**3. Data and Methods**

This study employs data from a Kenyan nationally representative digital credit survey conducted in 2017. The sample constitutes 3130 participants among Kenyan mobile phone users. The data are the property of the Central Bank of Kenya, Kenya National Bureau of Statistics, and Financial Sector Deepening Kenya, and the authors obtained permission to use the data for this study. Since the survey was conducted telephonically, the sample was drawn from mobile phone users in Kenya and was weighted to be representative of mobile phone owners in the country. Given that one can only use digital credit when one has access to a mobile phone, the subsample that reported to have used digital credit can be considered as representative of digital credit borrowers in the country. Out of the total sample (3130), about a third (1040) reported that they are digital credit users, and about 29% (304) of these digital credit users were identified as bettors. The key questions in the instrument that enabled us to assess the likelihood of being financially distressed included whether the participant was ever late in repaying a loan they took from the mobile phone, whether they received an SMS from the lender as a reminder for repayment on an overdue balance, and whether they were ever in a situation when payments were due on multiple loans at the same time and could not make all payments. For the welfare-undermining coping strategies, information on whether participants had to sell assets to pay loans or borrow to repay loans was gathered. Welfare outcomes included going without food or without medicine or medication that was needed.

Descriptive statistics are used in understanding the sample studied and the occurrence of betting and loan repayment behavior. The Chi-square test for association is used to ascertain if there is a relationship between the outcome variables and explanatory variables without controlling for other factors (Chamboko et al. 2017). Univariate logistic regression is used to determine if there is a relationship between the outcome variables and the individual explanatory variables. Further, multivariate logistic regression is used to check for an association between betting and the outcomes variables (proxy measures of financial distress, undesirable coping strategies, and welfare outcomes), controlling for education, age, gender, locality, and income. For the multivariate analysis component, the following specifications are implemented using the binary logistic regressions:



	- *SoldAssetsi* = *β*<sup>0</sup> + *β*1*Bettori* + *β*2*Agegroupi* + *β*3*Genderi* + *β*4*Educationi* +*β*5*Localityi* + *β*<sup>6</sup> *Incomegroupi* + *ε<sup>i</sup>*

$$\begin{array}{c} \textit{WithoutMeds}\_{i} = \beta\_{0} + \beta\_{1} \textit{Return}\_{i} + \beta\_{2} \textit{Agggroup}\_{i} + \beta\_{3} \textit{Gender}\_{i} + \beta\_{4} \textit{Enduction}\_{i} \\ \quad + \beta\_{5} \textit{Locality}\_{i} + \beta\_{6} \textit{Incomegroup}\_{i} + \ \varepsilon\_{i} \end{array}$$

where the main explanatory variable *bettor* is a binary (Yes/No) derived from a survey question "Have you tried any of the digital betting services?". A more detailed description of the other covariates in the models is provided in Table 1.


**Table 1.** Bivariate relationship between betting and socio-demographic variables.

The outcomes variables are derived from the survey questions and are defined as follows:

*LatePayment*: Have you ever been late in repaying a loan that you took from your phone? *ReceivedSMS*: Received SMS from the lender to encourage repayment on your overdue balance? *MultipleLoans*: Have you ever been in a situation when payments were due on multiple loans at the same time and you could not make all payments?

*SoldAssest*: Sold assets or belongings to pay loan?

*BorrowToPay*: Borrowed to pay loan?

*WithoutFood*: In the last 12 months, how often have (you) or your family gone without enough food to eat?

*WithoutMeds*: In the last 12 months, how often have (you) or your family gone without medicine or medical treatment that was needed?

The welfare variables *WithoutFood* and *WithoutMeds* required respondents to indicate the frequency at which they experienced the situations. Four options were provided, and from these, a binary variable indicating whether or not the person experienced the situation was constructed. All responses given as "often; sometimes or rarely" equal 1, and "never" equal 0.

#### **4. Results and Discussion**

As shown in Table 1, fifty-five percent of the participants were males, and 51% lived in rural areas. About 28% had primary or no formal education, 46% had secondary education, and 26% had tertiary education. Middle-aged people (26–45 years) dominated the sample (69%), while the remaining 31% was shared almost equally between those under 26 years and those above 45 years of age. In terms of income, more than half (59%) of the sample earned 10,000 shillings or less (1 US \$ ≈ 108 Kenyan shillings), 22% earned 10,001–20,000 shillings, 13% earned 20,001–40,000 shillings, and 7% earns 40,000 shillings or more. Table 1 also shows significant variation of betting by participants' gender, education, age, and income group (Chi-square tests). About 39% of males and 20% of females were bettors. Almost half of those with tertiary education, about a fifth of those with

secondary school, and almost a third of those with primary or no formal education were bettors, a pattern that suggests that betting is common among educated adults. Except for those in the age range of 46–55 years, for all other age categories, more than a fifth of the participants were bettors, and more than a quarter of each income group reported betting. Schmidt (2019, 2020) emphasized that gambling in Kenya is viewed as a legitimate and transparent way of earning a living and is motivated by limited employment and income-earning opportunities, but it is also viewed as a future-income-earning opportunity, especially among the affluent. These are all possible explanations why betting is high among the educated, as well as young and middle-aged adults, and cuts across all income groups.

The bivariate relationships between betting and financial distress, as well as undesirable coping strategies and selected welfare outcomes, are presented in Table 2. The prevalence of having multiple payments due at the same time and not being able to make all payments was 20%, that of receiving an SMS as a reminder for delayed due payment is 53%, and that of being late in repaying digital loans was 48%. In terms of undesirable coping mechanisms, the prevalence of borrowing to pay an existing loan was 16%, while that of selling assets/belongings to be able to repay the loan was 5%. With respect to welfare outcomes, the prevalence of having gone without food at some stage is 29%, while that of going without required medicine is 22%. Chi-square tests were employed to check on the associations of these covariates and outcomes. Bettors (57%) were significantly more likely to receive an SMS for their digital credit repayment from a lender to encourage repayment on overdue balance than non-bettors (51%) (*p* = 0.065). Bettors (54%) were also more likely than non-bettors (45%) to be late in repaying a loan taken through their phones, and this relationship is statistically significant (*p* = 0.007). The results also reveal that bettors (25%) were significantly more likely than non-bettors (17%) to have payments that were due on multiple loans at the same time and to be unable to make all payments (*p* = 0.004).

Regarding betting and coping mechanisms, the results show that bettors (8%) were more likely to sell assets or belongings to pay loans compared to non-bettors (4%), and this relationship is statistically significant (*p* = 0.015). However, bettors (18%) and nonbettors (15%) did not show any differences in terms of borrowing to repay loans (*p* = 0.211). The bivariate analysis between betting and selected welfare outcomes did not show any significance, although the percentages for going without food (30.9% vs. 29.6%) and without needed medicine or medication (23% vs. 21%) were higher for bettors compared to non-bettors.

Table 3 presents the univariate and multivariate association between betting and *digital credit repayment* outcomes. The results of interest are those from the multivariate regressions. After controlling for income, age, gender, location (rural/urban), and level of education, the results show that bettors were almost twice more likely than non-bettors to have payments due on multiple loans at the same time and could not make all payments, and this relationship was significant (odds ratio (OR) = 1.84, *p* = 0.002). Similarly, bettors were almost one and half times more likely than non-betters to receive an SMS from a lender encouraging repayment on the overdue balance, and this association is statistically significant (OR = 1.4, *p* = 0.043). Bettors were also one and a third significantly more likely than non-bettors to be late in repaying a digital loan (OR = 1.33, *p* = 0.072).

As is reported in Table 4, after controlling for income, age, gender, locality (rural/urban), and level of education, being a bettor is significantly associated with selling assets or belongings in order to pay loans. In fact, bettors are more than twice as likely as non-bettors to do so (OR = 2.39, *p* = 0.012).

When "being a bettor" is the explanatory variable for a binary logistic regression model and "going without food" is an outcome, and income, age, gender, locality (rural/urban), and level of education (Table 5) are controlled for, bettors are one and half times more likely than non-bettors to have gone without food at some point in the past 12 months, and this association is significant (OR = 1.56, *p* = 0.017). However, going without medication has no significant association with being a bettor.




*Int. J. Financial Stud.* **2021**, *9*, 10


**Table 3.** Association between betting and digital credit repayment.


**Table 4.** Association between betting and coping mechanism.

Notes: Inc = income group; level of significance: *p* < 0.01 \*\*\*, *p* < 0.05 \*\*, *p* < 0.10 \*.



Notes: Inc = income group; level of significance: *p* < 0.01 \*\*\*, *p* < 0.05 \*\*, *p* < 0.10 \*.

The risks associated with gambling in general are well-known, and Effertz et al. (2018) posit that the discussion about the gambling risks is as old as gambling itself. Research on online gambling is relatively recent, however, and Papineau et al. (2018) put the time frame of research focusing on online gambling and public health concerns as having started about twenty years ago, with the advancement in technology facilitating online gambling. In the current study, digital financial services, facilitated by accessibility of advanced technology-savvy gadgets such as smartphones and tablets, allow for ease of access to gaming and online applications for gambling. Effertz et al. (2018) argue that gaming and online applications for gambling are faster, more attractive, and less costly, yet they are more addictive when compared to traditional gambling opportunities. Zhang et al. (2018) find that mobile phones, especially smartphones, are the most commonly used platforms for online gambling among Asian individuals. Black et al. (2017) and Papineau et al. (2018) report gambling problems to have more adverse effects among online gamblers compared to offline gamblers. The current study reports such negative effects of digital betting on credit repayment, coping mechanisms, and welfare outcomes in a digitized developing economy.

The self-reported financial distress measures show that bettors have a higher likelihood of becoming financially distressed when compared to non-bettors. Mihaylova et al. (2013) and Håkansson and Widinghoff (2020) also report that online gambling is associated with problem gambling, overspending, and over-indebtedness. Online gambling as a behavioral addiction (Mallorquí-Bagué et al. 2017) is in our study found to be associated with negative outcomes. Participants in our study embraced financial innovations that are accessible via mobile phones and tablets, thus making finances accessible through digital means and at the same time have the opportunity to gamble. This puts them in a position to easily engage in online or digital gambling.

The current study also indicates that betting is associated with undesirable coping mechanisms as shown by the tendency to use assets or other belongings to repay loans among bettors. These undesirable coping mechanisms are exemplary of what Black et al. (2017) and Papineau et al. (2018) consider as extra burden impacts of online gambling on the lives of gamblers. Together, the association between betting and being financially distressed and engaging in undesirable coping mechanisms, as well as the risk of going without food among bettors, suggest an impaired quality of life among bettors. A study by Papineau et al. (2018) shows that online gambling impacts gamblers' work, relationships, mental and physical health, finances, and quality of life. Although financial innovations such as digital financial services improve peoples' lives, for the betting segment of the population, the negative effects of betting pose a considerable threat given that it enables problem gambling, Black et al. (2013) and other earlier researchers argue to be a public health problem that is costly to the society.

#### **5. Conclusions**

Digital financial services and, more importantly, mobile money, have become an important financial innovation to advance financial inclusion in developing and emerging economies. A growing body of literature also reports a positive and significant impact of digital financial services on household welfare outcomes. Nevertheless, to the growing betting segment of the Kenyan population, digital financial services have brought great convenience to betting by allowing easy access to digital credit that can be used for betting. Using survey data from Kenya, this study shows that digital betting is associated with undesirable outcomes on credit repayment, coping mechanisms, and the welfare of bettors. When controlling for socio-economic and demographic factors, bettors were shown to be more likely than non-bettors to be financially distressed, engage in welfare undermining coping strategies, and have inferior welfare outcomes. These findings suggest the need for educating the public about the possible effects of betting and gambling in general.

This study has some limitations. First, it only shows associations between betting and identified outcomes, and it does not infer causal relationships. As such, studies that can isolate the effects of betting on these outcomes using careful identification strategies are needed. The second limitation in the data is that there is no specific survey question that captures the amount of the wagers. For example, a 100 shillings wager every day, though higher in frequency, is less significant than a 2000 shillings wager three times a week. In addition, no information was gathered with respect to an increase in the amount of a wager over time. These data limitations can be addressed by developing a specific questionnaire that gathers detailed information with regard to gambling and welfare outcomes in Kenya.

**Author Contributions:** R.C. contributed to this article in terms of conceptualization, formal analysis, funding acquisation, methodology, visualization and reviewing and editing of the manuscript. S.G.'s contribution is on conceptualization, funding acquisition, methodology, writing up original draft and reviewing and editing the manuscript. All authors have read and agreed to the published version of the manuscript.

**Funding:** This project benefits from the research funds from the Institute for Intelligent Systems at the University of Johannesburg and the University of the Free State. The authors are very grateful.

**Acknowledgments:** The authors acknowledge the insightful comments from the unknown reviewers as well as the Central Bank of Kenya, Kenya National Bureau of Statistics & Financial Sector Deepening Kenya for allowing access to the data.

**Conflicts of Interest:** The authors declare that they have no conflict of interest.

#### **References**

Ajzen, Icek. 2011. The theory of planned behavior: Reactions and reflections. *Psychology & Health* 26: 1113–27. [CrossRef]

