  - Mobile account (%): Mean 11.11; Standard Deviation 13.63; Range 50.42.
  - Registered mobile money agents: Mean 138.14; Standard Deviation 192.72; Range 743.52.
- First-stage PCA: first principal component explains more than 70 percent of explanatory variables’ total variation in categories (examples):
  - Access (Traditional) PC1: 0.7982; PC2: 1.0000.
  - Usage (Traditional) PC1: 0.7759; PC2: 0.8986; PC3: 0.9623; PC4: 0.9849; PC5: 1.0000.
  - Access (Digital) PC1: 0.7884; PC2: 1.000.
  - Usage (Digital) PC1: 0.7495; PC2: 0.9311; PC3: 0.9774; PC4: 1.0000.
- Selected first-component loadings (first-stage PCA, examples preserved):
  - ATM per 100,000 population (X1): 0.7071 (PC1).
  - Bank per 100,000 population (X2): 0.7071 (PC1).
  - Account at an F.I. (%) (Y1): 0.4842 (PC1).
  - Mobile account (%) (P1): 0.5130 (PC1).
  - Use internet to pay (%) (P2): 0.3722 (PC1).
- Second-stage PCA: Traditional PC1: 0.8448; Digital PC1: 0.5435.
- Summary statistics of indices (52 countries, 104 observations):
  - Traditional Financial Inclusion Index — Access: Mean 0.296; Standard Deviation 0.233; Min 0; Max 1.
  - Traditional Financial Inclusion Index — Usage: Mean 0.34; Standard Deviation 0.247; Min 0; Max 1.
  - Digital Financial Inclusion Index — Access: Mean 0.527; Standard Deviation 0.267; Min 0; Max 1.
  - Digital Financial Inclusion Index — Usage: Mean 0.196; Standard Deviation 0.218; Min 0; Max 1.
  - Comprehensive Financial Inclusion Index — Overall: Mean 0.433; Standard Deviation 0.216; Min 0; Max 1.
- 2017 top/bottom examples (Appendix Table II.8):
  - Top 5: 1 Mongolia 1.00 (versus 2014: 1); 2 China 0.90 (versus 2014: 6); 3 Kenya 0.84 (versus 2014: 1); 4 Malaysia 0.83 (versus 2014: -1); 5 Ghana 0.81 (versus 2014: 25).
  - Bottom 5: 48 Congo, Dem. Rep. of 0.18 (versus 2014: -3); 49 Congo, Republic of 0.15 (versus 2014: -3); 50 Myanmar 0.15 (versus 2014: 0); 51 Madagascar 0.09 (versus 2014: 0); 52 Afghanistan 0.04 (versus 2014: 0).

### Correlations (selected exact figures from Appendix Table II.9)
- ATM per 100,000 vs Bank per 100,000: 0.532.
- Account at an F.I. vs Saving at an F.I.: 0.793.
- Account at an F.I. vs Debit card: 0.891.
- Mobile subscription vs Internet: 0.5728.
- Internet vs ATM per 100,000 population: 0.638.
- Mobile money agents vs Mobile subscription: -0.3003.
- Mobile money agents vs Internet: -0.303.
- Mobile money agents vs Mobile account: 0.553.
- Use internet to pay vs Account at an F.I.: 0.599.
- Mobile for utility vs Mobile account: 0.858.
- Mobile for wages vs Mobile account: 0.792.

### Key conclusions and policy-relevant implications
- Digital financial inclusion improved for most countries in 2014–2017, particularly in Africa and Asia and the Pacific.
- Comprehensive financial inclusion (traditional + digital) improved for most countries; in some cases improvements were entirely driven by digital means.
- Incorporating digital indicators provides a more accurate and differentiated picture of financial inclusion across countries.
- Data coverage limitations (sample size, years, lack of provider/service granularity) remain important constraints; the index is an initial step to be refined as new data become available.

*Source: wpiea2021090-print-pdf.*

### References ___________________________________________________________ 18

### References

### Introduction and key contribution
- Objective: construct a digital financial inclusion index covering 52 emerging markets and developing economics (EMDEs) for 2014 and 2017, and combine it with a traditional financial inclusion index to form a comprehensive aggregate index.
- Focus: payments aspect of financial inclusion enabled by digital financial services (DFSs); indices exclusively focus on the payments aspects of financial inclusion.
- Data sources used: World Bank Global Findex Database, IMF Financial Access Survey (FAS), International Telecommunication Union (ITU), and GSMA Mobile Money Dataset.
- Principal finding summary:
  - Digital financial inclusion increased between 2014 and 2017 across all countries in the sample, even where traditional financial inclusion stalled or declined.
  - Most countries saw increases in both access and usage dimensions.
  - Regional differences: Africa and Asia and the Pacific are in the lead on digital inclusion.
- Limitations noted:
  - Sample limited to 52 EMDEs and excludes advanced economies due to data availability.
  - Databases do not differentiate DFS providers (fintech firms vs. banks) nor the range of services a user has access to (only banks, only DFS, or both).
  - Lack of granular and long time-series data on digital financial inclusion indicators.

### Definitions (Box 1)
- Fintech: technology-enabled innovation in financial services that could result in new business models, applications, processes or products with an associated material effect on the provision of financial services.
- Financial inclusion: “access to and use of formal financial services,” capturing transactions, savings, credit, and insurance for individuals and firms.
- Digital financial inclusion or fintech-enabled financial inclusion: digital access to and usage of formal financial services, such as through mobile phone (both smart and non-smart phones) and computers (to access the internet). Includes services provided by fintech companies and financial institutions.
- Digital payment: payment executed digitally; includes payments using mobile phones, computer and/or internet; does not include card payments.
- Mobile banking: use of an application on a mobile device to access and execute banking services.
- Mobile money: pay-as-you-go digital medium of exchange and store of value using mobile money accounts, facilitated by mobile money agents; electronically stored monetary value available to a user to conduct transactions through a mobile device, issued upon receipt of funds, accepted as a means of payment by persons other than the issuer, and redeemable for cash. In most countries, traditional bank accounts accessible via electronic means are excluded from the definition of “mobile money.” Providers include mobile network operators and fintech companies.

### Methodology and index construction
- Indices constructed:
  - Access and usage sub-indices (stage 1).
  - Traditional financial inclusion index and digital financial inclusion index (stage 2).
  - Comprehensive aggregate financial inclusion index (stage 3), combining traditional and digital indices.
- Statistical method: three-stage principal component analysis (PCA) to determine weights on indicators and to mitigate bias toward highly correlated variables by estimating sub-indices in separate stages.
- Index comparability:
  - The three indices (traditional, digital and comprehensive) are constructed and normalized separately based on data for both 2014 and 2017.
  - Levels can be compared over time for each index but are not directly comparable across indices.
- Scope and rationale:
  - Focus on payments because payments often serve as the entry point to broader financial services; mobile money payment services play a greater role in low-income and lower middle-income countries.
  - Credit and insurance provided by alternative finance or mobile money remain small: total outstanding alternative finance is less than 0.1 percent of GDP in 2017 for most countries in the sample, except China (Cambridge Centre for Alternative Finance).
  - By contrast, the value of mobile money transactions ranged from around 20 percent of GDP in Bangladesh and Senegal to over 140 percent of GDP in Zimbabwe in 2018.

### Indicators and conceptual mapping
- Access measures:
  - Traditional: number of bank branches and ATMs per 100,000 adults (access to bank infrastructure).
  - Digital: mobile subscription per 100 people; % of population who has access to internet; number of registered mobile money agents per 100,000 adults.
- Usage measures:
  - Traditional: account holdings at financial institutions; active use of financial institution accounts for payments and receipt of wages.
  - Digital: mobile money account holdings; active use of mobile money for payments and receipt of wages.
- Weighting overview:
  - Example weights shown in Table 1 extract: access split includes 0.25 for bank infrastructure access (traditional) and 0.125 for access to digital infrastructure (digital); number of registered mobile money agents per 100,000 adults receives a staff estimate weight of 0.25 in the digital index excerpt.

### Advantages over prior measures
- Incorporates digital channels and multiple indicators rather than single indicators (e.g., mobile money account ownership).
- Combines data from multiple sources to capture DFS contributions from a multidimensional perspective.
- Distinguishes between digital and traditional financial inclusion for granular analysis of drivers of change.

### Uses and caveats for policymakers and researchers
- Usefulness: provides an analytical tool to assess progress in financial inclusion through both traditional and digital channels and inform appropriate measures.
- Cautions:
  - Results driven by available indicators and data coverage—limited country coverage (52 EMDEs) and two survey years (2014 and 2017).
  - Does not identify whether fintech broadens inclusion or provides alternative access for those already included because provider and service-range granularity are not available.

*Italic source attribution: wpiea2021090-print-pdf - References ___________________________________________________________ 18*

### 0.25 Usage

### 0.25 Usage

### Indicators of financial inclusion: stylized facts
- Access to financial institutions measured by ATMs per population saw a notable jump over the last decade especially in middle income countries. The level remains low for LICs, with the number of ATMs per population at less than 1/10th of the global average.
- Bank branches per population remained stable overall and saw a decline on average in high-income countries.
- The share of adults with financial institution accounts—measure of usage—roughly doubled in lower middle-income and low-income countries in 2011-17.
- Improvements in account holding for LICs are more pronounced in broader account holdings, likely reflecting the spread of mobile money products.
- More active measures of financial services use (share of population that saved or borrowed from a financial institution) saw more muted improvement.
- Mobile subscription increased sharply in LICs in the mid-2000s; mobile phones have become the main vehicle to access internet especially in countries with lower per capita income.
- The use of mobile phone for sending or receiving domestic remittances roughly doubled in lower middle income countries and LICs between 2014 and 2017.
- About half of the population received or sent domestic remittances using mobile phones in LICs in 2017.
- The share of population using the internet to pay bills or make purchases has increased across all countries and tends to be higher in countries with higher per capita income.
- The number of mobile money operators increased significantly over the last decade to over 250 in 2018.
- The number of active mobile money accounts almost tripled between 2013 and 2017 in lower-middle and low income countries.
- There are over 2.9 million active mobile money agents operating in 90 countries, and their cash-in/cash-out services account for 55 percent of the total value of mobile money transactions in 2017.
- The number of mobile money agent outlets is roughly the same as the number of ATMs globally.
- The value of mobile money transactions reached sizable amounts in 2018, ranging from around 20 percent of GDP in Bangladesh and Senegal, to over 140 percent of GDP in Zimbabwe.

### Data construction and coverage notes
- Missing data from IMF’s FAS on ATM per 100,000 adults and bank branches per 100,000 adults are interpolated using proxy variables (ATM per 10,000 km2 and bank branches per 10,000 km2). When proxy data are also not available, missing data are filled with the general past trend in the variable.
- The FAS includes annual data on Mobile Money transactions and volume, but the data are available for only a limited number of countries; these variables were excluded to retain as many countries as possible in the sample.
- A novel dataset on the number of mobile money agents across countries was constructed from various data sources. Regional aggregates and country-specific data from GSMA and IMF FAS were incomplete and supplemented by estimates based on mobile money service providers, GSMA, IFC Mobile Money Scoping country reports, CGAP, and other articles and reports.
- Mobile-related variables have the same weights as traditional bank-related variables in the final index, because weights come from the first principal component. This could lead to bias in final results in countries where DFSs have smaller presence compared to traditional financial services.

### Findings — Traditional Financial Inclusion Index
- Countries in Asia and the Pacific, Latin America and the Caribbean, and Emerging Europe in the sample have high degrees of traditional financial inclusion.
- Traditional financial inclusion is highly associated with levels of GDP per capita.
- Countries that rank high in traditional inclusion index in 2017 include Mongolia, China, Thailand, Brazil, Turkey, and Namibia.
- The majority of countries in the bottom quartile are in Africa.
- Traditional financial inclusion index remained broadly unchanged between 2014 and 2017 for most countries in the sample, reflecting the slow-moving nature of underlying access indicators.
- Changes in the underlying traditional access index are more concentrated around zero compared to the traditional usage index.
- Some countries in Asia and the Pacific and Latin America and the Caribbean with mid- to high-levels of traditional financial inclusion saw relatively large improvements in the index.
- Countries with low levels of traditional financial inclusion (primarily in Africa) saw limited improvements.
- Eight countries experienced a decline in levels of traditional financial inclusion (example countries include Nigeria and South Africa).

### Findings — Digital Financial Inclusion Index
- Countries in Africa and Asia and the Pacific regions in the sample are found to have high degrees of digital financial inclusion compared to other regions.

*Source: wpiea2021090-print-pdf*

### Appendix Table II.8). African countries,

### Appendix Table II.8). African countries,

### Progress in digital financial inclusion (2014–2017)
- Most of the 52 EMDEs in the sample saw an increase in the digital financial inclusion index between 2014 and 2017.
- Improvement was particularly large in African countries on average.
- Ghana, Benin, and Senegal were among the highest gainers.
- Digital inclusion improved in all countries except Panama.
- Both the access and usage dimensions increased for most countries; for a small number of countries the improvement was driven primarily by the increase in usage.

### Regional patterns and counts
- Sample composition by region:
  - Africa (19)
  - Asia and the Pacific (12)
  - Latin America and the Caribbean (13)
  - Middle East and Central Asia (6)
  - Emerging Europe (2)
- Africa featured prominently among countries in the top quartile of the index in 2017, led by Ghana, Kenya and Senegal.
- Countries in Latin America and the Caribbean rank around the middle, with Dominican Republic, Chile and Argentina among the highest for the region.
- Significant within-region differences are observed, especially in Africa (e.g., Ghana versus Nigeria) and Asia (e.g., Bangladesh versus Myanmar).

### Digital versus traditional financial inclusion
- Two clusters stand out when comparing digital and traditional indices:
  - High digital / low-to-medium traditional: mostly African countries (e.g., Ghana, Kenya, Senegal, Uganda, Rwanda), suggesting fintech can fill gaps in services provided by financial institutions.
  - High traditional / medium digital: includes Brazil, Romania, Panama and Guatemala, reflecting relatively well-developed banking sector penetration.
- Exceptions where both measures are high include China and Malaysia.
- There are cases where digital services substitute for traditional services (more common in lower per capita income countries) and cases where they complement traditional services (more common in higher per capita income countries).

### Cases where digital gains accompanied by declines in traditional inclusion
- Eight countries experienced an increase in digital inclusion index accompanied by a fall in traditional inclusion index:
  - Botswana
  - El Salvador
  - Mexico
  - Nigeria
  - Romania
  - Rwanda
  - South Africa
  - Zimbabwe
- Sub-component analysis indicates the decline in traditional inclusion was driven more by a fall in demand (usage) rather than access (supply) in all but two cases (Zimbabwe and Romania).
- This pattern could reflect substitution by technology-related financial services away from traditional financial institutions, and/or banks shifting toward technology-based delivery versus physical presences.

### Comprehensive Financial Inclusion Index: methodology and implications
- The comprehensive index incorporates digital financial inclusion indicators alongside traditional measures to give a more realistic estimate of differences in financial inclusion across countries.
- Index construction:
  - Uses World Bank Global Findex usage data for mobile money and online financial services, IMF FAS data on access to mobile money accounts, and supplementary data on access to mobile money agents.
  - Three-stage principal component analysis (PCA):
    1. Compute access and usage sub-indices.
    2. Combine these sub-indices into traditional and digital financial inclusion indices.
    3. Form a weighted combination of traditional and digital indices to produce the comprehensive measure.
- Incorporating digital indicators significantly improves the ranking of countries with high digital but low traditional inclusion (red cluster in Figure 18) and leads to declines for countries with strong bank infrastructure but low fintech adoption (purple cluster in Figure 18).
- Examples of ranking changes:
  - Kenya, Botswana and Jordan rank similarly on traditional financial inclusion; in the comprehensive index, Kenya ranks in the top group, Botswana remains similar, and Jordan is around the bottom 1/3.
  - Uganda ranks among the top quartile and Togo in the bottom quartile in the comprehensive measure, while both were in the bottom quartile in traditional financial inclusion.

### Comprehensive Financial Inclusion Index — country rankings (Table 2)
- Comprehensive Financial Inclusion Index ranking (country — ranking):
  - Mongolia 1
  - China 2
  - Kenya 3
  - Malaysia 4
  - Ghana 5
  - Namibia 6
  - Turkey 7
  - Thailand 8
  - Chile 9
  - Brazil 10
  - South Africa 11
  - Uganda 12
  - Rwanda 13
  - Senegal 14
  - Dominican Republic 15
  - Indonesia 16
  - Romania 17
  - Armenia 18
  - Zimbabwe 19
  - Sri Lanka 20
  - Bangladesh 21
  - Argentina 22
  - Panama 23
  - Botswana 24
  - Gabon 25
  - Cote d'Ivoire 26
  - Colombia 27
  - Benin 28
  - Peru 29
  - Guatemala 30
  - Zambia 31
  - Bolivia 32
  - Mexico 33
  - Tunisia 34
  - India 35
  - Honduras 36
  - Jordan 37
  - El Salvador 38
  - Philippines 39
  - Togo 40
  - Vietnam 41
  - Pakistan 42
  - Nicaragua 43
  - Cambodia 44
  - Nigeria 45
  - Cameroon 46
  - Mauritania 47
  - Congo, Democratic Republic of 48
  - Congo, Republic of 49
  - Myanmar 50
  - Madagascar 51
  - Afghanistan 52

### Key conclusions
- Digital financial inclusion improved for most countries in the 2014–2017 period, particularly in Africa and Asia and the Pacific on average.
- Comprehensive financial inclusion, combining traditional and digital measures, improved for most countries, although in some cases the improvement was entirely driven by digital means.
- Incorporating digital indicators yields a more accurate and differentiated picture of financial inclusion across countries.
- Data coverage limitations remain a challenge; the index is an important step toward more comprehensive measurement and will be refined as new data becomes available.

*Source: IMF staff calculations.*

### APPENDIX I: COMPOSITION  OF  VARIOUS FINANCIAL INCLUSION INDICES

### APPENDIX I: COMPOSITION  OF  VARIOUS FINANCIAL INCLUSION INDICES

### Overview of existing indices and coverage
- Multiple published financial inclusion indices summarized, varying by:
  - Coverage (examples: 94 countries 2004-2010; 82 countries 2011; 176 countries 1980-2013; 31 countries 2009-2012; others).
  - Methodology (UNDP weighted geometric average, two-stage PCA, PCA, equal weights, weighted geometric average with factor-analysis-derived weights).
  - Primary data sources: FAS (Financial Access Survey), World Bank Findex, Enterprise Survey, IMF WEO, ITU, GSMA, country mobile money provider data, IFC reports, news articles.
- Example indicator sets used across studies (preserve terminology exactly):
  - Access/Other dimensions (FAS): (1) bank branches per 100,000 population; (2) ATMs per 1000,000 population; (1) bank accounts per 1000 population; Credit and Deposits as a % of GDP.
  - Market/depth indicators: % of market capitalization outside of top 10 largest companies; % of value traded outside of top 10 traded companies; government bond yields; ratio of private to total debt securities; ratio of new corporate bond issues to GDP; total number of debt issuers.
  - Usage/Household and SME indicators (Findex, Enterprise Survey): accounts at a formal financial institution (% age 15+); saved at a financial institution in the past year (% age 15+); loan from a financial institution in the past year (% age 15+); has a credit card (% age 15+); has a debit card (% age 15+); ATM is main mode of withdrawal (% with an account, age 15+); % of firms with a bank loan/line of credit; % of firms with a checking or savings account; % of firms using banks to finance investment; % of working capital financed by banks; collateral needed for a loan in % of loan amount; % of firms not needing a loan; % of firms identifying cost of finance as a major constraint.
- Note: Source listings across studies include country counts such as "163 countries", "88 countries", "870 countries" (as presented in the source table heads).

### New comprehensive financial inclusion index: scope and approach
- Purpose:
  - New comprehensive index focused on payments services.
  - Collates fourteen indicators classified under two types: traditional and digital.
  - Distinguishes two dimensions within each type: access and usage.
- Sample and timing:
  - Covers 52 emerging market and developing economies (EMDEs).
  - Two years: 2014 and 2017 — data on all fourteen indicators publicly available for these years.
- Countries by region: Asia and the Pacific; Africa; Latin America and the Caribbean; Middle East and Central Asia; Emerging Europe (full country list provided in source).

### Data sources and variable selection
- Primary data sources used to construct the indices:
  - World Bank Global Findex ("Global Findex"), IMF’s World Economic Outlook (WEO), IMF’s Financial Access Survey (FAS), International Telecommunication Union (ITU), GSMA’s mobile money database.
- General selection criteria:
  - Variables chosen to represent access and usage aspects of payments-focused financial inclusion while retaining wide country coverage.

### Access indicators (definitions and data handling)
- Traditional access (traditional access index):
  - Indicators: ATMs per 100,000 adults; commercial bank branches per 100,000 adults.
  - Data source: IMF FAS.
- Digital access (digital access index):
  - Indicators: access to digital infrastructure (share of individuals with access to a mobile phone and access to the internet — ITU data); number of registered mobile money agents per 100,000 adults (IMF FAS; GSMA; staff estimates).
  - Mobile money agents density:
    - Primary data from IMF FAS for 38 sample countries (2009–2018).
    - Missing datapoints filled using publicly available provider websites, IFC mobile money reports, news articles; regional aggregate GSMA data used for consistency checks.
    - Point-in-time country data used to estimate time series for 2013-17 based on real GDP growth and adjusted for year of launch of services.
- Missing-data interpolation:
  - For missing FAS data on ATM per 100,000 adults and branches per 100,000 adults, proxies used (ATM per 10,000 km2 and bank branches per 10,000 km2) to interpolate. If proxy missing, filled with general past trend in the variable.

### Usage indicators (definitions and limitations)
- Traditional usage (traditional usage index) — Global Findex-based:
  - % of adults with an account at a financial institution; % who use this account for wage transfers and utility payments; % who save at a financial institution; % who have a debit card.
- Digital usage (digital usage index) — Global Findex-based:
  - % of adults with a mobile account; % who use internet to pay; % who use mobile phone to receive salaries or wages; % who use mobile phone to make utility payments.
- Sample limitation:
  - Lack of comprehensive digital usage data limits sample to 52 countries and two years (2014 and 2017).
- Note: FAS also includes annual data on mobile money transactions and volumes, but not comprehensively covered and therefore not included in the index.

### Index construction and weighting methodology
- Overall method:
  - A three-stage principal component analysis (PCA) constructs the comprehensive financial inclusion index.
  - Rationale: treat financial inclusion as an unobserved latent variable captured by correlated indicators; PCA quantifies variable importance by explaining variation.
- Three-stage PCA procedure:
  1. First stage: estimate sub-indices for access and usage separately for traditional and digital components:
     - Traditional access (FIA_Ta) from X1 = ATMs per 100,000 population and X2 = bank branches per 100,000 population.
     - Traditional usage (FIA_Tu) from Y1 = % adults with a financial institution account; Y2 = % adults who save at a financial institution; Y3 = % adults with debit cards; Y4 = % adults who received wages through a financial institution account; Y5 = % adults who use a financial institution account for utility payments.
     - Digital access (FIA_Da) from K1 = mobile subscription per 100 people; K2 = % population with access to the internet.
     - Digital usage (FIA_Du) from P1 = % adults with a mobile account; P2 = % adults who use internet to pay; P3 = % adults who use a mobile phone to receive wages; P4 = % adults who use a mobile phone to make utility payments.
     - First-stage equations shown in source (linear combinations with component loadings and error terms).
  2. Second stage: estimate traditional and digital financial inclusion indices using the access and usage sub-indices computed in stage one as explanatory variables.
  3. Third stage: compute comprehensive financial inclusion by using the two types (traditional and digital indices from stage two) as explanatory variables.
- PCA details:
  - Principal components are linear combinations ordered by explained variation; the first principal component (PC1) explains more than 70 percent of the explanatory variables’ total variation (Appendix Table II.4 referenced).

### Weights in the overall index (from Appendix Table II.2)
- Overall Financial Inclusion Index — variable weights (as presented):
  - Access to bank infrastructure: 0.25
    - Number of ATMs per 100,000 adults (IMF FAS)
    - Number of branches per 100,000 adults (IMF FAS)
  - Access to digital infrastructure: 0.125
    - Mobile subscription per 100 people (ITU)
    - % of population who have access to internet (ITU)
    - Number of registered mobile money agents per 100,000 adults (IMF FAS; GSMA; Staff est.) — listed under access with staff estimate
  - Usage (traditional): 0.25
    - % of adults with a financial institution account (WB Findex)
    - % of adults who save at a financial institution (WB Findex)
    - % of adults with debit cards (WB Findex)
    - % of adults who received wages through a financial institution account (WB Findex)
    - % of adults who use a financial institution account for utility payments (WB Findex)
  - Usage (digital): 0.125
    - % of adults who have a mobile account (WB Findex)
    - % of adults who use internet to pay (WB Findex)
    - % of adults who use mobile phone to receive salary or wages (WB Findex)
    - % of adults who use mobile phone to make utility payments (WB Findex)
- Note: “Weight” is defined in the source as the weight of the variable in the overall index of financial inclusion.

### Summary statistics (Appendix Table II.3 exact figures)
- Observations, means, standard deviations, and ranges for selected variables (note: "Obvs." = 104 for all listed variables):
  - Access (Traditional)
    - ATM per 100,000 population: Obvs. 104; Mean 32.17; Standard Deviation 29.49; Range 109.31
    - Bank branches per 100,000 population: Obvs. 104; Mean 11.47; Standard Deviation 7.84; Range 32.66
  - Usage (Traditional)
    - Account at a F.I. (%): Obvs. 104; Mean 40.57; Standard Deviation 21.25; Range 75.56
    - Saving at a F.I. (%): Obvs. 104; Mean 15.31; Standard Deviation 9.34; Range 35.52
    - Debit card (%): Obvs. 104; Mean 24.87; Standard Deviation 18.25; Range 65.05
    - F.I account for wages (%): Obvs. 104; Mean 7.03; Standard Deviation 6.15; Range 25.35
    - F.I account for utility (%): Obvs. 104; Mean 6.27; Standard Deviation 6.42; Range 25.02
  - Access (Digital)
    - Mobile subscription per 100 ppl.: Obvs. 104; Mean 105.16; Standard Deviation 31.03; Range 130.48
    - Internet (%): Obvs. 104; Mean 33.00; Standard Deviation 18.76; Range 66.26
  - Usage (Digital)
    - Mobile account (%): Obvs. 104; Mean 11.11; Standard Deviation 13.63; Range 50.42
    - Use internet to pay (%): Obvs. 104; Mean 8.11; Standard Deviation 7.77; Range 35.82
    - Mobile for wages (%): Obvs. 104; Mean 1.81; Standard Deviation 2.73; Range 11.59
    - Mobile for utility (%): Obvs. 104; Mean 3.22; Standard Deviation 4.25; Range 18.50
  - Mobile Money Agents
    - Registered mobile money agents: Obvs. 104; Mean 138.14; Standard Deviation 192.72; Range 743.52
- Data trimming:
  - The dataset is trimmed by the 2nd and 98th percentile to reduce influence of extreme values.
- Correlation matrix:
  - Referenced as Appendix Table II.9 in source (correlation matrix for selected variables).

*Source: IMF staff.*

### Appendix Table II.4. First-stage PCA: Cumulative variance explained by principal components

### Appendix Table II.4. First-stage PCA: Cumulative variance explained by principal components

### First-stage PCA: Cumulative variance explained (by category)
- Access (Traditional)
  - 푃푃푃푃1: 0.7982
  - 푃푃푃푃2: 1.0000
- Access (Digital)
  - 푃푃푃푃1: 0.7884
  - 푃푃푃푃2: 1.000
- Usage (Traditional)
  - 푃푃푃푃1: 0.7759
  - 푃푃푃푃2: 0.8986
  - 푃푃푃푃3: 0.9623
  - 푃푃푃푃4: 0.9849
  - 푃푃푃푃5: 1.0000
- Usage (Digital)
  - 푃푃푃푃1: 0.7495
  - 푃푃푃푃2: 0.9311
  - 푃푃푃푃3: 0.9774
  - 푃푃푃푃4: 1.0000

### First-stage PCA: Loadings (selected variables and first-component loadings)
- Access (Traditional) — No tatio n / 푃푃푃푃1 / 푃푃푃푃2
  - ATM per 100,000 population (푋1): 0.7071 / 0.7071
  - Bank per 100,000 population (푋2): 0.7071 / -0.7071
- Usage (Traditional) — No tatio n / 푃푃푃푃1 / 푃푃푃푃2 / 푃푃푃푃3 / 푃푃푃푃4 / 푃푃푃푃5
  - Account at an F.I (%) (푌1): 0.4842 / 0.1934 / 0.0217 / -0.5796 / -0.6259
  - Saving at an F.I (%) (푌2): 0.3954 / 0.7556 / 0.2523 / 0.4328 / 0.1473
  - Debit Card (%) (푌3): 0.4820 / -0.0465 / -0.3747 / -0.3772 / 0.6948
  - F.I account for wages (%) (푌4): 0.4551 / -0.3382 / -0.5087 / 0.5736 / -0.3012
  - F.I account for utility (%) (푌5): 0.4120 / -0.5245 / 0.7326 / 0.0735 / 0.1140
- Access (Digital) — No tatio n / 푃푃푃푃1 / 푃푃푃푃2
  - Electricity (%) (퐾1): 0.7071 / 0.7071
  - Internet (%) (퐾2): 0.7071 / -0.7071
- Usage (Digital) — No tatio n / 푃푃푃푃1 / 푃푃푃푃2 / 푃푃푃푃3 / 푃푃푃푃4
  - Mobile account (%) (푃1): 0.5130 / -0.4231 / 0.5674 / 0.4857
  - Use internet to pay (%) (푃2): 0.3722 / 0.8911 / 0.1412 / 0.2181
  - Mobile for wages (%) (푃3): 0.5356 / -0.1541 / -0.7974 / 0.2315
  - Mobile for utility (%) (푃4): 0.5580 / -0.0575 / 0.1496 / -0.8142

### Method: Constructing sub-indices from first principal component
- Index score definition (preserving original notation):
  - 푃푃푃푃푠푠푠푠푠푠푠푠푠푠 = Σ (퐿퐿푖 푥푥푖) over 푖 = 1 to 푛
  - Explanatory variables (푥푥) are standardized (mean = 0, standard deviation = 1).
  - Absolute loadings (퐿퐿) are taken from the first principal component (Table 5: column 3).
- Normalization:
  - Index scores are normalized between 0 and 1 across all countries and both years (2014 and 2017) using a global min-max procedure:
    - 푥푥푛푛푠푠푠푠 푛푛푎푎푛푛푖푖푠푠푠푠 푛푛 = (푥푥 − 푥푥푛푛푖푖푛푛) / (푥푥푛푛푎푎푚푚 − 푥푥푛푛푖푖푛푛)

### Weights and variable importance
- The percentage contribution (weighting) of each explanatory variable to sub-indices is derived from the loadings in the first principal component.
- Weightings are presented in Appendix Figure II.2 (chart referenced in source).

---

### Second-stage PCA

### Purpose and formulation
- A second-stage PCA combines access and usage sub-indices separately into:
  - Traditional financial inclusion index (퐹퐹퐹퐹_T)
  - Digital financial inclusion index (퐹퐹퐹퐹_F)
- Equations (preserving original notation):
  - (퐹퐹퐹퐹_T)_{iiii} = 훽1 (퐹퐹퐹퐹_T_a)_{iiii} + 훽2 (퐹퐹퐹퐹_T_u)_{iiii} + e_{iiii}
  - (퐹퐹퐹퐹_F)_{iiii} = 훼1 (퐹퐹퐹퐹_F_a)_{iiii} + 훼2 (퐹퐹퐹퐹_F_u)_{iiii} + μ_{iiii}
- 훼 and 훽 are the weights assigned to each sub-component (Appendix Figure II.3, left chart).

### Special treatment of mobile money agents
- For the digital financial inclusion index, "mobile money agents (per 100,000 adults)" is added at the second-stage PCA rather than the first-stage.
- Rationale:
  - Mobile money agent density is negatively correlated with access to internet and mobile subscription (Appendix Table II.9).
  - Including it in the first stage would assign a negative weight to mobile money agents, implying higher agent density leads to lower access to DFSs, which is counter-intuitive.
- Note in source: "Mobile money agents (per 100,000 adults) are added in this stage."

### Second-stage PCA: Cumulative variance explained
- Traditional financial inclusion index
  - 푃푃푃푃1: 0.8448
  - 푃푃푃푃2: 1.0000
- Digital financial inclusion index
  - 푃푃푃푃1: 0.5435
  - 푃푃푃푃2: 1.0000

---

### Third-stage PCA and comprehensive index

### Third-stage formulation
- The comprehensive financial inclusion index (퐹퐹퐹퐹) is computed by applying PCA to the traditional and digital financial inclusion indices:
  - 퐹퐹퐹퐹_{iiii} = 휔1 (퐹퐹퐹퐹_T)_{iiii} + 휔2 (퐹퐹퐹퐹_F)_{iiii} + 휔_{iiii}
- 휔 are the weights assigned to the two subcomponents (Appendix Figure II.3, right chart).
- The overall financial inclusion index is normalized between 0 and 1 (same min-max procedure as sub-indices).

---

### Summary statistics of indices (Appendix Table II.7)
- Traditional Financial Inclusion Index (52 countries, 104 observations)
  - Access: Mean 0.296, Standard Deviation 0.233, Min 0, Max 1
  - Usage: Mean 0.34, Standard Deviation 0.247, Min 0, Max 1
  - Traditional (overall): Mean 0.324, Standard Deviation 0.226, Min 0, Max 1
- Digital Financial Inclusion Index (52 countries, 104 observations)
  - Access: Mean 0.527, Standard Deviation 0.267, Min 0, Max 1
  - Usage: Mean 0.196, Standard Deviation 0.218, Min 0, Max 1
  - Digital (overall): Mean 0.349, Standard Deviation 0.204, Min 0, Max 1
- Comprehensive Financial Inclusion Index (52 countries, 104 observations)
  - Overall: Mean 0.433, Standard Deviation 0.216, Min 0, Max 1

---

### 2017 Rankings (Appendix Table II.8) — Top and bottom entries (selected)
- Top 5 in 2017 Comprehensive Index (value and change versus 2014)
  - 1 Mongolia 1.00 (versus 2014: 1)
  - 2 China 0.90 (versus 2014: 6)
  - 3 Kenya 0.84 (versus 2014: 1)
  - 4 Malaysia 0.83 (versus 2014: -1)
  - 5 Ghana 0.81 (versus 2014: 25)
- Bottom 5 in 2017 Comprehensive Index (value and change versus 2014)
  - 48 Congo, Dem. Rep. of 0.18 (versus 2014: -3)
  - 49 Congo, Republic of 0.15 (versus 2014: -3)
  - 50 Myanmar 0.15 (versus 2014: 0)
  - 51 Madagascar 0.09 (versus 2014: 0)
  - 52 Afghanistan 0.04 (versus 2014: 0)
- Note provided in source: “‘versus. 2014’ refers to the respective country’s change in ranking compared to 2014. Green shade suggests improvement in ranking from 2014 to 2017, whereas red shade indicates deterioration in country’s ranking.”

---

### Correlation matrix: selected correlations (Appendix Table II.9)
- Traditional Financial Inclusion: Access
  - ATM per 100,000 population vs Bank per 100,000 population: 0.532
- Traditional Financial Inclusion: Usage (selected)
  - Account at an F.I. vs Saving at an F.I.: 0.793
  - Account at an F.I. vs Debit card: 0.891
  - Debit card vs F.I account for wages: 0.884
- Digital Financial Inclusion: Access
  - Mobile subscription correlations (selected)
    - Mobile subscription vs Internet: 0.5728 (Internet vs Mobile subscription listed as 0.5728 when placed in matrix)
    - Mobile subscription vs ATM per 100,000 population: 0.5589
  - Internet vs ATM per 100,000 population: 0.638
- Digital Financial Inclusion: Usage (selected)
  - Use internet to pay vs Account at an F.I.: 0.599
  - Mobile for utility vs Mobile account: 0.858
  - Mobile for wages vs Mobile account: 0.792
- Mobile money agents correlations (selected)
  - Mobile money agents vs Mobile subscription: -0.3003
  - Mobile money agents vs Internet: -0.303
  - Mobile money agents vs Mobile account: 0.553
  - Mobile money agents vs Mobile for utility: 0.367
  - Mobile money agents vs Mobile for wages: 0.368

*Italic: Source — wpiea2021090-print-pdf - Appendix Table II.4. First-stage PCA: Cumulative variance explained by principal components*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021090-print-pdf.pdf_