Study . | Data . | Method . | Endogeneity & other issues . | Claimed result . |
---|---|---|---|---|
Probit/FE: Zero-1 dummy: for whether household i living in village j in district d uses mobile money services at time period t. Exact definition of “use” unclear. | Uganda Balanced panel of 838 households generated from the 3rd &4th rounds of household and community surveys in Uganda, 2009 & 2012 (RePEAT) project. | Probit regression; and linear probability model with household fixed effects district-by-time dummies; dummy for ownership of a mobile phone; and vector of household characteristics (age (and age squared), gender and education (years of schooling) of household head, dummy for migrant worker in household, distance to nearest mobile money agent, size of household, and household wealth (land size and total assets)). [Robust standard errors] | Household fixed effects and location-by-time dummies are used in a panel context, and many individual controls (including control for ownership of a mobile phone and a migrant worker) reducing potential endogeneity; possibly some household heterogeneity may remain. Yes. Mobile phone dummy used. | Cannot find a gender effect or an age effect for these rural adopters; distance to the agent is important as is wealth; and dummies for ownership of the phone and migrant worker are significant. |
OLS: Zero-1 dummy: for whether an individual uses mobile money; Frequency of mobile money transactions per user. Exact definition of “use” unclear. | Kenya, Tanzania and Uganda Repeated cross-sections. FinAccess data from Kenya (2006 and 2009); Finscope data for Tanzania and Uganda (2006 and 2009). (These are not panel data.) | OLS regressions vector of individual characteristics (dummies for urbanisation and the level of poverty, 3 age cohorts, education (primary/secondary/tertiary), marriage, and gender). [Robust standard errors] | There are endogeneity problems. Omission of measurable controls e.g., banking status, wealth and mobile phone ownership. Unobservables like spillover effects cannot be controlled for. But location-by-time fixed effects were not included for repeated cross-sections to control for (some) time- unobserved regional-level heterogeneity. The results are thus only suggestive. No. | They deduce for all three countries (limited significance in the less well-developed markets of Tanzania and Uganda) that adopters are younger, wealthier, better educated and urban dwellers. Analysis of frequency of mobile money transactions per user, yields similar findings. Cannot find a gender effect. |
OLS: log consumption per capita; DD: Binary variables for investment in agricultural categories (e.g., active farm or pesticides) or business categories (e.g., cattle trading). Treated individuals receive training about a new mobile money product, M-Kesh. | Mozambique Panel data (some analysed as cross-section) generated in rural provinces: Maputo-Province, Gaza, and Inhambane, March 2012 (102 rural Enumeration Areas: 51 locations in 3 regions randomly selected as treatment areas; the residual is control group). Administrative mobile money records combined with household survey data (3 years, 2012–14). [ : rural treatment locations required mCel coverage & 1 or more commercial banks; targeted individuals required a mobile phone number and a migrant family member in Maputo with mobile phone number.] [ : simple average of zero-1 indicators for mix of negative shocks: deaths, job loss, health problems, loss of valuables, agricultural losses.] | Randomized Controlled Trials (RCT). treated individuals receive training about a new mobile money product. OLS regression specification for consumption, comparing differences etc. comparing differences in outcomes for targeted and control individuals for 2013, 2014 and these years pooled. treatment dummy variable; province dummies; year dummies; and individual controls for age and gender. OLS regression specification for consumption and risk sharing, comparing outcomes for a cross-section in mid-2014. treatment dummy variable; a shock index; locational dummies; and individual controls for age and gender. • The shock dummy and M-money dummy are crossed to test if M-money users are better able to smooth risk. OLS Difference-in-Differences (DD) regressions for investment outcomes, comparing outcomes for 2013 and for 2014. treatment dummy variable; locational dummies; year dummies; and individual controls for age and gender. [Clustered standard errors] | The first stage of selection may not be random, and there are other problems of potential heterogeneity (see Deaton's critique, Box 2). Other selection criteria (see LHS) narrow the type of population which reduces generalizability. There is a problem of interpreting a treatment effect when intervention depends also on the type of training information provided (see ). The constructed shock index is misleading as it conflates shocks that raise and those that lower expenditure; a simple average is used. Absence of time-by-location dummies: yet are critical to control for heterogeneous effects across locations of the 2013 flood. They do not cross individual characteristics with the shock index (as in and ). Yes. Only individuals with phone numbers are selected. | No significance for the treatment dummy for consumption in the absence of shocks. The treated group increases consumption in response to a negative shock (e.g., health or funeral expenditures drawing on remittances); the control group has to reduce other expenditure. The negative coefficient for the treatment dummy suggests the treated group is spending less (perhaps because they are sending remittances to relatives or if there is a systematic difference between treated and untreated groups e.g., are poorer). Suggests improving rural households’ welfare as mobile money contributes to household consumption smoothing. No productive effects of remittances: for mobile money users, active farm investment and investment in cattle trading falls significantly, but household ownership of “safe asset” livestock is higher. Interpret as evidence that (informal) insurance from mobile money reduced the incentives for risky investment (given credit constraints). |
DD: three degrees of disaggregation: (i) total gross transfers of airtime received by all users in location at time . (ii) total gross transfers received by user in region at time (iii) total gross transfer of airtime sent to an individual , located in region at time , from another individual . MNO record of pre-paid airtime (a precursor of mobile money) transferred. | Rwanda Panel data. 2005–09, daily primary telecom operator's log of activity (50 billion transactions: calls, text messages, and airtime transfers and purchases), 1.5 million subscribers; 2005 Rwanda Demographic and Health Survey; 2009/2010 phone survey of 1000 individuals on household asset ownership and housing characteristics. | Panel Difference-in-Differences (DD) regressions an earthquake shock (i) shock dummy equal to 1 for location receiving a shock at time and 0 otherwise; time dummies; and location fixed effects. (ii) shock dummy equal to 1 for user in location receiving a shock at time and 0 otherwise; epicentre dummy for user near epicentre at any time; time dummies; and recipient fixed effects. (iii) as in (ii), but replacing the fixed effects by a fixed effect controlling for average intensity and direction of transfer flows between two users. Heterogeneity amongst individuals: add in (ii), the interactions of predicted measures of expenditure (to proxy for wealth) and of social connected-ness with the shock dummy, the epicentre dummy and a dummy capturing the day of a severe shock. Heterogeneity amongst sender-recipient pairs: add in (ii), the interactions of information on the geographic distance between i and j, and the history of transfers between them with the shock dummy, the epicentre dummy and a dummy capturing the day of a severe shock. [Clustered standard errors] | The earthquake shock is exogenous if unpredictable. Potential time variance in location could be tested for with broader location-by-time dummies than the epicentre-by-time dummy. There is imaginative use of fixed effects, and interaction effects with innovative wealth and social connectedness measures and others, to control for types of heterogeneity. There may be selection problems associated with social networks, see discussion in text. Selection is also induced when wealth itself determines the ownership of phones as in Rwanda in 2008, though in a sharing culture some may own only the SIM card and borrow a phone. Yes. Only individuals with phone numbers are selected. | 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 magnitude of these transfers is small in absolute terms. |
DD/IV: log of consumption per capita Households that used mobile money services at least once in the previous year. | Tanzania Panel data. Tanzania National Panel household panel survey (NPS) for 2008–9, 2010–11 and 2012–13, covers 3265 households in 26 districts containing 409 Enumeration Areas: 3 waves of data and a low attrition rate; and Finscope (2013) data. [Treatment groups are villages where mobile money is available.] [ : self-reported aggregate income shocks e.g., droughts or floods; or a constructed measure of rainfall deviations (> 1 standard deviation) from a 40 year mean, expressed as an absolute value.] | Panel Difference-in-Differences (DD) regressions a negative income shock M-money dummy equal to 1 for households that used mobile money services and 0 otherwise; a dummy for aggregate shock; household fixed effects, location-by-time dummies, a dummy for the proportion of mobile money users in a village; and household characteristics. • The shock dummy and M-money dummy are crossed to test if M-money users are better able to smooth risk. • The shock dummy and village M-money dummy are crossed to test if there are spillover effects. • The vector of household characteristics is crossed with the shock dummy. [ a rural dummy, age and education (years) of the household head, the size of household, a dummy for ownership of a mobile phone, some financial indicators, a wealth index constructed using principal component analysis, and a household head occupational dummy.] Instrumental Variables; as above [ distance to and cost of reaching the nearest mobile money agent, and the interactions of each with the shock] Propensity score model Matched users and non-users with similar characteristics. [Standard errors are clustered, village level] | The specification requires the shock to be random. If correlated with changes (given fixed effects) in observable household characteristics, shocks would not be random. A more precise rainfall measure would separate large positive from large negative deviations. Possibly restrictive to assume the social network for sharing is only village-wide, and constant. Time- unobservables are controlled for by household fixed effects. Village-by-time dummies average over individuals in villages, and eliminate some (not all) unobserved, village-level, time-varying heterogeneity (e.g., self-selection into villages by providers; localised “herd” effects and learning spillover; differential effects of rainfall by occupation across districts). But time-varying, unobservable, household heterogeneity may remain. The IV results do not reject their findings; but although the instruments are statistically exogenous, they were found to be weak, introducing bias. Yes. A mobile phone dummy used. | This study examines potential beneficial spillover effects of mobile money to the village community (which includes non-users) following an aggregate (co-variate) shock. The rainfall (or other) shock causes a drop in consumption of 6–11% for all households without mobile money use. For villages where at least one person uses mobile money, average village consumption is 4–10% higher (1% significance level and robust to the inclusion of fixed effects): signals positive spillover effects of mobile money to non-users in the village; For households with mobile money users (fixed effects included), their consumption is unaffected. There is no spillover benefit to the community for non-users. But for households using mobile money, consumption increases by 8–14% (at a 5% significance level), cancelling the effect of the negative shock, helping these households to smooth consumption. Benefits to both the users and community are highest in rural areas and decrease sharply with distance to the nearest mobile money agent. |
DD/IV: log annual per capita consumption for a household at a particular location and time. M-Pesa registrations from the telecommunications firm (at least one per household). | Kenya Panel data. Household panel survey conducted in Sep. 2008 (3000 HHs), Dec. 2009 (2017 of these HH) and Jun. 2010 (1595 HHs from 2008 sample, but 265 not interviewed in 2009). They construct a 2-period balanced panel of 2282 (or 2017 + 265) HHs, with attrition rate of ∼24%, controlling for round (time) dummies in regressions. Excluding Nairobi lowers the attrition rate to ∼18%. A March 2010 survey of nearly 7700 M-Pesa agents, who also reported when they began business. [ : negative shock could be covariate like a drought; or idiosyncratic like severe illness, job loss, fire, livestock death, and harvest or business failure.] | Panel Difference-in-Differences (DD) regressions a negative income shock M-money dummy equal to 1 for an M-Pesa user in the household in survey and 0 otherwise; a dummy for negative shock to income in last 6 months; household fixed effects; location-by-time dummies; rural-by-time dummies; and household characteristics. • The shock dummy and M-Pesa dummy are crossed to test if M-Pesa users are better able to smooth risk. • The vector of household characteristics is crossed with the shock dummy. [ : household demographics, household head years of education and occupation dummies (for farmer, business operator and professional), use of financial instruments (bank accounts, savings and credit cooperatives and rotating savings and credit associations), and a dummy for cell phone ownership.] Note that wealth is not included. Reduced form regressions as above, but without crossing vector of household characteristics with the shock dummy. • Simply substitute “access to an agent” for M-Pesa usage. Instrumental Variables as above. [ distance to the closest agent, the number of agents within 5 km of the household, and the interactions of each with the shock] [Standard errors are clustered, village level] | The specification requires the shock to be random. If correlated with changes (given fixed effects) in observable household characteristics, shocks would not be random. Self-reported wealth is not in the vector of characteristics. Time- unobservables are controlled for by household fixed effects. Location-by-time dummies average over individuals within locations, eliminating some (not all) unobserved, -level, time- heterogeneity. Ditto the inclusion of rural-by-time dummies. But time- unobservable heterogeneity may remain; also, if there are missing interaction effects from time-varying unobservables (e.g., wealth) that could help households to smooth risk, this may bias the role of M-Pesa in smoothing consumption. Their claim for validity of instruments relies on lack of systematic correlation between agent density and observable household characteristics that may help households to smooth risk (their Table 6C uses only correlations, however; see text on more comprehensive testing). There may still be correlation with observables or poorly-measured observables (e.g., wealth) that may help households to smooth risk. F tests suggest instruments are not weak; no tests are reported for whether they are exogenous. They do successfully conduct placebo tests. Yes. A mobile phone dummy used. | For Kenyans with access to mobile money, total consumption is unaffected by negative income shocks, while the consumption of non-users drops by 7% (significant at a 10% level). The effect is more evident for the bottom three quintiles of the income distribution. Same result for the impact of health shocks on total consumption; but food consumption is equally well-smoothed by users and non-users. Transactions cost savings mean users are better able to smooth consumption following negative income shocks, from the greater frequency, geographical diversity and size of mobile money remittances. Evidence suggests higher expenditure after negative shocks, rather than “stable” consumption, perhaps on repairs and medical treatment. The IV regressions reinforce the conclusions: improved access to agents improves a household's ability to smooth risk. The agent roll-out proved statistically to be uncorrelated with observables including self-reported wealth (though using only correlates, see LHS); in principle instrumenting could help to control for endogeneity. |
OLS: the outcome (measured in 2014) for household (or individual) i in location j for three categories of variable: (i) the log of average consumption per person in a household, the change in this variable, and the level of household poverty rates (consumption pc below $1.25 per day or “extreme poverty”, and below $2 per day); (ii) physical and financial wealth: the log of assets, the log of total financial savings, and presence of a bank account; and (iii) occupational choices: farming, business and sales, or secondary occupations. they proxy usage by the change in agent density (i.e., the number of agents within 1 km of the HH) between 2008 and 2010. | Kenya Panel data. Household panel survey conducted across 118 locations, in Sep. 2008 (3000 HHs), Dec. 2009, Jun. 2010, 2011 and 2014 (1608 HHs); the 2011 survey was targeted specifically toward attrited households from earlier rounds; Nairobi was dropped from the sample after 2011 (480 HHs); attrition from the original non-Nairobi sample, 2008–14, was 35%. A March 2010 survey of nearly 7700 M-Pesa agents, who also reported when they began business. | Panel OLS regressions the change in agent density between 2008 and 2010; location fixed effects; a dummy for gender of the household head in household level regressions (or for the individual in individual level regressions); and household (individual) characteristics. • The gender dummy and the change in agent density are crossed to estimate the marginal effect of an increase in agent density for females. • The change in agent density is crossed with household (or individual) characteristics to rule out cases where the gender effect was in fact driven by these other characteristics. [ used in the regressions (measured in 2008): age and age squared of the household head.] [ used in the interaction effect (measured in 2008): (i) for individual regressions: education; (ii) for household level regressions: education, wealth, and a dummy for the household being unbanked (education and wealth are dummy variables for whether the household is below the median value in the sample).] [Standard errors are clustered, location level] | Pre-dating the agent density proxy relative to 2014 outcomes intends to make it exogenous. There are 2 problems. It may be a poor proxy for of mobile money, as usage growth is catalysed 2010–14 (see text on statistics). The exogeneity assumption relies on lack of systematic correlation (using only correlations) with observable household characteristics possibly associated with future outcomes (see text on testing more comprehensively). There may also still be correlation with unobservables or poorly-measured observables (e.g., wealth) that affect outcomes. There is probably considerable unexplained heterogeneity in the regressions. House-hold fixed effects, location-by-time dummies, ownership of a mobile phone, wealth, education and possession of a bank account are excluded. More weight should be placed on the regression of the which serves to remove household fixed effects (though time-varying heterogeneity may still introduce bias). No. | Prior agent density (proxies access to M-Pesa) increased per capita consumption levels (in 2014) and reduced the level of poverty for two measures of poverty (in 2014). Effects are stronger for female-headed households for the levels of consumption and of extreme poverty. Consumption for male-headed households was negative; that of female-headed households was positive and statistically significant. (The result is robust to interactions between changes in agent density and other observable household characteristics.) Mobile money access (prior agent density) cannot explain the (level of) the log of assets. The regression of the log of total financial savings (including mobile money accounts) does not control for mobile phone ownership, wealth, marriage, income, education as in other savings studies, but only for gender, age and age squared of the household head. That said, “usage” promotes saving without a gender effect. With greater mobile money access (prior agent density), fewer report their major occupation as farming, for both genders, and more females report their main occupation to be in business, sales, or retail. The results are interpreted as saying mobile money has increased the efficiency of allocation of consumption over time, allowing allocation of labour to be more efficient, reducing poverty. |
two measures of “food security”: (1) Food consumption: OLS/IV: per capita aggregated food consumption expenditures (monthly per adult equivalents (AE): 7-day recall for regular purchases, 30-day recall for less frequent purchases); (2) Food Insecurity Indexes: OLS/IV: continuous Household Food Insecurity Access Scale (HFIAS), using weights from factor analysis Probit/Probit (IV): binary Food Insecurity Index (constructed on HFIAS data). Households that used mobile money services at least once in the previous year. | Uganda Cross-sectional survey of 482 households in 39 villages in two regions in November and December 2013. | OLS regressions/endogenous treatment effect models (for food consumption or the continuous food security index and treatment variable: mobile money usage dummy) OLS/Instrumental Variables regressions (for food consumption or the continuous food security index and treatment variables: continuous volume or frequency of transfer) Probit and Probit (IV) models (for the binary food security index and all treatment variables) M-money dummy equal to 1 for households that used mobile money services and 0 otherwise; or continuous variables for frequency of use of services or the volume transferred via mobile money. treatment variable; and household characteristics. [ : age, education (years) and gender of household head, household size, ratio of dependents (below 15 & above 65 years) to workforce (16–64 years), adult equivalent, land size, log value of farm equipment, dummy for household member(s) engaged in off-farm income activity, dummy for household-accessed credit, total livestock units, dummy for household ownership of a motorcycle and/or car, distance to output market and district dummies; : “the number of mobile phones owned”; “extension contact” for whether a household accessed information from an extension service; and “group membership” for community learning about agricultural and market information.] [ innovative instruments: household-specific mobile phone network connectivity & the size of the information exchange network of the household. These instruments were created through interviews] [Robust standard errors] | Only one IV result is reported: (i) OLS estimates are relied on; (ii) regression used for mobile money usage; OLS used for frequency of use and volumes transferred; (iii) ordinary probit estimates used. It is possible that the instruments are weak: no critical values are reported e.g., for the Cragg-Donald Wald F statistic. For the reported IV result, the level of significance of M-money dummy is low. The first instrument entails ownership of a phone and proxies for wealth, which may affect food security. The second instrument may be correlated with other information controls in the regression, and may signal a household with good connections and high status, affecting food security. Failure to find appropriate instruments would not legitimate the OLS results. Cross-sectional analyses are highly vulnerable to failure to control for household and village level heterogeneity. Yes. A version of a mobile phone dummy is used. | Mobile money use (10% significance level) increases food expenditure per AE by 9 percentage points; frequency of use and volumes transferred (both with 1% significance) increase food expenditure per AE by 1.9 percentage points and by 1 percentage point, respectively. Farm equipment and livestock units, mobile phone ownership and household size (negative effect), are important co-variates. Mobile money use and the volumes transferred (both with 1% significance) reduce food insecurity by 0.20 index points (1/5th of the standard deviation) and by 0.007 index points, respectively. Land size and ownership of a means of transport and livestock units are significant co-variates. Mobile money use reduces the probability of food insecurity by 10 percentage points (10% significance level). A one-unit increase in the volume of money transferred via mobile phone reduces the probability of food insecurity by 1.2 percentage points (5% significance). Land size, ownership of a means of transport, livestock units and group membership are significant co-variates. |
, DD: log of monthly real per capita household consumption: –Total consumption for a household at a particular location and time; –Disaggregated food, non-food and social expenditure (expenditure on ROSCAs, mutual funds, insurance and churches). ( , not mentioned in article, but see text footnote in ) Exact definition of “use” unclear. | Uganda Balanced panel of 838 households generated from the 3rd &4th rounds of household and community surveys in Uganda, 2009 & 2012 (RePEAT) project. | Panel Difference-in-Differences (DD) regressions the introduction of mobile money services M-money dummy equal to 1 for households that used mobile money services and 0 otherwise; household fixed effects; location-by-time dummies; dummy for household mobile phone possession; and household characteristics. [ household size, log of value of assets and land endowments, age, gender and education level of the household.] Instrumental Variables as above [ the log of the distance to the nearest mobile money agent] Propensity score model Matched users and non-users with similar characteristics. [Robust but not clustered standard errors] | There are issues with zeroes or small numbers in the log specification, see text; this may account for the disaggregated results. Household fixed effects control for all time- unobservables. Inclusion of location-by-time dummies averages over individuals within locations, and eliminates some (not all) unobserved, -level, time- heterogeneity. Thus, time- unobservable heterogeneity may remain. The specification requires agent roll-out to be random, which is questionable. The validity of the instrument relies on lack of systematic correlation between agent density and observable household characteristics that could affect household consumption (they refer to (do not report) only correlations). There may still be correlation with observables or poorly-measured observables (e.g., wealth) that may help households to smooth risk. F tests suggest instruments are not weak; no tests are reported for whether they are exogenous. They do successfully conduct placebo tests. The IV result (where the FE coefficient increases 4-fold) is problematic. Propensity scoring was used, though too little information is given to assess this properly. Yes. Mobile phone dummy used. | FE model: given the adoption of mobile money services, there is a 9.5% (at a 5% significance level) increase in total household per capita consumption; an insignificant coefficient for food consumption (most food is self-farmed); and a greatly higher 20% increase for non-food and 47% increase for social expenditure (both at a 5% significance level). IV model: total per capita consumption increases 4-fold upon adoption of mobile money (but with a 17% standard error). Propensity score methods for comparable households recover a coefficient of around 7% (at a 5% significance level) for overall consumption, but for food consumption remain insignificant. |
, FE/RE: outcome variables: - Total real household income (all net earnings from on-farm and off-farm sources, including remittances); –Per capita consumption; remittances received; –Proportion of coffee sold as shelled green beans allowing entry to higher-value markets; –Average coffee price received by farmers in the respective year. ( ) ( ) Households with at least one member who had a mobile money account and used services at least once in the previous year. | Uganda Unbalanced panel data from survey of smallholder coffee farmers; 2 randomly-selected robusta coffee-growing districts in Central Uganda [Round 1(2012) covered 419 households. Round 2 (2015) addressed a 6% attrition rate and also increased sample to 455 households. Unbalanced panel: 874 observations from 480 households. Mobile money questions only in 2015 Round] [ : per capita value of food and non-food goods & services; food consumption data from 7-day recall; non-food items monthly; all expenditure data converted to daily basis. Off-farm income: salaries, wages & pensions of household, land rents and capital earnings, and net profit from non-agricultural businesses.] | Panel fixed effects and random effects regressions M-money dummy equal to 1 for households that used mobile money services and 0 otherwise; year dummy to control for time fixed effects; dummy for mobile phone use; dummy for participation in certification schemes for sustainability standards; and household/farm characteristics. [ : education (years of schooling), age, and gender of the household head; land owned; value of other productive asset; distance to the next tarmac road; and a district dummy.] [Ordinary standard errors] | Consumption and income results are badly biased as they use inappropriate linear specifications, see text in . Log specifications should have been tested for the remaining two dependent variables, but these regressions are at least interpretable, see RHS. Unbalanced panels may introduce biases. Time fixed effects are included; but location-by-time dummies should also have been included to address potential, unobserved, time-varying heterogeneity at the district level. Yes. Mobile phone dummy used. | This study aims to explore the role of agricultural marketing and off-farm economic activities to promote welfare. We do not report the seriously biased consumption and income results. FE model: for mobile money users, the proportion of coffee sold as shelled beans increases by 19 percentage points (almost doubling), as less cash-constrained farmers are more willing to sell after drying and processing, and can transact with buyers from outside their location; mobile money users receive a 7% increase over the mean prices received by non-adopters through selling more of their coffee as shelled beans and having better access to buyers in higher-value markets. Important covariates in both cases are distance to road and sustainability certification, and additionally for coffee prices, productive assets (e.g., vehicles and transport equipment). |
, FE/RE: outcome variables: –Total real household income (the sum of all net earnings from on-farm and off-farm sources, including remittances); –Remittances received (all transfers from relatives and friends not residing in the household); –Transactions in agricultural input and output markets; and farm profits. ( in ) ( ) Households that used mobile money services at least once in the previous year. | Kenya Balanced panel data for end-2009 and end-2010, focusing on 320 households from banana-growing villages in the Central and Eastern Provinces of Kenya. | Panel fixed effects and random effects regressions M-money dummy equal to 1 for households that used mobile money services and 0 otherwise; year dummy to control for time fixed effects; and household/farm characteristics. [ : farm size (land owned), household size, the gender, age, and education (years of schooling) of the household head, the distance of the household to markets and roads, a ‘high-potential area’ dummy, which takes a value of one for regions with more fertile soils and higher amounts of rainfall, and zero otherwise, and a variable measuring the percentage of households using mobile phones at the village level to capture neighbourhood effects.] [Ordinary standard errors] Instrumental Variables as above [ the proportion of households using mobile money and the proportion of those owning a mobile phone at the village level] Propensity score model Matched users and non-users with similar characteristics. | The results are biased as they use inappropriate linear specifications, see text in . Not including a dummy for mobile money ownership means use of mobile money may be picking up this excluded factor. Location-by-time dummies should have been included to address potential, unobserved, time-varying heterogeneity at the village level. The wealth measure of land size is largely time-invariant over the short period of the study; a broader measure of less illiquid wealth is an essential control which could be time-variant over the sample. The exogeneity of the instruments with respect to income is in doubt, as they may proxy for wealth. Propensity scoring was used, though too little information is given to assess this properly. No. | FE models: the results are seriously biased because of several model misspecifications, see . They suggest that mobile money users have greater household income, higher remittances received, to apply more purchased farm inputs, market a larger proportion of their output, and have higher profits than non-users of this technology. The reported average treatment effects are implausibly large, e.g., a 40% income gain relative to the mean income of non-users, and a 35% profits gain over non-users. |
outcome & input variables: –Household agricultural input use (value of purchased inputs); –Agricultural commercialisation (ratio of the value of sales to the value of total production); –Farm incomes (value of agricultural revenue). Exact definition of “use” unclear. | Kenya Cross-sectional data, from a small survey of 379 multi-stage randomly selected farm households in 3 provinces of Kenya in March–April, 2010. [ : inputs included fertilizer, improved seed varieties, pesticides, and hired labour.] | Propensity score model Match treatment with controls (i.e., users of M-Money with non-users) that are similar in terms of their observable characteristics using 3 matching techniques. The differences in outcome variables between the matches are averaged to obtain the average treatment effect on the treated. [ : gender, age, distance to nearest mobile money agent, distance to nearest bank, household size, asset endowment variables, household non-farm income, current value of assets, land size, education, group membership and regional dummies.] | Biases and heteroscedasticity as in the above two papers, as logs were not used for the unscaled dependent variables, and for the relevant unscaled independent variables. Thus, larger farms or wealthier households are given undue emphasis when taking arithmetic means. At the least, geometric means should have been checked for robustness. Propensity scoring: reduction of the bias by 20% does not eliminate it. Moreover, it is assumed that observed characteristics will be correlated with unobserved characteristics; this is not necessarily the case, and cannot be proved. The generalizability from such a small sample is also in doubt. No. | The results are biased because of model misspecification, see . Propensity Score methods: they find that mobile money transfer services significantly increased the level of annual household input use by $42, household agricultural commercialization by 37% and household annual income by $224. |
OLS: various outcomes of interest (costs, uses of the cash transfer, food security and assets) of individual or household in village. Selected participants (see Col.4.) were given mobile money-enabled mobile phones. | Niger Cross-section or pooled cross-section. Household survey of 1152 recipients in 96 intervention villages: baseline in May 2010, follow-ups in Dec.2010 and May 2011 (main sample: 1082 households in Rounds 2 & 3); village-level survey; anthropometric data on children, for 691 households in May 2011; weekly price data in 45 markets, May 2010 to Jan.2011. [Most regressions use the Dec.2010 household data, straight after the transfer. When available, data for Dec.2010 & May 2011 are pooled and a linear time trend added.] | Randomized Controlled Trials (RCT). treated participants received cash transfer through mobile payments. Simple reduced form regression specification variously comparing differences in outcomes for the 3 channels in Dec.2010 or May 2011, or for pooled data from Dec.2010 and May 2011 rounds. indicator variables for participation in the M-money transfer program, and for whether a mobile phone was received; geographic fixed effects at the commune level; vector of household baseline covariates; presence of a seed distribution program at the village level. [ : age, raising livestock as an income source] [Clustered standard errors] | The first stage of selection may not be random, and there are other problems of potential heterogeneity (see Deaton's critique, Box 2). They do, however, control for household characteristics that differed between groups at baseline. Cost-savings rely on a well-established agent infrastructure. The results may not be generalizable. Yes, three channels: manual cash transfer; electronic cash transfer plus mobile money-enabled mobile phone given; & manual cash transfer, plus mobile money-enabled mobile phone given. | Transactions costs reduced, especially travelling and queuing time. Increased intra-household bargaining power for women. Increased diet diversity; better nutrition for children; women more likely to cultivate and market cash crops; fewer depleted durable and non-durable assets. No evidence of ‘leakage’. |
FE: Various outcomes of interest (saving, transfer and airtime purchase through M-Paisa, and welfare indicators such as consumption and self-reported happiness) of employees. Participants received mobile money-enabled mobile phones. | Afghanistan Panel data. Seven provinces, Jul. 2012 to April. 2013. Sample: 341 employees of Central Asia Development Group. Mobile operator Roshan transaction records, interviews, administrative records. Pre-baseline survey, baseline survey (before receipt of phones and training) and endline survey, and monthly phone surveys between the latter two. | Randomized Controlled Trials (RCT). treated participants received salaries through mobile payments. Simple fixed effects regression specification comparing outcomes in the endline and baseline rounds. indicator variables for a treated individual and for whether the observation was made after treatment, and the cross-effect of these two dummies; individual level fixed effects; survey wave fixed effects. [Clustered standard errors] | The first stage of selection may not be random, and there are other problems of potential heterogeneity (see Deaton's critique, Box 2). No individual controls were included. But fixed effects and survey wave effects would help control for heterogeneity. The results may not be generalizable from this special group of individuals; the time period of observation is short and sample size is small. Yes. Mobile phones provided to both treatment and control groups. | Significantly reduced net costs for disbursing firm; larger and more frequent airtime purchases and more spent in total by recipients; increased usage of mobile transfers and mobile savings by recipients, but with usage patterns differing by prior banking status and size of salary. Greater liquidity preference and savings withdrawal with increased perceptions of physical insecurity. No significant result obtained. |
Probit: Zero-1 dummy: for reported savings, credit and remittances; Tobit: log of annual savings, credit or remittances; OLS: log of annual savings, credit or remittances. Exact definition of ‘use’ unclear. | Uganda Cross-section of 820 households interviewed in 2014 on financial access and usage; household characteristics for same HHs from 4th round of household survey in Uganda, 2012 (RePEAT) project. | Probit regressions M-money dummy equal to 1 if at least one household member ‘used’ mobile money services and 0 otherwise; district dummies; and vector of household characteristics (household size, log of total asset value, age, gender and education (years of schooling) of household head, the log of distance to nearest mobile money agent). Tobit regressions the above, with additional characteristics (distance in logs to the nearest town not nearest mobile money agent; dummies for a migrant worker in household and a SACCO in district; and a land wealth variable). Variant regressions: (i) the residual from a first stage Probit regression for mobile money adoption is added to help control for endogeneity of mobile money and the log value of land is added; and (ii) the distance to the nearest mobile money agent is used as an exogenous measure of mobile money access. OLS regressions weighted by the propensity score : as for Probit regressions, plus additional characteristics (log value of land, log of distance to three other financial institutions and to district town). [Clustered standard errors] | Two approaches address endogeneity: adding residual from a first stage Probit regression for adoption in regressions; and propensity score matching. Little is significant beside the usage dummy (see RHS). The authors suggest this is because heterogeneity has been successfully removed. However, in cross-section it is very difficult to control for unobserved heterogeneity. Whether the significance of mobile money usage is indeed important or whether the coefficient is biased strongly upwards as it proxies for unobservables is unclear. No. | The authors suggest a role for mobile money in encouraging savings and as a channel for loans and remittances. Probit models: yield no significant variables at a 1% significance level, save for the (positive) mobile money usage dummy. Tobit models: yield no significant variables at a 1% significance level, save for the (positive) mobile money usage dummy. Partly controlling for the endogeneity of mobile money by adding the residual from a probit adoption regression: this is significant in the savings and credit regressions (the coefficient on mobile money usage remains stable). Assets promote savings and credit (10% significance level) in savings models without the residual; household size reduces savings (5% significance level). Propensity score matching models: nothing significant save for the (positive) mobile money usage dummy (coefficient on mobile money drops), and the value of assets (5% significance level) for savings. |
FE IV: a set of outcome variables including saving the proportion of individuals that use M-Pesa in a sub-location, but exact definition of ‘use’ unclear. | Kenya Balanced panel of (note: not of households), from combining the 2006 and 2009 FinAccess surveys. [Wealth measure constructed with principal component analysis applied to household assets and durable goods; grouping respondents by wealth quintile.] | First differenced, fixed effects Instrumental Variables regression a time fixed effect; a sub-location fixed effect; and vector of individual characteristics (education (level), gender, age, marriage rate and wealth (index and quantile dummies)). [ : 2006 perception responses (before introduction of M-Pesa) about riskier, slower and more costly transfer methods: the proportions of residents who identify the post office or a money transfer company or a friend as relatively more risky ] [Clustered standard errors] | Differenced specification removes biases due to time-invariant unobservables. The definition of the instruments is (see ). F tests suggest instruments are not weak; no tests are reported for whether they are exogenous. They do conduct some placebo tests. The instruments might be correlated with unobserved, time-varying characteristics of households that could be associated with the outcomes (e.g. ability, dynamism) and time-varying wealth if self-reported wealth is poorly measured and with (potentially) time-varying omitted variables like banking status. No. | Effect of M-Pesa adoption is to reduce both the use of informal savings groups and having to hide cash in secret places. |
OLS: binary dummy variables:—willingness to save and remit to migrants in Maputo; —willingness to save and remit using Mkesh (mobile money). Treated individuals receive training about a new mobile money product, MKesh. | Mozambique Experimental data generated in rural provinces: Maputo- Province, Gaza, and Inhambane, March 2012 (102 rural Enumeration Areas: 51 locations in 3 regions randomly selected as treatment areas; the residual is control group). Administrative mobile money records combined with household survey data (3 years, 2012–14). [ : rural treatment locations required mCel coverage & 1 or more commercial banks; targeted individuals required a mobile phone number and a migrant family member in Maputo with mobile phone number.] | Randomized Controlled Trials (RCT). treated individuals receive training about a new mobile money product. Simple OLS reduced form regression specification comparing differences in outcomes for targeted and control individuals for the years 2012, 2013, 2014 and for these years pooled. treatment dummy variable; province dummies; year dummies; and individual controls for age and gender. [Clustered standard errors] | The first stage of selection may not be random, and there are other problems of potential heterogeneity (see Deaton's critique, Box 2). Other selection criteria (see LHS) the type of population tested, which reduces the generalizability of results. There is a problem of interpreting a treatment effect when intervention depends also on the type of training information provided (see ). The results may not be generalizable. Remittances flow in the unusual rural to urban direction. Sample size is small and quantities saved/remitted are tiny. Yes. Only individuals with a phone number are selected. | Willingness to save and to remit through Mkesh increases for targeted individuals. The effect for savings is 23–25 percentage points and for remittances is 26–27 percentage points (both at a 1% significance level). Dissemination of Mkesh raised willingness to send money transfers regardless of transfer method, and at the margin Mkesh substituted traditional methods of saving. |
Probit: Zero-1 dummy: fo reported general savings; & zero-1 dummy; for reported M-Kesho savings (savings account with interest accessed via phone for mobile money users); OLS, IV: log of average monthly savings. M-Pesa registrations from the telecommunications firm. | Kenya Cross-section, survey conducted by the Financial Sector Deepening Kenya organization covering 6083 individuals, during Oct.-Nov.2010. [Total savings: M-Pesa, MKESHO/PESA PAP, KCB connect, bank account, SACCO account, ASCA, ROSCA, Microfinance Institution and ‘other’ means.] [Wealth index created using principal components analysis, grouping respondents by wealth quintile.] | Probit and IV Probit regressions for total savings & for M-Kesho savings M-money dummy for M-Pesa registration or instrument; and vector of individual characteristics (gender, age, age squared, marriage, education (unclear how measured), location (rural/urban), log of household income, and four wealth index quintiles). OLS & IV regressions as above. [ the fraction of respondents in the sub-location registered with M-Pesa.] [Ordinary standard errors] | Instrumenting for the endogenous M-Pesa usage dummy with a -level instrument, in both types of regression, averages over individuals within locations, and eliminates some but not all unobserved location-level heterogeneity. The results are suggestive only. There are no statistics examining the validity of the instruments. No. | Probit models: savings in general more likely if older, male, married, living in rural areas, with higher levels of education, reported income and wealth; with these controls, M-Pesa users are 32% more likely to report savings (at a 1% significance level). (Few used M-Kesho, but the same outcome was reached: wealthier, married, more educated, and male.) Instrumenting for M-Pesa usage drops the coefficient to 20% (at a 1% significance level). Using OLS: M-Pesa users save 12% more than those un-registered (at a 5% significance level). Using IV: the coefficient for M-Pesa users is not statistically significant. |
Logit: Zero-1 dummy: for whether an individual uses mobile money (receive, send or pay bills with mobile money or a combination of these) Households that used mobile money services at least once in the 12 months surveyed. | 35 countries Cross-section, using the World Bank's Global Findex survey (2011) usage micro-data; and constructed regulatory indices based on Porteous (2009), either equally-weighted or assigned weights through a Principal Components methodology. | Logit regression country fixed effects; the interaction of regulatory indexes with individual characteristics; and vector of individual/country characteristics. [ ] [Vector of individual characteristics: education (secondary schooling), gender, access to formal banking, age (and age squared) and income quintile). In some regressions, vector of country characteristics: log of GDP per capita, % unbanked population, % urban population, % population owning a mobile phone, concentration of banks, population density and total population.] [Ordinary standard errors] | The index is rather than . The index may be correlated with omitted country characteristics; most possible instruments for the index have the same potential problem. By using location fixed effects to reduce endogeneity, they are unable to include the index itself, but only its interaction with individual characteristics. No. | The interaction effects suggest: a regulatory framework that supports interoperability promotes higher usage among the poorest; and stronger consumer protection reduces usage by the poorest (costs) but promotes usage amongst the educated. |
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.
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 ).
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
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.
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.
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 .
Month: | Total Views: |
---|---|
October 2018 | 96 |
November 2018 | 60 |
December 2018 | 51 |
January 2019 | 61 |
February 2019 | 119 |
March 2019 | 74 |
April 2019 | 99 |
May 2019 | 322 |
June 2019 | 153 |
July 2019 | 148 |
August 2019 | 212 |
September 2019 | 178 |
October 2019 | 347 |
November 2019 | 402 |
December 2019 | 337 |
January 2020 | 397 |
February 2020 | 634 |
March 2020 | 497 |
April 2020 | 797 |
May 2020 | 441 |
June 2020 | 584 |
July 2020 | 699 |
August 2020 | 571 |
September 2020 | 690 |
October 2020 | 730 |
November 2020 | 723 |
December 2020 | 603 |
January 2021 | 545 |
February 2021 | 632 |
March 2021 | 960 |
April 2021 | 845 |
May 2021 | 818 |
June 2021 | 581 |
July 2021 | 717 |
August 2021 | 489 |
September 2021 | 554 |
October 2021 | 777 |
November 2021 | 871 |
December 2021 | 794 |
January 2022 | 681 |
February 2022 | 813 |
March 2022 | 1,028 |
April 2022 | 896 |
May 2022 | 871 |
June 2022 | 726 |
July 2022 | 610 |
August 2022 | 594 |
September 2022 | 513 |
October 2022 | 684 |
November 2022 | 649 |
December 2022 | 486 |
January 2023 | 576 |
February 2023 | 581 |
March 2023 | 705 |
April 2023 | 608 |
May 2023 | 602 |
June 2023 | 441 |
July 2023 | 479 |
August 2023 | 392 |
September 2023 | 424 |
October 2023 | 450 |
November 2023 | 524 |
December 2023 | 343 |
January 2024 | 535 |
February 2024 | 440 |
March 2024 | 615 |
April 2024 | 559 |
May 2024 | 530 |
June 2024 | 522 |
July 2024 | 449 |
August 2024 | 281 |
Citing articles via.
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
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.
151 Accesses
Explore all metrics
This study aims to analyze the keywords related to mobile banking by focusing on its development between 2003 and 2023. To do so, a systematic bibliometric analysis is undertaken, for which, the screening of Scopus database was done to identify the publication (institutions, most-prolific authors, year-wise, countries-wise and most-cited papers). Additionally, keyword occurrence analysis was also done using the VOSviewer software. The results of the study indicate that there are worldwide trends and increased production that have resulted in various changes. One prominent topic that consistently emerges in relation to mobile banking across different time periods is mobile telecommunication technologies. Through the demonstration of the origination of major terminologies in mobile banking, notable shifts in the evolution of the field's terminological framework were discernible. Hence, it is imperative to investigate the progress in forthcoming decades, specifically the impact of recent global events on the evolution of mobile banking usage worldwide. Furthermore, the current study adds substantial value to the existing body of literature by presenting a conceptual framework that might guide future research endeavors. The framework offers researchers the potential to investigate the many study streams in forthcoming studies. The present study represents a novel contribution in terms of its methodology. A thorough examination of research databases, such as Scopus and Google Scholar, reveals a lack of published literature that comprehensively and extensively addresses the topic of mobile banking (m-banking) in a multi-period context, particularly in an applied manner. Moreover, the present study addresses this research gap by the implementation of a bibliometric analysis and content analysis.
This is a preview of subscription content, log in via an institution to check access.
Subscribe and save.
Price includes VAT (Russian Federation)
Instant access to the full article PDF.
Rent this article via DeepDyve
Institutional subscriptions
Source: Scopus database ( https://www.scopus.com/results ) . Note: As of September 13, 2023
Source: The authors (from Scopus database)
Source: The authors (Scopus database)
Exploring the key drivers behind the adoption of mobile banking services.
Authors and affiliations.
Department of Business Management, HNB Garhwal Central University, Srinagar Garhwal, Uttarakhand, 246174, India
Kamlesh Kohli, Monika Kashyap & Mahendra Babu Kuruva
Department of Tourism Studies, School of Business Studies, Central University of Kerala, Kasargod, India
Sunil Tiwari
You can also search for this author in PubMed Google Scholar
Correspondence to Sunil Tiwari .
Conflict of interest.
All authors have approved the manuscript title “RECAPITULATION OF SCHOLARLY RESEARCH ON MOBILE BANKING FROM 2003 TO 2021: A BIBLIOMETRIC ANALYSIS” and agree with its submission to this journal. The current work has not been published and is not under consideration for publication elsewhere. Moreover, there is no conflict of any interests for this study.
Publisher's note.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
Reprints and permissions
Kohli, K., Kashyap, M., Kuruva, M.B. et al. Mobile banking: a bibliometric analysis. J Financ Serv Mark (2024). https://doi.org/10.1057/s41264-024-00267-7
Download citation
Received : 09 May 2023
Revised : 08 January 2024
Accepted : 16 January 2024
Published : 01 February 2024
DOI : https://doi.org/10.1057/s41264-024-00267-7
Anyone you share the following link with will be able to read this content:
Sorry, a shareable link is not currently available for this article.
Provided by the Springer Nature SharedIt content-sharing initiative
Please note you do not have access to teaching notes, literature review of mobile banking and individual performance.
International Journal of Bank Marketing
ISSN : 0265-2323
Article publication date: 2 October 2017
Most empirical studies of m-banking seek to understand the factors and motivations that influence the adoption or behaviour intention. The purpose of this paper is to focus on analysing and synthesising existing studies and make recommendations to researchers and practitioners.
Few papers focus on the m-banking use and individual performance, but on the determinants of adoption measures, instead. This research examines 64 journal articles published between 2002 and 2016 in top journals. Following a comprehensive review of the literature, the authors propose a research agenda.
The importance of use and individual performance has long been recognised by academics and practitioners in a variety of functional disciplines. The present review indicates that the topics of m-banking adoption and behavioural intention dominate the majority of research, but finds very few studies on post-adoption. The two most significant drivers of intentions to adopt m-banking are perceived ease of use and perceived usefulness. Considering several m-banking definitions, the authors propose a new, broader definition that takes into account the technological changes that have occurred over time. m-banking is a service or product offered by financial institutions that makes use of portable technologies.
This paper assembles this diverse body of knowledge into a coherent whole. The authors expect that this review will be of benefit to anyone interested in m-banking research and that it will help to stimulate further interest. In order to advance research in m-banking, future research should consider other theories uncovered in our findings.
Tam, C. and Oliveira, T. (2017), "Literature review of mobile banking and individual performance", International Journal of Bank Marketing , Vol. 35 No. 7, pp. 1044-1067. https://doi.org/10.1108/IJBM-09-2015-0143
Emerald Publishing Limited
Copyright © 2017, Emerald Publishing Limited
All feedback is valuable.
Please share your general feedback
Contact Customer Support
An official website of the United States government
The .gov means it’s official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.
The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.
Preview improvements coming to the PMC website in October 2024. Learn More or Try it out now .
Associated data.
The original contributions presented in the study are included in the article/supplementary material, further inquiries can be directed to the corresponding author/s.
Many firms in the modern world utilize m-banking systems to communicate with their consumers. The word m-banking refers to a widespread method of providing financial services and localization to customers. Since m-banking is important to both banks and users, it has been included in numerous literary works. As a result, embracing financial services via the m-banking platform is critical. This article's technique is mostly descriptive research that investigates common views, current situations, modern tactics, tangible emerging consequences, etc. The main objective here is to analyze the benefits of this study by investigating the past. Since this article analyzes what exists and is descriptive, the data is being retrieved by conducting a cross-sectional survey method about different features that are relevant by sampling the population. The main aim of this study is to explore the adoption of mobile banking technology by consumers. Based on the values of different variables such as affective commitment (AC), transaction convenience (TC), perceived ease of use (PEU), perceived reliability (PR), pre and post benefits (PPB), service, system, and information quality (SSIQ), bank trust (BT), and profitability (P), the inter-relationship between them and the adoption of m-banking technique by the users in banking technology. The model is investigated by examining the hypothesis and identifying the relationship that exists between these different parameters. A simple linear regression method is implemented using the Statistical Package for the Social Sciences (SPSS) software.
Nowadays, mobile technology plays a vital role in many industries such as banking on managing stocks and finances, business firms sell their products through applications and websites. Mobile banking services such as payment systems influence people's lives more dramatically than other innovations in the recent times ( 1 ). Mobile technology made banking services such as payment, transaction, and other services easy through the online mode instead of offline mode. Even the adoption of mobile banking services increased day-by-day by the consumers, still mobile payments do not become ordinary to many of the public ( 2 ). Due to the growth of mobile network technology, there is a need for security infrastructure for mobile payments. The security measures give trust to the consumers to adopt mobile banking at anytime from anywhere.
For achieving a successful outcome in business, mobile technology is recognized as competitive and creative equipment that provides better online transaction options ( 3 ). Recently, cellphones are being widely used as a tool by people for online shopping, banking, paying bills, etc. Our day-to-day routine lives have been changed through the adoption and development of mobile technology that modifies the way of communication, approaches for selling and buying goods and services, and also the mode of information collection. Such corresponding correlation arises in a constructive environment that has no limitation regarding the time or place ( 4 ).
The rapid development in mobile technology and the protection framework facilitate the people to perform m-banking transactions at any time from anywhere, whereas m-banking comprises financial service which is implemented directly or indirectly over a network of wireless telecommunication ( 3 ). In the financial sector, the services applied by mobile are referred to as mobile financial services (MFSs) they also include mobile banking and mobile payment. Several studies were conducted to examine the peoples' mindset in view of understanding the mode of adopting mobile banking as a unique and usual service ( 5 ). According to the statistics, the level of usage of smartphones has increased from 35 to 80% in 2020, also worldwide, the number of mobile phone beneficiaries is about 6.8 billion, though each mobile was fully activated with the internet. Thus, mobile banking helps in the development of bank access rates which influences bank perception ( 6 ). In the banking sector, it is observed that most of the customers have switched to smartphone, apps, or tablets which are highly utilized in their day-to-day life for shopping, entertainment, learning, socializing, and jobs, and these mobile-centric customers are preferred by more bankers ( 7 ). For example, a bank in Japan, Jibun Bank whose principal communication with customers is through the mobile channel and was the first bank that affords their complete banking services and products over mobile channels. With the help of cameras and mobile phone, customer has the facility to start a new account with their bank. This mode of banking enables the bank to improve their customer service over anywhere at any time basis. Also, the different mobile banking services are payment, web-related shopping, fund transfer, and web-related marketing. This Jibun Bank has reached more popular for its efficient services within the short duration of their mobile banking introduction ( 8 ). Through this, they have attained above 500,000 new bank accounts in their bank. In the initial days, the bank had the facility to send usual text messages related to money transfer and withdrawal and these services were updated to provide its customer a complete functional banking experience ( 9 ).
Mobile banking is adopted by many consumers as it is easy to open an account, pay bills, transfer funds, do mobile-based shopping, etc. Many banks in China and America who introduced mobile banking for all their services experienced high demand for their services through mobile banking ( 10 ). But many challenges were faced by these banks due to no face-to-face interactions. Trust is an important factor in mobile banking but lack of interaction led to the risk of uncertainty and anonymity. On the other hand, the consumers also suffer from uncertainty whether the bank is trustworthy are not.
During the last decades, mobile banking became relevant all over the world. The number of fund transfers through mobile banking also increased since last year, but mobile banking has many security problems. Sometimes, the security setting of mobile phones can be overridden by the virus in smartphones ( 11 ). Many banks open their banking applications, these apps should be updated frequently otherwise the consumers can be vulnerable to the attacks such as DDoS attacks, phishing, spoofing, corporate account takeover, and skimming ( 12 ).
The major contribution of the article is:
The remaining article is arranged accordingly. Section Literature Review presents the related work and the hypothesis is formulated in Section Hypothesis Development and Research Methodology. The results in terms of linear regression and descriptive analysis are provided in Section Results and the article is concluded in Section Conclusion and Discussions.
In the modern world, many firms use systems of mobile banking to communicate with the customers. The payment system in mobile banking services benefited people's lives. This article aims to analyze the effect of people using the technology of mobile to make banking services more accessible. Literature on mobile banking adoption, electronic banking services, mobile payment service user acceptance are reviewed.
Hamidi et al. ( 13 ) studied the influence of mobile banking adoption on consumer engagement and satisfaction utilizing the customer relationship management (CRM) system, which is the most essential aspect in banking industry. CRM is also seen as a critical role for enhancing client satisfaction in mobile banking. The statistical study performed evaluated the dialogue between the bank's customer sector and their client. The statistical analysis findings have a favorable influence on consumer interactions and satisfaction.
Geebren et al. ( 14 ) investigated the importance of consumer satisfaction in mobile eco-systems that used electronic banking services, particularly in developing nations. This entailed researching consumer satisfaction in mobile banking, with a focus on the importance of trust. To determine consumer satisfaction, structural modeling using partial least squares (PLS–SEM) methods were employed to examine the data, and trust demonstrated that customer contentment had a beneficial influence.
The relevance of an early trust theoretical model in mobile payment service user acceptance was highlighted by Gao et al. ( 15 ). The initial trust theoretical model highlighted the facilitators and barriers to user trust in m-payment services. The links in the original trust theoretical model were assessed using partial least squares structural modeling (PLS–SEM). The findings may be used in m-payment adoption research and practice in a variety of ways. In total, 52.3% of the difference in usage intention was explained by the current model.
Hentzen JK et al. ( 16 ) offered a mobile technology that allows for vital involvement, as well as an explanation of how a retirement app might assist people in planning for their post-retirement strategy. The available literature survey data were used to evaluate a sample of 440 Australian pension fund members. The findings show that consumers' perceived financial security, financial self-efficacy, retirement planning involvement, future consequences consideration, and perceived usefulness with a mobile retirement app have direct and indirect effects on their expected engagement through their goal to adopt the app.
Zhu et al. ( 17 ) investigated existing technology designs, including mobile banking, used by rural communities in six Chinese regions. According to the findings, interpersonal and mass communication channels have a bigger influence than organizational communication channels. Mobile banking should be examined since it can assist alleviate the lack of access to financial goods and financial infrastructure in rural areas.
Afeti et al. ( 18 ) developed a mobile payment technology for payment for micro-businesses. The study draws on the transaction cost theory and the task-technology fit (TTF) theory as the assumed lens. In total, 20 micro-businesses based on qualitative data were analyzed and the research findings denote those micro-businesses adoption of mobile payments results in strategic and operational benefits.
Alamoudi et al. ( 19 ) proposed that the mobile technology acceptance lookalike was changed as we investigated consumers' acceptance of mobile shopping in general stores by examining transaction convenience, usefulness, attitudes, ease of use, transaction speed, optimism, and personal innovativeness. A total of 351 respondents completed the questionnaire evaluation. Consumers are willing to use mobile shopping channels if the system is clear and straightforward to use.
Jebarajakirthy et al. ( 20 ) used a comprehensive moderated mediation framework to evaluate the influence of online convenience aspects on mobile banking uptake. Covariance-based structural equation modeling and the process macro are utilized to test these predictions. This study examines how convenience characteristics influence mobile banking adoption intentions.
Abdinoor et al. ( 21 ) studied the adoption of mobile financial services in Tanzania with the use of a technology acceptance model. To select the sample from data collection, a random sampling technique was used. The user and non-user of mobile financial services were included in the sample. Zhang et al. ( 3 ) investigated customers' use of mobile technology to help them with banking services and activities, as well as the variables that impact their adoption and engagement. Here, the analysis is done by the structural equation modeling technique to know the consumers' intentions toward mobile banking. The result examines the adoption of mobile banking apps to facilitate bank consumers' banking services.
Tiwart et al. ( 22 ) researched the variable that influences adoption in Commercial Bank Ethiopia. The author used factors such as perceived ease of use (PEOU), infrastructure (INF), security (SEC), trust, and e-banking adoption. Here, a structural equation model based on least partial square analysis is used. The result of this analysis proves that the trust of customers mediates between PEOU, INF, SEC, and e-banking adoption and realized the factors that guess the purpose and adoption of e-banking by Ethiopian customers.
The problem identified in mobile banking services is critical financial services through mobile banking because it has security problems.
Participant's demographic information.
This study is mainly conducted in China by taking the top e-banks that utilize m-banking. The questionnaire is distributed to both the users and staff of the bank. Within 6 months, 293 questionnaires were completed. These details were taken into consideration in this study. The details of the questionnaires were obtained online and also provided to the users on their official bank pages. The information obtained in the questionnaire is provided in Tables 1 , ,2 2 .
Demographic analysis.
Female | 54.50 | 121 |
Male | 45.49 | 101 |
1–20 | 3 | 25 |
20–30 | 59 | 120 |
30–40 | 15 | 14 |
>40 | 23 | 63 |
Education | ||
Master of science | 12 | 25 |
Bachelor of science | 15 | 32 |
Engineering | 20 | 57 |
Medicine | 18 | 60 |
Diploma | 40 | 85 |
Associate degree | 17 | 14 |
PhD | 5 | 7 |
Participant details.
State-owned | 3.25 |
Joint-stock | 2.14 |
Postal savings | 3.25 |
Agricultural bank | 2.52 |
Commercial city bank | 1.54 |
The information of the customers is mainly obtained based on their m-bank usage. The five main banks such as state-owned, joint-stock, postal savings, commercial city bank, and the agricultural bank were selected. Each participant was asked to say about their experience of their usage.
The amount of time spent by each user on average is provided in Figure 1A . As seen in Figure 1A , more than 39.38% of users are using more than 3 h a day and a minimum of is online Conceptual Model for less than 1 h. In total, 8.21% of the users only use the application for a limited amount of time (below 30 min).
Usage analysis of m-banking application (A) Users using m-banking during the day and (B) users' friends and references using mobile banking.
Figure 1B presents the percentage of people the users have referred for mobile banking. Based on Figure 1B , a total of 48.52% has more than 100 contacts who have been referred to the m-banking. The increase in m-banking contact can improve efficiency.
The main aim of this research is to identify the usage of mobile technology of consumers for the banking system. The hypothesis provided for each variable is explained in detail in this section and the conceptual model is provided in Figure 2 . The measurement instrument used for this study is presented in Table 3 .
Conceptual modeling structure.
Measurement instrument to analyze different parameters.
Bank trust | The bank helps me to use the mobile device of the consumer for accessing the services BT1: The m-banking is trustworthy BT2: The m-banking mainly satisfies the promises and commitments |
Pre and post benefit convenience | The m-banking resolves the problems in a fast manner Obtaining the follow-up service is easy The solutions provided by m-banking is fast and reliable The banking service is provided with little effort The services are easier to avail The problem is resolved fastly The time taken by it is reasonable |
Service, system, and information quality (SSIQ) | Convenient access of m-banking services Easy navigation of m-banking services Visual attractiveness of m-banking services Accurate information of m-banking services Up-to-date information of m-banking services Dependable services provided Personalized m-banking services Professional m-banking services Timely m-banking services |
Transaction convenience | Easier to complete the transaction Fast service access Little effort to complete the transactions |
Perceived ease of use | Usability of mobile devices for banking Interaction with the mobile devices for banking Skillset achieved in accessing mobile devices for banking |
Perceived reliability | To select the supporting technology for mobile banking Ensuring that the technology supporting m-banking does not fail The technology used is reliable |
Portability | Increased customer interaction Increased customer visit |
Service, system, and information quality (ssiq)".
System quality mainly measures the efficiency of the overall m-banking system in terms of service provider anonymity. The system quality of the m-banking mainly depends upon the ease of use, the attractiveness of the application, trust of the user, etc. If the user trusts the system quality, then they mainly select the system for future transactions. Service Quality mainly relies on the service provided by the m-banking system and it also evaluated the effectiveness of the service in terms of personalization, reliability, delay, etc. Hence, service quality is crucial to improve the quality of m-banking since poor-service quality can result in minimal user trust and satisfaction. Information quality mainly implies that high-quality information is reliable, complete, accurate, relevant, and accessible. The information quality is an important variable that improves the usage, trust, and satisfaction of the customers in the m-banking system.
Hypothesis (H6): The service, system, and information quality had a positive relationship in improving the m-banking quality and user satisfaction.
Affective obligations between business partners reflect mental reliance on others and are founded on emotions, loyalty, and dependency. The psychological bond in this example demonstrates an emotional commitment. Support for differences in particular data among new and long-term clients. Based on their differing expectations, new and loyal clients behave in the same way.
Hypothesis (H1): The interaction of customers helps to develop a good relationship with the affective commitment of a person to the bank which is focused on interaction.
The time and attempt to acquire some benefits of services is said to be benefit convenience. Some of the components of benefit convenience are fast service, timely services, and bank employee attitudes. Sometimes in banks, the consumer needs to visit repeatedly to avail specific service. But in m-banking, consumers can avail any services on the go from the home itself. So, the time and attempt to acquire the service became very easy. Therefore, benefit convenience shows a positive result in mobile banking adoption ( 23 ). The time and attempt to contact a service provider for a particular service is said to be post benefit convenience. In m-banking, the service provider can be contacted and solve the grievances within a few clicks through various means such as email, live chat, and toll-free numbers. But in offline banking, the consumer needs to visit the branch and wait for the service provider to solve the problem in service. So, the post benefit convenience shows a positive result in mobile banking adoption ( 20 ).
Hypothesis (H5): The pre- and post-benefit convenience are positively interrelated with m-banking.
The fast and simple way to complete the transaction is said to be transaction convenience. Acquiring the service of transaction in a lesser time is the main component of transaction convenience. Other components of transaction convenience are uninterrupted transaction, transaction confirmation, easy checkout, and price inconsistency. By using m-banking, the transaction can be done in any place at any time in a few clicks and can do several transactions simultaneously. But in offline banking, consumer needs to wait in a queue for availing a transaction service. Hence, transaction convenience shows a positive result in mobile banking adoption ( 24 ).
Hypothesis (H2): The transaction convenience is positively interrelated with m-banking.
Customer participation in an organization's mobile banking may increase customer value and profitability ( 25 ).
Hypothesis (H8): Profitability shows positive results in mobile banking adoption.
Perceived reliability is specified as the limit to which the individual has independently considered the technology as a faithful one. In order to offer a trustworthy technical service, the system reliability is important that enables the user to attain the aimed objective. An undependable technology that is employed by an individual seems to be low confidence and such technology utilization should be restricted ( 26 ). Research related to mobile technology and information system has revealed that a technology with perceived reliability possesses has an important impact on customers' perceived satisfaction, detected value, and observed quality. Also, customers' understanding level of trust is examined and analyzed as an antecedent of confidence in an organization through the mobile commerce setting. Though mobile is a wireless technology, it is easily susceptible to violations or attacks ( 27 ). Mostly, the customers were insisted to share their personal data while consuming and shopping services performed through mobile technology such as date of birth, debit and credit card details, funds, and address. The reliability of mobile technology should be improved thereby increasing customers' belief in securing their personal data. The postulated hypotheses are mentioned as follows:
Hypothesis (H4): Perceived reliability in m-banking helps to improve the consumer's trust and ease of use of banking services.
In this hypothesis, trust is considered as a faith of competence, integrity, ability, and benevolence that the individual has toward each other. Trust has become an important reducing feature of risk, whereas the e-commerce connection is considered as a naturally risky factor ( 28 ). Also, trust in banking is specified that it is linked to reliability and perceived privacy. Generally, the primary interaction between the customer and banker will significantly affect the trust progress in them.
Hypothesis (H7): Based on their usage of mobile devices and attitude, the consumer trust toward M-banking increases.
Perceived ease of use evaluates the increase in the amount of work required to acquire a job utilizing new technologies on an individual basis. In the effort condition, the simplicity of use of mobile innovation is critical for including customers in the co-creation of value experiences ( 29 ). Mobile technology's ability to increase advantages for bank clients has been widely shown, and its simplicity of use is a key factor in consumer acceptability. PEU has been linked to advanced tackling new technologies in the banking industry in several research. However, the launch was difficult, and it was claimed that customers' inclination to utilize mobile banking services is not intrinsically related to employment prospects. As a consequence, it indicates that study into PEU and its influence on consumer attitudes, as well as the interactive method, are both worthwhile endeavors ( 30 – 32 ).
Hypothesis (H3): Customers' views about the usage of mobile devices for facilitating financial services are favorably associated with their perceived ease of use (PEU).
The reliability of the scales was assessed using Cronbach's alpha values. Goodness-of-fit approaches were used to assess the overall model fit for confirmatory factor analysis (CFA). The RMSEA was 0.09, the comparative fit index (CFI) was 0.99, the goodness-of-fit (GFI) was 0.98, and the standardized root mean square residual (RMR) was 0.06. The X2 to df ratio was 4.14, the root mean square error of approximation (RMSEA) was 0.09, the CFI was 0.99, the GFI was 0.98, and the standardized RMR was 0.06. As indicated in Table 4 , the reliability coefficients of all constructs were more than 0.87, which is higher than the 0.9 thresholds. The extracted average variance was used to assess convergent validity (AVE). Because the AVE values ranged from 0.58 to 0.92, convergent validity was not an issue. The AVE scores were lower than the squared correlations between pairs of constructs, indicating discriminant validity in the data set. Table 4 shows the results of the Cronbach's alpha and AVE tests.
Hypothesis testing.
| ||||||||||
---|---|---|---|---|---|---|---|---|---|---|
Affective commitment (AC) | 0.78 | 0.58 | 0.77 | |||||||
Transaction convenience (TC) | 0.96 | 0.87 | 0.55 | 0.94 | ||||||
Perceived ease of use (PEU) | 0.97 | 0.88 | 0.62 | 0.68 | 0.95 | |||||
Perceived reliability (PR) | 0.98 | 0.91 | 0.51 | 0.52 | 0.56 | 0.96 | ||||
Pre and post benefits (PPB) | 0.98 | 0.92 | 0.55 | 0.74 | 0.66 | 0.53 | 0.97 | |||
Service, system, and information quality (SSIQ) | 0.95 | 0.82 | 0.47 | 0.56 | 0.66 | 0.44 | 0.74 | 0.92 | ||
Bank trust (BT) | 0.87 | 0.68 | 0.72 | 0.46 | 0.48 | 0.52 | 0.49 | 0.42 | 0.83 | |
Profitability (P) | 0.83 | 0.64 | 0.67 | 0.42 | 0.44 | 0.50 | 0.46 | 0.39 | 0.80 | 0.84 |
Regression tests are run when the allowable alpha coefficient for the scales has been determined. Table 5 shows the regression test results for each of the assumptions discussed in the following sections. Affective commitment (AC) and interaction were investigated, and the correlation coefficient 0.297 and likelihood of significance 0.000 indicate that there is a positive and significant relationship between the two. The regression test supports this positive relationship, showing that the interaction variable may be able to predict AC variation.
The results of the linear regression test.
-value | |||||
---|---|---|---|---|---|
Affective commitment (AC) | Interaction | H1 | 0.000 | 0.297 | Accepted |
Transaction Convenience (TC) | Interaction | H2 | 0.000 | 0.339 | Accepted |
Perceived ease of use (PEU) | Satisfaction | H3 | 0.000 | 0.333 | Accepted |
Perceived reliability (PR) | Interaction | H4 | 0.000 | 0.172 | Accepted |
Pre and post benefits (PPB) | Interaction | H5 | 0.000 | 0.242 | Accepted |
Service, system, and information quality (SSIQ) | Satisfaction | H6 | 0.000 | 0.430 | Accepted |
Bank trust (BT) | Interaction | H7 | 0.000 | 0.347 | Accepted |
Profitability (P) | Interaction | H8 | 0.000 | 0.230 | Accepted |
With a correlation value of 0.333 and a probability of significance of 0.000, the following test demonstrated a positive relationship between perceived ease of use and satisfaction (PEU). The correlation coefficient between the variables perceived reliability (PR) and interaction was 0.172, with a probability of significance of 0.000, demonstrating a positive link between the two variables. The correlation value of 0.430 and the probability of significance of 0.000 indicated a positive relationship between satisfaction and service, system, and information quality (SSIQ) ( 3 ).
The items in this questionnaire were answered using a five-point Likert scale, with the alternatives being: very low (1), low (2), medium (3), high (4), and very high (5). Table 6 shows the SD, middle value (MV), and lowest value (LV) of the descriptive data received from the surveys ( 3 ).
Descriptive data analysis.
Affective commitment (AC) | 1.208 | 2.52 | 5 | 1 |
Transaction convenience (TC) | 1.081 | 3.36 | 5 | 1 |
Perceived ease of use (PEU) | 1.061 | 3.21 | 5 | 1 |
Perceived reliability (PR) | 1.689 | 2.92 | 5 | 1 |
Pre and post benefits (PPB) | 1.393 | 2.67 | 5 | 1 |
Service, system, and information quality (SSIQ) | 1.206 | 2.79 | 5 | 1 |
Bank trust (BT) | 0.955 | 3.31 | 5 | 1 |
Profitability (P) | 1.220 | 3.10 | 5 | 1 |
Mobile banking is a cost-effective way to reach customers. The different activities that can be performed by the consumers in m-banking are checking their bank balance, making transactions, making investments, getting account statements, paying bills, etc. The motivation for user adoption and use of mobile banking is examined in this research. Positive connections between profitability, trust, and transaction convenience to utilize mobile banking have been found in previous studies. Users' confidence in using mobile banking services, enjoyment, utility of the system, etc., are not the same thing. The participants were unconcerned about the danger of fraud, system dependability, or perceived privacy while building and extending faith in their banks and mobile services. Because of today's technology-driven lifestyle, individuals are receptive to embracing new technologies that are compatible with their mobile phones.
To add value to practice, we must obtain convincing and widespread results from both users and mobile banking service providers. Like mobile phones, smartwatches have become more popular among people nowadays. With this wearable technology, banks have the greatest responsibility to fulfill hedonic and social needs. The purpose and use of this research are common to all consumers irrespective of various countries, populations, and socio-economic status. The variables used in this work are, namely, affective commitment (AC), transaction convenience (TC), perceived ease of use (PEU), perceived reliability (PR), pre- and post benefits (PPBs) service, system, and information quality (SSIQ), bank trust (BT), and profitability (P). Consumer behavior such as untrustworthiness, fraudulence, and so on, as well as mobile banking transaction risks such as financial, privacy, and cyber security, are important risk concerns that may be addressed in future research developments.
The problems and risks that arise by the consumers and banking sector for adopting mobile banking services would be reduced by the scientific researchers using various technologies and methodologies in the future. The results in this article show the different statistical surveys and believed that it is true for both mobile banking consumers and non-mobile banking consumers with various socioeconomic environments. But when we consider globally each country had its own rules and regulations regarding mobile banking transactions, services, and products. And these rules and regulations react differently in different countries.
Author contributions.
Both authors listed have made a substantial, direct, and intellectual contribution to the work and approved it for publication.
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
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.
Alexander Decker
Nadeem Maqbool
gazi hossain
Afr. J. Bus. Manage
Anton Abdulbasah Kamil
shakir shaik
IOSR Journals
Electronic Banking incorporates a multitude of platforms such as internet banking, automated teller services and mobile phone banking to deliver banking products to the customer. The proposed study attempts to throw light into pattern of customer use and non - use of banking technologies and the extent of their utilization.
IOSR Journals publish within 3 days
Mobile banking technology which is the third era of technological innovation of banking sector after phone and net banking and comparatively its growth is phenomenal when compared to the first two eras. Currently in India the Mobile Banking is growing fast because demonetization of economy, the customers are opted for online banking and M-banking facilities provided by the bank and the world's second largest subscriber base in mobile sector therefore this leads to increases the mobile banking users in india.The main objectives of the study is to highlight the theoretical background and current scenario of mobile banking services in Indian context and to examine the demographic profile of mobile banking users of SBI bank and to analysis the reasons for customers adoption of SBI mobile bank services and also to assess the customer usage of SBI mobile banking facilities. Finally this studycovers the customers satisfaction towards SBI mobile banking services in Mysuru city. The present study has been collected from primary data on the basis of issue questionnaire, the sample size for study was only 100 mobile banking users of SBI. For the purpose of analysis of data based on normality test applied for non-parametric tests such as, mean, standard deviation, Mann-Whitney U-Test and KruskalWallies Z-Test. Finally this study results Majority of the respondents has strongly agreed and opined the positive usage of mobile banking services provided by the SBI in Mysuru City.
Benson James Lyimo
The study sought to find out the effects of mobile banking to customer satisfaction among non-business customers in Tanzania, a case of ACB bank in Dar es salaam, the study adopted a case study design on a sample size of 50 respondents who were selected through simple random sampling from a target population, bank customers. Data was collected by use of questionnaires and the data analyzed by the aid of Statistical Package of Social Scientists Program (SPSS). The findings were summarized and data was presented using tables, charts and figures. Based on the study findings it was concluded that close to all those using mobile. The study findings showed that information technology infrastructure, installation costs and user perception affect the Adoption of the computerized Accounting System. The study suggests encouraging the widespread use of mobile banking; campaigns should be launched to disseminate the usefulness of the technology. Such as through televisions, radios and social networks like Facebook and twitter. The researcher highly recommends improvement of the service availability around the clock. This should involve reviewing of the service level agreement between the service providers and the mobile network operators so as to enhance customers to access their bank accounts anywhere and at any time. The researcher proposes the following recommendations: (i) Further research should be conducted in this area to explore the profitability associated with the technology. (ii) There is a need to explore more independent variables that can have an impact on customer satisfaction
Mohammad Anisur Rahman
The progress of Mobile Banking (M-banking) is unsatisfactory in terms of achieving a key objective, i.e. to reach the inaccessible unbanked customers at an affordable cost for financial inclusion. We identify the causes to this problem through the investigating the factors affecting the adoption of m-banking in remote areas of Bangladesh using a 236 primary sample of m-bank customers from seven geographical locations and across professions. We document a positive effect of perceived ease of use, trusts, and perceived usefulness, and a negative effect of user interface on adopting m-banking in rural Bangladesh. our findings provides significant policy implications for policy planners and the bank managers to enrich their policies and strategies for promoting financial inclusion as well as successful banking business in rural of Bangladesh.
Mir Abdullah Shahneaz
Loading Preview
Sorry, preview is currently unavailable. You can download the paper by clicking the button above.
International Journal of Islamic …
Mamun Rashid
Frank Addae-Aboagye
mamun rashid
Md. Mizanur Rahman
Sajid Hossain Khan
Kerathum Juma
Krishna Murari
13th Annual International Conference - Convergence 2018 on the theme "Applied and/or Interdisciplinary Research: Emerging Economy Perspectives"
Dr. Nazrul Islam , Rasel Mustafi
subrata koiry
Mirza Mohammad Shezan
Mohani Abdul
samia shabnaz
Journal of Internet Banking and Commerce
International Journal of Computer Theory and Engineering
Abdullah Kammani
Sadi Mohammad
SDIWC Organization
Muhsina Jannat
Md. Masudul Hassan
Journal of Advance Management and Accounting Research
Card Publication
Analyzing the effect of people utilizing mobile technology to make banking services more accessible.
Many firms in the modern world utilize m-banking systems to communicate with their consumers. The word m-banking refers to a widespread method of providing financial services and localization to customers. Since m-banking is important to both banks and users, it has been included in numerous literary works. As a result, embracing financial services via the m-banking platform is critical. This article's technique is mostly descriptive research that investigates common views, current situations, modern tactics, tangible emerging consequences, etc. The main objective here is to analyze the benefits of this study by investigating the past. Since this article analyzes what exists and is descriptive, the data is being retrieved by conducting a cross-sectional survey method about different features that are relevant by sampling the population. The main aim of this study is to explore the adoption of mobile banking technology by consumers. Based on the values of different variables such as affective commitment (AC), transaction convenience (TC), perceived ease of use (PEU), perceived reliability (PR), pre and post benefits (PPB), service, system, and information quality (SSIQ), bank trust (BT), and profitability (P), the inter-relationship between them and the adoption of m-banking technique by the users in banking technology. The model is investigated by examining the hypothesis and identifying the relationship that exists between these different parameters. A simple linear regression method is implemented using the Statistical Package for the Social Sciences (SPSS) software.
Nowadays, mobile technology plays a vital role in many industries such as banking on managing stocks and finances, business firms sell their products through applications and websites. Mobile banking services such as payment systems influence people's lives more dramatically than other innovations in the recent times ( 1 ). Mobile technology made banking services such as payment, transaction, and other services easy through the online mode instead of offline mode. Even the adoption of mobile banking services increased day-by-day by the consumers, still mobile payments do not become ordinary to many of the public ( 2 ). Due to the growth of mobile network technology, there is a need for security infrastructure for mobile payments. The security measures give trust to the consumers to adopt mobile banking at anytime from anywhere.
For achieving a successful outcome in business, mobile technology is recognized as competitive and creative equipment that provides better online transaction options ( 3 ). Recently, cellphones are being widely used as a tool by people for online shopping, banking, paying bills, etc. Our day-to-day routine lives have been changed through the adoption and development of mobile technology that modifies the way of communication, approaches for selling and buying goods and services, and also the mode of information collection. Such corresponding correlation arises in a constructive environment that has no limitation regarding the time or place ( 4 ).
The rapid development in mobile technology and the protection framework facilitate the people to perform m-banking transactions at any time from anywhere, whereas m-banking comprises financial service which is implemented directly or indirectly over a network of wireless telecommunication ( 3 ). In the financial sector, the services applied by mobile are referred to as mobile financial services (MFSs) they also include mobile banking and mobile payment. Several studies were conducted to examine the peoples' mindset in view of understanding the mode of adopting mobile banking as a unique and usual service ( 5 ). According to the statistics, the level of usage of smartphones has increased from 35 to 80% in 2020, also worldwide, the number of mobile phone beneficiaries is about 6.8 billion, though each mobile was fully activated with the internet. Thus, mobile banking helps in the development of bank access rates which influences bank perception ( 6 ). In the banking sector, it is observed that most of the customers have switched to smartphone, apps, or tablets which are highly utilized in their day-to-day life for shopping, entertainment, learning, socializing, and jobs, and these mobile-centric customers are preferred by more bankers ( 7 ). For example, a bank in Japan, Jibun Bank whose principal communication with customers is through the mobile channel and was the first bank that affords their complete banking services and products over mobile channels. With the help of cameras and mobile phone, customer has the facility to start a new account with their bank. This mode of banking enables the bank to improve their customer service over anywhere at any time basis. Also, the different mobile banking services are payment, web-related shopping, fund transfer, and web-related marketing. This Jibun Bank has reached more popular for its efficient services within the short duration of their mobile banking introduction ( 8 ). Through this, they have attained above 500,000 new bank accounts in their bank. In the initial days, the bank had the facility to send usual text messages related to money transfer and withdrawal and these services were updated to provide its customer a complete functional banking experience ( 9 ).
Mobile banking is adopted by many consumers as it is easy to open an account, pay bills, transfer funds, do mobile-based shopping, etc. Many banks in China and America who introduced mobile banking for all their services experienced high demand for their services through mobile banking ( 10 ). But many challenges were faced by these banks due to no face-to-face interactions. Trust is an important factor in mobile banking but lack of interaction led to the risk of uncertainty and anonymity. On the other hand, the consumers also suffer from uncertainty whether the bank is trustworthy are not.
During the last decades, mobile banking became relevant all over the world. The number of fund transfers through mobile banking also increased since last year, but mobile banking has many security problems. Sometimes, the security setting of mobile phones can be overridden by the virus in smartphones ( 11 ). Many banks open their banking applications, these apps should be updated frequently otherwise the consumers can be vulnerable to the attacks such as DDoS attacks, phishing, spoofing, corporate account takeover, and skimming ( 12 ).
The major contribution of the article is:
• To analyze the benefits of this study by investigating the past.
• To explore the adoption of mobile banking technology by consumers.
The remaining article is arranged accordingly. Section Literature Review presents the related work and the hypothesis is formulated in Section Hypothesis Development and Research Methodology. The results in terms of linear regression and descriptive analysis are provided in Section Results and the article is concluded in Section Conclusion and Discussions.
In the modern world, many firms use systems of mobile banking to communicate with the customers. The payment system in mobile banking services benefited people's lives. This article aims to analyze the effect of people using the technology of mobile to make banking services more accessible. Literature on mobile banking adoption, electronic banking services, mobile payment service user acceptance are reviewed.
Hamidi et al. ( 13 ) studied the influence of mobile banking adoption on consumer engagement and satisfaction utilizing the customer relationship management (CRM) system, which is the most essential aspect in banking industry. CRM is also seen as a critical role for enhancing client satisfaction in mobile banking. The statistical study performed evaluated the dialogue between the bank's customer sector and their client. The statistical analysis findings have a favorable influence on consumer interactions and satisfaction.
Geebren et al. ( 14 ) investigated the importance of consumer satisfaction in mobile eco-systems that used electronic banking services, particularly in developing nations. This entailed researching consumer satisfaction in mobile banking, with a focus on the importance of trust. To determine consumer satisfaction, structural modeling using partial least squares (PLS–SEM) methods were employed to examine the data, and trust demonstrated that customer contentment had a beneficial influence.
The relevance of an early trust theoretical model in mobile payment service user acceptance was highlighted by Gao et al. ( 15 ). The initial trust theoretical model highlighted the facilitators and barriers to user trust in m-payment services. The links in the original trust theoretical model were assessed using partial least squares structural modeling (PLS–SEM). The findings may be used in m-payment adoption research and practice in a variety of ways. In total, 52.3% of the difference in usage intention was explained by the current model.
Hentzen JK et al. ( 16 ) offered a mobile technology that allows for vital involvement, as well as an explanation of how a retirement app might assist people in planning for their post-retirement strategy. The available literature survey data were used to evaluate a sample of 440 Australian pension fund members. The findings show that consumers' perceived financial security, financial self-efficacy, retirement planning involvement, future consequences consideration, and perceived usefulness with a mobile retirement app have direct and indirect effects on their expected engagement through their goal to adopt the app.
Zhu et al. ( 17 ) investigated existing technology designs, including mobile banking, used by rural communities in six Chinese regions. According to the findings, interpersonal and mass communication channels have a bigger influence than organizational communication channels. Mobile banking should be examined since it can assist alleviate the lack of access to financial goods and financial infrastructure in rural areas.
Afeti et al. ( 18 ) developed a mobile payment technology for payment for micro-businesses. The study draws on the transaction cost theory and the task-technology fit (TTF) theory as the assumed lens. In total, 20 micro-businesses based on qualitative data were analyzed and the research findings denote those micro-businesses adoption of mobile payments results in strategic and operational benefits.
Alamoudi et al. ( 19 ) proposed that the mobile technology acceptance lookalike was changed as we investigated consumers' acceptance of mobile shopping in general stores by examining transaction convenience, usefulness, attitudes, ease of use, transaction speed, optimism, and personal innovativeness. A total of 351 respondents completed the questionnaire evaluation. Consumers are willing to use mobile shopping channels if the system is clear and straightforward to use.
Jebarajakirthy et al. ( 20 ) used a comprehensive moderated mediation framework to evaluate the influence of online convenience aspects on mobile banking uptake. Covariance-based structural equation modeling and the process macro are utilized to test these predictions. This study examines how convenience characteristics influence mobile banking adoption intentions.
Abdinoor et al. ( 21 ) studied the adoption of mobile financial services in Tanzania with the use of a technology acceptance model. To select the sample from data collection, a random sampling technique was used. The user and non-user of mobile financial services were included in the sample. Zhang et al. ( 3 ) investigated customers' use of mobile technology to help them with banking services and activities, as well as the variables that impact their adoption and engagement. Here, the analysis is done by the structural equation modeling technique to know the consumers' intentions toward mobile banking. The result examines the adoption of mobile banking apps to facilitate bank consumers' banking services.
Tiwart et al. ( 22 ) researched the variable that influences adoption in Commercial Bank Ethiopia. The author used factors such as perceived ease of use (PEOU), infrastructure (INF), security (SEC), trust, and e-banking adoption. Here, a structural equation model based on least partial square analysis is used. The result of this analysis proves that the trust of customers mediates between PEOU, INF, SEC, and e-banking adoption and realized the factors that guess the purpose and adoption of e-banking by Ethiopian customers.
The problem identified in mobile banking services is critical financial services through mobile banking because it has security problems.
Participant's demographic information.
This study is mainly conducted in China by taking the top e-banks that utilize m-banking. The questionnaire is distributed to both the users and staff of the bank. Within 6 months, 293 questionnaires were completed. These details were taken into consideration in this study. The details of the questionnaires were obtained online and also provided to the users on their official bank pages. The information obtained in the questionnaire is provided in Tables 1 , 2 .
Table 1 . Demographic analysis.
Table 2 . Participant details.
The information of the customers is mainly obtained based on their m-bank usage. The five main banks such as state-owned, joint-stock, postal savings, commercial city bank, and the agricultural bank were selected. Each participant was asked to say about their experience of their usage.
The amount of time spent by each user on average is provided in Figure 1A . As seen in Figure 1A , more than 39.38% of users are using more than 3 h a day and a minimum of is online Conceptual Model for less than 1 h. In total, 8.21% of the users only use the application for a limited amount of time (below 30 min).
Figure 1 . Usage analysis of m-banking application (A) Users using m-banking during the day and (B) users' friends and references using mobile banking.
Figure 1B presents the percentage of people the users have referred for mobile banking. Based on Figure 1B , a total of 48.52% has more than 100 contacts who have been referred to the m-banking. The increase in m-banking contact can improve efficiency.
The main aim of this research is to identify the usage of mobile technology of consumers for the banking system. The hypothesis provided for each variable is explained in detail in this section and the conceptual model is provided in Figure 2 . The measurement instrument used for this study is presented in Table 3 .
Figure 2 . Conceptual modeling structure.
Table 3 . Measurement instrument to analyze different parameters.
Service, system, and information quality (ssiq)".
System quality mainly measures the efficiency of the overall m-banking system in terms of service provider anonymity. The system quality of the m-banking mainly depends upon the ease of use, the attractiveness of the application, trust of the user, etc. If the user trusts the system quality, then they mainly select the system for future transactions. Service Quality mainly relies on the service provided by the m-banking system and it also evaluated the effectiveness of the service in terms of personalization, reliability, delay, etc. Hence, service quality is crucial to improve the quality of m-banking since poor-service quality can result in minimal user trust and satisfaction. Information quality mainly implies that high-quality information is reliable, complete, accurate, relevant, and accessible. The information quality is an important variable that improves the usage, trust, and satisfaction of the customers in the m-banking system.
Hypothesis (H6): The service, system, and information quality had a positive relationship in improving the m-banking quality and user satisfaction.
Affective obligations between business partners reflect mental reliance on others and are founded on emotions, loyalty, and dependency. The psychological bond in this example demonstrates an emotional commitment. Support for differences in particular data among new and long-term clients. Based on their differing expectations, new and loyal clients behave in the same way.
Hypothesis (H1): The interaction of customers helps to develop a good relationship with the affective commitment of a person to the bank which is focused on interaction.
The time and attempt to acquire some benefits of services is said to be benefit convenience. Some of the components of benefit convenience are fast service, timely services, and bank employee attitudes. Sometimes in banks, the consumer needs to visit repeatedly to avail specific service. But in m-banking, consumers can avail any services on the go from the home itself. So, the time and attempt to acquire the service became very easy. Therefore, benefit convenience shows a positive result in mobile banking adoption ( 23 ). The time and attempt to contact a service provider for a particular service is said to be post benefit convenience. In m-banking, the service provider can be contacted and solve the grievances within a few clicks through various means such as email, live chat, and toll-free numbers. But in offline banking, the consumer needs to visit the branch and wait for the service provider to solve the problem in service. So, the post benefit convenience shows a positive result in mobile banking adoption ( 20 ).
Hypothesis (H5): The pre- and post-benefit convenience are positively interrelated with m-banking.
The fast and simple way to complete the transaction is said to be transaction convenience. Acquiring the service of transaction in a lesser time is the main component of transaction convenience. Other components of transaction convenience are uninterrupted transaction, transaction confirmation, easy checkout, and price inconsistency. By using m-banking, the transaction can be done in any place at any time in a few clicks and can do several transactions simultaneously. But in offline banking, consumer needs to wait in a queue for availing a transaction service. Hence, transaction convenience shows a positive result in mobile banking adoption ( 24 ).
Hypothesis (H2): The transaction convenience is positively interrelated with m-banking.
Customer participation in an organization's mobile banking may increase customer value and profitability ( 25 ).
Hypothesis (H8): Profitability shows positive results in mobile banking adoption.
Perceived reliability is specified as the limit to which the individual has independently considered the technology as a faithful one. In order to offer a trustworthy technical service, the system reliability is important that enables the user to attain the aimed objective. An undependable technology that is employed by an individual seems to be low confidence and such technology utilization should be restricted ( 26 ). Research related to mobile technology and information system has revealed that a technology with perceived reliability possesses has an important impact on customers' perceived satisfaction, detected value, and observed quality. Also, customers' understanding level of trust is examined and analyzed as an antecedent of confidence in an organization through the mobile commerce setting. Though mobile is a wireless technology, it is easily susceptible to violations or attacks ( 27 ). Mostly, the customers were insisted to share their personal data while consuming and shopping services performed through mobile technology such as date of birth, debit and credit card details, funds, and address. The reliability of mobile technology should be improved thereby increasing customers' belief in securing their personal data. The postulated hypotheses are mentioned as follows:
Hypothesis (H4): Perceived reliability in m-banking helps to improve the consumer's trust and ease of use of banking services.
In this hypothesis, trust is considered as a faith of competence, integrity, ability, and benevolence that the individual has toward each other. Trust has become an important reducing feature of risk, whereas the e-commerce connection is considered as a naturally risky factor ( 28 ). Also, trust in banking is specified that it is linked to reliability and perceived privacy. Generally, the primary interaction between the customer and banker will significantly affect the trust progress in them.
Hypothesis (H7): Based on their usage of mobile devices and attitude, the consumer trust toward M-banking increases.
Perceived ease of use evaluates the increase in the amount of work required to acquire a job utilizing new technologies on an individual basis. In the effort condition, the simplicity of use of mobile innovation is critical for including customers in the co-creation of value experiences ( 29 ). Mobile technology's ability to increase advantages for bank clients has been widely shown, and its simplicity of use is a key factor in consumer acceptability. PEU has been linked to advanced tackling new technologies in the banking industry in several research. However, the launch was difficult, and it was claimed that customers' inclination to utilize mobile banking services is not intrinsically related to employment prospects. As a consequence, it indicates that study into PEU and its influence on consumer attitudes, as well as the interactive method, are both worthwhile endeavors ( 30 – 32 ).
Hypothesis (H3): Customers' views about the usage of mobile devices for facilitating financial services are favorably associated with their perceived ease of use (PEU).
The reliability of the scales was assessed using Cronbach's alpha values. Goodness-of-fit approaches were used to assess the overall model fit for confirmatory factor analysis (CFA). The RMSEA was 0.09, the comparative fit index (CFI) was 0.99, the goodness-of-fit (GFI) was 0.98, and the standardized root mean square residual (RMR) was 0.06. The X2 to df ratio was 4.14, the root mean square error of approximation (RMSEA) was 0.09, the CFI was 0.99, the GFI was 0.98, and the standardized RMR was 0.06. As indicated in Table 4 , the reliability coefficients of all constructs were more than 0.87, which is higher than the 0.9 thresholds. The extracted average variance was used to assess convergent validity (AVE). Because the AVE values ranged from 0.58 to 0.92, convergent validity was not an issue. The AVE scores were lower than the squared correlations between pairs of constructs, indicating discriminant validity in the data set. Table 4 shows the results of the Cronbach's alpha and AVE tests.
Table 4 . Hypothesis testing.
Regression tests are run when the allowable alpha coefficient for the scales has been determined. Table 5 shows the regression test results for each of the assumptions discussed in the following sections. Affective commitment (AC) and interaction were investigated, and the correlation coefficient 0.297 and likelihood of significance 0.000 indicate that there is a positive and significant relationship between the two. The regression test supports this positive relationship, showing that the interaction variable may be able to predict AC variation.
Table 5 . The results of the linear regression test.
With a correlation value of 0.333 and a probability of significance of 0.000, the following test demonstrated a positive relationship between perceived ease of use and satisfaction (PEU). The correlation coefficient between the variables perceived reliability (PR) and interaction was 0.172, with a probability of significance of 0.000, demonstrating a positive link between the two variables. The correlation value of 0.430 and the probability of significance of 0.000 indicated a positive relationship between satisfaction and service, system, and information quality (SSIQ) ( 3 ).
The items in this questionnaire were answered using a five-point Likert scale, with the alternatives being: very low (1), low (2), medium (3), high (4), and very high (5). Table 6 shows the SD, middle value (MV), and lowest value (LV) of the descriptive data received from the surveys ( 3 ).
Table 6 . Descriptive data analysis.
Mobile banking is a cost-effective way to reach customers. The different activities that can be performed by the consumers in m-banking are checking their bank balance, making transactions, making investments, getting account statements, paying bills, etc. The motivation for user adoption and use of mobile banking is examined in this research. Positive connections between profitability, trust, and transaction convenience to utilize mobile banking have been found in previous studies. Users' confidence in using mobile banking services, enjoyment, utility of the system, etc., are not the same thing. The participants were unconcerned about the danger of fraud, system dependability, or perceived privacy while building and extending faith in their banks and mobile services. Because of today's technology-driven lifestyle, individuals are receptive to embracing new technologies that are compatible with their mobile phones.
To add value to practice, we must obtain convincing and widespread results from both users and mobile banking service providers. Like mobile phones, smartwatches have become more popular among people nowadays. With this wearable technology, banks have the greatest responsibility to fulfill hedonic and social needs. The purpose and use of this research are common to all consumers irrespective of various countries, populations, and socio-economic status. The variables used in this work are, namely, affective commitment (AC), transaction convenience (TC), perceived ease of use (PEU), perceived reliability (PR), pre- and post benefits (PPBs) service, system, and information quality (SSIQ), bank trust (BT), and profitability (P). Consumer behavior such as untrustworthiness, fraudulence, and so on, as well as mobile banking transaction risks such as financial, privacy, and cyber security, are important risk concerns that may be addressed in future research developments.
The problems and risks that arise by the consumers and banking sector for adopting mobile banking services would be reduced by the scientific researchers using various technologies and methodologies in the future. The results in this article show the different statistical surveys and believed that it is true for both mobile banking consumers and non-mobile banking consumers with various socioeconomic environments. But when we consider globally each country had its own rules and regulations regarding mobile banking transactions, services, and products. And these rules and regulations react differently in different countries.
The original contributions presented in the study are included in the article/supplementary material, further inquiries can be directed to the corresponding author/s.
Both authors listed have made a substantial, direct, and intellectual contribution to the work and approved it for publication.
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
1. Patil P, Tamilmani K, Rana NP, Raghavan V. Understanding consumer adoption of mobile payment in India: Extending Meta-UTAUT model with personal innovativeness, anxiety, trust, and grievance redressal. Int J Inf Manage. (2020) 54:102144. doi: 10.1016/j.ijinfomgt.2020.102144
CrossRef Full Text | Google Scholar
2. Martin J. 7 reasons mobile payments still aren't mainstream. (2016).
3. Zhang T, Lu C, Kizildag M. Banking “on-the-go”: examining consumers' adoption of mobile banking services. Int J Quality Serv Sci . (2018) 10:279–95. doi: 10.1108/IJQSS-07-2017-0067
4. Sharma SK, Al-Muharrami S. Mobile banking adoption: key challenges and opportunities and implications for a developing country. In: Emerging Markets from a Multidisciplinary Perspective . Springer, Cham. (2018). p. 75–86. doi: 10.1007/978-3-319-75013-2_7
5. Riquelme HE, Rios RE. The moderating effect of gender in the adoption of mobile banking. Int J Bank Market. (2010) 28:328–41. doi: 10.1108/02652321011064872
6. Sanakulov N, Karjaluoto H. Consumer adoption of mobile technologies: a literature review. Int J Mobile Commun. (2015) 13:244–75. doi: 10.1504/IJMC.2015.069120
7. Shareef MA, Baabdullah A, Dutta S, Kumar V, Dwivedi YK. Consumer adoption of mobile banking services: An empirical examination of factors according to adoption stages. J Retail Consumer Serv. (2018) 43:54–67. doi: 10.1016/j.jretconser.2018.03.003
8. Hanafizadeh P, Behboudi M, Koshksaray AA, Tabar MJS. Mobile-banking adoption by Iranian bank clients. Telemat Inform. (2014) 31:62–78. doi: 10.1016/j.tele.2012.11.001
9. Srivastava M, Fernandes S. An empirical study to measure consumer adoption of mobile banking services. Int J Public Sector Perform Manage. (2022) 9:165–89. doi: 10.1504/IJPSPM.2022.119834
10. Narteh B, Yeboah-Asiamah E, Mackin EA. Analysis of young banked and unbanked customers' usage, satisfaction, trust and loyalty for mobile money services in Ghana. Int J Bus Syst Res. (2022) 16:40–64. doi: 10.1504/IJBSR.2022.119602
11. Nur T, Panggabean RR. Factors influencing the adoption of mobile payment method among generation Z: the extended UTAUT approach. J Account Res Organiz Econ. (2021) 4:14–28. doi: 10.24815/jaroe.v4i1.19644
12. Muhal MA, Luo X, Mahmood Z, Ullah A. Physical unclonable function based authentication scheme for smart devices in Internet of Things. In: 2018 IEEE International Conference on Smart Internet of Things (SmartIoT) . (2018) p. 160–165. IEEE. doi: 10.1109/SmartIoT.2018.00037
PubMed Abstract | CrossRef Full Text | Google Scholar
13. Hamidi H, Safareeyeh M. A model to analyze the effect of mobile banking adoption on customer interaction and satisfaction: a case study of m-banking in Iran. Telemat Inform. (2019) 38:166–81. doi: 10.1016/j.tele.2018.09.008
14. Geebren A, Jabbar A, Luo M. Examining the role of consumer satisfaction within mobile eco-systems: Evidence from mobile banking services. Comput Human Behav. (2021) 114:106584. doi: 10.1016/j.chb.2020.106584
15. Gao L, Waechter KA. Examining the role of initial trust in user adoption of mobile payment services: an empirical investigation. Inf Syst Front. (2017) 19:525–48. doi: 10.1007/s10796-015-9611-0
16. Hentzen JK, Hoffmann AO, Dolan RM. Which consumers are more likely to adopt a retirement app and how does it explain mobile technology-enabled retirement engagement? Int J Consumer Stud. (2021) 46:368–90. doi: 10.1111/ijcs.12685
17. Zhu Q, Lyu Z, Long Y, Wachenheim CJ. Adoption of mobile banking in rural China: Impact of information dissemination channel. Socio-Econ Plann Sci. (2021) 101011. doi: 10.1016/j.seps.2021.101011
18. Afeti EY, Owusu A. Impact of mobile payments on micro-business activities: a developing country experience. In: Digital Innovations, Business and Society in Africa . Springer, Cham. (2022). p. 75–95. doi: 10.1007/978-3-030-77987-0_4
19. Alamoudi H. Examining Retailing Sustainability in the QR Code-Enabled Mobile Payments Context During the COVID-19 Pandemic. Int J Customer Relationship Market Manage. (2022) 13:1–22. doi: 10.4018/IJCRMM.289210
20. Jebarajakirthy C, Shankar A. Impact of online convenience on mobile banking adoption intention: a moderated mediation approach. J Retail Consumer Serv. (2021) 58:102323. doi: 10.1016/j.jretconser.2020.102323
21. Abdinoor A, Mbamba UO. Factors influencing consumers' adoption of mobile financial services in Tanzania. Cogent Busin Manage. (2017) 4:1392273. doi: 10.1080/23311975.2017.1392273
22. Tiwari P. Electronic banking adoption in Ethiopia: an empirical investigation. SN Bus Econ. (2021) 1:1–28. doi: 10.1007/s43546-021-00114-0
23. Shen YC, Huang CY, Chu CH, Hsu CT. A benefit–cost perspective of the consumer adoption of the mobile banking system. Behav Inf Technol. (2010) 29:497–511. doi: 10.1080/01449290903490658
24. Trialih R, Astuti ES, Azizah DF, Mursityo YT, Saputro MD, Aprilian YA, et al. How mobile banking service quality influence customer satisfaction of generation x and y? In: 2018 International Conference on Information and Communication Technology Convergence (ICTC) . (2018) p. 827–32. IEEE. doi: 10.1109/ICTC.2018.8539720
25. Gupta SD, Raychaudhuri A, Haldar SK. Information technology and profitability: evidence from Indian banking sector. Int J Emerg Market . (2018) 13:1070–87. doi: 10.1108/IJoEM-06-2017-0211
26. Prabhu Kavin B, Ganapathy S. EC(DH)2: an effective secured data storage mechanism for cloud based IoT applications using elliptic curve and Diffie-Hellman. Int J Internet Technol Secur Transact. (2020) 10:601–17. doi: 10.1504/IJITST.2020.10029366
27. Prabhu Kavin B, Ganapathy S. A new digital signature algorithm for ensuring the data integrity in cloud using elliptic curves. Int Arab J Inf Technol. (2021) 18:180–90. doi: 10.34028/iajit/18/2/6
28. Prabakaran S, Ramar R, Hussain I, Kavin BP, Alshamrani SS, AlGhamdi AS, et al. Predicting attack pattern via machine learning by exploiting stateful firewall as virtual network function in an SDN network. Sensors. (2022) 22:709. doi: 10.3390/s22030709
29. Ramachandran V, Ramalakshmi R, Kavin BP, Hussain I, Almaliki AH, Almaliki AA, et al. Exploiting IoT and its enabled technologies for irrigation needs in agriculture. Water . (2022) 14:719. doi: 10.3390/w14050719
30. Adam MNM, Adam MNM, Idris AAA, Ali ASM, Khalid IO Evaluation Evaluation of the effect of time of fixation and microwave treatment on quality of fatty tissue fixation in breast cancer specimens. SPR . (2021) 2:445–51. doi: 10.52152/spr/2021.166
31. Mshvidobadze T. Bioinformatics as emerging tool and pipeline frameworks. SPR. (2021) 2:361–5. doi: 10.52152/spr2021.156
32. Dani R, Juyal D, Rawal YS. A Critical analysis of the restaurant industry's effect on environment sustainability. SPR . (2021) 2:385–92. doi: 10.52152/spr/2021.159
Keywords: mobile technology, banking, financial service, transaction convenience, mobile banking
Citation: Zhu J and Wang M (2022) Analyzing the Effect of People Utilizing Mobile Technology to Make Banking Services More Accessible. Front. Public Health 10:879342. doi: 10.3389/fpubh.2022.879342
Received: 19 February 2022; Accepted: 17 March 2022; Published: 29 April 2022.
Reviewed by:
Copyright © 2022 Zhu and Wang. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY) . The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
*Correspondence: Jiale Zhu, zhujiale8899@126.com
Disclaimer: All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.
Showing 1 through 3 of 0 Related Papers
There are two dueling trends defining 21st century payments: digitization and cybercrime .
And while the two trends are unrelated — particularly given that paper-based payments are more of a fraud risk than their digital counterparts — a new research paper published last week (Aug. 14) finds that, across the digital wallet landscape, the two trends are increasingly butting heads.
The paper, entitled “In Wallet We Trust: Bypassing the Digital Wallets Payment Security for Free Shopping,” posits that an ongoing reliance on outdated authentication methods, and the prioritizing of convenience over security, has left digital wallets vulnerable to attack and abuse by bad actors.
“The implications of these attacks are of a serious nature where the attacker can make purchases of arbitrary amounts by using the victim’s bank card, despite these cards being locked and reported to the bank as stolen by the victim,” the researchers wrote.
In an era where digital transactions are becoming the norm, the convenience of digital wallets is undeniable — and as these wallets become increasingly integral to daily transactions, their security is becoming a top priority.
By adopting emerging security imperatives like biometric authentication, tokenization, end-to-end encryption, multi-factor authentication (MFA) and AI-driven detection and prevention, digital wallet providers can build robust defenses against the growing array of cyber threats.
Read more : Debit Cards In Digital Wallets Gaining Ground Across Sectors
The convenience and efficiency offered by digital wallets are undeniable, but with this convenience comes the responsibility to safeguard sensitive financial information. After all, the ease of setting up and using these wallets, combined with the lack of stringent verification processes, has made them an attractive target for illicit activities.
Against this backdrop, end-user reliance on digital wallets is only increasing. What started as a convenient alternative to physical wallets has evolved into a key financial tool, housing sensitive data and enabling seamless transactions across various platforms.
PYMNTS Intelligence finds that digital wallet usage continues to grow across grocery, retail, restaurant and travel sectors, with debit becoming the preferred underlying payment method. In June 2024, consumers paying with digital wallets used stored debit cards in 55% of grocery transactions, as well as 52% of retail transactions, 62% of restaurant transactions and 46% of travel transactions.
As digital wallets continue to gain traction in the financial ecosystem, their security will be a critical factor in determining their long-term success.
Among the emerging security imperatives that are essential for the future of digital wallets, biometric authentication has rapidly gained traction as a reliable and secure method of verifying a user’s identity. Unlike traditional passwords or PINs, which can be easily stolen or forgotten, biometric data — such as fingerprints, facial recognition, and voice patterns — are unique to each individual, making them significantly harder to replicate or misuse.
For digital wallets, the integration of biometric authentication offers a dual advantage: it enhances security while improving user experience by streamlining access to accounts and transactions.
“If you do the facial scan immediately upfront … That means all these transactions will go through seamlessly and you no longer have to confirm your identity after the fact,” Mark Nelsen , senior vice president and global head of consumer payments at Visa , told PYMNTS.
Read more: 3 Big Ideas From PYMNTS Intelligence’s Digital Wallets UK Report
Tokenization is another powerful security measure that replaces sensitive data, such as credit card numbers, with a unique identifier or “token” that can be used in transactions without exposing the actual data. This technique minimizes the risk of data breaches, as the token itself is meaningless to unauthorized parties who may intercept it.
“We think tokenized payments have the propensity to penetrate some 70% of transactions as a whole,” Mehret Habteab , senior vice president of product and solutions at Visa Europe, told PYMNTS.
In the context of digital wallets, tokenization is particularly effective in securing payment data during transactions. When a user initiates a payment, their sensitive information is not directly transmitted. Instead, a token is generated and used to complete the transaction. This approach significantly reduces the risk of fraud, as even if a hacker gains access to the token, they cannot use it to access the user’s financial information.
And for digital wallets, other methods of end-to-end encryption that involve encrypting payment data, personal information, and other sensitive details at every stage of a transaction can help ensure that even if data is intercepted, it remains unintelligible to unauthorized parties.
Of course, sometimes the simple solutions are the most effective — and for digital wallets, MFA is particularly effective in preventing unauthorized access. Even if a hacker obtains a user’s password, they would still need the second form of authentication to gain access to the wallet. This added layer of security makes it significantly harder for cybercriminals to compromise accounts.
Looking ahead, the rise of artificial intelligence (AI) and machine learning has opened new avenues for enhancing cybersecurity in digital wallets. These technologies enable real-time threat detection and continuous monitoring, allowing digital wallet providers to proactively identify and respond to potential security risks before they escalate.
Trending news, the big story, featured news, partner with pymnts.
We’re always on the lookout for opportunities to partner with innovators and disruptors.
In neighborhoods experiencing discrimination from law enforcement, police are viewed both as part of the problem and the solution.
Is it possible for marginalized communities to work together to create thriving neighborhoods? Is there a role for police officers, who are often mistrusted, in this effort?
This was the subject of the latest research from Harvard Kennedy School’s Program in Criminal Justice Policy and Management (PCJ), led by faculty director Sandra Susan Smith.
Smith, the Daniel and Florence Guggenheim Professor of Criminal Justice and the Carol K. Pforzheimer Professor at the Radcliffe Institute, formed the Roundtable on Racial Disparities in Massachusetts Criminal Courts at the PCJ in 2021. Its goal is to influence future policies, practices, and procedures to eradicate the sources of racial inequity and racial disparity in Massachusetts’s courts.
The Roundtable, under the direction of Smith, has studied issues such as jury exclusion and reducing racial disparities through decriminalization. The latest study, This is What Thriving Communities Look Like , looked at four neighborhoods in the greater Boston area to better define what is meant by “reimagining public safety.”
Smith, with Amisha Kambath, a research assistant at the PCJ who is beginning Yale Law School this fall, and Noor Toraif, an assistant professor at UPenn, wanted to understand what it meant for people to feel and be safe in their neighborhoods. Working with focus groups in Roxbury, Dorchester, East Boston and South Boston—communities that have different racial and class compositions but share a troubled relationship with law enforcement—the team gained insights on what residents need to thrive and what role policing can play.
“Ever since the murder of George Floyd in 2020 and the conviction of the Minneapolis police officer who killed him, we have heard this plea to reimagine public safety,” the authors wrote. “In working with the residents of these Boston neighborhoods, we hoped to understand what a safe, thriving community looked like to them. And how law enforcement fits into that picture.”
As Smith revealed in a recent PCJ report the residents of these Boston communities report higher rates of police harassment and as a result feel a deep distrust of law enforcement.
“We also knew that creating a thriving community required deliberate and sustained engagements with residents themselves, with a particular emphasis on those most often targeted for aggressive and harmful police interventions but least considered when questions about how to achieve public safety arise,” the researchers wrote.
Sandra susan smith.
After a series of focus group sessions, which included participants from each of the four communities, people with diverse experiences and relationships to the state, including a range of age groups as well as formerly incarcerated residents, Kambath, Toraif, and Smith identified three themes:
Inherent in thriving communities, Smith’s team found, was a sense of community cohesion and freedom from harm: as one young adult respondent put it, “everyone kind of having each other’s back.”
The research team noted one striking response when talking about a sense of community.
“Rather than look to more affluent neighborhoods as examples of community cohesion,” the researchers wrote, “these residents most often looked back in time at their own neighborhoods where they were born and raised.”
Respondents pointed out how neighbors would sit on their porches to watch the street, cleaned the areas of trash, and actively engaged with one another in public spaces—all positive signs of a thriving community.
Other respondents noted that local events can enhance a sense of belonging. A formerly incarcerated resident described the East Boston farmer’s market, saying “Every Wednesday they pass out fresh vegetables to the public. It’s a thriving place. It’s a place to bring people together.”
The study indicated that the prevailing threat to safety was easy access to guns, a fear supported by Boston police statistics for the communities. None of the respondents could envision thriving without freedom from violence.
“Violence, experienced directly or indirectly and with or without guns, impacts the physical and mental health of residents,” said Smith. “It also erodes resident’s ability and willingness to engage in ways that contribute positively to their communities.”
The focus group discussions led to the research team identifying three themes for creating a thriving community: a healthy built environment, greater investments in capital development, and access to high quality physical and mental healthcare.
“Our participating residents focused on specific aspects of the built environment,” wrote Kambath, Toraif, and Smith, pointing to greater access to green spaces, parks and recreation. Many spoke of the need for more trees. The team identified prior research that noted the benefits of green space to help reduce violent crime and found that tree canopy coverage—areas shaded by trees—were especially low in two of the focus group communities.
Neighborhood gentrification—where poorer, urban communities are revitalized by wealthier citizens, often displacing the current residents—was also systemic in these areas, contributing to housing instability for longtime residents. In fact, Boston ranks third among highly gentrified cities in the country, with the four neighborhoods in this study at risk, according to a 2020 study by the National Community Reinvestment Coalition.
Investment in human capital was also top of mind, not only enhanced, free education for the area’s youth, but skill-building opportunities for adults as well. And as one resident argued, there was a sense that jobs and contracts in these neighborhoods should go to community members first.
Residents in all four neighborhoods identified a need for far more resources to address health concerns, especially mental health and substance abuse issues, but feared that more affluent communities with far less need for public support would be privileged in the city’s decision-making.
“When imagining safe and thriving communities,” Kambath, Toraif, and Smith noted, “Residents offered a wide range of views about the role police can play.” Many saw “police as protectors” equating public safety with police.
Others identified the police as flawed, unresponsive, and failing to treat residents with respect. Nevertheless, the research found residents wanted police to be protective and believed that through training they could become better public servants. As one East Boston resident said, “We need more police officers” to provide a sense of security.
And then some respondents viewed police as “violence workers.” “They terrorize our young people” is how one formerly incarcerated Roxbury-Dorchester resident put it.
"In a related report that the PCJ released in June based on analysis of a representative sample of Boston residents,” Smith wrote, “Our analysis of this Boston survey not only revealed that police harassment is predictive of distrust and eroded feelings of community safety, but police harassment is also predictive of symptoms of trauma and associated with chronic health issues.”
“These findings, which are consistent with prior research in this area, suggest that police harassment might be a direct and indirect contributor in making communities unsafe, thwarting residents’ efforts to thrive,” she continued.
The study, while sobering on many fronts, also offered opportunities for collaboration in these communities. And residents felt they had taken the first important steps, being heard, voicing concerns, and participating in solutions.
"What comes through powerfully from these focus group meetings is the level of sophistication that residents have in identifying the very real structural barriers to thriving their communities face and offering concrete, workable solutions to overcome these,” said Smith. Responses from residents even included photographs of what a thriving community meant to them.
“They are the experts that we should be listening to. In this way, they are no different than generations of residents who have come before them. It is well past time that we take their ideas seriously."
Banner photograph by Carlin Stiehl/The Boston Globe/Getty Images; inline photograph by Matthew J. Lee/The Boston Globe/Getty Images; portrait by Martha Stewart
Focus group participants submitted photographs that describe a thriving community. The second banner photo is from an East Boston young adult: “I just think it is cool, people can sit down with their friends, or just sit down.”
Three years after police reforms, black bostonians report harassment and lack of trust at higher rates than other groups, want a jury to be fair, impartial, and engaged in higher-quality deliberations diversify the jury pool, harvard researchers say., sandra susan smith aims to eradicate disparities in criminal courts.
Get smart & reliable public policy insights right in your inbox.
COMMENTS
systematic literature revie w which aims to examine the main theoretical models and factors of adoption that. researchers used to describe the acceptance of M-banking technolog y and predict ...
This article attempts to analyse the current status of research on mobile banking in order to identify the themes to be explored by future researchers. With the deft use of different inclusion and exclusion criteria and relevant and appropriate keywords, 79 journal articles published in renowned databases were selected for an in-depth review ...
During this Covid pandemic situation mobile banking plays a very important role in banking sector throughout the world. According to RBI report during March 2020 total value of 520199 crore of ...
self-service devices such as automated teller machines (ATM), telephone banking, internet. transactions, such as account transfers, paying bills, stock exchange transactions, and. (Shaikh and ...
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 ...
This study is a systematic review of mobile banking services. Its main objective is to provide a state-of-the-art review of this particular growing type of services. It inventories and assesses the most significant determinants of and barriers to consumers' adoption of mobile banking. Moreover, it identifies the most common consequences of this ...
The initial step involves a comprehensive scan of the data and the bibliometric analysis conducted on mobile banking research papers. In this phase, researchers have set specific criteria for document selection, including the use of Scopus as a database and the application of predetermined corpus curation criteria.
Research on digital banks (Shin, 2021) and mobile banking (Petrović et al., 2022) has demonstrated that users' satisfaction level depends on the ease of using the app. Empirical research has also confirmed a direct relationship between the ease of use and continued use of mobile banking apps (Singh & Srivastava, 2020). To compare the influence ...
This study aims to analyze the keywords related to mobile banking by focusing on its development between 2003 and 2023. To do so, a systematic bibliometric analysis is undertaken, for which, the screening of Scopus database was done to identify the publication (institutions, most-prolific authors, year-wise, countries-wise and most-cited papers). Additionally, keyword occurrence analysis was ...
The aim of this research is to study the factors impacting usage of mobile banking (mBanking) by consumers in India. The study adopts the Extended Unified Theory of Acceptance and Use of Technology (UTAUT2) with Social Value (SV), Monetary Value (MV), Emotional Value (EV), Quality Value (QV), Trust and the moderation effect of gender.
Banking institutions use advanced technologies to create various electronic channels that reduce the need for traditional banks. Mobile banking (M-banking) has become a new technology, new trend, and further interest (Cheng et al., 2006; Chin et al., 2003).Banks and other financial institutions allow their customers to manage business transactions by using mobile devices, such as smartphones ...
Kim et al. conducted a systematic literature review based on 54 academic research papers in the areas of MFS, financial inclusion and developments. Unlike the previous research efforts, where the synthesis of the literature was limited to a specific MFS domain (mobile payments, mobile banking) or region (Gulf Cooperative Council countries), the ...
The purpose of this study is to explore the key dimensions of mobile banking (m-banking) service quality. Four qualitative methods (focus group, in-depth interviews, critical incident technique, and netnography) were used for data collection. Inductive content analysis and Pareto analysis were used to explore the key themes (dimensions).
This paper explores the effect of mobile banking on the banking industry. It further seeks to investigate if banks improve financial performance as well as customer conversion and retention due to mobile banking. The research sifts through early entries in mobile banking features, data transfer technology evolution along with hand-held
Policy Research Working Paper 5664. Mobile banking is growing at a remarkable speed around the world. In the process it is creating considerable uncertainty about the appropriate regulatory response to this newly emerging service. This paper sets out a framework for considering the design of regulation of mobile banking.
Findings. The importance of use and individual performance has long been recognised by academics and practitioners in a variety of functional disciplines. The present review indicates that the topics of m-banking adoption and behavioural intention dominate the majority of research, but finds very few studies on post-adoption.
2. Round-the-clock availability even functional and holidays. 3. Provides a variety of banking and value-added service. 4. In Mobile banking there is not GPRS required; works only on voice ...
Mobile banking is a cost-effective way to reach customers. The different activities that can be performed by the consumers in m-banking are checking their bank balance, making transactions, making investments, getting account statements, paying bills, etc. The motivation for user adoption and use of mobile banking is examined in this research.
Research Proposal on "Measuring Customer's Perception to Mobile Money Transfer Services in Bangladesh (Mobile Banking)" Md. Idris Ali 3/7/19 Research Methodology ffResearch Title: Measuring Customer's Perception to Mobile Money Transfer Services in Bangladesh (Mobile Banking). Course Name: Research Methodology Course Code: AIS 4205 ...
Figure 1B presents the percentage of people the users have referred for mobile banking. Based on Figure 1B, a total of 48.52% has more than 100 contacts who have been referred to the m-banking.The increase in m-banking contact can improve efficiency. Conceptual Model. The main aim of this research is to identify the usage of mobile technology of consumers for the banking system.
DOI: 10.24036/jtip.v16i2.626 Corpus ID: 265638158; Analysis of Mobile Banking Acceptance in Indonesia using Extended TAM (Technology Acceptance Model) @article{Andhika2023AnalysisOM, title={Analysis of Mobile Banking Acceptance in Indonesia using Extended TAM (Technology Acceptance Model)}, author={Imam Andhika}, journal={Jurnal Teknologi Informasi dan Pendidikan}, year={2023}, url={https ...
The study investigates the challenges, technology, and future research agenda of digital banking. The paper follows the manifestation of Kitchenham's SLR protocol. Six databases were used to ...
And while the two trends are unrelated — particularly given that paper-based payments are more of a fraud risk than their digital counterparts — a new research paper published last week (Aug ...
transactions using a mobile phone" Mobile banking is the act of doing financial transactions on a. mobile device (cell phone, tablet, etc ).and using software usually called an app provided by ...
The research team noted one striking response when talking about a sense of community. "Rather than look to more affluent neighborhoods as examples of community cohesion," the researchers wrote, "these residents most often looked back in time at their own neighborhoods where they were born and raised."
In this paper, we aim to determine customer perception about mobile banking services of banks. Customer has different views on mobile banking services provided by their service providers.