literature review on money market

• Search Menu
• Browse content in A - General Economics and Teaching
• Browse content in A1 - General Economics
• A12 - Relation of Economics to Other Disciplines
• A14 - Sociology of Economics
• Browse content in B - History of Economic Thought, Methodology, and Heterodox Approaches
• Browse content in B4 - Economic Methodology
• B41 - Economic Methodology
• Browse content in C - Mathematical and Quantitative Methods
• Browse content in C1 - Econometric and Statistical Methods and Methodology: General
• C18 - Methodological Issues: General
• Browse content in C2 - Single Equation Models; Single Variables
• C21 - Cross-Sectional Models; Spatial Models; Treatment Effect Models; Quantile Regressions
• Browse content in C3 - Multiple or Simultaneous Equation Models; Multiple Variables
• C38 - Classification Methods; Cluster Analysis; Principal Components; Factor Models
• Browse content in C5 - Econometric Modeling
• C59 - Other
• Browse content in C8 - Data Collection and Data Estimation Methodology; Computer Programs
• C80 - General
• C81 - Methodology for Collecting, Estimating, and Organizing Microeconomic Data; Data Access
• C83 - Survey Methods; Sampling Methods
• Browse content in C9 - Design of Experiments
• C93 - Field Experiments
• Browse content in D - Microeconomics
• Browse content in D0 - General
• D02 - Institutions: Design, Formation, Operations, and Impact
• D03 - Behavioral Microeconomics: Underlying Principles
• D04 - Microeconomic Policy: Formulation; Implementation, and Evaluation
• Browse content in D1 - Household Behavior and Family Economics
• D10 - General
• D12 - Consumer Economics: Empirical Analysis
• D14 - Household Saving; Personal Finance
• Browse content in D2 - Production and Organizations
• D22 - Firm Behavior: Empirical Analysis
• D24 - Production; Cost; Capital; Capital, Total Factor, and Multifactor Productivity; Capacity
• Browse content in D3 - Distribution
• D31 - Personal Income, Wealth, and Their Distributions
• Browse content in D6 - Welfare Economics
• D61 - Allocative Efficiency; Cost-Benefit Analysis
• D62 - Externalities
• D63 - Equity, Justice, Inequality, and Other Normative Criteria and Measurement
• Browse content in D7 - Analysis of Collective Decision-Making
• D72 - Political Processes: Rent-seeking, Lobbying, Elections, Legislatures, and Voting Behavior
• D73 - Bureaucracy; Administrative Processes in Public Organizations; Corruption
• D74 - Conflict; Conflict Resolution; Alliances; Revolutions
• Browse content in D8 - Information, Knowledge, and Uncertainty
• D83 - Search; Learning; Information and Knowledge; Communication; Belief; Unawareness
• D85 - Network Formation and Analysis: Theory
• D86 - Economics of Contract: Theory
• Browse content in D9 - Micro-Based Behavioral Economics
• D91 - Role and Effects of Psychological, Emotional, Social, and Cognitive Factors on Decision Making
• D92 - Intertemporal Firm Choice, Investment, Capacity, and Financing
• Browse content in E - Macroeconomics and Monetary Economics
• Browse content in E2 - Consumption, Saving, Production, Investment, Labor Markets, and Informal Economy
• E23 - Production
• E24 - Employment; Unemployment; Wages; Intergenerational Income Distribution; Aggregate Human Capital; Aggregate Labor Productivity
• Browse content in E4 - Money and Interest Rates
• E42 - Monetary Systems; Standards; Regimes; Government and the Monetary System; Payment Systems
• Browse content in E5 - Monetary Policy, Central Banking, and the Supply of Money and Credit
• E52 - Monetary Policy
• E58 - Central Banks and Their Policies
• Browse content in E6 - Macroeconomic Policy, Macroeconomic Aspects of Public Finance, and General Outlook
• E60 - General
• E61 - Policy Objectives; Policy Designs and Consistency; Policy Coordination
• E62 - Fiscal Policy
• E65 - Studies of Particular Policy Episodes
• Browse content in F - International Economics
• Browse content in F0 - General
• F01 - Global Outlook
• Browse content in F1 - Trade
• F10 - General
• F11 - Neoclassical Models of Trade
• F13 - Trade Policy; International Trade Organizations
• F14 - Empirical Studies of Trade
• F15 - Economic Integration
• Browse content in F2 - International Factor Movements and International Business
• F21 - International Investment; Long-Term Capital Movements
• F22 - International Migration
• F23 - Multinational Firms; International Business
• Browse content in F3 - International Finance
• F32 - Current Account Adjustment; Short-Term Capital Movements
• F34 - International Lending and Debt Problems
• F35 - Foreign Aid
• F36 - Financial Aspects of Economic Integration
• Browse content in F4 - Macroeconomic Aspects of International Trade and Finance
• F41 - Open Economy Macroeconomics
• F42 - International Policy Coordination and Transmission
• F43 - Economic Growth of Open Economies
• Browse content in F5 - International Relations, National Security, and International Political Economy
• F50 - General
• F52 - National Security; Economic Nationalism
• F53 - International Agreements and Observance; International Organizations
• F55 - International Institutional Arrangements
• Browse content in F6 - Economic Impacts of Globalization
• F61 - Microeconomic Impacts
• F63 - Economic Development
• F66 - Labor
• Browse content in G - Financial Economics
• Browse content in G0 - General
• G01 - Financial Crises
• Browse content in G1 - General Financial Markets
• G10 - General
• G15 - International Financial Markets
• G18 - Government Policy and Regulation
• Browse content in G2 - Financial Institutions and Services
• G20 - General
• G21 - Banks; Depository Institutions; Micro Finance Institutions; Mortgages
• G22 - Insurance; Insurance Companies; Actuarial Studies
• G23 - Non-bank Financial Institutions; Financial Instruments; Institutional Investors
• G28 - Government Policy and Regulation
• Browse content in G3 - Corporate Finance and Governance
• G32 - Financing Policy; Financial Risk and Risk Management; Capital and Ownership Structure; Value of Firms; Goodwill
• G33 - Bankruptcy; Liquidation
• G38 - Government Policy and Regulation
• Browse content in H - Public Economics
• Browse content in H1 - Structure and Scope of Government
• H11 - Structure, Scope, and Performance of Government
• Browse content in H2 - Taxation, Subsidies, and Revenue
• H20 - General
• H23 - Externalities; Redistributive Effects; Environmental Taxes and Subsidies
• H25 - Business Taxes and Subsidies
• H26 - Tax Evasion and Avoidance
• H27 - Other Sources of Revenue
• Browse content in H3 - Fiscal Policies and Behavior of Economic Agents
• H31 - Household
• Browse content in H4 - Publicly Provided Goods
• H41 - Public Goods
• H43 - Project Evaluation; Social Discount Rate
• Browse content in H5 - National Government Expenditures and Related Policies
• H52 - Government Expenditures and Education
• H53 - Government Expenditures and Welfare Programs
• H54 - Infrastructures; Other Public Investment and Capital Stock
• H55 - Social Security and Public Pensions
• H56 - National Security and War
• H57 - Procurement
• Browse content in H6 - National Budget, Deficit, and Debt
• H60 - General
• H61 - Budget; Budget Systems
• Browse content in H7 - State and Local Government; Intergovernmental Relations
• H71 - State and Local Taxation, Subsidies, and Revenue
• H75 - State and Local Government: Health; Education; Welfare; Public Pensions
• H77 - Intergovernmental Relations; Federalism; Secession
• Browse content in H8 - Miscellaneous Issues
• H83 - Public Administration; Public Sector Accounting and Audits
• H84 - Disaster Aid
• Browse content in I - Health, Education, and Welfare
• Browse content in I0 - General
• I00 - General
• Browse content in I1 - Health
• I10 - General
• I12 - Health Behavior
• I15 - Health and Economic Development
• I18 - Government Policy; Regulation; Public Health
• Browse content in I2 - Education and Research Institutions
• I20 - General
• I21 - Analysis of Education
• I22 - Educational Finance; Financial Aid
• I24 - Education and Inequality
• I25 - Education and Economic Development
• I28 - Government Policy
• Browse content in I3 - Welfare, Well-Being, and Poverty
• I30 - General
• I31 - General Welfare
• I32 - Measurement and Analysis of Poverty
• I38 - Government Policy; Provision and Effects of Welfare Programs
• Browse content in J - Labor and Demographic Economics
• Browse content in J0 - General
• J01 - Labor Economics: General
• J08 - Labor Economics Policies
• Browse content in J1 - Demographic Economics
• J10 - General
• J11 - Demographic Trends, Macroeconomic Effects, and Forecasts
• J12 - Marriage; Marital Dissolution; Family Structure; Domestic Abuse
• J13 - Fertility; Family Planning; Child Care; Children; Youth
• J15 - Economics of Minorities, Races, Indigenous Peoples, and Immigrants; Non-labor Discrimination
• J16 - Economics of Gender; Non-labor Discrimination
• J17 - Value of Life; Forgone Income
• J18 - Public Policy
• Browse content in J2 - Demand and Supply of Labor
• J21 - Labor Force and Employment, Size, and Structure
• J22 - Time Allocation and Labor Supply
• J23 - Labor Demand
• J24 - Human Capital; Skills; Occupational Choice; Labor Productivity
• J26 - Retirement; Retirement Policies
• J28 - Safety; Job Satisfaction; Related Public Policy
• Browse content in J3 - Wages, Compensation, and Labor Costs
• J38 - Public Policy
• Browse content in J4 - Particular Labor Markets
• J48 - Public Policy
• Browse content in J5 - Labor-Management Relations, Trade Unions, and Collective Bargaining
• J58 - Public Policy
• Browse content in J6 - Mobility, Unemployment, Vacancies, and Immigrant Workers
• J61 - Geographic Labor Mobility; Immigrant Workers
• J62 - Job, Occupational, and Intergenerational Mobility
• J63 - Turnover; Vacancies; Layoffs
• J68 - Public Policy
• Browse content in J8 - Labor Standards: National and International
• J88 - Public Policy
• Browse content in K - Law and Economics
• Browse content in K2 - Regulation and Business Law
• K23 - Regulated Industries and Administrative Law
• Browse content in K3 - Other Substantive Areas of Law
• K34 - Tax Law
• Browse content in K4 - Legal Procedure, the Legal System, and Illegal Behavior
• K40 - General
• K42 - Illegal Behavior and the Enforcement of Law
• Browse content in L - Industrial Organization
• Browse content in L1 - Market Structure, Firm Strategy, and Market Performance
• L11 - Production, Pricing, and Market Structure; Size Distribution of Firms
• L14 - Transactional Relationships; Contracts and Reputation; Networks
• L16 - Industrial Organization and Macroeconomics: Industrial Structure and Structural Change; Industrial Price Indices
• Browse content in L2 - Firm Objectives, Organization, and Behavior
• L20 - General
• L23 - Organization of Production
• L25 - Firm Performance: Size, Diversification, and Scope
• L26 - Entrepreneurship
• Browse content in L3 - Nonprofit Organizations and Public Enterprise
• L33 - Comparison of Public and Private Enterprises and Nonprofit Institutions; Privatization; Contracting Out
• Browse content in L5 - Regulation and Industrial Policy
• L51 - Economics of Regulation
• L52 - Industrial Policy; Sectoral Planning Methods
• Browse content in L9 - Industry Studies: Transportation and Utilities
• L94 - Electric Utilities
• L97 - Utilities: General
• L98 - Government Policy
• Browse content in M - Business Administration and Business Economics; Marketing; Accounting; Personnel Economics
• Browse content in M5 - Personnel Economics
• M53 - Training
• Browse content in N - Economic History
• Browse content in N3 - Labor and Consumers, Demography, Education, Health, Welfare, Income, Wealth, Religion, and Philanthropy
• N35 - Asia including Middle East
• Browse content in N5 - Agriculture, Natural Resources, Environment, and Extractive Industries
• N55 - Asia including Middle East
• N57 - Africa; Oceania
• Browse content in N7 - Transport, Trade, Energy, Technology, and Other Services
• N77 - Africa; Oceania
• Browse content in O - Economic Development, Innovation, Technological Change, and Growth
• Browse content in O1 - Economic Development
• O10 - General
• O11 - Macroeconomic Analyses of Economic Development
• O12 - Microeconomic Analyses of Economic Development
• O13 - Agriculture; Natural Resources; Energy; Environment; Other Primary Products
• O14 - Industrialization; Manufacturing and Service Industries; Choice of Technology
• O15 - Human Resources; Human Development; Income Distribution; Migration
• O16 - Financial Markets; Saving and Capital Investment; Corporate Finance and Governance
• O17 - Formal and Informal Sectors; Shadow Economy; Institutional Arrangements
• O18 - Urban, Rural, Regional, and Transportation Analysis; Housing; Infrastructure
• O19 - International Linkages to Development; Role of International Organizations
• Browse content in O2 - Development Planning and Policy
• O20 - General
• O22 - Project Analysis
• O23 - Fiscal and Monetary Policy in Development
• O24 - Trade Policy; Factor Movement Policy; Foreign Exchange Policy
• O25 - Industrial Policy
• Browse content in O3 - Innovation; Research and Development; Technological Change; Intellectual Property Rights
• O31 - Innovation and Invention: Processes and Incentives
• O32 - Management of Technological Innovation and R&D
• O33 - Technological Change: Choices and Consequences; Diffusion Processes
• O34 - Intellectual Property and Intellectual Capital
• O38 - Government Policy
• Browse content in O4 - Economic Growth and Aggregate Productivity
• O40 - General
• O41 - One, Two, and Multisector Growth Models
• O43 - Institutions and Growth
• O47 - Empirical Studies of Economic Growth; Aggregate Productivity; Cross-Country Output Convergence
• Browse content in O5 - Economywide Country Studies
• O55 - Africa
• O57 - Comparative Studies of Countries
• Browse content in P - Economic Systems
• Browse content in P1 - Capitalist Systems
• P14 - Property Rights
• Browse content in P2 - Socialist Systems and Transitional Economies
• P26 - Political Economy; Property Rights
• Browse content in P3 - Socialist Institutions and Their Transitions
• P30 - General
• Browse content in P4 - Other Economic Systems
• P43 - Public Economics; Financial Economics
• P48 - Political Economy; Legal Institutions; Property Rights; Natural Resources; Energy; Environment; Regional Studies
• Browse content in Q - Agricultural and Natural Resource Economics; Environmental and Ecological Economics
• Browse content in Q0 - General
• Q01 - Sustainable Development
• Browse content in Q1 - Agriculture
• Q10 - General
• Q12 - Micro Analysis of Farm Firms, Farm Households, and Farm Input Markets
• Q13 - Agricultural Markets and Marketing; Cooperatives; Agribusiness
• Q14 - Agricultural Finance
• Q15 - Land Ownership and Tenure; Land Reform; Land Use; Irrigation; Agriculture and Environment
• Q16 - R&D; Agricultural Technology; Biofuels; Agricultural Extension Services
• Q17 - Agriculture in International Trade
• Q18 - Agricultural Policy; Food Policy
• Browse content in Q2 - Renewable Resources and Conservation
• Q25 - Water
• Browse content in Q3 - Nonrenewable Resources and Conservation
• Q33 - Resource Booms
• Browse content in Q4 - Energy
• Q43 - Energy and the Macroeconomy
• Browse content in Q5 - Environmental Economics
• Q51 - Valuation of Environmental Effects
• Q52 - Pollution Control Adoption Costs; Distributional Effects; Employment Effects
• Q54 - Climate; Natural Disasters; Global Warming
• Q56 - Environment and Development; Environment and Trade; Sustainability; Environmental Accounts and Accounting; Environmental Equity; Population Growth
• Q57 - Ecological Economics: Ecosystem Services; Biodiversity Conservation; Bioeconomics; Industrial Ecology
• Q58 - Government Policy
• Browse content in R - Urban, Rural, Regional, Real Estate, and Transportation Economics
• Browse content in R1 - General Regional Economics
• R11 - Regional Economic Activity: Growth, Development, Environmental Issues, and Changes
• R12 - Size and Spatial Distributions of Regional Economic Activity
• R13 - General Equilibrium and Welfare Economic Analysis of Regional Economies
• R14 - Land Use Patterns
• Browse content in R2 - Household Analysis
• R20 - General
• R23 - Regional Migration; Regional Labor Markets; Population; Neighborhood Characteristics
• R28 - Government Policy
• Browse content in R3 - Real Estate Markets, Spatial Production Analysis, and Firm Location
• R38 - Government Policy
• Browse content in R4 - Transportation Economics
• R40 - General
• R41 - Transportation: Demand, Supply, and Congestion; Travel Time; Safety and Accidents; Transportation Noise
• R48 - Government Pricing and Policy
• Browse content in R5 - Regional Government Analysis
• R52 - Land Use and Other Regulations
• Browse content in Y - Miscellaneous Categories
• Y8 - Related Disciplines
• Browse content in Z - Other Special Topics
• Browse content in Z1 - Cultural Economics; Economic Sociology; Economic Anthropology
• Z13 - Economic Sociology; Economic Anthropology; Social and Economic Stratification
• Advance articles
• Author Guidelines
• Open Access
• About The World Bank Research Observer
• About the World Bank
• Editorial Board
• Advertising and Corporate Services
• Self-Archiving Policy
• Dispatch Dates
• Journals on Oxford Academic
• Books on Oxford Academic

Article Contents

Some background on mobile money and its role in financial inclusion, the economics of mobile money: the micro-view, empirical research, mobile money and the economy: a review of the evidence.

• Article contents
• Figures & tables
• Supplementary Data

Janine Aron, Mobile Money and the Economy: A Review of the Evidence, The World Bank Research Observer , Volume 33, Issue 2, August 2018, Pages 135–188, https://doi.org/10.1093/wbro/lky001

• Permissions Icon Permissions

Mobile money is a recent innovation that provides financial transaction services via mobile phone, including to the unbanked global poor. The technology has spread rapidly in the developing world, “leapfrogging” the provision of formal banking services by solving the problems of weak institutional infrastructure and the cost structure of conventional banking. This article examines the evolution of mobile money and its important role in widening financial inclusion. It explores the channels of economic influence of mobile money from a micro perspective, and critically reviews the empirical literature on the economic impact of mobile money. The evidence convincingly suggests that mobile money fosters risk-sharing, but direct evidence of the promotion of welfare and saving is still mostly rather less robust.

“ Leapfrog ”: to improve a position by going past others quickly or by missing some stages of an activity or process.” [Cambridge Business English Dictionary, CUP]

Mobile money is novel : it was barely heard of a decade ago. 1 Yet it has transformed the landscape of financial inclusion, spreading rapidly in developing and emerging market countries (see figure 1 ), and “leapfrogging” the provision of formal banking services. The poor are especially vulnerable to risk (e.g., from illness, unemployment, death of family members, or natural disasters). Enhancing financial inclusion of the unbanked urban and rural poor—a goal of the G20 group of countries—can help to diversify risk. Financial inclusion policy has focused on extending access to formal banking services, but progress has been thwarted by cost and market failure challenges.

Number of Live Mobile Money Services for the Unbanked by Region

Source : Data from the GSMA State of the Industry Report ( 2017 ).

Note : The first mobile money system was launched in the Philippines in 2001, and M-Pesa was launched in 2007.

The new technology helps overcome problems from weak institutional infrastructure and the cost structure of conventional banking. Small size, volatility, informality, and poor governance place constraints on the commercial viability of financial institutions in developing countries ( Beck and Cull 2013 ), see figure 2 . The poor mostly cannot afford the minimum balance requirements and regular charges of typical bank accounts. Mobile phone technology has the advantage that consumers themselves invest in a mobile phone handset, while the (scalable) infrastructure is already in place for the widespread distribution of airtime through secure network channels (see figure 3 ). By adopting mobile money, under-served citizens gain a secure means of transfer and payment at a lower cost, and safe and private storage of funds. Mobile money has filled a lacuna, and has “changed the economics of small accounts” ( Veniard 2010 ). 2

Provision of Banking Infrastructure

Source : G20 Financial Inclusion Indicators database, World Bank and IMF Financial Access Survey.

Note : This shows the position shortly after the adoption of mobile money in Kenya.The first five regions refer to “developing only”.

Fixed Telephone and Mobile-cellular Subscriptions: 2005 and 2017

Source : ITU World Telecommunication, ICT Indicators database.

Note : Subscriptions are per 100 inhabitants. “Mobile phone subscribers” refer to active SIM cards rather than individual subscribers.

The technological innovation has helped ameliorate the perennial asymmetric information constraint faced by conventional banks in lending to the collateral-less poor. 3 The movement of cash into electronic accounts gives a record, for the first time for the unbanked, of the history of their financial transactions in real time. By using algorithms, these records can provide evolving individual credit scores for the unbanked. 4 After a designated period of usage and once a score is available, registered users of mobile money may obtain a pathway to formal banking services accessed only through a mobile phone: to interest-bearing savings accounts that can protect assets; to credit extension to invest in livelihoods; and insurance products that reduce risk.

Apart from reducing asymmetric information, the impact of enhancing transparency through electronic records is far-reaching. Tax collection could be improved by the rise of more visible spending, quite apart from the greater ease of tax collection via mobile money payments. The increased transparency of records protects customers’ rights and fosters trust in business, promoting the growth of efficient payments networks. Mobile money should make international transactions more readily traceable and therefore facilitate identification and better control of money laundering. If the high cost of remittances were reduced by mobile money, this could attract more official remittances, and re-channel “informal” remittances through official channels, raising recorded remittances. 5 In essence, mature mobile money systems and the records they produce help foster the “formalization” of the economy, integrating informal sector users into business networks, formal banking and insurance, and linking them to government through social security, tax, and secure wages payments. However, there are legal data privacy considerations concerning access to and use of mobile money records which have barely begun to be addressed.

The channels through which mobile money can affect the economy are many and complex, and not necessarily well-understood. A burgeoning body of empirical literature has attempted to quantify the possible economic gains for different countries of access to secure financial services through mobile money (e.g., improved risk-sharing, food security, consumption, business profitability, saving, and use of cash transfers), and the factors driving the adoption of mobile money. Demonstrating welfare and risk-sharing gains from mobile money across countries could bolster the case for significant government and donor support, as well as investment.

Unfortunately, interpreting the evidence on the economic impact of mobile money is not straightforward. The empirical literature is burdened by a range of sometimes serious problems with data, methodology, and identification, which some authors underestimate or choose to ignore. Work on mobile money faces “selection” problems since both the “roll-out” of mobile money by Mobile Network Operators (MNO) and their agents and the adoption or usage of mobile money by individuals may be influenced by other factors such as education, wealth, and changes in technology preference. There is mixed success using various methods and data sets in dealing with the resultant ambiguous causality. Although various studies establish statistically significant relationships, they frequently do not test the robustness of their results to different model specifications, measurement errors, and bias due to the possible omission of variables. Furthermore, in practice it is difficult to generalize from these models.

This article introduces the phenomenon of mobile money and its role in financial inclusion. It examines possible channels for the economic influence of mobile money, and reviews the new empirical literature on mobile money, both to obtain a better understanding of the linkages involved and to critically assess the sometimes strong claims made by the authors. Lessons are distilled for improved practice in the future empirical analysis of mobile money.

In economies with deep financial markets like the United States, mobile payments or transfers are predominantly linked with pre-existing bank accounts; mobile payments are rapidly gaining market share after a slow start, catalyzed by new technology and commercial partnerships (e.g., Apple Pay). This is distinct from mobile money payments or transfers in largely cash-based developing or emerging countries, where most users are unbanked. Yet as mobile money systems evolve and smartphones become ever cheaper in less advanced countries, the range of financial services could expand to link with products managed by formal financial institutions such as banks and insurance companies. This will ultimately blur the distinctions between mobile banking and mobile money. Survey evidence suggests that security concerns about mobile payments have diminished in the United States, shaped by industry efforts to enhance security (e.g., Federal Reserve 2016 ). There may be a technological spillover to less advanced economies, and biometrics may allay security concerns (though there are caveats about their use in poor countries). This could catalyze a transformation to a virtually cashless economy, and possibly a new role for some banks beyond traditional payments. 6

The term “financial inclusion” is of recent vintage, and has gained currency with policymakers, most prominently in the Maya Declaration of 2011, when 80 regulatory institutions from 76 countries collectively endorsed a set of financial inclusion principles. The G20 has backed the Maya declaration, promoted indicators to measure “financial inclusion”, and the G20 Summit in 2017 prominently endorsed digital approaches to financial inclusion. Mainstream definitions of financial inclusion share the goal of participation in the formal financial sector, which has severely constrained progress to inclusion. Until recently, the use of electronic mobile money has not been counted as part of financial inclusion under most definitions. Mobile money's role is seen as a pathway for registered users to formal sector financial inclusion via products (insurance, credit and a bank savings account) accessed through a mobile phone.

Aron (2017) argues that a revised definition of financial inclusion should encompass tiers of semi-formal inclusion, and not focus on comprehensive formal banking sector inclusion. Mobile money has transformed the lives of poor consumers who can hold recorded cash privately in non-bank electronic accounts and perform financial transfers easily and cost effectively. Fast-spreading and cheaper smartphones (and recycled smartphone handsets) potentially offer access to sophisticated features and a spectrum of financial services for huge numbers of illiterate people through well-designed applications ( Villasenor 2013 ). Such users may not embrace the formal sector products even if they become available, for example, if they qualify for credit, the loans may be small and not adequate to purpose, creating a disincentive to participate. Moreover, the actual number of informal users may be far higher than is formally reported. South Asia has close to 90% of the global unregistered mobile money customers, using an over-the-counter (OTC) model where the challenges and costs of establishing identity in registering were circumvented in favor of a drive for early market share ( Scharwatt et al. 2015 ). 7 In practice, the proliferation of mobile money services and the sheer numbers of new users actively signed up has become integral to achieving ambitious targets under the 2011 Maya Declaration. A revised set of G20 indicators in 2016 has raised the prominence of mobile money, reducing the bias to formality.

In box 1 , the Kenyan mobile money system M-Pesa is summarized and serves to explain the “nuts and bolts” of a profitable mobile money system. Instead of bank branches, mobile money systems rely on a large network of agents. These are linked under various contractual arrangements with a parent MNO, usually in partnership with a prudentially-regulated bank. 8 The nature of agent network structures and the design of the individual agent contracts are crucial for the successful development of mobile money systems ( Aron 2017 ). The typical authorized agents of the mobile money services provider are shops or outlets staffed by small business owners. 9 Mobile money systems were initially dominated by domestic money transfers, but have expanded into a broader payments platform for utility bills, rent, taxes, school fees, and retail payments. Business usage is expanding rapidly through special networks for the payment of suppliers, wages payments, and potentially pensions. Government usage for the payment of wages and social security has lagged, though the cost savings or reductions, especially in insecure environments, could be significant.

Kenya's mobile money system originated in 2005 as an experiment for loan payments via mobile phones in micro-credit schemes, in a public-private partnership between DFID (UK), the Kenyan Government, and Vodafone. In March 2007, Safaricom, the Kenyan subsidiary of Vodafone, launched a commercial payments service, M-Pesa, with the slogan “send money home”, exploiting the proliferation of mobile phone ownership. A decade later, there were six operators, though Safaricom controlled 65% of the market. The FinAccess (2013) survey revealed that 67% of the adult population used financial services in 2013 versus 41% in 2009, driven by mobile money. There were 27 million registered M-Pesa customers by 2017, of whom 19 million were (30-day) “active”. M-Pesa revenue grew by 33% to Kshs 55bn (US$536m) in the year to Mar. 2017, over one-quarter of Safaricom's total service revenue. The Bank of Kenya recorded in 2015, for all operators, a monthly value of transactions of Kshs 227.9bn (US $2.2bn), or about one-half of average monthly GDP.

In August, 2014, the National Payment System Regulations were issued under the National Payment System Act, providing a legal framework for mobile money. These regulations formalized and extended prudential and market conduct requirements for mobile money providers as previously articulated in simple letters of no-objection from the Central Bank of Kenya (CBK). The CBK has duties of oversight, inspection, and enforcement. There are mechanisms for consumer protection, redress, and confidentiality of data.

In Kenya, banks and non-banks, including mobile network operators (MNOs), may provide mobile money services. The net deposits from customers have to be invested in prudentially-regulated banks for safe-keeping in “Trust” accounts, which back 100% of the money of the participants in the mobile money service; the banks are required to satisfy fiduciary responsibility in all transactions concerning the Trust funds ( Greenacre and Buckley 2016 ). No investment of Trust funds is allowed; the funds are strictly separated from the service provider's own accounts and safeguarded from claims of its creditors. Safaricom's Trust account interest income is covenanted to charity.

The early agent exclusivity arrangement for M-Pesa was formally outlawed in July 2014; the CBK ordered Safaricom to open the agent network to other operators to improve competition and to lower fees for customers. Interoperability of platforms was implemented in April 2018; before this, users of mobile money services had to affiliate with multiple mobile providers.

By 2017, there were 136,000 M-Pesa agents countrywide (compared with about 2.43 commercial bank branches per 1,000 km 2 in 2013, or 1,410 total branches). Establishing an agency network and the training and payment of agents is a considerable early investment by operators to develop the market. Retail cash agents transact with their own cash and electronic money in their own M-Pesa accounts to meet customer demand. Wholesale agents (banks or non-bank merchants) are allowed higher limits on electronic money stored in their M-Pesa accounts; they perform a liquidity management service for retail agents, who typically transact daily with wholesalers. Retail agents open accounts observing identity checks required by anti-money laundering legislation, and the cash provision function spans in-store cash merchants to street-based merchants. M-Pesa agents are compensated from transaction fees charged to customers.

Mobile phone users purchase a SIM card with the mobile money “app” for their phone, register with a retail agent using a national identity card and acquire an electronic mobile money account. They deposit money into the account by giving cash to the agent, and receive, in return, equivalent value “electronic money” via their mobile phone. To withdraw money, they transfer electronic money via their mobile phone to the cash merchant's mobile money account, and receive cash in return. Electronic money can be transferred instantly from a customer's account to any other individual, whether registered or not, without using formal bank accounts. The transactions are authorized and recorded in real time. A secure text message (SMS) with a code is sent to the recipient, authorizing a retail agent to transfer money from the remitter's account into cash for the designated recipient. The maximum allowed account balance is Ksh 100,000 (US $970), the maximum daily transaction is Ksh 140,000, the maximum per transaction is Ksh 70,000, and the minimum allowed transfer is Ksh1 (US 10cents). The main transactions are non-bank payments services such as buying airtime, paying bills and school fees, and domestic transfers.

Depositors do not receive interest on their electronic accounts and bear the risk of loss of value through inflation. They pay the cost of transferring and withdrawing money, but there is no charge for depositing money. The graduated withdrawal fee pays for the cost of the M-Pesa account, ranging from about 0.5% for large transfers to 20% for the smallest. The costs of transfer are 10% for the smallest transfers, falling to 0.5% at transfers of Kshs 20,000, and to 0.16% for Kshs 70,000. Costs are greater to transfer to unregistered users.

Safaricom has pioneered a business payments platform and this is an important growth area for the company. The “Lipa na M-Pesa” business network has built a critical mass of consumers using retail payments providing dedicated business till numbers and low transaction fees, and it enables bulk disbursements such as promotional payments or salary payments. For Safaricom, customer-to-business payments accounted for 10.5% of the average monthly value of all payments in 2016.

M-Shwari is a savings and loan product operated entirely from the mobile phone, launched in 2012 by partners Safaricom and Commercial Bank of Africa. By 2016, (30-day) active customers numbered 3.9m, with Kshs 8.1bn on deposit. Customers can move funds between their M-Pesa account and M-Shwari bank savings account (with no minimum balances or charges, and paying graduated interest rates of 2% to 5%). The new Lock Box service pays higher interest rates for fixed deposits. M-Pesa subscribers of 6 months standing can apply for an M-Shwari loan without fees or paperwork. An initial credit score and loan limit is calculated using an algorithm from the stream of recorded financial actions. Loan disbursement and repayment is via M-Pesa, without loan interest charges, but with a facility fee of 7.5%. Loan sizes range from US $1 to US $235 with a 30-day term but can be rolled over at a monthly fee of 7.5% (this resembles an interest rate at a high annual compounded rate of 138 percent). Progressively larger loans can be extended when a loan is successfully repaid. By 2016, there was Kshs 7.4 bn on loan; non-performing loans numbered 1.93% of the portfolio, with an average loan size of Ksh 4,000 ($39).

In 2015, an M-Pesa health micro-insurance product, launched during the previous year, was discontinued through failure to gain traction. The annual premium (Kshs12,000) had bought family cover worth Kshs 290,000 for maternity, dental and optical care, and hospital and funeral expenses. In late 2015, M-Tiba (“mobile care”), a dedicated health savings “wallet” to pay for care at selected affordable health providers, was launched by Safaricom with two partners, enabling users to save and pay for healthcare. Donors and insurers can use M-Tiba for targeted products including vouchers, managed funds, and low-cost health insurance.

Kenya received an estimated US $1.7bn of international remittances in 2016 (World Bank Migration Brief 27). In 2014, Safaricom partnered with MoneyGram to enable remittances from over 90 countries worldwide to be sent to M-Pesa users, and now has similar agreements with Western Union and several other partners. In 2015, Vodafone and MTN announced an interconnection of mobile money services enabling affordable regional remittances between M-Pesa customers in Kenya, Tanzania, Democratic Republic of Congo, and Mozambique, and MTN Mobile Money customers in Uganda, Rwanda, and Zambia. In 2016, Vodafone partnered with HomeSend (a joint venture created by MasterCard, eServGlobal and BICS) to extend remittances for M-Pesa users in Africa, Albania, and Romania.

Governments could securely pay policemen and other officials their wages; the national revenue authority could accept payments for taxes, licenses, and fines, and municipalities for parking payments; and public transport could use mobile money payments. Delivery of social welfare or aid with mobile payments could reduce “leakage” and ghost recipients. Some of these are a reality in Kenya, with M-Pesa and Airtel, through pilots or fully-functioning systems, but government salary and social payments have lagged relative to Afghanistan, Tanzania, and Malawi. Donor and commercial initiatives increasingly use the technology; for example, affordable solar energy-powered electricity systems in rural areas can be fully purchased remotely on a pay-as-you-use basis using mobile payments (M-Kopa Solar launched in 2014 in Kenya).

Vodafone has concentrated on the proliferation of its mobile money platform in markets that are heavy cash users. M-Pesa is used in several countries other than Kenya, by order of roll-out: Tanzania, Fiji, South Africa, Fiji, Democratic Republic of Congo, India (launched in 2013), Mozambique, Egypt, Lesotho, Romania (2014), Albania (2015), and Ghana (2015).

A fast-growing product is international remittance through mobile money channels. The size of officially-recorded remittance flows to developing countries and the high transactions costs suggest that the potential gains from transparent and cheaper methods of remittance are significant. 10 Security concerns present a challenge because of poor compliance to international law at the receiving end. If the local compliance challenge can be overcome, mobile money (bound by “know your client” legislation and electronic recording of transactions) should facilitate remittances to war-torn countries with weak governance and limited or no functional banking, like Somalia.

The novelty of mobile money and its recent introduction in many countries means few studies have examined the economics of mobile money. 11 The mobile money storage and payments system, and its further linkages to bank savings accounts, micro-insurance, and credit via algorithmic credit scores, could affect households and businesses through several different channels. Mobile money potentially helps ameliorate several areas of market failure in developing economies. 12

Reducing Transactions Costs

Mobile money reduces the transactions costs of sending and receiving money, especially given the inadequate and expensive transport infrastructure. Jack and Suri , (2014) observe that in Kenya, where families and social networks are widely-dispersed from internal migration, remittances on average travel 200 km. 13

Transactions costs include the transport costs of travel, for example, to a bank, utility company, or government office; the travel time and the waiting time in long queues; the coordination costs between individuals, between firms and suppliers or customers, and between government and individuals, which can be extensive in time and money lost; and the costs of delays and “ leakages ” through corruption or middlemen, acting like a tax (or complete loss through theft from insecure methods of money transfer) . There is also an opportunity cost to lost money and time. The money could have been invested, spent, or saved; the time could have been spent in productive activities. The automated delivery of cash transfers, wages, social security funds, and private remittances by electronic transfer increases the certainty of the timing of cash receipts, which helps planning. This further reduces coordination costs, the costs of delays, and hence the opportunity costs.

Reducing Asymmetric Information and Improved Transparency

Recording financial transactions creates greater financial transparency and reduces asymmetric information. Asymmetric information and the fixed costs of servicing an account lie at the heart of the failure of the formal banking sector to advance credit to poor customers who lack collateral and financial histories. Moving cash from under the mattress into an electronic account turns it into recorded cash. Every deposit, withdrawal, transfer, or payment transaction through mobile money creates a recorded financial history. Linking algorithmic credit scores and the granting of small loans was discussed above (see box 1 ).

An electronic record of payments potentially protects consumers against theft, fraud, and misinformation. Such protection can reduce transaction costs for consumers and increase the use of business through trust. For example, Radcliffe and Voorhies (2012) note how the “anonymity of cash” may inhibit trust between traders and new vendors. Greater transparency through records can help regulate the service, including the dissemination and posting of information on transactions costs to promote competition. Recorded transfers with appropriate ID documentation (“know your customer”) also facilitates cheaper international remittance transfers.

Changing the Nature of Saving and Increasing Savings through Digital Means

There are several motives for saving. Life-cycle motives compensate for differences in timing between income and expenditure streams, and these include saving for education, leisure, marriage, consumer durables, housing purchases, retirement, and funeral expenses. Precautionary motives (buffer stock saving) reflect the uncertainties of future income and expenditures, and include saving for unemployment, illness, accidents, natural disasters, and risks associated with old age. Finally, there is saving for a bequest motive, to give gifts in one's lifetime or to leave a legacy to heirs. Saving thus helps to allocate consumption over time, and to reduce risk.

For the unbanked poor, their “immersion in physical cash creates considerable frictions in their financial lives” ( Radcliffe and Voorhies 2012 ). Cash-based households have informal saving options, which carry risks of theft or “liquidation”: cash under the mattress; accumulation of assets such as jewelry or livestock; and storing savings with informal savings groups. The loss of savings in this manner is common. Mobile money electronic accounts offer the safe storage of cash, though without the payment of interest.

Another advantage is privacy. Compared with cash receipts, the reduced observability of the timing and sizes of mobile transfers and the accumulated electronic balances could protect savings for the recipient ( Aker et al. 2016 ). Moreover, in an economic psychology literature on how the poor could be encouraged to accumulate savings, for example, the use of “commitment” savings accounts ( Dupas and Robinson 2013 ), mobile money accounts offer a practical template.

Risk and Insurance

Living standards of the poor are at risk of multiple communal shocks including flooding, droughts, pestilence, other natural disasters, sometimes conflict, and medical epidemics, as well as idiosyncratic shocks including theft, damage to the homestead, illness, and death in the family. There are very limited opportunities for insuring against these risks. Formal insurance is typically absent, but family, clan, and network ties can create informal insurance networks, ameliorating such risks by periodic transfers and monitored by trust relationships amongst members of the network ( De Weerdt and Dercon 2006 ). Jack and Suri (2011) suggest several ways by which mobile money can facilitate risk-spreading. For example, the geographic reach of networks can enlarge, while timely transfers of money can arrest serious declines that may be impossible or hard to reverse. The mobile money technology allows small and more frequent transfers of money that allow a more flexible management of negative shocks. Thus, informal insurance networks may function more effectively. In turn, more efficient investment decisions can be made, improving the risk and return trade-off. Where mobile money develops sufficiently to allow access to micro-insurance (see box 1 ), there is potentially an additional buffer against negative shocks.

Incomplete Property Rights, Changing Family Dynamics and Changing Social Networks

Women or minority groups may face limitations in their opportunities and their access to property, an aspect of inequality often resulting in more widespread economic inefficiencies. Mobile money could change bargaining power within the family. Greater privacy may influence both inter-household allocations ( Jakiela and Ozier 2016 ) and intra-household allocations ( Duflo and Udry 2004 ). If the nature of expenditure by gender differs ( Chattopadhyay and Duflo 2004 ), there could be welfare changes in the household ( Aker et al. 2016 ).

Little research has been done on network formation or dissolution, or on migration and remittance decisions using network data ( Chuang and Schechter 2015 ). Mobile money could change the nature of social networks. The cohesion of a network could be strengthened or weakened. The size of networks could be expanded with the greater geographical reach of the transfer mechanism. Morawczynski and Pickens (2009) note the greater autonomy of rural Kenyan women as they can more easily solicit funds from their husbands and other contacts in the city. The reduced transactions costs of remittances might create a more liberal attitude to migration from the homestead ( Jack and Suri 2011 ), though distant migrants are also less observable and accountable. Johnson (2014) stresses the continued importance of rotating credit schemes for perpetuating trust and coordination in communities. There is evidence of substitution away from these schemes due to mobile money ( Mbiti and Weil 2016 ), but also evidence that the schemes themselves use the mobile money transfer and storage mechanism ( Wilson, Harper, and Griffith 2010 ).

Improving other Aspects of Economic Efficiency

The combination of better communication and coordination with mobile phones and instantaneous mobile payments could improve business planning and efficiency. Indeed, mobile payments facilitate trade. Access to credit, informally and through banking services linked to mobile money, can improve investment decisions. Improved risk sharing and cheaper, secure, long-range remittances can expand the scope of labor decisions to encompass higher-risk but higher-return occupations, or migration to higher-return labor markets ( Suri and Jack 2016 ). There could be better allocation of savings and labor within the household and in businesses, and more efficient investment decisions affecting agriculture and business, and education and skills. Returns to investment could rise, with a feedback to greater savings.

“Perhaps the ‘holy grail’ of demand side data is the impact question. How can we understand whether branchless banking services are making a positive difference in client's lives?” McKay and Kendall (2013) .

The rapid global growth of payments, transfers, and international remittances speaks of mobile money providers satisfying a demand for financial services not previously adequately met. This revealed preference suggests a net welfare improvement. Moreover, positive externalities imply a larger total than private benefit, as greater connectedness in the system occurs with each adoption. But are empirical studies able to measure economic benefits, as well as local if not system-wide externalities?

Given its novelty, few academic studies have examined the economics of mobile money. The bulk of empirical work employs survey data at the household- or firm level. To reach robust conclusions on the economic benefits, the bar is set very high for empirical analysis. First, it is important to analy z e the appropriate data , but often this is hard to achieve. Second, there are considerable methodological challenges in the empirical work, so that results need to be carefully assessed, and not taken at face value. An analytical typology table summarizes the empirical studies ( table 1 ). A more in-depth analysis of the studies is presented in Aron (2017) .

A typology of Micro-empirical Studies on the Economic Effects of Mobile Money

Source : Constructed by the author from sourced papers in column 1.

Notes : 1. Disentangle technology/service: Some RCT studies are able to disentangle the mobile money services delivery from ownership of a mobile phone by providing new phones to both treatment and control groups, or by considering only participants with a mobile phone number. Other studies achieve this by introducing a dummy for ownership of a mobile phone into regressions. 2. Definition of M-money usage: For the unwary, there are definitional ambiguities using both telecoms and self-reported data, see section on Challenges for Data. If individuals own multiple, valid SIM cards with different providers, this will exaggerate users. If registered customers are inactive (and globally two thirds of registered accounts are inactive with a generous 90 day definition), this will exaggerate the participation. On the other hand, there is undercounting of overall usage where unregistered customers intensively use an over-the counter service, as in South Asia.

Challenges for Data

Definitional ambiguities could cause mis-counting when measuring mobile money “usage”. If the precision of the variable is compromised, measurement bias is introduced into regressions (see table 1 , column 1). Using the number of mobile money accounts or the number of registered customers may induce multiple counting of the same individual if several accounts are held with different providers. If registered customers are inactive (and globally two-thirds of registered accounts are inactive with a generous 90-day definition), this will exaggerate the true participation (see figure 4 ). Where unregistered customers intensively use the service, as in over-the-counter (OTC) services, overall usage will be underestimated.

Registered and Active Total Accounts

Source : Data from the GSMA State of the Industry report ( 2017 ).

Some data are unobservable. Empirical regressions will be mis-specified when omitting hard-to-measure variables linked to mobile money, such as spillover learning effects in the community, and technological and quality changes. Important

“observables”, such as education (where quality is not assessed) and wealth are typically poorly measured in household surveys, which may exacerbate the biases.

Institutional and political regime changes also affect the uptake of mobile money. For example, adoption is enhanced with more liberal registration requirements below a low threshold of use. In Côte d'Ivoire, the cessation of conflict and onset of greater growth and stability from 2012 was a key to driving mobile money adoption ( Pénicaud and Katakam 2014 ). There are likely to be shifts over time in the relevance of particular determinants, for example, cheaper, more capable smartphones widen access and ownership. Shifts can be proxied by carefully-dated dummy variables; interaction of these dummies with explanatory variables introduces non-linearities and tests whether the effects of the variables alter with regime changes.

Data may be proprietorial, and it may be difficult to design surveys optimally in advance. Against these difficulties, if privacy concerns can be overcome, new access to a rich seam of “big” data on the administrative mobile money transactions from both businesses and individuals presents an enormous research opportunity. Mobile money transactions data could have a wealth of potential applications of which four examples follow: to help forecast hard-to-gauge household assets and expenditure that otherwise rely on self-reported data (this has been done using mobile phone data, see Blumenstock, Cadamuro, and On 2015 ); to derive proxies for migration patterns from geotagged data ( Blumenstock 2012 ); to link GPS data with administrative data to examine price discrimination schemes ( Economides and Jeziorski 2016 ); and to explore evolving social networks with changing remittances ( Aron 2017 ; Aker and Blumenstock 2015 ).

Challenges for Empirical Methods

The quantitative empirical work on mobile money falls into two categories: studies which assess the determinants of the adoption of mobile money (i.e., where a proxy for usage of mobile money is the dependent variable) and studies of the effects of mobile money on micro-economic outcomes (i.e., where usage of mobile money is not the dependent variable). Examples of the latter include whether mobile money promotes improved risk-sharing, food security, consumption, business profitability, saving, and effective use of cash transfers.

Research on mobile money faces two “selection” problems, raising the problem of endogeneity in empirical analysis. 14 The “roll-out” of mobile money by MNOs and their agents may not be random if they select into areas on the basis of household and village characteristics. For instance, there will be an upward bias on the effect of mobile money on consumption if the wealth of a village determines agent selection into that village (and that wealth is not controlled for in regressions). It is difficult to disprove self-selection by the agents toward more profitable locations. Several authors contend there is little statistical correlation between agent “roll-out” and household observable characteristics that might have been associated with future outcomes; but they use partial correlates only, which is not decisive. In Jack and Suri (2014) , such bivariate correlations between agent density at 1 km, 2 km, or 5 km and a range of observables also include location-by-time and rural-by-time fixed effects. 15 But this is rather different from trying to explain agent density with a full range of the variables and all relevant interaction effects to prove it is exogenous or “unpredictable”. Moreover, it does not rule out correlation between agent roll-out and unobservables or poorly-measured observables (such as wealth) that also affect outcomes.

One factor suggesting that roll-out may have been non-random is that Jack and Suri (2014) themselves suggest the following: “. . .many of the agents had business relationships with Safaricom prior to the advent of M-PESA, and about 75 percent report sales of cell phones or Safaricom products as their main business.” As Aker and Blumenstock (2015) imply for the prior telecom infrastructure, “. . . decisions regarding expansion of ICT infrastructure and ICT-based programs are typically driven by private sector or policy criteria.” Thus, even if the bias is likely to be low for Kenya, there may be greater selectivity biases in countries such as Niger, Tanzania, and Uganda, with relatively less developed technological infrastructure.

A second selection problem is undisputed: the adoption of mobile money by individuals is influenced by factors both observable (e.g., education, wealth, urban dwelling, and the use of banking services) and unobservable (e.g., susceptibility to risk, community learning spillover effects, and changes in technology preference) that may be correlated with mobile money use.

Given the selection problems, the dominant empirical methodologies are Randomized Controlled Trials (RCT), quasi-experiments with a Difference-in-Differences estimation strategy or the non-parametric method of Propensity Score Matching, and Instrumental Variables (see box 2 ). The choice amongst methods is not uncontroversial. The methods have differing degrees of success in dealing with heterogeneity at the individual or household level. 16 A consideration is whether results can be “scaled-up” or “transported” to allow generalization to other contexts. Since institutional structures, regulation and demand patterns differ across countries, generalizations of evidence need to be made cautiously (e.g., generalizability may depend on the extent and quality of the agent network). Econometric modelling difficulties imply that the conclusions drawn are often suggestive only.

Common in medical research, RCT was little used in economics before 2003, and has generated heated debate. This critique is pertinent to the reliability and generalizability of mobile money RCT studies. An RCT evaluates whether a specific, controlled change has a discernible impact on a treated group relative to a control group. RCTs focus on small interventions that apply in certain contexts so that inferences for other settings, or even scaling up based on the results, may be invalid. Identifying a causal connection in one situation might be specific to that trial and not a general principle; even the direction of causality can depend on the setting. Deaton (2010) argues that there are actually two stages of selection. In the first, a group is chosen from the entire population that will in the second stage be randomly divided into the treated and control groups. The first stage is not random, but may be determined by convenience or politics, and therefore may not be representative of the entire population. Deaton and Cartwright (2016) further argue that randomization does not guarantee that the treatment and control groups are identical except for the treatment, that is, it does not guarantee that other causal factors are balanced across the groups at the point of randomization. a The studied populations in RCTs are typically very small, so an outlier in the experimental group can have a large distortionary effect. Further, the trial or intervention itself ( Gillespie 1991 ), and the nature and quality of information provided about the intervention, can affect behavior. Standard errors are often erroneously computed and spurious inferences are made, as t-statistics for estimated average treatment effects from RCTs do not in general follow the t-distribution.

A second approach, more widely-used in mobile money research, tests specific theoretical hypotheses using a Difference-in-Differences (DD) estimation, which mimics an experimental approach by comparing differences in the changes of a control and a treated group after an intervention (here, the adoption of mobile money). The restrictive assumption is made that in the absence of the intervention, the average change in the outcome for the affected and control groups would have been the same. This is the “parallel or common trends” assumption. The DD estimates typically derive from an Ordinary Least Squares (OLS) regression for repeated cross-sections or for a panel of data on individuals (appropriately sampled to avoid selection bias) for one or more periods before and after an intervention. A dummy variable is included for the intervention and a set of control variables. The method has the appeal of simplicity, and when the interventions are approximately random, conditional on the time and location fixed effects, and also on household fixed effects in the context of household panels, it can reduce the (time- invariant ) endogeneity problems from comparing heterogeneous individuals. b What remains is time- variant , unobserved household heterogeneity. This may be partially mitigated with appropriate controls for time-variant household characteristics (demographics, for instance) and location-by-time fixed effects (accounting for only part of the time- variant , unobserved heterogeneity, since these dummies average over households in a location). c Further problems arise when the intervention is not random, when the linear assumption under OLS is inappropriate, and from serial correlation problems exaggerating levels of significance in standard errors when several years of data are involved ( Bertrand, Duflo, and Mullainathan 2004 ). One useful test of the DD strategy is the placebo test; it uses data from prior periods before the intervention, and the DD is redone aiming for a close-to-zero placebo effect for the included intervention.

Several mobile money studies present supplementary evidence from Propensity Score matching methods. These methods mimic characteristics of an RCT in the context of an observational (or non-randomized) study, using non-parametric rather than regression techniques to estimate the effects of an intervention (e.g., use of mobile money) on outcomes between treated and control groups. Where baseline characteristics of treated subjects often differ systematically from those of untreated subjects, Propensity Score matching can match samples of subjects who are as similar as possible on observed (pre-treatment) characteristics. Differences in post-treatment outcome variables between the matches are averaged and are attributed to the treatment. There are two crucial assumptions for the validity of the technique. There should be no hidden bias from unobserved heterogeneity and the criteria for adequate balance should be clear and satisfied. However, conditioning on the Propensity Score need not balance unmeasured covariates; and even the balance-checking between measured co-variates is problematic because the criteria for adequate balance are ill-defined (see Hill (2008) on the “rampant lack of good practice”, and Austin (2011) ).

IV can be used for consistent estimation when correlation between explanatory variable/s and the error term is suspected. An endogenous variable is replaced by the predicted value from a set of instruments that are strongly correlated to the explanatory variable (informative or strong), but uncorrelated with the errors (valid or exogenous). Finding credible exogenous instruments for mobile money usage is a challenge. Several instruments have been used in the mobile money empirical literature but statistical tests tend to find them weak, which may introduce bias. d Instruments based on agent density and network connectivity assume that the roll-out of mobile money and network coverage itself was “random”.

See a non-technical version at: http://voxeu.org/article/limitations-randomised-controlled-trials , Nov. 2016.

A dummy variable is included for every household or entity (bar one entity).

A national time effect is a common effect across time experienced by all regions , for example, from macro-fluctuations. But disaggregating to two regions, North and South say, where North is less affected by drought, then interacting both regional dummies with time allows their differential response over time to be captured. With location-by-time fixed effects (without a national time effect), there is a location (e.g., district, region, or country) dummy for each year (bar one location and one year).

Instruments used for mobile money usage ( table 1 ) are as follows: the log of the distance to the closest agent and the number of agents within 5 km of the household ( Jack and Suri 2014 ), the distance to and cost of reaching the nearest mobile money agent ( Riley 2018) , and the log of the distance to the nearest mobile money agent ( Munyegera and Matsumoto 2016a ); the fraction of respondents in the sub-location registered with M-Pesa ( Demombynes and Thegeya 2012 ) and the proportion of households using mobile money and for those owning a mobile phone at the village level ( Kikulwe, Fischer, and Qaim 2014 ); household-specific mobile phone network connectivity and the size of the information exchange network of the household ( Murendo and Wollni 2016 ); and 2006 survey responses (before M-Pesa was introduced) about riskier, slower, and more costly transfer methods ( Mbiti and Weil 2016 ).

Many studies fail to “disentangle” the adoption of the technology (the phone) from adoption of the service (mobile money) it provides ( Aker et al. 2016 ). How and whether the different studies address this to reduce bias is explicitly clarified in table 1 (column 4). Whether clustered standard errors are reported ( Bertrand, Duflo, and Mullainathan 2004 ) is noted in column 3 of table 1 .

To explore the factors that determine the adoption of mobile money (i.e., where a proxy for usage is the dependent variable), Probit or Tobit regressions or OLS regressions are commonly used. The principal empirical problem is the identification of causal relationships. This encompasses biases introduced by poorly measured determinants, omitted observable variables, and omitted unobservables. Examples of hard-to-measure unobservables are the following: spillover effects; technological and quality changes of the handset and services; the quality of agents and trust in the system; and the effects of advertising campaigns and incentives to register. 17 , 18 Non-linearities are crucial in adoption empirics (e.g., adoption can be catalyzed by the cessation of conflict), but are typically ignored. Network effects also matter since a critical mass of users and a critical mass of reliable agents fosters sustainable adoption.

Given these challenges, it is unsurprising that studies of adoption in different countries have been conducted by non -economists focused largely on qualitative aspects, or have examined mobile money adoption correlations with firm and household surveys ( Aker and Mbiti 2010 ). 19 These studies find that adopters of mobile money are more likely to be younger, wealthier, better educated, have a bank account, own a mobile phone and reside in urban areas. One convincing econometric study has supported these links ( Munyegera and Matsumoto 2016a ) and deserves attention; this panel study removes time-invariant household heterogeneity with household fixed effects and some time-variant household heterogeneity with location-by-time dummies in a panel context in rural Uganda. 20 These authors include many individual controls (e.g., control for ownership of a mobile phone, distance to the nearest mobile money agent and a migrant worker in the family) further helping to reduce endogeneity. 21 The authors find no gender effect or age effect for rural adopters, but distance to the nearest mobile money agent proved important, as did education and wealth; both the dummies for the ownership of the phone and the migrant worker are significant (all with a 1% significance). It is still possible that there is some time-variant household heterogeneity that is not controlled for, as location-by-time dummies only address an average over households in a location. 22

Private Mobile Money Transfers and Risk Sharing

Amongst the most convincing analyses of the impact of mobile money are the panel data studies using a Difference-in-Differences approach that explore how mobile money has fostered improved risk-sharing amongst informal networks after large shocks. The proposed mechanism operates via lower transaction costs (compared to alternatives) for money transfer, influencing the size, frequency, and (sender) diversity of domestic remittances. The intervention is a negative shock, and such shocks are probably random. 23 The focus is not on the direct effect of mobile money usage on outcome variables like consumption, but rather on the interaction of mobile money usage with the shock (while controlling for household characteristics to interact with the shock). This puts less emphasis on the endogeneity of the mobile money usage dummy. The best of these studies fully exploit the panel data to remove sources of unobserved time-invariant household heterogeneity using household fixed effects (see box 2 ), include location-by-time dummies and rural-by-time dummies to help control for time- varying heterogeneity according to location or the rural-urban divide, and (mostly) include appropriate controls.

All the reviewed risk-sharing studies disentangle the impact of the mobile phone technology from the transfer mechanism, either by considering only participants with a mobile phone number (though this introduces a new selection criterion), or by introducing a dummy for ownership of a mobile phone into the regressions.

A sophisticated study by Blumenstock, Eagle, and Fafchamps (2016) uses a Difference-in-Differences approach to analyze the transfer of airtime: the authors call it a “rudimentary form of mobile money” but it is not convertible for cash. These authors exploit the random timing and location of earthquakes in Rwanda in a natural experiment to identify covariate economic shocks. 24 Their study relies solely on administrative telecoms data and lacks survey measures of welfare or wealth. 25 The link between risk-sharing and money transfer is instead implied, given the consistency between observed patterns of transfers and the characteristics of their theoretical models of reciprocal risk sharing. All regressions include a shock dummy and time fixed effects. Location fixed effects in regional-level regressions are replaced by recipient fixed effects in individual-level regressions, and by a fixed effect controlling for the average intensity and direction of transfer flows between two users in dyadic regressions. In extended regressions these authors allow for heterogeneity between individuals and different types of sender-recipient pairs, and cross the characteristics with shock dummies (see table 1 ).

Blumenstock, Eagle, and Fafchamps (2016) find, perhaps surprisingly, that as well as geographical proximity, transfers to victims near the epicentre after the Lake Kivu earthquake of 2008 are determined by a past history of reciprocity between individuals, and the transfers decrease in the wealth of the sender and increase in the wealth of the recipient. The opposite would be obtained in the case of charity or altruism. There are possible selection issues. Selection is induced because wealth itself determines the ownership of phones in Rwanda in 2008 ( Blumenstock and Eagle 2012 ). Further, the wealth of the recipient is likely be correlated with the size of his or her geographical network. Ideally, the differences in such networks should be controlled for, as airtime does not in this sense have the same utility in times of disaster for the wealthy and the poor.

A path-breaking study by Jack and Suri (2014) exploring risk sharing and mobile money finds total consumption of Kenyan mobile money users is unaffected by a range of negative (self-reported) income shocks, while that of non-users drops by 7% (with 10% significance). 26 The effect is more evident for the bottom three quintiles of the income distribution. A similar result is found when isolating the impact of health shocks on total consumption. 27 A Difference-in-Differences approach is applied to a panel specification controlling for household fixed effects, location-by-time dummies, and rural-by-time dummies. There is a dummy for a negative shock to income in the last six months, and a dummy for an M-Pesa user in the household, and the two dummies are crossed to test whether M-Pesa users are better able to smooth risk. An included vector of controls (though not including wealth, see table 1 ) is crossed with the shock dummy to help control for correlations of M-Pesa with observables that might help smooth risk.

For Tanzanian mobile money users, a very similar set-up by Riley (2018) takes matters a stage further by examining the potential beneficial spillover effects (local externalities) of mobile money to the village community (which includes non-users) following an aggregate shock (either a self-reported shock such as droughts or floods, or a measure of rainfall deviations from a long-term mean, see table 1 ). 28 The regressions include a dummy for mobile money use by an individual in a village, and one for the proportion of mobile money users in a village, so that there are three interaction effects with the shock dummy, including its interaction with the vector of controls. Unlike in Jack and Suri (2014) , wealth, expected to be time-varying, is here included as a control.

Riley (2018) finds that there are spillover effects in the absence of a shock, as mobile money users share remittances with the village, resulting in per capita consumption of everyone in the village increasing. After an aggregate shock, however, households using mobile money benefit from an 8% to 14% increase in consumption (with 5% significance) compared with non-users, cancelling the effect of the negative shock on users; but there are no spillover effects to the community of non-users. The benefits to users and to communities (in the absence of a shock) are found to be highest in rural areas and to decrease sharply with distance to the nearest mobile money agent. The included district-by-time dummies are important in helping to control for heterogeneity from the self-selection into districts by mobile money services providers, for localized spillover effects, and for unobservable differential effects of rainfall (e.g., for different occupations by district).

All three studies conduct placebo tests supporting the common trends assumption of the DD specification. In Riley (2018) , Propensity Scoring was used to try to match users and non-users with similar characteristics, confirming results. Attempts by both Riley (2018) and Jack and Suri (2014) to apply the IV technique (see box 2 ) and instrument the usage dummy and its interaction with the shock are less successful, typically with weak instruments based on agent rollout data such as agent density (see box 2 ). The IV regressions do not contradict the conclusions, but in Riley (2018) , although a Sargan-Hansen test determines the instruments are valid (exogenous), they are found by Cragg-Donald Wald F statistic tests to be statistically weak, which may potentially introduce a large bias. The former test is missing in Jack and Suri (2014) .

Using data from a survey of nearly 7,700 M-Pesa agents, Jack and Suri (2014) also compare consumption responses in reduced-form panel regressions with fixed effects, substituting “access to an agent” for M-Pesa usage, and claim that the results reinforce their conclusions. However, the crucial assumption of exogeneity of the agent density proxy rests only on bivariate correlations, discussed critically above.

It remains possible that time-variant household heterogeneity (e.g., changing risk preference or changing technology preference) may still confound the results. One specific example of time-variation in characteristics would be where in the first wave of the panel, a fifteen-year old is not in work, but by the second wave, three years later, she is working, which affects her ability to purchase a mobile phone and use mobile money. It would be important to control properly for age structure in this case. More difficult to deal with is systematic unobserved heterogeneity from interaction effects. If there are missing interaction effects from time-varying unobservables or time-varying excluded observables (e.g., wealth) that could help households to smooth risk, then the effect of M-Pesa in smoothing consumption could be exaggerated. For instance, there could be an upward bias if a household that is wealthier in the second period is better placed to withstand a negative income shock; or if households wealthier in the second period than the first tend to experience smaller negative income shocks.

Mobile Money Transfers and Welfare

Far less satisfactory are the (non-RCT) welfare studies reviewed, where results are generally judged unreliable by this survey. Endogeneity problems for the usage dummy are center stage, and the use of instrumentation and other methods to mitigate it by removing as many sources of heterogeneity as possible are not always convincing.

Of the six studies, only three disentangle the impact of the mobile phone technology from the transfer mechanism by including a dummy for ownership of a mobile phone into regressions: Munyegera and Matsumoto (2016a) , Murendo and Wollni (2016) , and Sekabira and Qaim (2016) . One cross-sectional study faces serious problems of controlling for unobserved heterogeneity ( Murendo and Wollni 2016 ). Two panel studies use inappropriate linear specifications that are likely to introduce heavy biases ( Sekabira and Qaim (2016) and Kikulwe, Fischer, and Qaim (2014) ), see discussion in Aron (2017) . A fourth study employs propensity scoring with a very small cross-sectional sample, but is subject to unobserved heterogeneity ( Kirui, Okello, and Njiraini 2013 ). The full critical analyses of these studies can be found in Aron (2017) , and details are summarized in table 1 .

The two remaining studies use panel data. Of these two, one fully exploits Ugandan panel data to control for heterogeneity where possible (see table 1 ), and claims an increase of 9.5% (with 5% significance) in the monthly real per capita household consumption for mobile money users ( Munyegera and Matsumoto 2016a ). The Difference-in-Differences specification requires the mobile money intervention to be random, which is questionable. Their IV regression to address this problem shows the above coefficient in the regression for consumption increasing four-fold , which casts doubt on the results. Similar to Jack and Suri (2014) , the authors rely on bi variate correlations only to validate the agent density-based instrument. Using fixed effects regressions, the authors find a similar coefficient for food consumption as for total consumption, but greatly higher coefficients for non-food. Given the ambiguous results, propensity score methods are applied to try to match comparable households, and weighted regressions are run for total and food consumption. This recovers a coefficient of around 7% (at the 5% level) for overall consumption, but the coefficient for food consumption is poorly measured. Too little information is given to properly evaluate the method, however (see box 2 ).

The other panel study, by Suri and Jack (2016) , argues strongly for a causal role for mobile money on welfare. 29 The effect of mobile money in Kenya is explored for categories of outcomes, measured in 2014 (see table 1 ). Unlike the other studies in this sub-section, these authors use the change in agent density between 2008 and 2010 to proxy or substitute for mobile money usage (i.e., they are not using agent density as an instrument in an IV regression). 30 By pre-dating the proxy relative to 2014 outcomes, the authors hope to make their proxy exogenous. There are two problems with this. First, the measure may not be highly correlated with later usage (which is like having a weak instrument in an IV regression). Second, the crucial assumption of exogeneity of the agent density proxy rests on bivariate correlations conducted in Jack and Suri (2014) . That being said, placebo tests support the common trends assumption of the DD specification.

To estimate the marginal effect of an increase in agent density for females, a gender dummy and the change in agent density are crossed. The change in agent density is also crossed with household (or individual) characteristics to rule out cases where the gender effect was in fact driven by these other characteristics.

Suri and Jack (2016) do not use household fixed effects or location-by-time dummies, but control only for location fixed effects—upon which a great deal then rests to try to mop up household heterogeneity. There are controls for age and gender, but controls such as dummy for ownership of a mobile phone, household physical and financial wealth, education, and possession of a bank account are excluded. Their analysis is at its most convincing in a differenced specification for consumption (their table 1 ), which at least then effectively excludes household time-invariant fixed effects through differencing (the level regressions are likely to have considerable unexplained heterogeneity). Nevertheless, even in the differenced specification, time-varying heterogeneity from unobservables (and omitted wealth) may still introduce bias. With these caveats in mind, we present their results for consumption. These authors find that for households using mobile money, consumption growth for male-headed households was negative, while that of female-headed households was positive and statistically significant. They suggest that the latter could be driven by increased labor or capital income, or by transfers between individuals with different propensities to consume. They draw implications for the reduction of poverty (affecting 2% of Kenyan households), and shifts in occupations out of farming, particularly for female-headed households. However, if there is unobserved heterogeneity of the type discussed above, for example, if wealth which is not controlled for is correlated with mobile money services, then they may be over-estimating the reduction in poverty.

Of the few RCT studies reviewed, see table 1 , some deal with very small transfers and small and specialized samples, and results are not easily generalizable. Two papers exploring the impacts of public or employer mobile money cash or wages transfers are Aker et al. (2016) and Blumenstock et al. (2015b) . Both identify cost savings from reduced transactions costs for the disbursing party. But there are different results for the recipient: there are cost savings in Aker's study based in Niger, and possible cost increases in the Blumenstock et al. study in the more insecure environment in Afghanistan. Both studies disentangled mobile money delivery from ownership of a mobile phone, providing new phones to treatment and control groups.

The impressive RCT study on household welfare by Aker et al. (2016) finds improvements in household welfare after drought for the recipients of cash transfers through mobile money accounts in Niger, one of the world's poorest countries. Intra-household bargaining power for women was promoted and their productivity improved through reduced transport costs, and reduced travelling and queuing time. 31 Recipients were more likely to cultivate and market cash crops conventionally grown by women, and had fewer depleted durable and non-durable assets. Household and child diet diversity was 9% to 16% higher among households who received mobile transfers, mostly due to increased consumption of beans and fats (1% significance level), and children consumed one-third more of a meal per day (5% significance level). These authors emphasize that the mobile money “infrastructure” has to be working well to reap the benefits. Repeating such RCT studies across many locations, cultures, continents, and time periods may help reinforce the conclusions and generalizability.

Given the short time period of observation and the small sample size, the Blumenstock et al. (2015b) study, which was able to distinguish changes in the saving behavior of recipients of wage transfers in Difference-in-Differences estimates of the treatment effect, was not able to find improvements in welfare indicators such as consumption and self-reported satisfaction.

Analyses of Savings Behavior

There are several qualitative studies with localized implications for saving behavior. For instance, Wilson, Harper, and Griffith (2010 ) describe how members of informal savings groups in Nairobi find it cost- and time-effective to move their cash (especially with larger savings) into a group M-Pesa account each week from the deposit collector's own account. Further, Jack and Suri (2011) find that by 2009, 90% of early adopters used M-Pesa for saving (amongst other savings instruments and use of cash) for reasons of improved security, greater privacy, increased ease of use, reduced transactions costs, and precautionary saving against emergencies.

Three non-RCT studies encompassing a variety of techniques all suggest the beneficial influence of mobile money on reported savings by method, and on saving flows ( table 1 ). Two of these studies use cross-sectional survey data ( Demombynes and Thegeya 2012 and Munyegera and Matsumoto 2016b ), and one makes a balanced panel of locations, not individuals ( Mbiti and Weil 2016 ). None of these studies disentangles the technology from the service it provides by controlling for the ownership of a mobile phone. Attempts to instrument the mobile money dummy are not successful in these studies, but an approach employing the residual of an adoption regression by Munyegera and Matsumoto (2016b) is supportive, though in a cross-sectional context. No robust and conclusive results are reached, therefore. There are serious concerns with how the saving flow is measured and from the implications of the use of log specifications (see details in Aron (2017) ).

Probit regressions for saving by Demombynes and Thegeya (2012) with various controls ( table 1 ), find reported saving by any method is more likely for older individuals who are male, rural, married, and with higher levels of education, reported income, and wealth. With these controls, and instrumenting for M-Pesa usage, M-Pesa users are 20% more likely to report having savings (1% significance). The instrument (the fraction of respondents in the sub-location registered with M-Pesa) averages over individuals within locations, and eliminates only some unobserved district-level heterogeneity. This caveat suggests that the result is indicative only. The authors also apply IV estimation to the log of average monthly saving (a flow) on similar controls and with the same instrument (see table 1 ). The coefficient for M-Pesa usage is not statistically significant. It is unclear whether the endogeneity is severe and the instrument is so successful in dealing with it that mobile usage is not relevant to saving, or whether it is simply a poor instrument for M-Pesa usage.

A related exercise for Uganda using Probit regressions for reported saving yields no significant variables at the 1% significance level, save for the mobile money usage dummy ( Munyegera and Matsumoto 2016b ). The specification is not comparable to that of Demombynes and Thegeya (2012) , which included log income (highly significant), wealth quintiles, and marital status for a far larger survey ( table 1 ). Whether the significance of mobile money usage for Uganda is indeed important or whether the coefficient is biased strongly upwards as it proxies for unobservables is unclear. The log of annual saving (a flow) is modelled in Tobit regressions, with similar controls. 32 Two approaches are adopted to help address endogeneity (though not the IV approach). A residual from a first-stage Probit regression for mobile money adoption is added to the Tobit, and is significant at the 1% level. The coefficient on the mobile money usage dummy remains fairly stable, and is positive and significant, which is a supportive test. Second, to reduce observable (time-invariant) household heterogeneity, propensity-score matching is applied (though with scant information on methods used and robustness). These authors run OLS regressions weighted by the propensity score with various controls ( table 1 ), but nothing proves significant except the mobile money usage dummy and the value of assets (at the 5% level). The authors suggest this is because heterogeneity has been successfully removed and suggest a role for mobile money in encouraging saving. The conclusions require the proverbial “large pinch of salt” because despite the authors’ heroic attempts, in cross-section it is very difficult to control for unobserved heterogeneity, and the propensity result is also subject to unobserved heterogeneity concerns (see box 2 ).

A potentially interesting finding from the quantitative work of Mbiti and Weil (2016) is that adopting M-Pesa reduces both the use of informal savings groups and the need to hide cash in secret places. These authors use a first-differenced IV regression for saving methods with various controls ( table 1 ), the differenced specification removing biases due to any time-invariant unobservables. However, it is difficult to draw firm conclusions as the set of instruments used is not intuitive (see Aron (2017) ); and biases might arise from correlation with unobserved, time- varying characteristics of households.

Two RCT studies were the only saving studies that disentangled the mobile technology from the service it provides ( table 1 ). One RCT experimental study ( Batista and Vicente 2016 ) uses cross-sectional data and narrows the type of population tested in its selected sample; it is subject to the problem of interpreting a treatment effect when the intervention depends also on the type of training information provided. Both aspects limit the generalizability of the finding that mobile money increases the willingness to save, though the narrowing of selection helps deal with heterogeneity. A second RCT panel study controlling for individual and survey wave fixed effects, based in Afghanistan ( Blumenstock et al. 2015b ), was applied to a small and specialized sample. Increased usage of mobile savings differed by the prior banking status and size of salary of recipients, and liquidity preference and savings withdrawal rose with perceptions of physical insecurity. However, recipients had to incur the costs of finding liquid agents (where adequate mobile network and agent coverage actually existed), and some had privacy concerns for security reasons. Again, the results are suggestive but not generalizable.

One cross-country study tries to relate “enabling” regulation to the usage of mobile money for 35 countries. Gutierrez and Singh (2013) use self-constructed ( de jure ) regulatory indices in a logit regression controlling for both country characteristics and individual (micro-) characteristics. 33 , 34 By using location (country) fixed effects to reduce omitted variable bias, these authors are unable to include the indices themselves, but only their interaction with individual characteristics. 35 The interaction effects nevertheless yield some plausible insights. A regulatory framework that supports interoperability appears to promote higher usage among the poorest. Stronger consumer protection appears to reduce usage by the poorest, perhaps through raised costs, while amongst the educated, greater consumer protection promotes usage. But heterogeneity remains present in the cross-section, and the direct effect of regulation could only be tested if a panel of Global Findex usage data should become available.

The main contribution of this survey has been to explore the channels of economic impact and to critically survey a new body of economic research in order to answer the following question: Are empirical studies able to measure the economic benefits and local if not system-wide externalities? As a reality check for policy-makers, there is an important role for micro-studies in evaluating the often optimistic assumptions underlying macro-studies that link digital finance and economic growth and inequality. These include assumptions about the barriers to adoption, the welfare impact, the uptake of diversified services including credit, and the government's tax take. For instance, a highly optimistic study by McKinsey (2016) applies a proprietary general equilibrium macroeconomic model to macro-data for seven countries and extrapolate the results globally for all emerging market countries; these authors predict that adoption and use of digital finance (banking in general) could increase the GDP of all emerging economies by 6%, or $3.7 trillion, by 2025.

The survey has distilled lessons for improved practice in the empirical analysis of mobile money. Studies should demonstrate that they take the data issues seriously, including correctly measuring the usage of mobile money, or else providing caveats. It is important to disentangle phone ownership from usage of its services, such as mobile money. The survey suggests that studies do grapple with unobserved heterogeneity but often not sufficiently. The wary policy-maker should give the greater weight to micro-studies using balanced panel data , and which apply their considerable potential advantages for control of time-invariant and some time-variant (e.g., by location) heterogeneity (see box 2 ). Ideally, these should include appropriate controls for potentially time-variant household characteristics (e.g., demography, wealth , having a migrant worker in the family, and being formally banked) and location-by-time dummy proxies. Such a panel approach is probably “as good as it gets” in terms of ameliorating biases from unobserved heterogeneity. Some residual time-variant unobservable heterogeneity may still confound results, but in shorter time periods the bias is likely to be small. In areas where mobile money is fairly new, panel survey data collection should be encouraged. Controlling for heterogeneity and finding exogenous instruments in cross-sectional studies is a heroic exercise: these studies are likely to be compromised and unreliable.

Finding credible exogenous instruments for the endogenous mobile money usage measure in instrumental variable (IV) methods has proved highly challenging. Most are based on agent density and network connectivity, assuming the “random roll-out” of mobile money, and of network coverage. Statistical F tests often find the instruments weak, leading to potentially biased results. An increasing trend is to present propensity score analysis to reinforce the results when IV results prove ambiguous. However, more detail and clarity on evaluation and assumptions is required given the debate and controversy in the literature, so that the propensity score application is transparent and not a black box result.

Given drawbacks with all the techniques, it would be most satisfactory if studies could apply and contrast a range of techniques. 36 Applying a best practice approach to panel data both with and without fixed effects can ascertain the size and direction of the bias of OLS methods. The bias may be positive or negative; authors need to consider the direction of the bias, since then OLS methods can provide useful upper or lower bounds on estimates. Not controlling for unobserved heterogeneity and a lack of instrumenting or weak instruments probably results in an upward bias of the importance of mobile money for the level of consumption or saving. But, if looking at interactions with a negative shock, there is more likely to be a bias to zero; hence, the micro-studies could be under-stating the absolute size of the beneficial effect of mobile money on risk-sharing. 37 And while Suri and Jack (2016) characterize the risk-sharing result as more short-term in nature, if illness and death are prevented by improved insurance of this type, then there are long-term implications as well. With a range of techniques, the potential biases of IV methods and of the propensity score matching can also be ascertained. Where there is an under-statement of the bias, this qualitatively strengthens policy conclusions from noisy micro-studies.

Another problem, universally neglected by the surveyed studies, is non-constant parameters, e.g., because of spillover effects and technological improvements. By its nature, the evolution of mobile money entails regime changes. These shifts introduce potential non-linearities that need to be tested for in both micro- and macro-work. The changes could result in earlier estimates being an underestimate of later effects. Structural breaks can mean the findings of studies can be hard to generalize. The micro-studies ignoring spillover effects may be picking up only part of an effect, and hence may be a poor guide to the economy-wide effect of a policy.

Robustness testing and testing of the validity of instruments (their strength and exogeneity) are patchy over the studies. 38 Researchers should try harder to illuminate those dimensions where welfare improvements are greatest by checking for differences in responses between more and less affluent households and other types of non-linearity (e.g., urban versus rural, by occupation, and by education level), and by gender ( Suri and Jack (2016) ). Areas for future research, where there has been little quantitative work as yet, include building on Riley (2018) in exploring community spillover effects, and on Jack, Ray, and Suri (2013) and Blumenstock et al. (2016) on little-studied network effects, as well as on timely investigation of the new products of digital credit ( Francis, Blumenstock, and Robinson 2017 ) and insurance through mobile money channels.

Focusing on the studies that apply best practice, the most convincing evidence is from the panel studies of Riley (2018) and Jack and Suri (2014) , suggesting that mobile money fosters improved risk-sharing amongst informal networks in Kenya and Uganda after large shocks, through lower transaction costs of domestic transfer. On mobile money adoption, the Ugandan panel study of Munyegera and Matsumoto (2016a) deserves attention, supporting widespread qualitative evidence that education and wealth matter, but these authors found no gender or age effect for rural adopters. Generalizability of all these results may depend on the extent and quality of the agent network. Though all the non-RCT studies claim the beneficial influence of mobile money on reported savings (by saving method), and on saving flows, the results are compromised by a lack of balanced panel data and appropriate instruments, and no robust and conclusive results can be reached. RCT studies in Mozambique and Afghanistan suggest that saving did not increase, though the saving method switched to mobile money; these studies use small and specialized samples and are probably not generalizable. Far less satisfactory are the (non-RCT) welfare studies reviewed, where results are generally judged unreliable by this survey. A Ugandan panel study suggests an improvement in consumption for mobile money users ( Munyegera and Matsumoto 2016a ); the IV regression casts doubt on the claimed result, but it is supported by a propensity score analysis. A panel study for Kenya by Suri and Jack (2016) is at its most convincing in a differenced specification for consumption; consumption growth for male-headed households was negative and of female-headed households was positive with access to mobile money, but the result is tempered by probable bias from the limited control of heterogeneity. The RCT study by Aker et al. (2016) found the receipt of cash transfers through mobile money accounts promoted intra-household bargaining power for women and their productivity in Niger, with reduced transactions costs. Child nutrition improved and increased diet diversity for the household, with fewer depleted durable and non-durable assets than for control groups. The generalizability of this study is uncertain and depends on a functioning agent network. Repeating such RCT studies across many locations, cultures, continents, and time periods may help reinforce the conclusions and generalizability. 39

Digital finance is one of few areas where there has been a real revolution in services and leapfrogging over deficient traditional infrastructure. However, improved access to financial services is compromised by economic obstacles, significant amongst which are corruption, a lack of electricity generation, and appalling road infrastructure. 40 Complementary action is required to address such problems. The micro-studies show how difficult it is to quantify outcomes accurately and to extrapolate from individual studies of different countries, scaling up the effects to make policy pronouncements. Given the lack of complementary inputs, there could be strong returns to scale in the short-run from mobile money, but not in the long-run, given the constraints. On the other hand, the micro-benefit established by several studies could be multiplied greatly through spillover effects in the presence of well-functioning general infrastructure and transparency (lack of corruption)—especially if mobile money itself reduced corruption.

Atkinson (2015) has argued that economic inequality is often aligned with differences in access to, use of, or knowledge of information and communication technologies. This author stressed that researchers, firms, policymakers and governments have the possibility to shape the direction and path of technological change. Aid agencies, other donors, charitable foundations, and international agencies have played a key role in the beneficial growth of mobile money and the associated financial inclusion ( Aron 2017 ). Creative coalitions and the investment in multi-stakeholder partnerships can prompt deeper change, learning, and practical action. An important application is for academic research on mobile money. Poor quality data and sub-optimal data collection and analysis severely compromise the conclusions that can be reached from empirical work. A concerted attempt by donors, regulators such as central banks, the regulated MNOs, and academics could harness the appropriate data for timely best practice analysis. If anonymizing procedures were accepted, then the benefits from research analysis using anonymized disaggregated data could be reaped. The survey has highlighted the best practice techniques that when applied to empirical analysis could reach more reliable conclusions and bolster the case for significant government and donor support, and commercial investment.

Janine Aron is a Senior Research Fellow at the Institute for New Economic Thinking at the Oxford Martin School, University of Oxford, UK; Centre for the Study of African Economies, Department of Economics, Manor Road Building, Oxford OX1 3UQ. This work was supported by the Gates Foundation (grant number MQRYDE00), the Open Society Foundations and the Oxford Martin School. Special thanks go to John Muellbauer (Nuffield College, Oxford University). The author also thanks Chris Adam (Oxford University), Tony Atkinson (Oxford University), John Duca (Federal Reserve Bank of Dallas, USA), Colin Mayer (SAID Business School, Oxford University), Ggombe Kasim Munyegera (IGC), David Porteous (Bankable Frontier Associates), Emma Riley (Oxford University), Federico Varese (Oxford University) and Sebastian Walker (IMF) for their helpful comments.

The phenomenal growth since 2007 of Kenya's M-Pesa system has brought mobile money to international prominence (“M” is for mobile, and “pesa” is Swahili for money), see box 1 .

Prior to mobile money in Kenya, there were fewer than three bank branches per 100,000 people. Saving was mostly in the form of cash under the mattress. Domestic transfers used scarce post office branches, or insecure intermediaries such as bus-drivers. International remittances were received expensively via money transfer companies or Hawala.

Rotating savings and credit associations and cooperatives address the problem of asymmetric information, allowing small accumulated sums by groups to help individual members spread risk. The related micro-credit movement offers collateral-free loans to marginalized borrowers at near-market interest rates. However, assessing such micro-finance in a long-running evaluation in India, Banerjee et al. (2015) conclude it has had limited success.

The FICO scores in the United States, decisive in 90% of U.S. lending decisions by 2015, are created in a similar manner (Financial Times 2015).

The official remittances statistics would improve as well as the economic management of remittances. In highly dollarized economies (see Corralesa et al. (2016) for the extent of this phenomenon in Africa), mobile money through lower transactions costs may reduce currency substitution, thereby deepening their financial systems.

On the merits of a cashless economy, including fighting corruption and money-laundering, see Rogoff (2016) .

Registration aids financial inclusion toward formal sector products. By contrast, an OTC transaction is conducted through an agent's account on behalf of the customer.

Regulation of mobile money is discussed in detail in Aron (2017) , especially prudential regulation by the central banks; see also Di Castri (2013) .

Third party merchants are not “agents” in a strict legal sense of having the legal authority to act for the service provider—this depends on the local regulation requirements.

Remittances to developing countries are projected to reach US$444 billion in 2017. The true size of remittances, including unrecorded flows, is likely to be significantly larger (World Bank Migration and Development Brief no. 27).

The following authors have examined aspects of the economics of mobile money: Mas and Klein (2012) , Jack, Suri, and Townsend (2010) , Jack and Suri (2011) and Weil, Mbiti, and Mwega (2012) .

See Karlan et al. (2016) on market failure in a more general context of financial services.

Mobile money halves the cost of sending compared to Western Union, and is about a third lower than the postal bank or bus delivery cost, excluding transportation or time costs (see also Morawczynski (2009) ).

An endogeneity problem in econometrics occurs when an explanatory variable is correlated with the error term as a result of simultaneous causality, omitted variables, and/or measurement error. There are several statistical methods that aim to correct the resulting bias in the regression estimates (see box 2 ).

The log of wealth is one of the observables and there is weak evidence for a correlation with wealth.

Heterogeneity refers to variation across individual units of observation, some of which can be observed (e.g., age and education), and some of which is difficult to measure (e.g., changing technological preferences). Thus, omitted heterogeneity is an omitted variable, and hence a kind of endogeneity (see box 2).

On agent quality, see Balasubramanian and Drake (2015) .

Work in progress by Blumenstock and co-authors explores the negative effects of violence on the adoption of mobile money in Afghanistan. Available at: http://www.jblumenstock.com/ .

Not on adoption per se but with implications for adoption, Economides and Jeziorski (2016) match administrative transactions data with GPS data in Tanzania, quantifying motivations for usage, such as willingness to pay to avoid walking with cash or to avoid storing money at home to alleviate criminal risk.

These authors take two approaches, and find similar results, using first a Probit regression, and then a linear probability model with fixed effects. The mobile money “usage” measure in the dependent variable does not match the preferred definition of active (90-day) users, however, and this could bias the results.

Note that agent density may not be exogenous.

The results of a related study on adoption by Weil, Mbiti, and Mwega (2012) should be regarded as suggestive, and of supporting correlations, see Aron (2017) and table 1 . The study cannot control for individual fixed effects and suffers from an omission of controls.

This is a reasonable assumption if unexpected shocks are reported, and not systematically correlated with most household characteristics. Though unlikely in a short time frame, if shocks are correlated with changes in unobservable household characteristics then they would not be random.

Idiosyncratic shocks affect individuals or households; covariant shocks affect groups of households, communities, regions, or even entire countries.

The average amount transferred over the two-month period is small at around US $1; the total additional influx (explicit transfers to all 15 cellular towers within 20 km of the epicentre) measured about US $84.

Food consumption, however, appears to be equally well-smoothed by both users and non-users in the sample.

User households can finance health care expenditures from remittances without compromising other consumption, but non-users must reduce non-medical spending for this; see also Suri, Jack, and Stoker (2012) .

A broadening of networks is likely ( Chuang and Schechter 2015 ), but Riley (2018) more restrictively assumes the sharing social network is village-wide, rather than across villages by lineage, for instance, and that it is constant over time.

“…Thus, although mobile phone use correlates well with economic development, mobile money causes it,” ( Suri and Jack (2016) , my italics).

Agent density is defined as the number of agents within 1 km of the household. This change variable approximates to the level of agent density in 2010, as agent density would have been low in 2008.

Cash-transfer recipients were temporarily able to conceal the arrival of the transfer, increasing bargaining power.

This technique serves to censor observations at zero as the lower limit since households not using financial services will not yield an outcome.

De facto rather than de jure regulations should enter an index, so that it is the quality or performance of the existing regulations that matter rather than merely their existence ( Aron 2000 ).

The data are from Global Findex, and regulatory categories favor openness and certainty ( Porteous 2009 ).

The indices may be correlated with omitted country characteristics; most possible instruments have the same problem.

Several authors apply a range of techniques, for example, Riley (2018) .

For instance, if wealthy households are more likely to adopt mobile money but have less need of the insurance than the poor when a negative shock strikes or are less likely to experience a large negative shock than the poor, then there is a bias toward zero.

Riley (2018) , Blumenstock et al. (2016) , and Jack and Suri (2014) are amongst rarer examples that test robustness, and present clear assumptions and caveats for the techniques.

The challenge of scalability for RCT studies is addressed in Banerjee et al. (2016) . Deaton and Cartwright (2016) recommend a route to precision through prior information (which is excluded by randomization) and controlling for those factors that are likely to be important. Then, they argue, there is a better chance of “transporting” results more generally to other contexts.

Special Report: Business in Africa, The Economist, April 2016.

Aker J. C. Mbiti I. M. . 2010 . “Mobile Phones and Economic Development in Africa.” Journal of Economic Perspectives 24 ( 3 ): 207 – 32 .

Google Scholar

Aker J. C. Blumenstock J. E. . 2015 . In “The Economics of New Technologies in Africa.” In The Oxford Handbook of Africa and Economics , Vol. 2 , edited by Monga C. Lin J. Y. , 353 – 71 . Oxford : Oxford University Press .

Google Preview

Aker J. C. Boumnijel R. McClelland A. Tierney N. . 2016 . “Payment Mechanisms and Anti-Poverty Programs: Evidence from a Mobile Money Cash Transfer Experiment in Niger.” Economic Development and Cultural Change 65 ( 1 ): 1 – 37 .

Aron J. 2017 . “‘Leapfrogging’: A Survey of the Nature and Economic Implications of Mobile Money.” CSAE Working Paper Series 2017-2, Centre for the Study of African Economies, University of Oxford .

Aron J. 2000 . “Growth and Institutions: A Review of the Evidence.” World Bank Research Observer 15 ( 1 ): 99 – 135 .

Atkinson A. B. 2015 . Inequality: What Can Be Done? Cambridge, MA : Harvard University Press .

Austin P. C. 2011 . “An Introduction to Propensity Score Methods for Reducing the Effects of Confounding in Observational Studies.” Multivariate Behavioral Research , 46 (3) : 399 – 424 .

Balasubramanian K. Drake D. F. . 2015 . “Service Quality, Inventory and Competition: An Empirical Analysis of Mobile Money Agents in Africa.” Working Paper 15–059 , Harvard Business School , Cambridge, MA .

Banerjee A. Duflo E. Glennerster R. Kinnan C. . 2015 . “The Miracle of Microfinance? Evidence from a Randomized Evaluation.” American Economic Journal: Applied Economics 7 ( 1 ): 22 – 53 .

Banerjee A. Banerji R. Berry J. Duflo E. Kannan H. Mukerji S. Shotland M. Walton M. . 2016 . “From Proof of Concept to Scalable Policies: Challenges and Solutions, with an Application.” Working Paper 16-11, Department of Economics, Massachusetts Institute of Technology , Boston, MA .

Batista C. Vicente P. C. . 2016 . “Introducing Mobile Money in Rural Mozambique: Evidence from a Field Experiment.” CSAE Conference 2016: Economic Development in Africa, Oxford. Available at: https://editorialexpress.com/conference/CSAE2016/program/CSAE2016.html#75 , 22 March, 2016 .

Beck T. Cull R. . 2013 . “Banking in Africa.” Policy Research Working Paper 6684 , World Bank Washington, DC .

Bertrand M. Duflo E. Mullainathan S. . 2004 . “How Much Should We Trust Difference-in-Differences Estimates?” Quarterly Journal of Economics 119 ( 1 ): 249 – 75 .

Blumenstock J. 2012 . “Inferring Patterns of Internal Migration from Mobile Phone Call Records: Evidence from Rwanda.” Information Technology for Development 18 ( 2 ): 107 – 25 .

Blumenstock J. E. Eagle N. . 2012 . “Divided We Call: Disparities in Access and Use of Mobile Phones in Rwanda.” Information Technology and International Development 8 ( 2 ): 1 – 16 .

Blumenstock J. E. Cadamuro G. On R. . 2015a . “Predicting Poverty and Wealth from Mobile Phone Metadata.” Science 350 ( 6264 ): 1073 – 6 .

Blumenstock J. E. Callen M. Ghani T. Koepke L. . 2015b . “Promises and Pitfalls of Mobile Money in Afghanistan: Evidence from a Randomized Control Trial.” The 7th IEEE/ACM International Conference on Information and Communication Technologies and Development (ICTD '15), 15–18 May, 2015, Singapore .

Blumenstock J. E. Eagle N. Fafchamps M. . 2016 . “Airtime Transfers and Mobile Communications: Evidence in the Aftermath of Natural Disasters.” Journal of Development Economics 120 : 157 – 81 .

Chattopadhyay R. Duflo E. . 2004 . “Women as Policy Makers: Evidence from a Randomized Policy Experiment in India.” Econometrica 72 ( 5 ): 1409 – 43 .

Chuang Y. Schechter L. . 2015 . “Social Networks in Developing Countries.” Annual Review of Resource Economics 7 : 451 – 72 .

Corralesa J.-S. Imamb P. A. Weberc S. Yehouea E. . 2016 . “Dollarisation in Sub-Saharan Africa.” Journal of African Economiesn 25 ( 1 ): 28 – 54 .

De Weerdt J. Dercon S. . 2006 . Risk-Sharing Networks and Insurance Against Illness . Journal of Development Economics 81 ( 2 ): 337 – 56 .

Deaton A. 2010 . “Instruments, Randomization, and Learning About Development.” Journal of Economic Literature 48 ( 2 ): 424 – 55 .

Deaton A. Cartwright N. . 2016 . “Understanding and Misunderstanding Randomized Controlled Trials.” NBER Working Paper No. 22595, National Bureau of Economic Research , Cambridge, MA .

Demombynes G. Thegeya A. . 2012 . “Kenya's Mobile Revolution and the Promise of Mobile Savings.” Policy Research Working Paper No. 5988 , World Bank , Washington, DC .

di Castri S. 2013 . “Mobile Money: Enabling Regulatory Solutions.” GSMA, February https://www.gsma.com/mobilefordevelopment/wp-content/uploads/2013/02/MMU-Enabling-Regulatory-Solutions-di-Castri-2013.pdf .

Duflo E. Udry C. . 2004 . “Intrahousehold Resource Allocation in Cote d'Ivoire: Social Norms, Separate Accounts and Consumption Choices.” NBER Working Papers 10498, National Bureau of Economic Research , Cambridge, MA .

Dupas Pascaline Robinson J. . 2013 . “Why Don't the Poor Save More? Evidence from Health Savings Experiments.” American Economic Review 103 ( 4 ): 1138 – 71 .

Economides N. Jeziorski P. . 2016 . Mobile Money in Tanzania . Mimeo. Stern School of Business , New York (forthcoming, Marketing Science) .

Federal Reserve . 2016 . Survey of Consumers and Mobile Financial Services 2016 . Washington : Board of Governors of the Federal Reserve System .

Francis E. Blumenstock J. Robinson J. . 2017 . “Digital Credit In Emerging Markets: A Snapshot of the Current Landscape and Open Research Questions.” Center for Effective Global Action, Bill and Melinda Gates Foundation. http://www.digitalcreditobservatory.org/uploads/8/2/2/7/82274768/dco_landscape_analysis.pdf .

Gillespie G . 1991 . Manufacturing Knowledge, A History of the Hawthorne Experiments . Cambridge : Cambridge University Press .

Greenacre J. Buckley R. . 2016 . “Using Trusts to Protect Mobile Money Customers.” Singapore Journal of Legal Studies (July) 59 – 78 .

GSMA (Global System for Mobile Communications) . 2017 . State of the Industry report Mobile Money . www.gsma.com/mobilemoney .

Gutierrez E. Singh S. . 2013 . “What Regulatory Frameworks Are More Conducive to Mobile Banking? Empirical Evidence from Findex Data.” Policy Research Working Paper 6652 , World Bank , Washington, DC .

Hill J. 2008 . “Comments on ‘A Critical Appraisal of Propensity-Score Matching in the Medical Literature between 1996 and 2003’ edited by P. Austin.” Statistics in Medicine 27 (12) : 2055 – 61 .

Jack W. Ray A. Suri T. . 2013 . “Transaction Networks: Evidence from Mobile Money in Kenya.” American Economic Review 103 ( 3 ): 356 – 61 .

Jack W. Suri T. . 2011 . “Mobile Money: The Economics of M-Pesa.” NBER Working Paper No. 16721 , National Bureau of Economic Research , Cambridge, MA .

Jack W. Suri T. . 2014 . “Risk Sharing and Transactions Costs: Evidence from Kenya's Mobile Money Revolution.” American Economic Review 104 ( 1 ): 183 – 223 .

Jack W. Suri T. Townsend R. . 2010 . “Monetary Theory and Electronic Money: Reflections on the Kenyan Experience.” Federal Reserve Bank of Richmond Economic Quarterly 96 ( 1 ): 3 – 122 .

Jakiela P. Ozier O. . 2016 . “Does Africa Need a Rotten Kin Theorem? Experimental Evidence from Village Economies.” Review of Economic Studies 83 ( 1 ): 231 – 68 .

Johnson S. 2014 . “Compelling Visions of Financial Inclusion in Kenya: The Rift Revealed by Mobile Money Transfer.” Working Paper No. 30, Bath Papers in International Development and Well-Being, University of Bath .

Karlan D. Kendall J. Mann R. Pande R. Suri T. Zinman J. . 2016 . “Research and Impacts of Digital Financial Services.” NBER Working Paper No. 22633, National Bureau of Economic Research , Cambridge, MA .

Kikulwe E. M. Fischer E. Qaim M. . 2014 . “Mobile Money, Smallholder Farmers, and Household Welfare in Kenya.” PLoS One 9 ( 10 ): e109804 . doi:10.1371/journal.pone.0109804 .

Kirui O. K. Okello J. J. Njiraini G. W. . 2013 . “Impact of Mobile Phone-Based Money Transfer Services in Agriculture: Evidence from Kenya.” Quarterly Journal of International Agriculture 52 : 141 – 62 .

Mas I. Klein M. . 2012 . “A Note on Macro-Financial Implications of Mobile Money Schemes.” Working Paper No. 188 , Frankfurt School of Finance & Management , Frankfurt .

Mbiti I. Weil D. N. , 2016 . “Mobile Banking: The Impact of M-Pesa in Kenya.” In National Bureau of Economic Research: African Successes: Modernization and Development , edited by Edwards S. Johnson S. Weil D. , 247 – 93 . Chicago : University of Chicago Press .

McKay C. Kendall J. . 2013 . “The Emerging Landscape of Demand-Side Data in Branchless Banking.” CGAP . http://www.cgap.org/blog/emerging-landscape-demand-side-data-branchless-banking .

McKinsey . 2016 . “Digital Finance for All: Powering Inclusive Growth in Emerging Economies.” McKinsey Global Institute , McKinsey & Company . https://www.mckinsey.com/featured-insights/employment-and-growth/how-digital-finance-could-boost-growth-in-emerging-economies .

Morawczynski O . 2009 . “Exploring the Usage and Impact of ‘Transformational’ Mobile Financial Services: The Case of M-Pesa in Kenya.” Journal of Eastern African Studies 3 ( 3 ): 509 – 25 .

Morawczynski O. Pickens M. . 2009 . Poor People Using Mobile Financial Services: Observations on Customer Usage and Impact from M-Pesa . Washington, DC : Consultative Group to Assist the Poorest .

Murendo C. Wollni M. . 2016 . “Mobile Money and Household Food Security in Uganda.” Global Food Discussion Papers No. 76 . Georg-August-University of Göttingen , Göttingen .

Munyegera G. K. Matsumoto T. . 2016a . “Mobile Money, Remittances, and Household Welfare: Panel Evidence from Rural Uganda.” World Development ( 79 ): 127 – 37 .

Munyegera G. K. Matsumoto T. . 2016b . “Banking on the Cell-Phone: Mobile Money and the Financial Behaviour of Rural Households in Uganda.” CSAE Conference 2016: Economic Development in Africa, Oxford. https://editorialexpress.com/conference/CSAE2016/program/CSAE2016.html#75 , 22 March, 2016 .

Pénicaud C. Katakam A. . 2014 . State of the Industry 2013 . GSMA Mobile Financial Services for the Unbanked . GSMA . https://www.gsma.com/mobilefordevelopment/wp-content/uploads/2014/02/SOTIR_2013.pdf .

Porteous D. 2009 . “Mobilizing Money through Enabling Regulation.” Innovations 4 ( 1 ): 75 – 90 .

Radcliffe D. Voorhies R. . 2012 . “A Digital Pathway to Financial Inclusion.” (December 11, 2012). Available at: https://ssrn.com/abstract=2186926 .

Riley E. 2018 . “Mobile Money and Risk Sharing Against Aggregate Shocks.” Journal of Development Economics 135 : 43 – 58 .

Rogoff K. S. 2016 . The Curse of Cash . Princeton, NJ : Princeton University Press .

Scharwatt C. Katakam A. Frydrych J. Murphy A. Naghavi N. . 2015 . State of the Industry 2014 . GSMA Mobile Financial Services for the Unbanked . GSMA . https://www.gsma.com/mobilefordevelopment/wp-content/uploads/2015/03/SOTIR_2014.pdf .

Sekabira H. Qaim M. . 2016 . “Mobile Money, Agricultural Marketing, and Off-Farm Income in Uganda.” Global Food Discussion Paper No. 82 , University of Göttingen , Göttingen .

Suri T. Jack W. . 2016 . “The Long-Run Poverty and Gender Impacts of Mobile Money.” Science 354 ( 6317 ): 1288 – 92 .

Suri T. Jack W. Stoker T. M. . 2012 . “Documenting the Birth of A Financial Economy.” Proceedings of the National Academy of Sciences 109 ( 26 ): 10257 – 62 .

Veniard C. 2010 . “How Agent Banking Changes the Economics of Small Accounts.” Brief written for the Global Savings Forum, Bill & Melinda Gates Foundation, Seattle, November https://docs.gatesfoundation.org/documents/agent-banking.pdf .

Villasenor J. 2013 . “Smartphones for the Unbanked: How Mobile Money Will Drive Digital Inclusion in Developing Countries.” The Brookings Institution, Issues in Technology Innovation 24 : 1 – 12 . https://www.brookings.edu/research/smartphones-for-the-unbanked-how-mobile-money-will-drive-digital-inclusion-in-developing-countries/ .

Weil D. N. Mbiti I. Mwega F. . 2012 . “The Implications of Innovations in the Financial Sector on the Conduct of Monetary Policy in East Africa.” Working Paper 12/0460, International Growth Centre, London .

Wilson K. Harper M. Griffith M. (eds). 2010 . Financial Promise for the Poor: How Groups Build Microsavings . Kumarian Press .

Email alerts

Citing articles via.

• Recommend to your Library

Affiliations

• Online ISSN 1564-6971
• Print ISSN 0257-3032
• Copyright © 2024 World Bank
• About Oxford Academic
• Publish journals with us
• University press partners
• What we publish
• New features  
• Open access
• Institutional account management
• Rights and permissions
• Get help with access
• Accessibility
• Advertising
• Media enquiries
• Oxford University Press
• Oxford Languages
• University of Oxford

Oxford University Press is a department of the University of Oxford. It furthers the University's objective of excellence in research, scholarship, and education by publishing worldwide

• Copyright © 2024 Oxford University Press
• Cookie settings
• Cookie policy
• Privacy policy
• Legal notice

This Feature Is Available To Subscribers Only

Sign In or Create an Account

This PDF is available to Subscribers Only

For full access to this pdf, sign in to an existing account, or purchase an annual subscription.

To read this content please select one of the options below:

Please note you do not have access to teaching notes, monetary policy and market interest rates: literature review using text analysis.

International Journal of Development Issues

ISSN : 1446-8956

Article publication date: 17 August 2021

Issue publication date: 25 August 2021

This paper aims to examine the relationship between monetary policy and market interest rates. This paper examines the efficiency of interest rate channel used in monetary regulation as well as implementation of monetary policy under low interest rates. This paper examines and reviews the scientific literature published over the past 30 years to determine primary research areas, to summarize their results and to identify appropriate measures of monetary policy to be used in practice in changing economic environment.

Design/methodology/approach

This paper reviews 94 studies focused on the relationship between monetary policy and market interest rates in terms of meeting the goals of macroeconomic regulation. The articles are selected on the basis of Scopus citation and bibliometric analysis. A major feature of this paper is the use of text analysis (data preparation, frequency of terms and collocations use, examination of relationships between terms, use of principal component analysis to determine research thematic areas). Using the method of principal component analysis while studying abstracts this paper reveals thematic areas of the research. Thus, the conducted text analysis provides unbiased results.

First, this paper examines the whole complex of relationships between monetary policy of central banks and market interest rates. Second, this research reviews a wide range of literature including recent studies focused on specific features of monetary policy under low and negative rates. Third, this study identifies and summarizes the thematic areas of all the researches using text analysis (transmission mechanism of monetary policy, efficiency of zero interest rate policy, monetary policy and term structure of interest rates, monetary policy and interest rate risk of banks, monetary policy of central banks and financial stability). Finally, this paper presents the most important findings of the studied articles related to the current situation and trends on the financial market as well as further research opportunities. This paper finds the principal results of studies on significant issues of monetary policy in terms of its efficiency under low interest rates, influence of its instruments on term structure of interest rates and role of banking sector in implementation of transmission mechanism of monetary policy.

Research limitations/implications

The limitation of the review is examining articles for the study period of 30 years.

Practical implications

Central banks of emerging economies should apply the instruments and results of the countries' monetary policies reviewed in this paper. Using text analysis this paper reveals the main thematic areas and summarizes findings of the articles under study. The analysis allows presenting the main ideas related to current economic situation.

Social implications

The findings are of great value for adjusting the monetary policy of central banks. Also, these are important for people because these show the significant role of monetary policy for the economic growth.

Originality/value

Using text analysis this paper reveals the main thematic areas (transmission mechanism of monetary policy, efficiency of zero interest rate policy, monetary policy and term structure of interest rates, monetary policy and interest rate risk of banks, monetary policy of central banks and financial stability) and summarizes findings of the articles under study. The analysis allows defining the current ideas relevant to the monetary policy of developing countries. It is important for central banks because it examines the monetary policy problems and proposes optimal solutions.

• Monetary policy
• Interest rates
• Interest rate channel

Acknowledgements

The authors would like to thank the Associate Professor Ludmila Vinogradova for assistance in translation and Fedor Fedorov for help in text analysis.

Fedorova, E. and Meshkova, E. (2021), "Monetary policy and market interest rates: literature review using text analysis", International Journal of Development Issues , Vol. 20 No. 3, pp. 358-373. https://doi.org/10.1108/IJDI-02-2021-0049

Emerald Publishing Limited

Copyright © 2021, Emerald Publishing Limited

Related articles

We’re listening — tell us what you think, something didn’t work….

Report bugs here

All feedback is valuable

Please share your general feedback

Join us on our journey

Platform update page.

Visit emeraldpublishing.com/platformupdate to discover the latest news and updates

Questions & More Information

Answers to the most commonly asked questions here

• Publications
• Peer Review Reports

Thematic Review on Money Market Fund Reforms: Peer review report

Addressing vulnerabilities in money market funds is a key element of the FSB’s work programme to enhance the resilience of the non-bank financial intermediation sector.

Money market funds (MMFs) are important providers of short-term financing for financial institutions, corporations, and governments. MMFs are also used by retail and institutional investors to invest excess cash and manage their liquidity.

MMFs are subject to two broad types of vulnerabilities that can be mutually reinforcing: they are susceptible to sudden and disruptive redemptions, and they may face challenges in selling assets, particularly under stressed conditions. The prevalence of this liquidity mismatch, which crystallised during the March 2020 market turmoil, may depend in individual jurisdictions on market structures, use, and characteristics of MMFs.

In 2021, the FSB published a report with policy options to address MMF vulnerabilities by imposing on redeeming investors the cost of their redemptions; enhancing the ability to absorb credit losses; addressing regulatory thresholds that may give rise to cliff effects; and reducing liquidity transformation. This peer review takes stock of the measures adopted or planned by FSB member jurisdictions in response to that report, including those jurisdictions’ evidence-based explanation of relevant MMF vulnerabilities and policy choices made. The review does not assess the effectiveness of those policy measures, as this will be the focus of separate follow-up work by the FSB in 2026.

Press Release

27 february 2024 fsb review finds uneven implementation of money market fund reforms, related information, 16 august 2023 thematic peer review on money market fund reforms: summary terms of reference and request for public feedback, 11 october 2021 policy proposals to enhance money market fund resilience: final report.

• Terms and Conditions
• Privacy Notice
• Cookie Notice

Academia.edu no longer supports Internet Explorer.

To browse Academia.edu and the wider internet faster and more securely, please take a few seconds to  upgrade your browser .

Enter the email address you signed up with and we'll email you a reset link.

• We're Hiring!
• Help Center

INDIAN STOCK MARKET -REVIEW OF LITERATURE

Related Papers

Bonfring International Journal

In India, the history of capital markets dates back to the 18th century when East India Company securities traded the country. The present study is largely based on the available secondary data. The statistical data regarding growth of the capital markets was available from various websites. Capital markets help to channelize surplus funds into productive use. Generally, this market trades mostly in long-term securities. The important divisions of the capital market are stock market, bond market and primary, secondary markets. Primary markets deal with the trade of new issues of stocks and other securities, whereas secondary market deals with the exchange of existing or previously-issued securities. Our finding is that during the first and second five year plans, the Government emphasized on the development of agriculture and public undertakings. The Public sector undertaking was healthier than Private undertakings, but shares were not listed in the stock exchange. More over controller of Capital Issue (CCI) closely supervised everything. A number of investors were interested to invest their savings in debentures instead of company deposits. We conclude that Capital markets were not well organized and developed during the British rule. But in the present scenario, we find that Capital markets are well developed after the introduction of SEBI. Through provision of long term loans, the capital market brings about effective functioning of various sectors of the economy. A sound and efficient capital market is one of the most instrumental factors in the economic development of a nation.

Dr Shruti Mishra

India had an internet user base of about 462 million as of Dec 2016 and buoyed by Internet penetration in rural areas, the number of web users in India will see a two-fold rise at 730 million by 2020. Despite being the second-largest user base in world, only behind China (650 million, 48% of population), the penetration of e-commerce is low compared to markets like the U.S (266 million, 84%), or France (54 M, 81%), but is growing at an unprecedented rate, adding around 6 million new entrants every month. The industry consensus is that growth is at an inflection point. In India, cash on delivery is the most preferred payment method, accumulating 75% of the e-retail activities. Demand for international consumer products (including long tail items) is growing much faster than in-country supply from authorized distributors and e-commerce offerings. This paper will give the scenario of present and future of e-commerce business taking consideration of demonetization in country.

MANISH SHAW

José G. Vargas-Hernández

BRICS are being considered as “biggest richest innovative countries association”. Their role and contribution are of paramount significance in the present globalize world. Emergence of BRICS as an economic bloc is also seen as an alternative to EU. Keeping the increasing significance of the bloc; the aim of this work is to identify strategies that enable emerging economies of the BRICS group play a bigger role globally, these structural changes have been made in their political and economic reforms, which together with the business sector has been progress scorch significant breakthrough towards international market, since society and organizations are functional, and institutional organizations for the development of a society and the economy. The role of BRICS bank may go a long way in the promotion of development in BRICS in particular and world in general.

International Journal of Recent Advances in Multidisciplinary Research

Raja Mannar Badur

Indian financial market has seen an extraordinary volatility in the last few years. Since the year 2002, Indian market has grown from a much volatile condition to growth phenomena, from a SENSEX point of 5500 in December 2003 to 13,787 in December 2006 and crossed the mark of 20,000 in the year 2007 and again in 2013. Due to various reasons, the stock market has also experienced drastic decline to even less than 8,000 points in 2008. It is not because of only the domestic market but also the international investors. There are many other variables which contribute to the positive growth of the stock market. FIIs investment is considered to be one of the biggest push after the economic fundamentals. There is no doubt that the liberalisation of the FII flows into the Indian Capital Market since 1993 has had a considerable impact on Indian stock market. The present paper is an attempt to explore the FDIs investment behaviour and its relationship with GDP, SENSEX and NIFTY movement. Further, an attempt is made to develop an understanding of the dynamics of the trading behaviour of FDIs and effect on the Indian stock market. The study is covers the period, financial year 2000-2001 to 2016-17on GDP, BSE Sensex and Nifty and FII activity. It provides the evidence of significant positive correlation between FDI activity and effects on Indian Capital Market. The analysis also finds that the movements in the Indian Capital Market are fairly explained by the FDI net inflows

Arvind Kumar

IOSR Journals

Abstract: Financial market is a place which provides a place for investment and helps in enhancing the income in terms of return. The main aim of financial market is to create cash flow in the market, so that individuals can take investment decision without any fear. Every investor would like to get required rate of return with minimum risk. To attain the objective of high return with minimum risk, various instruments, practices and strategies have been devised and developed in the recent past. After privatization and globalization financial market has entered into a new phase of global integration and liberalization. On the one hand integration of the Indian capital market with global market open the boundaries for investment to everyone, which also helps in increasing the cash flow, on the other hand there has increased in financial risk as the frequent changes in the interest rates, currency exchange rate and stock prices. To overcome from the increased financial risk a risk minimizing tool were launched by NSE during the year 2001, and that tool was Derivatives. This study helps in analyzing the facts behind launching of financial derivative by NSE India and how derivatives help in the growth of share market in India. The case will cover introduction, contextual note, various arguments and the results, remaining problems and new ingenuities regarding financial derivatives of NSE India. Key Words: Financial Market, Return, Risk, Globalization, Privatization, Derivatives

Rakesh Gupta

Hypothesis of Market Efficiency is an important concept for the investors who wish to hold internationally diversified portfolios. With increased movement of investments across international boundaries owing to the integration of world economies, the understanding of efficiency of the emerging markets is also gaining greater importance. In this paper we test the weak form efficiency in the framework of random

Prafulla Kumar Sinha

RELATED PAPERS

srilaxmi challagundla , Dr. Abdul Majeeb Pasha. Shaik

eDITOR'S nOTE tHE tEAM

Pramod Yadav

Dipankor Coondoo

Samuel Onyuma, PhD

SSRN Electronic Journal

Manoj Kumar

•   We're Hiring!
•   Help Center
• Find new research papers in:
• Health Sciences
• Earth Sciences
• Cognitive Science
• Mathematics
• Computer Science
• Academia ©2024