Thesis MSc Finance
Course code: EBM866B20
Author: Querien Ornstein S2503603
q.j.e.ornstein@student.rug.nl
Supervisor: Prof. Dr. B.W. Lensink
The Effect of National Culture on the Adoption of Mobile Money in Emerging Markets
ABSTRACT
Mobile money has enormous potential for financial inclusion, but the variance in adoption rates differs enormously. The central question in this thesis is how national culture affects mobile money adoption rates. I use the four Hofstede dimensions, Power Distance, Uncertainty Avoidance, Individualism and Masculinity to estimate the effects of culture mobile money adoption. The Global Findex Database is used to measure mobile money adoption, from which both the aggregated data and the microdata are extracted. The two data types allow me to assess the effects of culture in two ways. In the first model, the effects of culture on the national mobile money adoption are estimated. Using Multiple Imputation replace the missing culture dimensions, the coverage of the Hofstede data is extended from 43
countries to 76 countries. In the second model, the effect of national culture on the individual probability to adopt mobile money is estimated. I find small, but significant evidence for two out of the four dimensions. Uncertainty Avoidance has a negative effect on the individual probability to adopt money. This implies that in high uncertainty avoidance cultures, providers must ensure and clearly communicate the security of their agent- and infrastructure networks.
Keywords: Financial inclusion, Mobile money, National culture, Hofstede dimensions, Multiple imputation, Multilevel logit
I. Introduction
According to the president of the World Bank Group, Jim Yong Kim, the key to eradicating poverty is to achieve universal access to financial services. Technological development plays a crucial role in achieving this goal (World Bank, 2016). Especially mobile technology provides great potential for financial inclusion. More people have access to a mobile phone than the amount of people having an account at a financial institution. Mobile phones are considered a bridge between the unbanked and access to financial services, in the form of mobile money. A basic mobile money service requires little technological advancement. Transactions can be done using any phone that can send and receive SMS texts using the cellular network (Kapoor, Morduch & Ravi, 2007). Users put money on their accounts by purchasing some form of credit by local retailers or ‘agents’, as opposed to depositing funds at a bank branch. The same retailers also buy the credits back from consumers in exchange for cash (Jack & Suri, 2011; Hughes & Lonie, 2007).
The low technological requirements combined with the fact that the frequency of local agents is generally higher than the number of bank branches in an area1, lowers barriers to use the service and increases accessibility and availability to financial services.
Evidence shows the positive potential of mobile money. One of the biggest success stories was the launch of a mobile money service in Kenya called M-PESA in 2007. In 2009, 70% of the households in Kenya had adopted this service (Jack & Suri, 2011). Jack and Suri (2011; 2016) estimate that the introduction of mobile money has lifted 194,000 families in Kenya out of poverty and has improved the welfare and the position of women in Kenya.
This success is not observed in each country where mobile money is introduced and the adoption rates of mobile money differ greatly per country. Evans and Pirchio (2014) show that in most cases, either a service grows explosively and succeeds or a service does not lift off at all. Where models comparable to M-PESA has succeeded in other countries such as Paraguay, Honduras and El Salvador, such models have failed in other countries such as South Africa (Evans & Pirchio, 2014; Mas & Radcliffe, 2011).
Many scholars take interest in the disparities in success of mobile money and have examined the facilitating conditions of uptake of mobile money, for example by studying the facilitating conditions of the success in Kenya (see, e.g. Jack & Suri, 2016; Mbiti & Weil, 2015; Jack &
Suri, 2014; Kikulwe, Fischer & Qaim, 2014; Buku & Meredith, 2013; Demombynes & Thegeya, 2012; Hughes & Lonie, 2007).
Literature on the adoption of mobile money has two main focuses. The first focus is identification of the effects of macro-economic conditions, such as the regulatory framework and the infrastructural requirements in a nation (e.g. Souter, Ferguson and Neubert (2019). This approach is useful to compare uptake rates across countries, but these approaches often fail to incorporate demand (Donovan, 2014).
The second focus is identification of the individual characteristics of potential customers. A traditional model to identify the adoption probability with respect to a new technology is the Technology Acceptance Model (TAM). The TAM approach uses a survey instrument to quantify a respondent’s opinion regarding the perceived ease of use and the perceived usefulness of a technology, based on the respondent’s experience with a specific product. These two constructs lead to a measured of behavioral intent to use, which in turn predicts actual use (Davis, Bagozzi and Warshaw, 1989). Many scholars have applied this model on the adoption of mobile money (e.g. Wilson & Mbamba, 2017; Chauhan, 2015; Marumbwa, 2014; Chitungo & Munongo, 2013; Lule, Omwansa & Waema; 2012; Tobbin & Kuwornu, 2011; Mas & Morawczynski, 2009). While these studies are successful in predicting the uptake of mobile money, the findings are often limited to the application of the technology that was used in the study.
National culture is a concept often used to explain aggregate (consumer) behavior in a country and to compare differences in behavior between countries. Because measures of national culture are based on the values of individuals and are constructed of commonalities within countries and the differences between countries, it is a valuable tool to identify structural differences in behavioral patterns. Evidence shows that national culture explains many phenomena, for example individual spending patterns, the entrepreneurial development of a country and the adoption of new technologies (e.g. Petersen, Kushwaha & Kumar, 2015; Dwyer, Mesak & Hsu, 2005; Hayton, George & Zahra, 2002). It is often recognized that cultural aspects play a significant role in the mobile money deployment process, both as a barrier and as an enabling factor (Bankole et al., 2011). Based on this tentative evidence for the role of culture, in this thesis I aim find to what extent national culture explains the differences between mobile money adoption rates.
country is analyzed or the number of countries compared is often very small (Wamuyu, 2014; Potnis, 2014). I contribute to current literature by examining the effects of national culture on mobile money adoption over a large amount of countries and providing an empirical model for comparison. Moreover, I contribute to literature by proposing an instrument to capture the unobservable selection effects that are often ignored in mobile money adoption studies (Aron, 2017).
This study highlights the cultural differences in mobile money acceptance. This is relevant, because which improves understanding about the demand potential of an application. For mobile money providers, it is crucial to assess demand potential before entering a market. Donovan (2014) states that one of the complications of the mobile money market that providers must convince both agents and customers to sign up in order to generate a viable network. Understanding the potential cultural barriers can influence the value proposition offered by mobile money service providers.
The paper is structured in the following way. In section II, several theories on national culture are discussed. Building onto the concepts from literature, I formulate the hypotheses. In section III, the variables included in the model are presented. The methodology and estimation procedure is described. In section IV, the data used to test the hypotheses are described. Section V the results of the analysis are show and discussed. In the final section VI, the concluding remarks, limitations and implications for further research are stated.
II. Theoretical framework II.I. National culture
Culture is an extensively studied concept in many areas of research, varying from social anthropology to accounting. In terms of the scope of this thesis, I limit the exploration of theories on culture to main theories frequently cited in studies related to finance, business and economics. In the subsequent section, I will discuss different prominent theories and their accompanying methodologies. Based on this discussion of these theories, I will discuss the rationale behind choosing the framework by Hofstede. In section II.II, the empirical evidence with respect to the effects of national culture will be discussed. Building on the evidence provided, I will formulate the hypotheses and summarize this is in a conceptual framework.
II.I.I. Hofstede’s cultural dimensions theory
marketing, and his work has been cited over 9,000 times in peer-reviewed journals (Steenkamp, 2001). According to Hofstede, culture is defined as the collective programming of the mind that distinguishes the members of one group from the others (Hofstede, 2001; 1980). In his first work, Hofstede distinguished four dimensions of national culture: Power distance, Uncertainty Avoidance, Individualism and Masculinity (Hofstede, 1980). In later years, Hofstede extended his framework with two additional dimensions: Long-term versus Short-term Orientation and Indulgence versus Restraint (Hofstede & Minkov, 2001).
Power distance can be defined as ‘the degree to which the less powerful members of a society accept and expect that power is distributed unequally’ (Hofstede, 1980). This dimension relates to the need for justification of a hierarchical order. In societies where power distance is low, people value an equal distribution of the power and will most probably demand justification where power is distributed unequally.
Uncertainty avoidance is defined as ‘the degree to which the members of a society feel uncomfortable with uncertainty and ambiguity.’ (Hofstede, 1980). Cultures that score high on uncertainty avoidance are more dependent on formalized policies and generally have a higher intolerance for change. In low uncertainty avoidance cultures, young people are respected more than in high uncertainty avoidance cultures. It is also generally more accepted for people to question the things they are taught (Hofstede, 2001).
A highly individualist society can be defined as ‘a preference for a loosely-knit social framework in which individuals are expected to take care of only themselves and their immediate families’. Conversely, people in a highly collectivist society place more value on a ‘tightly-knit framework in society in which individuals can expect their relatives or members of a particular ingroup to look after them in exchange for unquestioning loyalty’ (Hofstede, 1980). The level of individualism refers to the degree to which an individual in a society prefers to act as an individual rather than members of a social group. In very individualistic societies, social ties are weaker as people are primarily concerned with their own self-interest, but this must not be confused with egocentrism. People in individualist societies do care for others, but generally because they want to and not because they feel obliged to. This also shows in the dynamics with immediate family (Hofstede, 2001).
preference for cooperation, modesty, caring for the weak and quality of life. Society at large is more consensus-oriented.’.
The Hofstede dimensions are not without flaws, and there are various other important frameworks that must be considered.
While it must be kept in mind that the entirety of a nation’s culture cannot be represented by a limited set of constructs, the dimensions as determined by Hofstede do provide a simple, practical and usable way to integrate culture into a study (Soares, Farhangmehr & Shoham, 2007)
II.I.II. Other studies on national culture
In order to provide a more comprehensive view on national culture, I discuss two other important theories on national culture. The first is the framework established in the Global Leadership and Organisational Behavior Effectiveness (GLOBE) project. The GLOBE-project started as a replication study of the Hofstede dimensions and lead to a recalculation and extension of the Hofstede dimensions. In the study, culture is defined as the ‘shared motives, values, beliefs, identities, and interpretations or meanings of significant events that result from common experiences of members of collectives that are transmitted across generations.’ (House et al., 2004).
Based on surveys among 17,000 middle managers in 62 cultures, researchers of the GLOBE-project identified nine dimensions of national culture. The first six dimensions based on Hofstede’s four-dimension framework. These dimensions are Power Distance, Uncertainty Avoidance, Institutional Collectivism, In-Group Collectivism, Gender Egalitarianism, Assertiveness. The authors added three additional dimensions: Performance Orientation, Future Orientation and Humane Orientation. These dimensions were each scored based on two sub-dimensions, practices and values. Scores on practices represent the actual situation, whereas scores on values represent the culture dimension as they should be ideally according to respondents. The scores on the dimensions correlate moderately to highly with the original Hofstede dimensions. The exact correlations are shown in Appendix I, table 6.
Countries with high secular-rational values place less importance on religion and tradition. Controversial topics such a divorce, abortion, euthanasia and suicide are considered more acceptable. The second dimension regards the values Survival versus Self-expression. Survival values imply that these nations place emphasis on economic and physical security. These societies are characterized by low levels of trust and tolerance. Societies scoring high on self-expression values are more concerned with environmental and social issues, such as the tolerance with respect to foreigners and gender equality. The dimensions scores are calculated based on each wave of the World Values Survey. Plotting the country scores on the first dimension on the y-axis and the second dimension on the x-axis leads to the creation of a cultural map of the world, in which clusters are identified based on observable characteristics such as language, religion or law system (Inglehart & Welzel, 2013). The most recent map is shown in Appendix I, figure 7. A simplified conclusion of their framework is that developing countries are situated at the bottom left of the plane, and with their development move towards the upward-right corner of the plane.
When comparing the aim, scope and purpose of the three models, Hofstede is the most suitable approach to assess the effect of culture on mobile money adoption. The aim of the GLOBE-study is to identify the characteristics of leadership while Inglehart and Welzel model the change over time. In this thesis, I am interested in the aggregate behavior of a representative sample of the world population. Hofstede (2001) argues that the dimensions are constant over time and if they change, they do so constantly and slowly. Furthermore, the data coverage for Hofstede is larger compared to the other two frameworks in emerging countries. The Inglehart-Welzel framework is not appropriate for this specific analysis, because it mainly considers the change of culture and how countries move from one situation at the other. Because most countries of interest in this thesis are situated in the lower left corner, at this specific point in time, there is little variance to explain phenomena such as mobile money.
II.II Empirical evidence and hypotheses II.II.I. Uncertainty avoidance
uncertainty avoidance are expected to embrace new technologies. The opposite relationship is found when examining the empirical evidence on this relationship. In societies where there is a low level of uncertainty avoidance, people are more accepting of new solutions to problems and are more comfortable with non-conformity to social norms, which implies a general higher propensity to adopt new services (Shane, Venkataraman & MacMillan, 1995). Hofstede and De Mooij (2002) demonstrate that cultures with low uncertainty avoidance were among the first to adopt the internet, while high uncertainty avoidance cultures lagged in their adoption of the internet. Yenigurt and Townsend (2003) argue that low uncertainty avoidance cultures are generally entrepreneurial and op en to adopt new products. The authors indeed find this effect, and find a significant effect between national culture on the adoption of internet usage, cellular phone usage and PC usage, moderated by socio-economic variables. Apart from the penetration rates of new products, uncertainty avoidance affects the innovativeness of individuals. Steenkamp, Hofstede and Wedel (1999) used a consumer-specific innovation measure to evaluate the effect of national culture on innovativeness, accounting for the respondent’s personal values. They show that uncertainty avoidance has a negative impact on the innovativeness of an individual, even accounting for the personal values of an individual. A similar study by Yaveroglu and Donthu (2002) argue that societies characterized by high uncertainty are more rigid and more resistant towards change. They test their hypothesis by calculating then correlation between the dimensions of national culture with the innovation coefficient derived from the Bass model for innovation diffusion. Their results are consistent with their expectations. The level of uncertainty avoidance correlates negatively with the innovation coefficient. The imitation coefficient however is positive, which indicates that societies with high lower distance will adopt after- and if the innovation proves to be successful. A similar result is found in the study by Lee, Trimi and Kim (2013). They find that cultures with low uncertainty avoidance, people are more prone to adopt in the first stage of the innovation launch.
The evidence on the effect of uncertainty avoidance on technology adoption shows that uncertainty avoidance has a different effect depending on the newness and availability of the product. When a service is newly available and there is no sufficient proof of its usefulness or reliability, high uncertainty avoidance cultures are likely to refuse. The effect is different when there is a measure of certainty with respect to the performance of the new service. I expect to find that if mobile money is not widely available in a country, the adoption rate is lower in countries with high uncertainty avoidance and higher in countries with low uncertainty avoidance.
H1b: Mobile money availability has a negative moderating effect on the relationship between uncertainty avoidance and the adoption of mobile money.
II.II.II. Individualism versus collectivism
Of the six dimensions determined by Hofstede, Individualism has shown to have in general the most predictive power (Beugelsdijk, 2012). Taylor and Wilson (2012) have studied the effect of individualism on scientific progress and technological innovation. They measure innovation in terms of the citations-weighted technology patents per capita and the citations-weighted scientific publications (per capita) and control for wealth, military spending, trade openness, education and R&D spending. They authors find a significant and positive effect of individualism on the and technological innovativeness of a nation. Yaveroglu and Donthu (2002) study the effect of culture on the innovation adoption behavior of individuals. The authors argue that in countries with high individualism, people trust their own judgement and are therefore independent in their decision whether or not to adopt an innovative technology. They show that the coefficient for innovation is significantly and positively correlated with individualism. The opposite is true for the coefficient of imitation. This coefficient correlates negatively with the adoption of technology. Similar results are shown by Lee et al (2013). The authors apply two types of innovation diffusion patterns with respect to mobile phone adoption. They find that in a highly collectivist culture such as South-Korea, uptake is initially low, but increases rapidly when consumers start to imitate each other.
In emerging countries, another effect might be of importance. In developing countries, a key source of external income is in the form of remittances. The act of sending domestic remittances is one of the drivers of demand fueling the growth of mobile money (Ricardo Martinez-Cid, Gonzalo Pernas, 2017; Munyegera & Matsumoto, 2016; Mbiti & Weil, 2011). Mas and
Morawczynski (2009) argue that the deep-rooted cultural practice of sending home money is indeed one of the enablers of M-PESA in Kenya. One of the drivers for this obligation to send money to family is collectivism. Triandis, Brislin and Hui (1988) pose that collectivist societies experience a high level of family obligations, manifesting itself into moving or migrating under unfavorable conditions to be able to provide for the remainder of the family. Given this evidence, I expect that the relationship between domestic remittances as a main driver for mobile money demand and actual mobile money adoption is stronger in collectivist countries compared to individualist societies.
H2b: Individualism/collectivism has a negative moderating effect on the relationship between domestic remittances and the adoption of mobile money.
II.II.III. Power distance
pagers and PCs). Their findings show that power distance has a significant negative effect on cell phone adoption and internet adoption. On an individual level, power distance has a negative effect on the relationship between the perceived usefulness and the attitude towards new IT applications (Alshare, Mesak, Grandon & Badri, 2011).
In high power distance countries, I expect that, because individuals are less trusting of institutions and more reluctant towards new technology, mobile money adoption is lower. In low power distance countries, where the trust in institutions is higher and people expect others not to be primarily driven by their own self-interest, I expect to find a higher uptake of mobile money.
H3: Power distance has a negative effect on mobile money adoption. II.II.IV. Masculinity versus femininity
The masculinity versus femininity dimension can manifest itself in different ways. The study by Doney et al. (1998) shows that in masculine societies, people are generally expected to be more driven by opportunism and primarily care about their own self-interest. They argue that in masculine cultures, people expect others to differ in their capabilities and that it therefore differs who can actually deliver on their promises. Husted (1999) finds that higher masculinity is correlated with a higher level of corruption. This could imply that in a masculine society, people are more cautious to engage in new financial practices, because there is no general trust in the intentions of agents. Conversely, masculinity might fuel innovativeness. Laukkanen and Cruz (2012) examine the effects of culture on mobile banking adoption. They perform a logistic regression analysis and find a significant and positive effect of masculinity on mobile banking adoption. The authors find that respondents associate mobile banking with technological advancements and the possession of a new gadget. This could imply that new (financial) technology is used as a symbol of status. Steenkamp (1999) finds that consumers in masculine societies are more innovative compared to more feminine societies. Although the evidence on this dimension is mixed, I expect to find that higher masculinity and the higher level of innovativeness contributes to a higher uptake of mobile money.
III. Data and methodology III.I. Data
III.I.I. Mobile money adoption
In order to measure the uptake of mobile money, data from the Global Findex Database is used. The aim of the Global Findex Database is to measure financial inclusion and to examine the potential and effects of FinTech innovations. The database aims to collect data on indicators on individual financial behavior and indicators can be used to provide insights on how adults save, borrow, make payments and manage risk. The Global Findex database is compiled using nationally representative surveys of more than 150,000 adults from over 140 countries (Demircic-Kunt et al, 2018). Data has been collected every three years since 2011. Mobile money is measured on two levels. For the scope of this thesis, mobile money is defined as a monetary value stored on a mobile device, that can be used either to conduct transactions or is accepted as a means of payment (Di Castri, 2013). I measure the rate of adoption of mobile money using the indicator from the Global Findex survey. In the wave of 2017, respondents were asked:
“In the PAST 12 MONTHS, have you, personally, used a mobile phone to make payments, to buy things, or to send or receive money using a service such as [local example of mobile money from GSMA database, like M-PESA]?” (Global Findex Database, 2018).
III.I.II. National culture
I use the Hofstede dimensions as a measure for national culture. The original four dimensions were determined in 1984 using IBM survey data, which covers more than 100,000 employees from the 50 biggest countries. Hofstede later added two more dimensions, which will not be included in this thesis because of their limited data availability and lack of empirical evidence. The total scope of the dataset was extended up until 93 countries in 2010. Data on the culture dimensions in publicly available at Hofstede Insights.
III.I.III. Domestic Remittances
last 12 months and 0 otherwise. Whether the respondent has received domestic remittances in the past 12 months is coded likewise.
III.I.IIII. Mobile Money Availability
Taylor and Horst (2013) argue that in order to examine mobile money adoption the development stage of the market must be included in the model. I use mobile money agent density as a measure of the extent to which a mobile money service has spread over a country (Evans & Pirchio, 2015). Agent Density is measured as the number of active mobile money agent outlets per 100,000 adults. I choose to use active mobile money outlets over registered mobile money outlets because the latter might lead to an overstatement of the actual availability (Aron, 2017).
Control variables
I include control variables in my analysis to ensure internal validity of the analysis. The control variables included are determinants of mobile money adoption as found in previous literature. I distinguish three categories of control variables: respondent characteristics; respondent’s financial and infrastructural characteristics; and country characteristics. A complete overview the of the data source per variable is provided in appendix II.
Respondent characteristics
There is a strong body of evidence with respect to the effect of respondent characteristics on mobile money adoption. In this analysis, I have included age, gender, within-country income quintile, whether a respondent is in the workforce and education, following the approach of comparable studies (see, e.g. Munyegera & Matsumoto, 2016a; Munyegera & Matsumoto, 2016b; Kikulwe et al, 2014; Gutierrez & Singh, 2013; Demombynes & Thegeya, 2012).
Financial and infrastructural characteristics
The availability of a financial system is powerful determinant for mobile money adoption. I use commercial bank density to control for the extent to which a formal financial system is established in a country. Evans and Pirchio (2015) find that mobile money is more successful in countries with limited banking systems and limited transportation. The uptake of mobile money is lower in countries where there are ample bank outlets. In nations where the market for financial services is more fragmented, the market for mobile money is more mature. According to GSMA intelligence, the relationship between the financial account ownership and mobile money adoption in a country has the form of an inverted U-shape. In countries with a medium level of financial institution accounts, the uptake of mobile money is highest.
National characteristics
Population density is positively related to the number of mobile money accounts. The reason for this result is that with higher population density, less agents are needed in order to reach the same number of people compared to countries where people live further away.
I include population density and urbanization rate to control for the effect that countries where a larger percentage of the population lives in rural areas, it is harder to reach this group. Higher urbanization and population density facilitate the creation a viable agent network (Heyers & Mas, 2011). I further control for GDP per capita.
III.II. Sample selection III.II.I. National dataset
The final sample on a national level was generated in the following manner. The mobile money dataset recorded 93 countries with mobile money uptake, 17 of these countries had little or no information on any of the other (control) variables. Multiple Imputation is not suitable for cases where virtually all information is lacking, therefore, these countries were excluded. The macroeconomic sample after imputation consisted of 76 complete observations.
Multiple imputation
(Van Buuren, 2018). Multiple imputation estimates the values of the missing cases in such a way, that the sample as a whole retains is original statistical properties.
To impute a dataset reliably, the entire model must be included. If the variables include with the model are not related to the reason of missingness, the model can be supplemented using auxiliary variables2 can be added. For this study, I can be quite certain that the reason for missing values for culture cannot be attributed to the other (control) variables in the model and I have therefore supplemented the model using several auxiliary variables. These variables correlate with the culture dimensions and can therefore improve the quality of the imputation. A description of the auxiliary variables and their correlations with the culture variables can be found in appendix II.
I adhere to the imputation process as specified by Rubin (1987). There are three steps in multiple imputation. The first step is to impute the missing values multiple times using a specific imputation algorithm. Imputation results in m datasets, which are then used to estimate the coefficients m times. In the third step, the results from all separate estimations are pooled, by calculating the mean, variance and confidence interval. Multiple imputation therefore incorporates and adjusts for the uncertainty with respect to the imputed variable.
In this thesis, I have used the Classification and regression trees (CART) algorithm (Breiman et al, 1984). The CART algorithm is a machine learning algorithm and a robust and flexible method for multiple imputation. CART methods are able to handle a different distributions and outliers. Using the CART algorithm, imputed values are drawn m times from a distribution. In this case, I have imputed the data five times with 50 iterations per draw. For each of the five imputed datasets, potential missing values were drawn 50 times from a distribution. In the seconds stage, for each of the datasets, the coefficients were estimated five times. The third stage is to pool the results of these five analyses. This ensures reliable estimates and confidence intervals. A description of variables included in the multiple imputation model and presentation of model fit is provided in Appendix X
III.II.II. Microdata
The microeconomic sample was constructed using the Global Findex Microdata from the 43 countries with both mobile money uptake data and values for the culture dimensions. This resulted in a dataset of 49,837 observations. I limited the sample to respondents age 15 and over. This resulted in a database of 49,666 observations. Following the approach of
Sommet and Morselli (2017), I have mean-centered the variables before performing the analyses.
III.II. Methodology
In this study, the aim is to quantify and identify the effect of culture on the uptake of mobile money. In this section, I elaborate on how I have quantified this relationship. I consider and test two approaches frequently used to model the adoption of mobile money. The first approach is to compare adoption rates on a across countries and assess whether culture explains the difference in these uptake rates. The second approach is to examine the effect of culture on the individual adoption probability. In this section, I will discuss both methods and formulate the model.
III.II.I. The effect of culture on the national mobile money adoption rate
The aim of my analysis is to quantify the effect of national culture on mobile money adoption. In the first analysis, I compare the national adoption rates across countries using a linear regression analysis. Equation 1 denotes the regression model specified to test this relationship.
𝑌" = 𝛽0+ 𝑿𝒊𝜷 + 𝒁𝒊𝜸 + 𝜀" (1)
In equation 1, 𝑌" represents the mobile money uptake rate per country i, with 𝑖 = 1 … 𝑁. 𝛽3 represents the intercept. 𝑿𝒊 represents a vector of the four culture dimensions for country i and 𝒁𝒊
represents a vector of the control variables for country i. 𝜷 and 𝜸 represent the corresponding vectors of estimated coefficients. In this model, I assume that the Hofstede culture dimensions are linearly related to the mobile money uptake rate.
To estimate the interaction effect between Uncertainty Avoidance and Agent Density, I extend the model with an interaction term using the approach of Baron and Kenny (1986). The model including the interaction is specified in equation 2. In this equation, the 𝐴𝐷" represents the agent density in country i. 𝛽6 represents the estimated coefficient of agent density. (𝑈𝐴𝐼")(𝐴𝐷") denotes the interaction term and 𝛽; denotes the corresponding estimated coefficient.
As mentioned briefly, the dataset used for this model is the complete imputed dataset. For reliable estimates, the coefficients are estimated five times, using the five complete imputed datasets. The results of the five iterations are pooled according to the procedure specified by Van Buuren (2018, p. 41). I estimate the coefficients using Ordinary Least Squares.
III.II.II. The effect of culture on individual mobile money adoption rate
I specify a second model to examine the effects of culture on the individual mobile money adoption probability. In order to do so, I must incorporate the fact that the variables included in the model are measured on two different levels. Failure to incorporate the country clustering in the model will lead to a violation of the assumption of independence of the residuals (Bressoux, 2010). I specify a multilevel binomial logit model with two levels using the procedure by Sommet and Morselli (2017), to estimate the effects of culture on the adoption probability of mobile money. The lower-level (level 1 variables) are the individual characteristics of respondents. The higher-level variables (level 2 variables) are the culture dimensions and the country-specific controls. The parameters for these the independent variables are estimated with fixed effects.
The model can be expressed in two steps shown in the equations (3) and (4). 𝑌"<∗ = 𝛽
3<+ 𝜷><𝒙𝒊𝒄+ 𝜖"< (3) 𝛽3< = 𝛾33+ 𝜸3>𝒁𝒄 + 𝛿3< (4) Where 𝑌"<∗ represents the observed choice of the respondent whether or not to adopt mobile money, 𝒙𝒊𝒄 denotes a vector of the respondent characteristics with 𝜷< representing the corresponding vector of estimated coefficients and 𝜖"< representing the random intercept variance. The random intercept variance indicates the log odds deviation per country given the log odds of the overall intercept that an individual adopts mobile money. In equation (3), intercept 𝛽3< represents the fixed slope for country-effects, and is predicted as formulated in equation (4). In this equation, 𝛾33 represents the fixed intercept, overall all countries and respondents. 𝒁𝒄 represents a vector of culture variables and a control variable on country-level, with corresponding vector of estimated coefficients 𝜸3>. 𝛿3< represents the country-specific variance. Substituting equation (4) in equation (3) and rewriting the vector of coefficients of level 1 effects as 𝛾>3 leads to reformulation of the model as stated in equation (5).
𝑌"<∗ = 𝛾
Equation (5) can ultimately be transformed into choice probability 𝜋"<, expressed in equation (6)
𝜋"< = 𝐹(𝛾33+ 𝜸𝟎𝟏𝒁𝒄+ 𝜸𝟏𝟎𝑿𝒊𝒄 + 𝛿3<) (6) The coefficients are estimated using a Generalized Least Squares estimator, which uses maximum likelihood estimation.
IV. Descriptive statistics
In this section, the characteristics of the dataset will be explored in more detail. First, I pay attention to the main variables of analyses. Subsequently, a description of the control variables included in the analysis will be provided.
IV.I. Mobile money adoption rates
To analyze the uptake of mobile money, I have compared the uptake rates over the two samples. The comparison is visualized in figure 2. The boxplot in figure 2 shows that there is one outlier. This outlier is the mobile money adoption rate in Kenya.
I have performed a Welch Two-sample t-test to evaluate if the samples are significantly different. The test-statistic shows that the two samples are not significantly different. Excluding the countries for which there are no original Hofstede dimensions available does not change the representation of the distribution
of mobile money. The descriptive statistics in table 1 are based on the 76-country sample. The summary statistics show that in Sub-Saharan Africa, the Middle East & North Africa, East Asia & the Pacific and South-Asia, uptake is the highest in lower middle income countries. In the regions of Latin America & Caribbean and Europe & Central Asia, uptake is higher in upper middle income countries. Mobile money adoption is by far the highest in Sub-Saharan Africa.
Figure 2 - Sample distribution comparison of mobile money adoption rates
Table 1 – Descriptive statistics of mobile money per region and income group
Region Income n Mean Std. dev. Min Max
East Asia & Pacific Lower middle income 6 6.57% 7.69% 0.69% 21.90% Upper middle income 2 9.57% 1.85% 8.26% 10.88%
Total 8 7.32% 6.68% 0.69% 21.90%
Europe & Central Asia Lower middle income 3 5.02% 4.12% 2.20% 9.76% Upper middle income 3 7.26% 7.93% 2.38% 16.41%
Total 6 6.14% 5.79% 2.0% 16.41%
Latin America & Caribbean Low income 1 13.52% NA 13.52% 13.52% Lower middle income 5 4.58% 2.04% 2.13% 7.12% Upper middle income 10 7.04% 8.06% 2.42% 28.88%
Total 16 6.67% 6.69% 2.13% 28.88%
Middle East & North Africa Lower middle income 4 1.38% 0.65% 0.64% 2.04% Upper middle income 2 15.26% 15.61% 4.21% 26.30%
Total 6 6.01% 10.02% 0.64% 26.30%
South Asia Low income 1 0.91% NA 0.91% 0.91%
Lower middle income 4 8.14% 9.02% 1.99% 21.25%
Total 5 6.69% 8.45% 0.91% 21.25%
Sub-Saharan Africa Low income 19 23.04% 13.16% 0.32% 50.58% Lower middle income 9 25.82% 21.93% 4.01% 72.93% Upper middle income 5 27.21% 16.37% 5.63% 43.58%
Total 33 24.43% 15.94% 0.32% 72.93%
IV.II. National culture
terms of socio-economic development. It is therefore according to expectations that countries in this sample are generally more collectivist. The world average for Uncertainty Avoidance is 65.4, which is close to an average score of 63.5 in my sample. The same is true for Masculinity. The world average for Masculinity is 49.1, which is the same as the average value of masculinity in this sample. A comparison between the distribution of the culture dimensions included in this sample and entire Hofstede dataset is visualized in figure 3.
Table 2 – Descriptive statistics of Hofstede culture dimensions
Culture variables Mean St. dev.
Power distance Country score on power distance, on a scale from 1-100. 73.7 (12.3) Uncertainty Avoidance Country score on uncertainty avoidance, on a scale from 1-100. 63.5 (18.0) Masculinity Country score on masculinity versus femininity. On a scale from
1-100, -low scores (<50) indicate a tendency towards femininity, high scores (>50) indicate a tendency towards masculinity.
49.1 (13.7)
Individualism Country score on individualism versus collectivism. On a scale from 1-100, -low scores (<50) indicate a tendency towards collectivism, high scores (>50) indicate a tendency towards individualism.
27.3 (14.3)
Figure 3 – Boxplot of the Hofstede culture dimensions
The full table including the control variable descriptions and corresponding descriptive statistics of the control table is provided in appendix IV, table 2.
All Sample All Sample All Sample All Sample
V. Results
V.I. Results linear regression
V.II. Results multilevel binary logit
In table 4 the results for the multilevel binary logit are shown. I have provided the coefficients for three models. Model (i) and (ii) are baseline models including the interaction term for remittances and individualism. Model (iii) is the full model which includes all control variables. I have also estimated an empty model, for comparison of model fit and calculation of the intra-class correlation, and a baseline model, which only includes the culture variables. The results for these analyses are provided in Appendix V, table 10.
Based on the null-model, I have determined the intra class correlation coefficient. For this dataset, the intra class correlation is 0.3850, which implies that 38.5% of the between-country variance can be explained solely through the country differences in mobile money adoption. I do not find significant effects for the direct effect of the culture variables in model (i) through (iii). I do find slightly significant negative effects for the interaction effect between individualism and whether or not the respondent has sent or received domestic remittances. This effect is small, but robust; it holds after accounting for the effect of the other control variables. The interaction effect can be visualized as shown in Appendix V, figure 10 and 11.
Table 3 – Linear regression results
(i) Baseline model
(ii) Baseline model
with interaction (iii) Full model
Variable name 𝛽 SE 𝛽 SE 𝛽 SE Intercept 0.457* (0.180) 0.2543* (0.128) -0.022 (0.159) Power Distance -0.002 (0.002) -0.0022 (0.001) -0.001 (0.001) Masculinity -0.001 (0.001) 0.0005 (0.001) 0.001 (0.001) Individualism -0.000 (0.001) -0.0003 (0.001) -0.001 (0.001) Uncertainty Avoidance -0.002 (0.001) -0.0007 (0.001) -0.001 (0.001) Uncertainty Avoidance x Agent Density 0.0004 (0.000) -0.001 (0.002)
Control variables Agent density 0.0000** (0.0000) 0.00036*** (0.0001) % of population banked 0.1382 (0.0986) Mobile subscriptions 0.0006 (0.0000) Bank density -0.0002 (0.0019) GDP per capita -0.0000 (0.0000) Population density -0.0001 (0.0001) Urbanization -0.0010 (0.0013) Region
Europe and Central Asia 0.0530 (0.0704)
America and the Carribean 0.0839 (0.0811)
East and North Africa 0.0999 (0.0790)
South Asia 0.0527 (0.0686)
Sub-Saharan Africa 0.1744*** (0.050)
Table 4 – Results multilevel logit
i) Interaction 1 ii) Interaction 2 Full Model – no interaction
Variable name 𝛽 SE 𝛽 SE 𝛽 SE Intercept -2.098*** (0.197) -2.099*** (0.197) -2.64*** (0.224) Power Distance -0.0217 (0.019) -0.0219 (0.019) -0.0216 (0.020) Masculinity -0.0109 (0.015) -0.0108 (0.015) -0.0122 (0.016) Uncertainty Avoidance -0.0178 (0.011) -0.0179 (0.011) -0.0012 (0.013) Individualism 0.0032 (0.015) -0.0032 (0.015) -0.0002 (0.016) Received remittances x Individualism -0.0062** (.015)
Sent remittances x Individualism -0.0062** (0.002) -0.0049** (0.002)
Respondent characteristics
Received domestic remittances 1.2344*** (0.037) 1.2292*** (0.037) 0.949*** (0.038) Sent domestic remittances 1.0044*** (0.037) 1.050*** (0.037) 0.778*** (0.040)
In the workforce 0.4193*** (0.043)
Gender (female) -0.2675*** (0.038)
Mobile phone 0.9477*** (0.057)
Age -0.0175*** (0.001)
Account at a financial institution 0.7334*** (0.044)
Education
(reference= secondary education)
Primary education or less -0.294*** (0.044)
Tertiary education or more 0.604*** (0.065)
Income Quintile (reference = Middle 20%) Poorest -0.285*** (0.068) Second -0.121* (0.064) Fourth 0.026 (0.058) Richest 0.218*** (0.055) Country characteristics GDP per capita -0.0001* (0.000) AIC 23072.9 23072.8 20634.0 BIC 23151.8 23151.8 20826.5 Log likelihood -11527.4 -11527.4 10295.0
V.III. Results of the control variables
I have controlled for several characteristics of both countries and individual respondents. On a national level, agent density is significantly related to mobile money adoption. With each additional agent per 100,000 adults, the percentage of people adopting mobile money increases with 0.34%. An average increase of 10 agents therefore implies that the national adoption rate increases with 3.4%. I have also controlled for the region. The region that differs significantly is Sub-Saharan Africa. On average, countries in Sub-Saharan Africa have an adoption rate of 17.44% higher compared to the other countries. In the multilevel logit, I find a negative significant effect of GDP per capita. The probability an individual adopts mobile money in a specific country decreases with each increase of GDP per capita.
Respondent-specific characteristics are of great importance on the probability to adopt mobile money. People who are more likely to adopt mobile money are male, young, in the workforce, have tertiary education or more, are in the highest income quintile, own a mobile phone and have an account at a financial institution. Furthermore, people who either send or receive domestic remittances are more likely to adopt mobile money. These findings are all conform the findings on previous studies with respect to mobile money adoption (see, e.g. ).
Figure 5 - Predicted marginal probabilities for uncertainty avoidance
VI. Conclusion and discussion VI.I. Conclusion
I must conclude that the predictive power of the Hofstede dimensions of national culture is quite low. I have summarized the results with respect to acceptance of my hypotheses in table 5. For the dimensions of power distance and masculinity, no effect is found and I must reject my hypotheses. A possible explanation for the lack of effect for power distance is that the variability for this dimension is very low in my sample compared to the full set of countries as scored by Hofstede. As seen in figure 3, the exclusion of high-income countries eliminated countries that score low on power distance. Another possible explanation is the observed effect of difference in effects for high power distance versus low power distance. This could point that there is another interaction effect that I have not identified in this study.
new technology, as it represents some sort of status or achievement. On the other hand, highly feminine societies are associated with strong community ties and caring for family, which might drive the demand for mobile money and improve the network effects.
I identify small but significant effects for the other two dimensions tested. Uncertainty avoidance has a small, significant effect in the baseline logit model. I partly accept hypothesis H1a, because the result is not robust when comparing the effect to the full model. Combined with the predicted marginal probabilities, I conclude that a high level of uncertainty avoidance in a country decreases the probability that an individual in that country will adopt mobile money, but the evidence with respect to a the effect low level of uncertainty is inconclusive. In countries with a very high tendency towards uncertainty avoidance, such as Benin, Guatemala, Costa Rica and Iraq, more effort must be placed in convincing potential customers about the security and reliability of the mobile network.
Furthermore, I find a small, but significant negative interaction effect in for individualism on the relationship between both sending and receiving remittances and the adoption of mobile money. This implies that in countries with high individualism, this effect is weaker. In more collectivist countries, the people who send or receive domestic remittances have a larger probability to adopt mobile money. This could be explained through the fact that in collectivist countries, people feel more obliged to take care of family and therefore are more recipient to new ways that allows them to do so. An implication of the result is that mobile money has a
Table 5 – Hypothesis and confirmation
Effect Hypothesis Model
Linear regression Multilevel binary logit
Uncertainty Avoidance
Direct effect H1a Reject Partly accept
Interaction with availability H1b Reject NA
Individualism
Direct effect H2a Reject Reject
Interaction with remittances H2b NA Accept
Power Distance
Direct effect H3 Reject Reject
Masculinity
Direct effect H4 Reject Reject
higher probability to succeed in more collectivist countries where there is a strong tradition of sending or receiving domestic remittances.
When considering the main question asked in this thesis, whether or not national culture affects mobile money adoption, the answer remains unclear. I have observed that the predicted marginal probabilities with respect to countries of high power distance, high uncertainty avoidance and high collectivism behave more uniformly in terms of mobile money adoption probability compared to countries where there is low power distance, low uncertainty avoidance and low collectivism. While the causal evidence is of limited size, I do conclude that my analysis lays the groundwork for further analysis of the role of culture.
VI.I Limitations of this study
There are several limitations that affect this study. The first is related to the use of the Hofstede dimensions as a measure for culture. Limitations are related to both the coverage of the Hofstede dimensions as well as to the instrument itself. The Hofstede culture dimensions have a limited coverage in Sub-Saharan Africa, and in this region, the mobile money usage is highest. I have tried to overcome the mismatch in data availability by using multiple imputation techniques. While multiple imputation is a state-of-the-art way to handle missing data, the values do not resemble the actual values per country. Therefore, I cannot form a conclusion regarding the effect on culture on mobile money adoption in individual countries.
Suggestions for further research
VII. Literature
Akour, I., Alshare, K., Miller, D., & Dwairi, M. (2006). An exploratory analysis of culture, perceived ease of use, perceived usefulness, and internet acceptance: The case of
Jordan. Journal of Internet commerce, 5(3), 83-108.
Alshare, K. A., Mesak, H. I., Grandon, E. E., & Badri, M. A. (2011). Examining the
moderating role of national culture on an extended technology acceptance model. Journal of Global Information Technology Management, 14(3), 27-53.
Aron, J. (2018). Mobile money and the economy: a review of the evidence. The World Bank Research Observer, 33(2), 135-188.
Bagchi, K., Hart, P., & Peterson, M. F. (2004). National culture and information technology product adoption. Journal of Global Information Technology Management, 7(4), 29-46. Bagozzi, R. P. (2007). The legacy of the technology acceptance model and a proposal for a paradigm shift. Journal of the association for information systems, 8(4), 3.
Bahia, K. & Muthiora, B. (2019). The Mobile Money Regulatory Index. GSM Association, retrieved from:
https://www.gsma.com/mobilefordevelopment/wpcontent/uploads/2019/03/The-Mobile-Money-Regulatory-Index-1.pdf
Baskerville, R. F. (2003). Hofstede never studied culture. Accounting, organizations and society, 28(1), 1-14.
Baron, R. M., Kenny, D. A., 1986. The moderator-mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. Journal of personality and social psychology 51, 1173-1182.
1469-Buku, M. W., & Meredith, M. W. (2013). Safaricom and M-Pesa in Kenya: financial inclusion and financial integrity. Washington Journal of Law, Technology and Arts, 8, 375.
Cardon, P. W., & Marshall, B. A. (2008). National culture and technology acceptance: The impact of uncertainty avoidance. Issues in Information Systems, 9(2), 103-110
Chauhan, S. (2015). Acceptance of mobile money by poor citizens of India: Integrating trust into the technology acceptance model. info, 17(3), 58-68.
Chitungo, S. K., & Munongo, S. (2013). Extending the technology acceptance model to mobile banking adoption in rural Zimbabwe. Journal of Business Administration and Education, 3(1).
Clausen, B., Kraay, A., & Nyiri, Z. (2011). Corruption and confidence in public institutions: Evidence from a global survey. The World Bank Economic Review, 25(2), 212-249.
Davis, J. H., & Ruhe, J. A. (2003). Perceptions of country corruption: Antecedents and outcomes. Journal of Business Ethics, 43(4), 275-288.
De Mooij, M., & Hofstede, G. (2011). Cross-cultural consumer behavior: A review of research findings. Journal of International Consumer Marketing, 23(3-4), 181-192.
De Mooij, M., & Hofstede, G. (2002). Convergence and divergence in consumer behavior: implications for international retailing. Journal of retailing, 78(1), 61-69.
Demirguc-Kunt, A., & Klapper, L. (2012). Measuring financial inclusion: The global findex database. The World Bank.
Di Castri, S. (2013). Mobile money: Enabling regulatory solutions. Available at SSRN 2302726.
Dwyer, S., Mesak, H., & Hsu, M. (2005). An exploratory examination of the influence of national culture on cross-national product diffusion. Journal of International
Marketing, 13(2), 1-27.
Erumban, A. A., & De Jong, S. B. (2006). Cross-country differences in ICT adoption: A consequence of Culture?. Journal of world business, 41(4), 302-314.
Etim, A. S. (2014). Mobile banking and mobile money adoption for financial inclusion. Research in Business and Economics Journal, 9, 1.
Evans, D. S., & Pirchio, A. (2014). An empirical examination of why mobile money schemes ignite in some developing countries but flounder in most. Review of Network
Economics, 13(4), 397-451.
Getz, K. A., & Volkema, R. J. (2001). Culture, perceived corruption, and economics: A model of predictors and outcomes. Business & society, 40(1), 7-30.
GSMA, “State of the Industry Report: Mobile Money” (2015);
www.gsma.com/mobilefordevelopment/wp-content/ uploads/2016/04/SOTIR_2015.pdf
Gutierrez, E., & Singh, S. (2013). What regulatory frameworks are more conducive to mobile banking? empirical evidence from findex data. The World Bank.
Hayton, J. C., George, G., & Zahra, S. A. (2002). National culture and entrepreneurship: A review of behavioral research. Entrepreneurship theory and practice, 26(4), 33-52.
Hess. 2018. The Global Findex Database 2017: Measuring Financial Inclusion and the Fintech Revolution. Washington, DC: World Bank.
House, R. J., Hanges, P. J., Javidan, M., Dorfman, P. W., & Gupta, V. (Eds.). (2004). Culture,
Hofstede, G. (2001). Culture’s consequences: Comparing values, behaviors, institutions, and organizations across nations (2nd ed.). Thousand Oaks, CA: Sage.
Hofstede, G. (2011). Dimensionalizing cultures: The Hofstede model in context. Online readings in psychology and culture, 2(1), 8.
Hughes, N., & Lonie, S. (2007). M-PESA: mobile money for the “unbanked” turning cellphones into 24-hour tellers in Kenya. Innovations: technology, governance, globalization, 2(1-2), 63-81.
Hung, C. L., & Chou, J. C. L. (2014). Examining the cultural moderation on the acceptance of mobile commerce. International Journal of Innovation and Technology Management, 11(02), 1450010
Husted, B. W. (1999). Wealth, culture, and corruption. Journal of international business studies, 30(2), 339-359.
Ifinedo, P. (2009). Information technology security management concerns in global financial services institutions: is national culture a differentiator?. Information Management &
Computer Security, 17(5), 372-387.
Igbaria, M., Zinatelli, N., Cragg., and Cavaye, A. (1997). Personal Computing Acceptance Factors in Small Firms: A Structural Equation Model. MIS Quarterly 21(3), 279-302.
Inglehart, R., & Welzel, C. (2005). Modernization, cultural change, and democracy: The human development sequence. Cambridge University Press.
Jack, W., & Suri, T. (2011). Mobile money: The economics of M-PESA (No. w16721). National Bureau of Economic Research.
Kapoor, M., Morduch, J., & Ravi, S. (2007). From microfinance to m-finance innovations case discussion: M-PESA. Innovations: Technology, Governance, Globalization, 2(1-2), 82-90
Kendall, J., Maurer, B., Machoka, P., & Veniard, C. (2011). An emerging platform: From money transfer system to mobile money ecosystem. Innovations: Technology, Governance, Globalization, 6(4), 49-64.
Kikulwe, E. M., Fischer, E., & Qaim, M. (2014). Mobile money, smallholder farmers, and household welfare in Kenya. PloS one, 9(10), e109804.
Lee, S. G., Trimi, S., & Kim, C. (2013). The impact of cultural differences on technology adoption. Journal of world business, 48(1), 20-29.
Lule, I., Omwansa, T. K., & Waema, T. M. (2012). Application of technology acceptance model (TAM) in m-banking adoption in Kenya. International Journal of Computing & ICT Research, 6(1).
Makulilo, A. B. (2015). Privacy in mobile money: Central Banks in Africa and their Regulatory Limits. International Journal of Law and Information Technology, 23(4), 372-391.
Marumbwa, J. (2014). Exploring the moderating effects of socio-demographic variables on consumer acceptance and use of mobile money transfer services (MMTs) in Southern Zimbabwe. American Journal of Industrial and Business Management, 4(2), 71.
Mas, I., & Morawczynski, O. (2009). Designing mobile money services lessons from M-PESA. Innovations: Technology, Governance, Globalization, 4(2), 77-91.
Mbiti, I., & Weil, D. N. (2015). Mobile banking: The impact of M-Pesa in Kenya. In African Successes, Volume III: Modernization and Development (pp. 247-293). University of Chicago Press.
Minkov, M. (2012). Cross-cultural analysis: the science and art of comparing the world's modern societies and their cultures. SAGE publications.
Minkov, M. (2018). A revision of Hofstede’s model of national culture: old evidence and new data from 56 countries. Cross Cultural & Strategic Management, 25(2), 231-256.
Munyegera, G. K., & Matsumoto, T. (2016). Mobile money, remittances, and household welfare: panel evidence from rural Uganda. World Development, 79, 127-137.
Petersen, J. A., Kushwaha, T., & Kumar, V. (2015). Marketing communication strategies and consumer financial decision making: The role of national culture. Journal of
Marketing, 79(1), 44-63.
Porteous, D. (2009). Mobilizing money through enabling regulation. Innovations: Technology, Governance, Globalization, 4(1), 75-90.
Potnis, D. D. (2014). Examining mobile banking in developing nations from pro‐poor “context, culture, and community” perspective. Proceedings of the American Society for information science and technology, 51(1), 1-4.
Santos, H. C., Varnum, M. E., & Grossmann, I. (2017). Global increases in individualism. Psychological science, 28(9), 1228-1239.
Soares, A. M., Farhangmehr, M., & Shoham, A. (2007). Hofstede's dimensions of culture in international marketing studies. Journal of business research, 60(3), 277-284.
Soutter, L., Ferguson, K., & Neubert, M. (2019). Digital Payments: Impact Factors and Mass Adoption in Sub-Saharan Africa. Technology Innovation Management Review, 9(7).
Srite, M., & Karahanna, E. (2006). The role of espoused national cultural values in technology acceptance. MIS quarterly, 679-704.
Steenkamp, J. B. E. (2001). The role of national culture in international marketing research. International marketing review.
Steenkamp, J. B. E., Ter Hofstede, F., & Wedel, M. (1999). A cross-national investigation into the individual and national cultural antecedents of consumer innovativeness. Journal of marketing, 63(2), 55-69.
Suri, T. and Jack, W. (2016). The long-run poverty and gender impacts of mobile money. Science, 354(6317), 1288-1292.
Taras, V., Steel, P., & Kirkman, B. L. (2016). Does Country Equate with Culture? Beyond Geography in the Search for Cultural Entities. Management International Review. 56(4): 455- 472.
Taylor, E. B., & Horst, H. A. (2013). From street to satellite: Mixing methods to understand mobile money users. In Ethnographic praxis in industry conference proceedings (1 (9), 88-102.
Taylor, M. Z., & Wilson, S. (2012). Does culture still matter? The effects of individualism on national innovation rates. Journal of Business Venturing, 27(2), 234-247.
Triandis, H. C., Brislin, R., & Hui, C. H. (1988). Cross-cultural training across the
individualism-collectivism divide. International journal of intercultural relations, 12(3), 269-289.
Wilson, V., & Mbamba, U. O. (2017). Acceptance of mobile phone payments systems in Tanzania: Technology acceptance model approach. Business Management Review, 20(2), 15-25.
World Bank. (2018, April 19). Financial Inclusion on the Rise, But Gaps Remain, Global Findex Database Shows [Press release]. Retrieved from
https://www.worldbank.org/en/news/press-release/2018/04/19/financial-inclusion-on-the-rise-but-gaps-remain-global-findex-database-shows.
World Bank (2016). World Development Report 2016: Digital Dividends. Washington, DC: World Bank. Retrieved from: https://www.worldbank.org/en/publication/wdr2016.
Yaveroglu, I. S. and N. Donthu. (2002). “Cultural Influences on the Diffusion of New Products,” Journal of International Consumer Marketing, 14(4), 49–63.
Appendix I – Additional exploration of differences and similarities of theories on national culture
Table 6 - Correlation matrix culture dimensions
GLOBE Project Hofstede (2001)
PDI UAI IDV MAS LTO
Appendix II – Multiple imputation
Figure 8 a-f - Density plots of imputed values
Figure 9 - Description of the CART algorithm (Van Buuren, 2018, p86)
(f)
(a) (b) (c) (d)
Appendix III – Data sources multiple imputation and analyses Table 7 – Data sources per variable
Variable name Data source Providers
Mobile money adoption Mobile money Adoption– country level
Global Findex Database World Bank
Culture
Uncertainty Avoidance Hofstede Hofstede Insights
Masculinity Hofstede Hofstede Insights
Individualism Hofstede Hofstede Insights
Power distance Hofstede Hofstede Insights
Country characteristics
GDP per capita World Development Indicators 2017 World Bank Population World Development Indicators 2017 World Bank Population density World Development Indicators 2017 World Bank Urbanization World Development Indicators 2017 World Bank
Infrastructure
Percentage banked Global Findex Database 2017 World Bank Bank density Financial Access Survey 2017 IMF Cell phone penetration World Telecommuncation/ICT
Development database 2017
International Telecommuncation Union
Mobile money agent density Financial Access Survey 2017 IMF Respondent characteristics
Age Global Findex Microdata Database World Bank Gender Global Findex Microdata Database World Bank Education Global Findex Microdata Database World Bank Workforce Global Findex Microdata Database World Bank Income Quintile Global Findex Microdata Database World Bank Mobile phone ownership Global Findex Microdata Database World Bank Account at a financial
institution
Global Findex Microdata Database World Bank
Sends remittances Global Findex Microdata Database World Bank Receives remittances Global Findex Microdata Database World Bank
World Governance Indicators World Bank
Human Development Index United Nations Development Programme
Human Freedom Index Cato Institute, the Fraser Institute, and the Liberales Institut at the Friedrich Naumann Foundation for Freedom.
Human Capital Index World Bank
Appendix IV – Additional descriptive statistics
Table 8 Variable definitions and descriptive statistics
Variable name Definition Country-level
sample
Individual-level sample
Countries in sample 76 43
Country characteristics
GDP per capita Gross Domestic Product per capita (x 1000) in US dollars
11.0 (14.44)
10.96 (8.04)
Population Country population (x1,000,000) 56.66 (158.336)
92.49 (221.20) Population density Country population per squared kilometer. 228.28
(910.69)
155.56 (208.42) Urbanization Percentage of population residing in an urban
area 52.34 (21.01) 54.89 (20.69) Infrastructure
Percentage banked Percentage of the population with an account at a financial institution
44.05% (22.8%)
-
Bank density Number of commercial bank branches per 100,000 adults
12.11 (10.58)
-
Cell phone penetration Number of mobile subscriptions and active prepaid accounts per 100,000 adults.
101.48 (32.70)
-
Mobile money agent density
Number of active mobile money agent outlets per 100,000 adults 235.98 (215.98) - Respondent characteristics Mean3
Age Respondent age in years - 38.24
(16.62)
Gender 1 if female, 0 if male. - 0.554
Primary education or lower
1 if respondent has finished primary education or lower, 0 otherwise
- 0.464
Variable name Definition Country-level
sample
Individual-level sample
Secondary education 1 if respondent has finished secondary education, 0 otherwise
Tertiary education or higher
1 if respondent has finished tertiary education or higher, 0 otherwise
- 0.084
Workforce 1 if respondent is currently in the workforce, 0 otherwise
- 64.8%
Income Quintile Respondent income, measured as the within-country income quintile
- Lowest 20% - 0.172 Second 20% - 0.179 Middle 20% - 0.190 Fourth 20% - 0.209 Highest 20% - 0.249 Mobile phone ownership
1 if respondent has a mobile phone, 0 otherwise
- 0.770
Account at a financial institution
1 if respondent has an account at a financial institution, 0 otherwise
-
0.478 Sends remittances 1 if respondent has sent remittances in the
past year, 0 otherwise.
- 0.190
Receives remittances 1 if respondent has received remittances in the past year, 0 otherwise.
- 0.230
Appendix V – Additional results
Table 9 - Comparison coefficients for complete and imputed data
(i) Incomplete data (ii) Imputed data
