## _wp1173

## Source details

**Canonical URL:** [_wp1173](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp1173.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp1173.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp1173.pdf.json)

---

### Introduction and Purpose
- Assess role of ICT development, especially mobile phone penetration, in promoting financial inclusion and economic growth in Africa.
- Focus on 44 African countries between 1988 and 2007.
- Key objectives:
  - Analyze impact of mobile phone penetration on economic growth rates in Africa.
  - Add indicators of financial inclusion (number of deposits per head; number of loans per head across all financial intermediaries) to test whether mobile penetration influences growth via financial inclusion.
  - Investigate interactions among mobile penetration, financial inclusion, and growth in countries where mobile financial services take hold.
- Data and methodology highlights:
  - Control variables: initial levels of GDP, human capital development, government consumption, institutional development, and ICT variables.
  - Estimator: System Generalized Method of Moment (GMM) (Blundell and Bond (1998)) with Windmeijer (2005) small sample robust correction.
  - Variables averaged over four years; sample period divided into five subperiods: 1988–1991, 1992–1995, 1996–1999, 2000–2003, and 2004–2007.
  - Also test effects of prices of a three-minute fixed and mobile telephone local call on economic growth.

### Stylized Facts and Context
- Mobile cellular network coverage (2002–2007):
  - sub-Saharan Africa: around 55 percent of the population
  - MENA countries: more than 80 percent of the population
- Financial infrastructure and exclusion:
  - Average number of bank branches in sub-Saharan Africa in 2007: less than 700
  - Average number of ATMs in sub-Saharan Africa in 2007: less than a thousand
  - Financial exclusion (2009): Mozambique: 88 percent; Botswana: 41 percent
- GSMA projection: 1.7 billion unbanked customers with mobile phones by 2012
- Mobile financial services existence during sample: Zambia since 2001, South Africa since 2004, and Kenya since 2007
- ICT penetration threshold cited: 40 lines per 100 inhabitants (Roller and Waverman, 2001)

### Main Empirical Findings — ICT and Economic Growth
- ICT development, notably mobile phone penetration, contributes to economic growth in Africa.
- Mobile phones and fixed lines act as substitutes; effect of mobiles is higher in lower-income countries.
- Part of the growth effect of mobile penetration operates through improved financial inclusion.
- Specific marginal effects (from baseline specification, column 5, Table 2):
  - an additional 10 percentage point increase in the mobile penetration rate could lead to a 0.7 percentage point increase in real GDP growth
  - an additional 10 percentage point increase in the fixed penetration rate could lead to a 1.1 percentage point increase in real GDP growth
- Table 2 reported coefficients (selected, preserved):
  - Total telephone subscribers (fixed + mobile) per head: 0.116 [0.027]***
  - Fixed telephone lines per head: 0.700 [0.169]***
  - Mobile telephone subscribers per head: 0.056 [0.017]*** and 0.073 [0.009]***
  - Price of a 3-minute fixed telephone local call: -0.401 [0.033]*** and -0.356 [0.021]***
  - Price of 3-minute mobile local call: -0.085 [0.020]*** and -0.050 [0.011]***
  - Observations range: 157 to 190; Number of countries range: 39 to 44
  - Hansen test (prob.) examples: 0.54, 0.60, 0.61; AR2 (prob.) examples: 0.23, 0.17, 0.34

### Nonlinearities, Interaction Effects, and Heterogeneity
- Mobile and fixed telephones are substitutes in Africa:
  - Interaction term Fixed × Mobile telephone subscribers per head: -2.548 [0.088]***
- Marginal impact of mobile penetration decreases with GDP per capita:
  - Mobile telephone subscribers per head × GDP per capita: -0.012 [0.007]*
- Internet effects conditional on income:
  - Internet users per 100 inhabitants: -0.579 [0.193]***
  - Internet users per 100 inhabitants × GDP per capita: 0.350 [0.039]***
- Robustness across subsamples and averaging:
  - Subsampling (98 percent draws): coefficient on mobile close to full-sample; at 95 percent average coefficient 0.37 vs. 0.56 (full sample)
  - Recursive estimates: coefficient on mobile remains positive and significant but marginal impact declines as mobile penetration increases
  - Averaging over 3-year periods: coefficient on mobile remains positive and significant at 1 percent
  - Five-year averages: reduce sample by 20 percent without affecting result quality
  - Restricting to sub-Saharan Africa excluding Mauritius, Seychelles, South Africa: mobile coefficient positive, significant, larger magnitude
  - Developing countries sample: mobile development stimulates growth; African marginal impact not statistically different from other developing economies

### Financial Inclusion as a Transmission Channel
- Financial inclusion measures used:
  - number of deposits per head (deposits at commercial banks, cooperatives, microfinance institutions, specialized state financial institutions)
  - number of loans per head (loans by the aforementioned financial institutions)
- Main findings (Table 5):
  - Number of deposits per head coefficients: 0.658 [0.154]*** and 0.087 [0.005]***
  - Mobile × Number of deposits per head coefficients: 0.103 [0.034]*** and 0.130 [0.078]*
  - Mobile × Number of deposits per head × Mobfi: 1.392 [0.119]***
  - Number of loans per head coefficients: 3.540 [0.870]***, 0.919 [0.043]***, 1.100 [0.100]***
  - Mobile × Number of loans per head coefficients: 1.756 [0.311]*** and 0.570 [0.344]*
  - Mobile × Number of loans per head × Mobfi: 7.419 [0.889]***
  - Private Credit/GDP coefficient: 0.050 [0.022]**; Mobile × Private Credit/GDP: 0.096 [0.056]*; Mobile × Private Credit/GDP × Mobfi: 0.107 [0.021]***
  - Observations: 184–190; Number of countries: 42–44; Hansen test (prob.) examples: 0.60, 0.59, 0.56; AR2 (prob.) range: 0.20–0.43
- Interpretation:
  - Financial inclusion (deposits and loans per head) is a positive and significant determinant of growth.
  - Controlling for financial inclusion reduces the coefficient on mobile penetration, indicating partial mediation.
  - Interaction terms show mobile penetration enhances the growth impact of financial inclusion.
  - Presence of mobile financial services (Mobfi dummy = 1 for Zambia, South Africa, Kenya in sample) amplifies the joint contribution of mobile penetration and financial inclusion to growth.

### Annex 1 — Mobile Financial Services (MFS): Models and Case Evidence
- MFS models:
  - Bank-based model: customers have direct contractual relationship with bank; transactions handled by retailers outside bank branches (examples: Brazilian model, South African models MTN and WIZZIT, Indian model).
  - Nonbank-based model: no direct contractual relationship with a bank; electronic value recorded by MNO or stored-value issuer (examples: M-PESA in Kenya, Orange Money in Ivory Coast).
  - Nonbank providers cannot intermediate repayable deposits; customer funds typically held in trust accounts; interest on pooled client funds cannot benefit MNOs.
- M-PESA case (preserved statistics and features):
  - Launch: March 2007
  - As of January 2010:
    - more than 9 million registered users (40 percent of Kenyan adults and 23 percent of the population)
    - about $320 million per month of person-to-person transfers
    - 16,900 agents
    - 75 companies using M-PESA for collections; electricity company accounts for 20 percent of their customers paying through M-PESA
  - Operational notes: registration requires a formal ID card; cash/e-float management by agents; cash in trust accounts held with two commercial banks; partnership developments (e.g., Equity Bank withdrawal access)
- Regulatory considerations:
  - Prudential regulation and consumer protection differ by model; nonbank-based providers face limits on deposit intermediation
  - AML/CFT implications: ID requirements can impede inclusion for unbanked customers lacking formal ID
  - Interest on pooled client funds: regulatory and policy choices discussed (e.g., donating interest)

### Annex 2 — Assessing Mobile Phone Penetration on Financial Inclusion (2000–2007)
- Objective: test whether mobile telephone subscribers per head affect number of deposits per head and number of loans per head.
- Estimation approach:
  - Dependent variables: deposits per head; loans per head (data available for 2003 and 2007).
  - Explanatory variables: mobile telephone subscribers per head; GDP per head (log); population density (log); banks’ overhead cost (percent of assets); institutions; number of bank branches per km2.
  - Random-effects estimator chosen (Hausman test probability > 0.10).
  - Explanatory variables averaged over 2000–2003 and 2004–2007.
- Key regression coefficients from Table 6 (Mobile Phone Development and Financial Inclusion, 2000-2007) — coefficients with standard errors in brackets:
  - Mobile telephone subscribers per head:
    - Deposits (1): 0.382 [0.161]**
    - Deposits (2): 0.101 [0.036]***
    - Loans (3): 0.211 [0.084]**
    - Loans (4): 0.071 [0.011]***
  - GDP per head (log):
    - Deposits (1): 0.132 [0.047]***
    - Deposits (2): 0.093 [0.032]***
    - Loans (3): 0.015 [0.016]
    - Loans (4): -0.001 [0.010]
  - Population density (log):
    - Deposits (1): 0.098 [0.035]***
    - Deposits (2): 0.065 [0.030]**
    - Loans (3): 0.007 [0.010]
    - Loans (4): -0.023 [0.010]**
  - Banks’ overhead cost (percent of assets):
    - Deposits (2): -1.405 [0.682]**
    - Loans (4): 0.227 [0.202]
  - Institutions:
    - Deposits (2): 0.95 [0.142]***
    - Loans (4): 0.279 [0.073]***
  - Number of bank branches per km2:
    - Deposits (2): 0.005 [0.001]***
    - Loans (4): 0.002 [0.000]***
- Sample sizes and diagnostics:
  - Deposits (1): Observations 49; Number of countries 36; R2 (between) 0.47; Hausman test (prob) 0.69
  - Deposits (2): Observations 39; Number of countries 31; R2 (between) 0.68; Hausman test (prob) 0.34
  - Loans (3): Observations 46; Number of countries 34; R2 (between) 0.37; Hausman test (prob) 0.99
  - Loans (4): Observations 36; Number of countries 29; R2 (between) 0.55; Hausman test (prob) 0.57
- Interpretation:
  - Mobile phone development is positively associated with financial inclusion (both deposits and loans per head), robust to controls for GDP per head, population density, bank efficiency, institutions, and branch coverage.
  - Bank overhead costs negatively associated with deposits per head, indicating banking efficiency matters.
  - Institutions and bank-branch coverage positively associated with both deposits and loans per head.
  - Some puzzling signs (e.g., population density negative for loans when branch density included) possibly reflect differences in formal vs. informal lending and measurement issues.

### Policy Implications and Recommendations
- Promote domestic and foreign investment in ICT and development of the ICT sector.
- Reduce cost of communications to stimulate ICT diffusion and growth; increased competition in the telecommunication industry is one option.
- Carefully weigh tax policy on mobile communications because higher taxes can raise communication costs and potentially lower growth.
- Foster greater interaction between ICT and financial sectors to boost financial inclusion; address challenges:
  - security concerns;
  - compliance with AML/CFT rules and ID requirements.
- Support regulatory frameworks that permit safe scaling of mobile financial services while protecting customer funds (consider prudential and consumer-protection approaches appropriate to bank-based and nonbank-based models).

*Source: _wp1173 — 3. Variable Definition and Sources; Literature review; Stylized facts; Econometric specification; Main results; Annex 1 and Annex 2 (extracted from the source PDF content provided).*

### References .............................................................................................................

### _wp1173 - References ............................................................................................................................................ 22

### Tables
- 1. Transmission Channels from ICT to Growth ..................................................................................... 6
- 2. Impact of Mobile and Fixed Telephone Penetration on Economic Growth in Africa ...................... 34
- 3. Testing for Interaction Terms and other Forms of ICT .................................................................... 35
- 4. Robustness Tests .............................................................................................................................. 36
- 5. Mobile Penetration, Financial Inclusion, and Economic Growth in Africa ..................................... 37
- 6. Mobile Phone Development and Financial Inclusion, 2000-2007 ................................................... 41

### Figures
- 1. Correlation between Mobile Penetration, Growth and Financial Inclusion ..................................... 25
- 2. Financial Inclusion and Mobile Penetration by Type of Financial Institutions ................................ 26
- 3. Trends in ICT Use: International Comparison ................................................................................. 26
- 4. ICT Growth in Developing Countries .............................................................................................. 27
- 5. Trends in Fixed versus Mobile Phone Subscribers: International Comparison................................ 27
- 6. Trends in Mobile Penetration: International Comparison ................................................................ 28
- 7. Trends in Prepaid versus Postpaid Phone Subscribers: International Comparison .......................... 28
- 8. Access to Financial Services in Selected African Countries ............................................................ 29
- 9. Bank Loans and Deposits: International Comparison ...................................................................... 29
- 10. Bank Branches and ATMs .............................................................................................................. 30
- 11. Financial Infrastructure Gap in Developing Countries ................................................................... 31
- 12. Comparing Trends in Bank Credit and Mobile Penetration ........................................................... 31
- 13. Testing for the Stability of the Coefficient on Mobile Penetration Using Random Samples ......... 32
- 14. Recursive Estimates of the Coefficient on Mobile Penetration ...................................................... 32
- 15. Quality of the Prediction ................................................................................................................ 33

### Annexes
- 1. Mobile Financial Services ................................................................................................................ 38
- 2. Assessing the Effect of Mobile Phone Penetration on Financial Inclusion ...................................... 40

### Appendices
- 1. List of the Sample Countries ............................................................................................................ 42
- 2. Summary Statistics and Correlation Matrix ..................................................................................... 43

### 3. Variable Definition and Sources .....................................................................................

### 3. Variable Definition and Sources

### Introduction
- Rapid diffusion of information and communication technologies (ICT) in African countries observed since the 1990s, with mobile phone penetration overcoming fixed line coverage.
- Mobile cellular network coverage:
  - around 55 percent of the population in sub-Saharan Africa (between 2002 and 2007)
  - more than 80 percent of the population in Middle East and North African (MENA) countries (between 2002 and 2007)
- Financial infrastructure gap in Africa evidenced by very low numbers of bank branches and automated teller machines (ATMs) (Beck, Demirguc-Kunt, and Martinez Peria, 2007).
  - The average number of bank branches in sub-Saharan Africa was less than 700 in 2007.
  - The average number of ATMs was less than a thousand in 2007.
- Financial exclusion and informal finance (FinMark, 2009):
  - 88 percent of the population in Mozambique financially excluded or using informal finance (2009)
  - 41 percent in Botswana (2009)
  - Bank penetration lower than 10 percent in some regions of Africa
- GSMA projection:
  - 1.7 billion unbanked customers with mobile phones by 2012
- Branchless banking and mobile financial services are increasingly used to address the financial infrastructure gap.
- Purpose of paper:
  - Assess role of ICT development, especially mobile phone penetration, in promoting financial inclusion and economic growth.
  - Analyze impact of mobile phone penetration on economic growth rates in Africa.
  - Add indicator(s) of financial inclusion (number of deposits per head; number of loans per head across all financial intermediaries) to test whether mobile penetration influences growth via financial inclusion.
  - Investigate interactions among mobile penetration, financial inclusion, and growth in countries where mobile financial services take hold.
- Data and methodology overview:
  - Focus on 44 African countries between 1988 and 2007.
  - Control variables include initial levels of GDP (to account for convergence), human capital development, government consumption, institutional development, and mobile telephony variables such as penetration rates.
  - Use System Generalized Method of Moment (GMM) estimator to address reverse causality and endogeneity.
  - Also test effects of prices of a three-minute fixed and mobile telephone local call on economic growth rates.
- Key empirical findings summarized in the introduction:
  - ICT development, notably mobile phone penetration, contributes to economic growth in Africa.
  - Mobile phones and fixed lines are substitutes; effect of mobiles is higher in lower-income countries.
  - Part of growth effect of mobile penetration operates through improved financial inclusion.
  - Effect of financial inclusion on growth is enhanced where mobile financial services are available.

*Italic line omitted here per pipeline; final attribution provided below.*

### Literature Review

#### Theoretical Background — Macroeconomic impact of ICT development
- ICT characteristics that support their role in growth:
  - (i) ICT are omnipresent in most business sectors,
  - (ii) ICT improve continuously and therefore reduce costs for the users,
  - (iii) ICT contribute to innovation and to the development of new products and processes.
- Network externalities:
  - Value of a telephone line increases exponentially with the number of users.
  - Roller and Waverman (2001) estimate ICT affect economic growth only when penetration rate reaches 40 lines per 100 inhabitants.
- ICT as social overhead capital (SOC) (Waverman, Meschi, and Fuss, 2005): ICT generate social returns larger than private returns, akin to infrastructure such as education, health, and roads.
- Haacker (2010): growth impacts from falling ICT equipment prices are weaker in low- and middle-income countries because ICT equipment is often imported, but capital deepening benefits can still be large.
- Direct and indirect economic benefits (Datta and Agarwal, 2004):
  - Direct effects from supply side: domestic output and employment creation; capital accumulation; increased government revenues.
  - Indirect effects from ICT use: improved firms’ productivity; better and larger markets; deepened financial inclusion; contribution to rural development.
- Transmission channels from ICT to growth (Table 1):
  - Direct effects of ICT
    - From supply side: Contribute to domestic output and employment creation; Spur capital accumulation; Increase government revenues.
    - From ICT use: Improve firms' productivity; Affect balance of payments; Favor better and larger markets.
  - Indirect effects of ICT
    - Deepen financial inclusion; Contribute to rural development.
- ICT and FDI:
  - ICT supply attracts FDI, portfolio and venture capital; improves market efficiency via wider dispersion of market information.
- Productivity gains and reduced transaction costs:
  - ICT enable flexible firm structures and locations, better intrafirm communication, reduced unproductive travel time, improved logistics.
  - ICT reduce costs of retrieving information, increase arbitrage abilities, and facilitate price discovery—improving market functioning and trade opportunities.
- Financial inclusion channel:
  - Mobile telephony reduces transaction costs of providing financial services and supports branchless banking.
  - ICT enable better information flows and use of depositor data to analyze creditworthiness, improving access to credit and deposit facilities.
  - Improved financial inclusion can stimulate private investment and economic growth.
- Rural development:
  - Voice applications reduce isolation, improve bargaining power of farmers, eliminate middlemen, and enable development of nonagricultural activities (ecolodges, women-owned microbusinesses).
- Opportunity costs and threshold effects:
  - Potential negative impacts from opportunity costs if ICT investment displaces spending in education and health (Heeks, 1999).
  - Share of household income devoted to mobile services in developing countries is rising and may reduce budgets for food, health, education.
  - Threshold/network effects may prevent low-income countries from reaping ICT benefits if minimum ICT levels are required.

#### Empirical Studies
- Early cross-country findings:
  - Hardy (1980): telephone impact on growth significant; radio rollout impact not significant. Used previous-year values to partially address reverse causality but did not control for country-specific effects.
  - Norton (1992): improvement in telecommunications reduces transaction costs; includes initial-year stock of telephones to address reverse causality; finds positive and significant association with growth.
- Roller and Waverman (2001):
  - Use structural model endogenizing telecom investment jointly estimated with macro growth equation for 21 OECD and 14 developing countries over 20 years.
  - Find little impact once simultaneity and fixed effects are controlled for, and only a positive causal link after a critical mass of telecom infrastructure is reached—evidence of nonlinearity and network externalities.
- Extensions including mobile phones:
  - Sridhar and Sridhar (2004) apply Roller and Waverman’s framework and extend analysis to mobile phones, estimating systems that endogenize growth and telecom penetration, and disaggregate effects for fixed lines and mobile phones in 63 developing countries (1990 onwards in source).

### Key Statistics and Data Features (as reported)
- Regional mobile coverage (2002–2007):
  - sub-Saharan Africa: around 55 percent of the population
  - MENA countries: more than 80 percent of the population
- Financial exclusion (2009):
  - Mozambique: 88 percent of the population financially excluded or using informal finance
  - Botswana: 41 percent of the population financially excluded or using informal finance
- Banking infrastructure (2007):
  - average number of bank branches in sub-Saharan Africa: less than 700
  - average number of ATMs in sub-Saharan Africa: less than a thousand
- Projected unbanked mobile phone users:
  - 1.7 billion unbanked customers with mobile phones by 2012
- Study sample and period:
  - 44 African countries between 1988 and 2007
- ICT penetration threshold cited:
  - 40 lines per 100 inhabitants (Roller and Waverman, 2001)

### Main Analytical Conclusions Reported
- ICT development, and notably mobile phone penetration, contributes to economic growth in Africa.
- Mobile phones and fixed lines function as substitutes; mobile effects are larger in lower-income countries.
- Part of the growth effect of mobile penetration operates through improved financial inclusion (measured by number of deposits per head and number of loans per head across all financial intermediaries).
- The positive effect of financial inclusion on growth is strengthened in countries where mobile financial services are available.

*Source: _wp1173 - 3. Variable Definition and Sources (section content provided).*

### 2001. They find that the elasticity of aggregate national output with respect to main

### _wp1173 - 2001. They find that the elasticity of aggregate national output with respect to main

### Literature findings on telecommunications and growth
- Waverman, Meschi, and Fuss (2005) — 92 countries, 1980–2003:
  - Mobiles in developing countries play the same role as fixed lines played in the 1970s and 1980s in OECD countries.
  - In developing countries, mobile phones are substitutes for fixed lines; in developed countries they are complements for fixed lines.
  - Impacts on growth are positive and significant—twice as large as their impacts in developed countries.
  - Hypothesis: mobile phone rollout has greater effects on economic growth in developing countries because mobiles have more network effects and have more effects on mobility than in developed countries.
  - Price and income elasticities of mobile phone demand are superior to 1 in developing countries.
- Kathuria, Uppal and Mamta (2009) — 19 Indian states, 2000–2008:
  - Higher mobile penetration rates are associated with faster growth.
  - A critical mass at a penetration rate of 25 percent amplifies the impact of mobile phones on growth via network effects.
  - Telecom networks are highly subject to network effects: growth impact larger when a significant threshold network size is achieved.
  - Substantial variation found across urban and rural areas and between rich and poor households in cities.
- Lee, Levendis, and Gutierrez (2009) — sub-Saharan Africa focus:
  - Use of generalized method of moments (GMM) to correct potential endogeneity between growth and telephone expansion.
  - Marginal impact of mobile telecommunication services is greater where landline phones are rare.
  - Study did not test for the price effect of telecommunications on growth.
  - Channels such as financial inclusion were not investigated.
  - Authors note potential statistical shortcomings: System GMM estimator may be inappropriate for annual data if variables are not stationary.
- Cross-sectional studies summary:
  - Reverse causality is difficult to address.
  - Network externalities in telecommunication infrastructure lead to higher growth effects.
  - In developing countries, mobile phones and fixed lines appear substitutes rather than complements.
  - Few studies specifically focused on the African continent despite significant mobile rollouts.

### Stylized facts — Growth of ICT in Africa and other developing countries
- Rapid spread of ICT in developing countries:
  - Since the 1990s the average number of telephone subscribers, personal computer users, and Internet users per 100 inhabitants has increased, including in sub-Saharan Africa and South Asia.
- Robust growth of telephone subscribers:
  - Telephone technology (fixed and mobile) use is widespread relative to personal computers and Internet; Internet use increased sharply but telephone subscription growth rates are high.
  - In some regions like Africa, Internet users per 100 inhabitants decreased after a dramatic increase in the 1990s.
- Africa experienced the highest growth rates of telephone subscribers:
  - Growth rate of telephone subscribers in Africa was the highest of all developing regions at the end of the 1990s and beginning of the 2000s (except for South Asia in the 2000s), albeit from low levels.
  - Possible partial explanation: two waves of telecom privatization in Africa — first between 1995 and 1997 and second between 2000 and 2001.

### Stylized facts — Growth of mobile technology in Africa
- Mobile subscriptions overtaking fixed lines:
  - In all developing regions the number of mobile phone subscribers per 100 inhabitants is now above the number of fixed telephone lines per 100 inhabitants.
  - Mobile penetration and coverage have increased rapidly, probably reflecting liberalization and privatization policies and lack of wired infrastructure.
- Positive growth perspectives in Africa:
  - People consider investment in mobile technology necessary despite significant share of earnings spent on it.
  - Willingness to pay for mobile technology is higher than in higher income countries, and price elasticity of demand is also higher (Grace, Kenny, and Qiang, 2003; Waverman, Meschi, and Fuss, 2005).
  - Example: Tcheng and others (2007) note that in some African countries such as Namibia, Ethiopia, and Zambia households spend up to 10 percent of their monthly income on telephone expenses, whereas the average is 3 percent in developed countries.
  - Telecommunications sector expected to continue to grow rapidly; limits to growth are uncertain (Tcheng, Huet, and Romdhane, 2010).
- Preference for prepaid contracts:
  - In most developing countries prepaid contracts are more common than postpaid contracts.
  - In African countries many customers do not earn regular income and are unbanked, making them less capable of affording fixed costs associated with postpaid contracts.
  - Fixed lines require monthly payments regardless of use; prepaid mobiles account for variability of consumption but at a higher cost.
  - Prepaid mobile contracts have succeeded relative to postpaid offers despite higher provider costs (e.g., scratch card production and sales).

### Stylized facts — Opportunities from mobile financial services
- High exclusion from formal financial services in Africa:
  - Large share of population is financially excluded or uses informal financial services.
- Formal financial services dominated by banks:
  - Deposits are more common than loans in all regions.
  - Number of loans and deposits per head is relatively low in sub-Saharan Africa and to a lesser extent in MENA, but the average size of loans and deposits relative to GDP per capita is high, suggesting high propensity to save constrained by lack of access.
- Lack of financial infrastructure in developing countries:
  - Total number of bank branches and ATMs is very low in developing countries compared to high-income countries.
  - Sub-Saharan Africa and MENA among the lowest.
- Mobile financial services as a potential solution:
  - Mobile cellular coverage is close to that of high-income countries while bank penetration and financial inclusion remain low.
  - Mobile financial services seen as an opportunity to reach unbanked customers and a new profit source for mobile network operators.
  - GSMA and Mobile Money for the Unbanked (MMU) deployment tracking (as of end-2007): three African countries operating mobile financial services — M-PESA in Kenya, WIZZIT and MTN Mobile Money in South Africa, and CELPAY in Zambia; services include household saving, bill payments, and money transfers.
  - Mobile financial services schemes are growing more sophisticated through partnerships between mobile operators and microfinance institutions.
  - Common mobile financial services: domestic money transfers, air time top ups, bill payments; strong desire for savings (Rasmussen, 2010).
  - International money transfer and loan repayments via mobile phones becoming widely used.
  - After M-PESA success (with more than 9 million subscribers), mobile network operators seek growth in Africa via socially-oriented services.
  - Between 2008 and 2010 GSMA and MMU tracking reports 16 African countries launched financial services via mobile phones — Uganda, Tanzania, Ghana, Cote d’Ivoire, Rwanda, Democratic Republic of Congo, Nigeria, Sierra Leone, Malawi, Niger, Somalia, Morocco, Madagascar, Egypt, and Senegal.
  - Competitive schemes also started in Kenya (ZAP and YUCASH), South Africa (Community Banking, Mopay, Send Money from FNB), and Zambia (Mobile Transactions).
  - Prospects for mobile financial services depend on stakeholder long-term strategies, appropriate service design, and government ability to foster innovation and channel payments.

### Econometric specification and data
- Sample and period:
  - Panel of 44 African countries; data from 1988 through 2007.
  - Deployment of mobile phones began in the 1990s, informing the study period.
  - Variables are averaged over four years to focus on long-term growth and avoid stationarity issues with annual data.
  - Sample period divided into five subperiods: 1988–1991, 1992–1995, 1996–1999, 2000–2003, and 2004–2007.
- Model framework:
  - Standard endogenous growth model following Barro (1991) and Waverman, Meschi, and Fuss (2005).
  - Dynamic panel data model with temporal and individual dimensions and a lagged dependent variable.
  - Variables: y_i,t is log of real per head GDP; X_i,t is set of growth determinants other than lagged per capita GDP; η_i is unobserved country-specific effect; ε_i,t is error term; i and t represent country and time period respectively.
  - Estimation approach improves on Lee, Levendis, and Gutierrez (2009) by using four-year averages and a wider range of ICT variables.
- Control variables used in estimations:
  - Initial level of real GDP per capita (conditional convergence).
  - Primary school enrollment rate (human capital).
  - Other controls: inflation, government consumption, and institutional development.

*Source: _wp1173 - 2001. They find that the elasticity of aggregate national output with respect to main*

### Appendix 2 presents descriptive statistics for all the variables. Data are obtained mainly from

### _wp1173 - Appendix 2 presents descriptive statistics for all the variables. Data are obtained mainly from

### Data and variables
- Data are obtained mainly from the International Monetary Fund, the World Bank, and the International Telecom Union databases.
- Appendix 3 contains variable definitions and sources (referenced in the source).
- Financial inclusion measures used:
  - number of deposits per head (including deposits at commercial banks, cooperatives, microfinance institutions, and specialized state financial institutions);
  - number of loans per head (including loans by the aforementioned financial institutions).
- Period covered: 1988–2007.
- Mobile financial services existed during the sample in: Zambia since 2001, South Africa since 2004, and Kenya since 2007.
- Data limitations noted: number of deposits and loans per head data are available for 2003 and 2007; the study assumes the average level holds throughout the period.

### Estimation approach and identification
- Estimator: System GMM estimator developed by Blundell and Bond (1998) (henceforth BB).
  - BB chosen because it "performs better than Arellano and Bond’s estimator when the autoregressive coefficient is relatively high, and the number of periods is small."
  - BB requires series (y_i,1, y_i,2, ..., y_i,T) to be mean stationary, i.e., have a constant mean η_i/(1−α) for each country i.
- Instrument strategy:
  - Keep number of instruments to the minimum to avoid overfitting (Roodman, 2009).
  - For lagged real GDP per capita: use first difference lagged one period as instruments for equations in levels.
  - For equations in first difference: use the first lagged value as instrument.
  - For other variables assumed endogenous: use the second lagged value as instruments.
  - Adopted the two-step System GMM with Windmeijer (2005) small sample robust correction.
- Validity and tests:
  - Validity of BB estimators checked by Hansen-Sargan test p-values of overidentifying restrictions.
  - Also check Arellano-Bond test p-values for AR(2) serial correlation.

### Baseline regression results (impact of macro and institutional covariates)
- High government consumption and macroeconomic instability captured by high inflation rates dampen economic growth in African countries.
- Human capital accumulation favors growth.
- Legal environment (civil and political liberty indexes) is not significant in baseline regressions; rule of law would give better results but reduces sample size by a quarter.
- Evidence of growth convergence among African countries: countries with lower initial income tend to grow faster conditional on controls.

### Main findings — ICT and economic growth
- Penetration rates of fixed and mobile telephones have a significant and positive impact on economic growth in Africa (consistent with Hardy (1980), Roller and Waverman (2001)).
- Results hold when including fixed and mobile penetration rates alternatively or jointly.
- Marginal effects (specification in column 5, Table 2):
  - an additional 10 percentage point increase in the mobile penetration rate could lead to a 0.7 percentage point increase in real GDP growth;
  - an additional 10 percentage point increase in the fixed penetration rate could lead to a 1.1 percentage point increase in real GDP growth.
- The marginal impact of the fixed penetration rate appears stronger than that of the mobile penetration rate in Africa, suggesting scope to further improve the contribution of mobile phone development to economic growth.
- Communication costs:
  - Price of a 3-minute fixed or mobile telephone local call is negatively associated with economic growth.
  - Example: In Ghana, cost of a 3-minute mobile telephone local call dropped by 62 percent, from US$1.18 in 1999 to US$0.46 in 2006; this would yield a 3.6 percent increase in real income over eight years.
  - Egypt and Mozambique experienced a drop in mobile communication cost by a similar magnitude during the same period.

### Nonlinearities and interaction effects (Table 3)
- Mobile and fixed telephones are substitutes in Africa:
  - Interaction term between mobile and fixed penetration rates is negative and significant; marginal impact of mobile development on growth stronger where fixed penetration is low.
- Marginal impact of mobile penetration decreases with GDP per capita:
  - Interaction between mobile penetration and GDP per capita shows diminishing growth returns to mobile telephone development.
- Computer and Internet use:
  - Coefficient on computer use is positive but not statistically significant (likely due to very low penetration rates).
  - Internet access appears to have a positive effect on economic growth only when GDP per capita is high enough ("Internet traps").

### Robustness tests
- Subsampling (random draws):
  - Selecting 98 percent of observations and repeating 250 times: average coefficient on mobile penetration remains very close to full-sample coefficient; distribution base widens.
  - Selecting 95 percent of observations: average coefficient on mobile penetration is 0.37, compared to 0.56 for the full sample.
  - Further reducing subsample size moves average coefficient closer to zero and increases distribution tail — indicates country heterogeneity.
- Recursive estimates (observations ranked by increasing mobile penetration rate):
  - Coefficient on mobile penetration remains positive and significant, but marginal impact declines as mobile penetration rate increases.
  - Suggests heterogeneity and possible lack of network effects at current penetration levels.
- Predictive performance:
  - Model predicts dependent variable well within a 10 percent confidence interval for all countries in the sample.
- Additional robustness:
  - Averaging data over a 3-year period increases sample size by 8 percent; coefficient on mobile penetration remains positive and significant at 1 percent.
  - Five-year averages reduce sample by 20 percent without affecting result quality.
  - Restricting to sub-Saharan African countries excluding Mauritius, Seychelles, South Africa: coefficient on mobile remains positive and significant and has larger magnitude.
  - Running regression on a sample of developing countries: mobile telephone development stimulates economic growth; marginal impact for African countries not statistically different from other developing economies.

### Financial inclusion as a transmission channel
- Financial inclusion measures (number of deposits per head, number of loans per head) are positive and significant determinants of growth (columns 2 and 4, Table 5).
- Controlling for financial inclusion:
  - Coefficient on mobile penetration drops, suggesting part of ICT’s positive impact on growth is channeled through financial inclusion.
- Interaction effects:
  - Interaction between mobile penetration rate and number of deposits per head: positive and significant (column 3, Table 5) — mobile penetration enhances the growth impact of deposits per head.
  - Interaction between mobile penetration rate and number of loans per head: positive and significant (column 5, Table 5) — similar reinforcement for loans.
- Mobile financial services effect:
  - Dummy Mobfi = 1 if mobile financial services operating (Zambia, South Africa, Kenya during sample), 0 otherwise.
  - Crossing Mobfi with the interaction term between financial inclusion and mobile penetration shows that where mobile financial services exist, mobile penetration further enhances the contribution of financial inclusion to growth (columns 6 and 7, Table 5).
- Robustness and alternative measures:
  - Results robust to removing outliers based on residual size thresholds (greater than two standard deviations, one standard deviation, one-half standard deviation).
  - Replacing financial inclusion with financial development (ratio of private credit to GDP): financial development positively correlated with growth; interaction between financial development and mobile penetration is positive and significant (column 8, Table 5).
  - Effect stronger in countries where mobile financial services are available.

### Policy implications and recommendations
- Encourage domestic and foreign investment in ICT and promote development of the ICT sector.
- Drive down the cost of communications to stimulate ICT diffusion and spur growth; increased competition in the telecommunication industry is suggested as one option.
- Weigh tax policy carefully: increasing tax on mobile communications can raise revenue but also increase communication costs and potentially lower growth.
- Promote greater interaction between the ICT and financial sectors to boost financial inclusion while addressing mobile banking challenges:
  - security concerns;
  - compliance with AML/CFT rules.
- Mobile banking experience in Kenya, Zambia, and South Africa demonstrates potential to reduce the financial infrastructure gap and lack of access to financial services.

### Conclusion (summary of empirical results)
- Using System GMM on a sample of African countries during 1988–2007, ICT development (mobile and fixed telephone penetration and communication costs) contributes to economic growth in Africa.
- Financial inclusion (measured by number of deposits and loans per head) is conducive to economic growth and is one channel through which mobile phone development affects growth.
- Interaction between mobile phone penetration and financial inclusion is positive and significant; in countries where mobile banking services are available, the joint impact on growth is stronger.
- Mobile phone rollout is highlighted as an important source of growth for African countries and as a mechanism to improve financial inclusion.

*Source: _wp1173 - Appendix 2 presents descriptive statistics for all the variables. Data are obtained mainly from (PDF chapter/section).*

### REFERENCES

### _wp1173 - REFERENCES

### Bibliographic references (selected)
- Barro, Robert J., 1991, ―Economic Growth in a Cross Section of Countries,‖ The Quarterly Journal of Economics, 106(2), pp. 407–43.  
- Beck, Thorsten, Asli Demirguc-Kunt, and Maria Soledad Martinez Peria, 2007, ―Reaching Out: Access to and Use of Banking Services Across Countries,‖ Journal of Financial Economics, 85(1), pp. 234–66.  
- Blundell, Richard, and Stephen Bond, 1998, ―Initial Conditions and Moment Restrictions in Dynamic Panel Data Models,‖ Journal of Econometrics, 87(1), pp. 115–43.  
- Collins, Daryl, Jonathan Morduch, Stuart Rutherford, and Orlanda Ruthven, 2009, Portfolios of the Poor: How the World's Poor Live on $2 a Day (Princeton, N.J.: Princeton University Press).  
- Haacker, Markus, 2010, "ICT Equipment Investment and Growth in Low- and Lower-Middle-Income Countries," IMF Working Paper 10/66 (Washington: International Monetary Fund).  
- Mas, Ignacio, and Dan Radcliffe, 2010, ―Mobile Payments Go Viral: M-PESA in Kenya,‖ Technical Report (Bill & Melinda Gates Foundation).  
- Tarazi, Michael, and Paul Breloff, 2010, ―Nonbank E-Money Issuers: Regulatory Approaches to Protecting Customer Funds,‖ Technical Report 63 (Washington: CGAP).  
(Full list of references appears in the source document; above are illustrative citations preserved exactly as presented.)

### Figures and tables (inventory and notes)
- Figures listed include:  
  - Figure 1. Correlation among Mobile Penetration, Growth, and Financial Inclusion  
  - Figure 2. Financial Inclusion and Mobile Penetration by Type of Financial Institution  
  - Figure 3. Trends in ICT Use: International Comparison  
  - Figure 4. ICT Growth in Developing Countries  
  - Figure 5. Trends in Fixed versus Mobile Phone Subscribers: International Comparison  
  - Figure 6. Trends in Mobile Penetration: International Comparison  
  - Figure 7. Trends in Prepaid versus Postpaid Phone Subscribers: International Comparison  
  - Figure 8. Access to Financial Services in Selected African Countries (Percent of total population)  
  - Figure 9. Bank Loans and Deposits: International Comparison  
  - Figure 10. Bank Branches and ATMs  
  - Figure 11. Financial Infrastructure Gap  
  - Figure 12. Comparing Trends in Bank Credit and Mobile Penetration  
  - Figure 13. Testing for the Stability of the Coefficient on Mobile Penetration Using Random Samples  
  - Figure 14. Recursive Estimates of the Coefficient on Mobile Penetration  
  - Figure 15. Quality of the Prediction  
- Sources for the figures: Beck, Demirguc-Kunt, and Martinez Peria (2007), International Telecommunication Union, FinMark (2009), and authors’ calculations.  
- Notes for the figures: MFIs = microfinance institutions; SSFIs = specialized state financial institutions; automated teller machines (ATMs); SSA = sub-Saharan Africa; ECA = Europe and Central Asia; MENA = Middle East and North Africa; SA = South Asia; LAC = Latin America and Caribbean; EAP = East Asia and Pacific; LICs = low-income countries; LMC = lower middle income countries; UMC = upper middle income countries.

- Tables listed include (selected):  
  - Table 2. Impact of Mobile and Fixed Telephone Penetration on Economic Growth in Africa (columns (1)–(8) with coefficients, standard errors, observations, number of countries, Hansen test (prob.), AR2 (prob.))  
  - Table 3. Testing for Interaction Terms and other Forms of ICT (columns (1)–(3))  
  - Table 4. Robustness Tests (columns (1)–(5))  
  - Table 5. Mobile Penetration, Financial Inclusion, and Economic Growth in Africa (columns (1)–(8))  
- Notes for the tables: Standard errors in brackets; * significant at 10%; ** significant at 5%; *** significant at 1%.

### Key empirical findings and statistics (as presented)
- Table-related summary points (values preserved as in the source):  
  - Table 2 reports an Initial GDP (log) coefficient ranging across columns: -0.230, -0.190, -0.190, -0.090, -0.100, -0.040, 0.010, -0.010 with associated standard errors and significance levels indicated in brackets.  
  - Table 2 includes estimates for: Total telephone subscribers (fixed + mobile) per head coefficient 0.116 [0.027]*** in one specification; Fixed telephone lines per head coefficient 0.700 [0.169]*** in another; Mobile telephone subscribers per head coefficients 0.056 [0.017]*** and 0.073 [0.009]*** across specifications. Price of a 3-minute fixed telephone local call coefficients include -0.401 [0.033]*** and -0.356 [0.021]***; Price of 3-minute mobile local call includes -0.085 [0.020]*** and -0.050 [0.011]***. Observations range from 157 to 190; Number of countries range from 39 to 44; Hansen test (prob.) values reported include 0.54, 0.60, 0.60, 0.61, 0.59, 0.37, 0.32, 0.78. AR2 (prob.) values reported include 0.23, 0.17, 0.20, 0.34, 0.31, 0.45, 0.27, 0.19.  
  - Table 3 reports: Fixed telephone lines per head coefficients 0.747 [0.058]*** and 0.060 [0.089]; Mobile telephone subscribers per head coefficients 0.300 [0.024]*** and 0.112 [0.037]***; Fixed × Mobile telephone subscribers per head coefficient -2.548 [0.088]***; Mobile telephone subscribers per head × GDP per capita coefficient -0.012 [0.007]*; Internet users per 100 inhabitants coefficient -0.579 [0.193]*** and Internet users per 100 inhabitants × GDP per capita coefficient 0.350 [0.039]***. Observations 128–189; Number of countries 41–43; Hansen test (prob.) 0.75, 0.75, 0.95; AR2 (prob.) 0.34, 0.38, 0.45.  
  - Table 4 (robustness) reports Mobile telephone subscribers per head coefficients 0.071 [0.010]***, 0.078 [0.032]**, 0.226 [0.006]***, 0.104 [0.019]***, 0.104 [0.018]*** across columns; Observations 150–451; Number of countries 36–114; Hansen test (prob.) 0.93, 0.21, 0.53, 0.03, 0.06; AR2 (prob.) 0.96, 0.17, 0.24, 0.92, 0.92. Column (3) is the sample of sub-Saharan African countries, excluding Mauritius, Seychelles, and South Africa. Columns (4) and (5) are sample of developing countries. Column (1) data averaged over 3-year periods; (2) data average over 5-year periods. Africa is a dummy variable taking 1 for African countries and 0 otherwise.  
  - Table 5 reports interactions among Mobile telephone subscribers per head and financial inclusion measures: Number of deposits per head coefficients 0.658 [0.154]*** and 0.087 [0.005]*** in different specifications; Mobile × Number of deposits per head coefficients 0.103 [0.034]*** and 0.130 [0.078]*; Mobile × Number of deposits per head × Mobfi coefficient 1.392 [0.119]***; Number of loans per head coefficients include 3.540 [0.870]***, 0.919 [0.043]***, 1.100 [0.100]***; Mobile × Number of loans per head coefficients 1.756 [0.311]*** and 0.570 [0.344]*; Mobile × Number of loans per head × Mobfi coefficient 7.419 [0.889]***; Private Credit/GDP coefficient 0.050 [0.022]** and Mobile × Private Credit/GDP coefficient 0.096 [0.056]*; Mobile × Private Credit/GDP × Mobfi coefficient 0.107 [0.021]***. Observations 184–190; Number of countries 42–44; Hansen test (prob.) reported across columns include 0.60, 0.59, 0.56, 0.66, 0.44, 0.50, 0.60, 0.73; AR2 (prob.) range 0.20–0.43. Notes: Mobfi is a dummy variable taking 1 for countries where financial services on mobile telephones are available, and 0 otherwise.

### Annex 1 — Mobile Financial Services (MFS): models, operational features, and empirical statistics
- Definitions and models:  
  - MFS belong to branchless banking services enabling banks to provide remote financial services. Branchless banking services can be additive (in addition to services offered to existing customers) or transformational (tailored for customers not reached profitably by branch-based services). The source states: "Whereas mobile financial services generally are additive in developed countries, they are mainly transformational in developing countries with the objective of improving financial inclusion of unbanked people."  
  - Two main models described:  
    - Bank-based model: customers have a direct contractual relationship with the bank; transactions handled by retailers outside bank branches. Examples: Brazilian model, South African models (MTN and WIZZIT), Indian model.  
    - Nonbank-based model: no direct contractual relationship between the user and a bank; users give cash to the retailer in exchange for an electronic record of value on a virtual account kept through the MNO or an issuer of stored value card. Examples: M-PESA in Kenya and Orange Money in Ivory Coast. Nonbank providers are not subject to prudential regulation and cannot intermediate repayable deposits; they earn profits through transaction charges or savings in airtime distribution costs, reduced churn, etc. (Tarazi, and Breloff, 2010).  

- M-PESA (case study) — operational features and statistics (values preserved):  
  - Service: mobile payment and transfer service provided by Safaricom offering an electronic wallet (e-wallet) stored on cell phones. Users can deposit into and withdraw money from their e-wallet at an M-PESA agent, transfer electronic money to other users, and buy prepaid airtime via text messages.  
  - Regulatory features: registration requires a formal ID card (noted as part of AML and CFT regulations). No minimum account balance; there is a maximum account balance and daily transactions and withdrawals are allowed only within a predetermined band (CGAP, 2007). Cash collected in exchange for e-float is held by M-PESA Trust Company, Ltd., in trust accounts with two commercial banks; any interest earned cannot benefit Safaricom or be passed through to customers because that would constitute banking activity. Discussions with the Central Bank of Kenya on what to do with the interest are noted. Tarazi and Breloff (2010) mention negotiations to donate interest earned on trust accounts to charity.  
  - Growth and usage statistics:  
    - Since its launch in March 2007, M-PESA experienced extraordinary growth.  
    - As of January 2010, there were more than 9 million registered users of M-PESA (40 percent of Kenyan adults and 23 percent of the population), of which the majority is active.  
    - About $320 million per month of person-to-person (P2P) transfers have occurred.  
    - M-PESA has 16,900 agents who are Safaricom dealers or other entities such as petrol stations (Mas and Radcliffe, 2010).  
    - As of January 2010, 75 companies are now using M-PESA to collect payments from their customers; the biggest user is the electricity company (20 percent of their customers pay through M-PESA).  
    - In January 2010, a partnership with Equity Bank allows M-PESA customers to withdraw money (but not deposit) at any Equity Bank ATM regardless of whether they are clients of Equity Bank.  
    - A savings service called M-Kesho (mobile-phone based deposit accounts) is being launched; partnerships between M-PESA and commercial banks aim to integrate non-bank based MFS with conventional banking to improve financial inclusion.  
  - Agents and liquidity: agents handling customer deposits and withdrawals receive a commission, hold e-float purchased from Safaricom or customers on their cell phones, and must maintain cash on premises to meet withdrawals. Managing agents’ liquidity is essential for M-PESA (Mas and Ng’weno, 2009).

### Regulatory and implementation considerations (from Annex)
- Prudential regulation and consumer protection: bank-based models provide a direct contractual relationship and subject providers to prudential regulation, which can reassure customers and regulators. Nonbank-based providers are not subject to prudential regulation and cannot intermediate repayable deposits; regulatory approaches to protecting customer funds are discussed in Tarazi and Breloff (2010).  
- AML/CFT implications: ID card requirements for registration are part of AML and CFT regulations; while less onerous than bank account opening documentation, ID requirements can still impede mobile banking growth and financial inclusion because targeted unbanked and low-income customers rarely possess formal ID.  
- Interest on pooled client funds: interest earned on pooled trust accounts cannot be retained by the nonbank MNO; discussions on options (including donation to charity) are reported.

*Source: _wp1173 - REFERENCES (source PDF content provided).*

### Annex 2. Assessing the Effect of Mobile Phone Penetration on Financial Inclusion

### Annex 2. Assessing the Effect of Mobile Phone Penetration on Financial Inclusion

### Objective and Motivation
- Evaluate whether mobile phone development (mobile telephone subscribers per head) affects financial inclusion, measured by number of deposits and loans per head.
- Build on Kendall, Mylenko, and Ponce (2010) by explicitly modeling financial inclusion and using a panel regression (data for 2003 and 2007).
- Rationale: Table 5 suggested mobile penetration’s effect on growth weakened once financial inclusion was controlled for, implying financial inclusion could be a channel through which mobile penetration influences growth.

### Model and Data
- Dependent variables: financial inclusion indicators measured as number of deposits per head and number of loans per head.
- Key explanatory variable: mobile telephone subscribers per head.
- Main controls: GDP per head (log) and population density (log).
- Additional controls: banks’ overhead cost (percent of assets) to capture banking efficiency, institutions (quality of legal environment), number of bank branches per km2 (geographical coverage).
- Country-specific effects accounted for; random-effects estimator used due to limited time dimension and because Hausman test probability > 0.10 (random effects appropriate).
- Explanatory variables averaged over two periods: 2000–2003 and 2004–2007. Financial inclusion indicators available for 2003 and 2007.
- Sample information (from Appendices): Observations, number of countries, and summary statistics available for relevant variables (see Appendix 1 and Appendix 2 in source).

### Main Regression Results (Table 6: Mobile Phone Development and Financial Inclusion, 2000-2007)
- Number of deposits per head (columns (1) and (2)) and number of loans per head (columns (3) and (4)).
- Coefficients, standard errors in brackets, significance: * significant at 10%; ** significant at 5%; *** significant at 1%.

- Mobile telephone subscribers per head:
  - Deposits (1): 0.382 [0.161]** 
  - Deposits (2): 0.101 [0.036]*** 
  - Loans (3): 0.211 [0.084]** 
  - Loans (4): 0.071 [0.011]*** 

- GDP per head (log):
  - Deposits (1): 0.132 [0.047]*** 
  - Deposits (2): 0.093 [0.032]*** 
  - Loans (3): 0.015 [0.016] 
  - Loans (4): -0.001 [0.010] 

- Population density (log):
  - Deposits (1): 0.098 [0.035]*** 
  - Deposits (2): 0.065 [0.030]** 
  - Loans (3): 0.007 [0.010] 
  - Loans (4): -0.023 [0.010]** 

- Banks’ overhead cost (percent of assets):
  - Deposits (2): -1.405 [0.682]** 
  - Loans (4): 0.227 [0.202] 

- Institutions (quality of legal environment):
  - Deposits (2): 0.95 [0.142]*** 
  - Loans (4): 0.279 [0.073]*** 

- Number of bank branches per km2:
  - Deposits (2): 0.005 [0.001]*** 
  - Loans (4): 0.002 [0.000]*** 

- Constant terms:
  - Deposits (1): -1.008 [0.347]*** 
  - Deposits (2): -0.802 [0.227]*** 
  - Loans (3): -0.104 [0.111] 
  - Loans (4): 0.034 [0.070] 

- Sample sizes and model diagnostics:
  - Deposits (1): Observations 49; Number of countries 36; R2 (between) 0.47; Hausman test (prob) 0.69
  - Deposits (2): Observations 39; Number of countries 31; R2 (between) 0.68; Hausman test (prob) 0.34
  - Loans (3): Observations 46; Number of countries 34; R2 (between) 0.37; Hausman test (prob) 0.99
  - Loans (4): Observations 36; Number of countries 29; R2 (between) 0.55; Hausman test (prob) 0.57

### Interpretation of Findings
- Mobile phone development is strongly positively correlated with financial inclusion for both deposit and loan measures, and this relationship holds after controlling for GDP per head, population density, banks’ overhead cost, institutions, and branch coverage.
- Banks’ overhead cost (higher costs) is negatively associated with number of deposits per head (significant), suggesting banking efficiency matters for deposit access.
- Better institutions and greater bank-branch coverage positively enhance financial inclusion (both deposits and loans).
- GDP per head, bank efficiency, and population density are positively associated with access to deposits; results for access to loans are less clear.
- Population density shows a negative sign for loans when branch density is included; possible explanation: commercial banks avoid geographically concentrated lending to limit collective risk, whereas less formal institutions (microfinance) use geographically concentrated group lending—financial inclusion indicators may understate access to less formal services, contributing to the negative sign.

### Additional Notes and Tests
- The Hausman test probabilities reported support the random-effects estimator choice (probabilities all > 0.10).
- Tests of inflation and banking concentration (as in Kendall, Mylenko, and Ponce (2010)) found neither significant for financial inclusion.
- Related evidence: Household survey-based findings (Beck and others (2010)) indicate ownership of a cell phone increases likelihood of using financial services in Kenya.

*Source: Annex 2. Assessing the Effect of Mobile Phone Penetration on Financial Inclusion (WP1173).*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp1173.pdf_