## 2.1 Descriptive statistics

## Source details

**Canonical URL:** [2.1 Descriptive statistics](https://www.imf.org/-/media/files/publications/wp/2023/english/wpiea2023238-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2023/english/wpiea2023238-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2023/english/wpiea2023238-print-pdf.pdf.json)

---

### Mobile money penetration and activity
- Value of mobile money transactions accounted for 94 percent of GDP in 2021.
- 53.2 percent of individuals reported having used mobile money in the past 12 months (Observations: 2,830).
- Frequency of mobile money use among users:
  - Daily: 3.2 percent
  - Once a week: 17.4 percent
  - Several times a month but not weekly: 26.3 percent
  - Less than once a month: 43.5 percent

### Perceptions and behavior toward cash
- 68.8 percent of respondents perceive cash as riskier than cards and machines (Observations: 2,392).
- 73.2 percent dislike carrying cash (Observations: 2,766).
- 82.4 percent prefer to pay in cash because everyone uses cash (Observations: 2,716).

### Financial behavior and inclusion indicators (sample-level)
- Owns mobile phone: 51.6 percent (Observations: 3,002).
- Sent remittances in the past 12 months: 38.1 percent (Observations: 2,835).
- Received remittances in the past 12 months: 43.9 percent (Observations: 2,835).
- Female respondents: 53.8 percent (Observations: 2,999).
- Age (years, average): 36.0 (Observations: 2,999).
- Completed primary education: 42.4 percent (Observations: 2,994).
- Completed secondary education: 7.9 percent (Observations: 2,994).
- Completed university: 2.9 percent (Observations: 2,994).
- Married: 65.0 percent (Observations: 3,002).
- Rural: 76.1 percent (Observations: 3,002).
- Has access to a bank account: 10.4 percent (Observations: 3,002).

### Contrast between mobile money users and non-users (selected indicators; N=1516 vs N=1314)
- Ownership and demographics:
  - Owns mobile phone: 81.8 vs 18.4 (p-value of diff = 0.000***)
  - Age (years, average): 33.7 vs 37.6 (p-value of diff = 0.000***)
  - Completed primary education: 62.3 vs 22.0 (p-value of diff = 0.000***)
  - Completed secondary education: 15.1 vs 1.1 (p-value of diff = 0.000***)
  - Completed university: 5.8 vs 0.0 (p-value of diff = 0.000***)
  - Female: 62.4 vs 69.0 (p-value of diff = 0.001***)
  - Rural: 60.4 vs 85.5 (p-value of diff = 0.001***)
  - Access to a bank account: 17.7 vs 1.7 (p-value of diff = 0.001***)
  - Annual income (in million Ugandan Shilling): 5.5 vs 1.8 (p-value of diff = 0.000***)
- Financial behavior:
  - Borrows money: 48.8 vs 35.8 (p-value of diff = 0.000***)
  - Saves money: 60.2 vs 39.5 (p-value of diff = 0.000***)
  - Sends money: 67.8 vs 5.1 (p-value of diff = 0.000***)
  - Receives money: 78.6 vs 6.9 (p-value of diff = 0.000***)
- Behavior towards cash (note differing Ns where reported):
  - Prefers to pay in cash: 81.8 vs 86.1 (p-value of diff = 0.001***) — N=1474 for MM users and N=1237 for non-users.
  - Perceives cash as risky: 73.7 vs 63.2 (p-value of diff = 0.000***) — N=1363 for MM users and N=1024 for non-users.
  - Dislikes cash: 77.9 vs 68.5 (p-value of diff = 0.000***) — N=1499 for MM users and N=1262 for non-users.
  - Prefers face-to-face banking: 75.7 vs 82.3 (p-value of diff = 0.000***) — N=1450 for MM users and N=1177 for non-users.

### Data and sample design (summary)
- Data: 2018 Uganda FinScope Survey commissioned by Financial Sector Deepening Uganda (FSD Uganda); sample nationally representative of individuals aged 16 years and older; estimated adult population 18.6 million.
- Sampling: three-stage stratified sampling — 320 enumeration areas selected probability proportional to size; 10 households randomly selected per EA; one adult randomly selected per household.
- Survey analysis: individual survey weights and svyset in Stata used to account for survey design.
- Key variable constructions:
  - Mobile money user: dummy = 1 if answered "yes" to "have you used mobile money in the past 12 months?"
  - "Cash is risky" and "I dislike carrying large amounts of cash": each a dummy = 1 if answered "yes".

*Source: wpiea2023238-print-pdf - 2.1 Descriptive statistics (FinScope Uganda 2018; IMF Working Paper).*

### 2.2 Mobile money, financial inclusion, and retail payment facts in Uganda

### Mobile money context and user experience
- Mobile money introduced in Uganda in 2009 and is regulated by the central bank; services fully provided by private sector operators.
- As of September 2023, there were at least seven such operators.
- Mobile money in Uganda is fully backed by bank deposits and can be exchanged for legal tender (IMF Survey on CBDC and Digital Payments).
- As of December 2022, around 25 million registered mobile accounts.
- FinScope (2018) user perceptions:
  - 81% had been using mobile money for more than a year.
  - Main reasons to start using mobile money: send money to others 25.7%; receive money from others 58.4%; other reasons 15.8%.
  - 63.9% experienced network failures in the previous 12 months; 36.1% did not.
  - 9.0% experienced loss of money in the last 12 months when using mobile money; 91.0% did not.
  - Can receive money from someone who uses another network: Yes 44.2%; No 24.4%; Don't know 31.4%.
  - It is cheaper to send money to someone in the same network as you: Yes 58.4%; No 9.8%; Don't know 31.8%.
- Interpretation: users are familiar with the technology, primarily use it to transfer remittances, report network failures but low incidence of money loss, and report interoperability and cross-network cost challenges.

### Financial inclusion and retail payment metrics
- Diffusion comparisons (2018–2022):
  - Number of debit cards increased by 20 percent.
  - Number of credit cards increased by 25 percent.
  - Number of active mobile money accounts increased by 73 percent.
- By end-2022 counts:
  - Total number of debit cards: about 3.1 million.
  - Total number of credit cards: about 10 thousand.
  - Active mobile money accounts: over 25 million.
- Savings and lending channels (FinScope):
  - 23 percent of Ugandan savers save through mobile money.
  - 11 percent save through traditional banks.
  - Over 60 percent save through informal mechanisms or by keeping cash at home.
  - Those who save through mobile money are most likely to save once than once a month.
  - Lending through mobile money accounts accounts for only 2 percent of savers.
- Payments:
  - 28 percent of adults use digital payments (mostly mobile money) for goods and services including groceries and school fees.
- Urban/rural contrasts:
  - Access to bank accounts: Rural 6.6 percent; Urban 22.6 percent.
  - Mobile money usage: Rural 46.8 percent; Urban 73.4 percent.
  - Interpretation: mobile money acts as an alternative to bank accounts, especially significant for rural households.

### Macro-level trends (selected)
- Outstanding balances on active mobile money accounts (Percent of GDP):
  - Grew from 0.07 percent of GDP in 2011 to 0.8 percent of GDP in 2022.
- Currency in circulation (Percent of M2):
  - Declined from about 30 percent of M2 in 2001 to around 24 percent in 2010.
  - Following introduction of mobile money in 2009, currency in circulation initially volatile; broadly declining path since 2015.
- Remittances (US$ million), 1999–2022:
  - Grew from about USD 450 million in 2007 to USD 1.4 billion in 2019 before temporarily declining amid the pandemic.
- Outstanding loans from commercial banks (Percent of GDP), 2010–2022:
  - Both total credit and household credit from commercial banks increased over time.
- Caution: macro trends do not necessarily imply causality between mobile money expansion and changes in currency in circulation, remittances, or borrowing; causal assessment addressed in micro-level analysis.

### Empirical strategy (overview)
- Two main estimations:
  - Equation (1) Probit for binary outcomes:
    - PROB{Y_i = 1} = Φ(α_0 + α_1 MM_i + α_2 X_i)
    - Y_i dummies: (i) perceives cash as risky, (ii) dislikes carrying cash, (iii) receives remittances, (iv) sends remittances, (v) saves money, (vi) borrows money.
    - MM_i = 1 if used mobile money in last 12 months.
    - X includes age, educational attainment, gender, marital status, urban/rural location, access to a bank account, and income.
    - Parameter of interest: α_1.
  - Equation (2) OLS for logged monetary amounts:
    - Z_i = β_0 + β_1 MM_i + β_2 X_i + ε_i
    - Z_i is natural log of amount received, sent, saved, or borrowed.
    - Parameter of interest: β_1.
- Endogeneity addressed with Propensity Score Matching (PSM) in three steps:
  - Step 1: Estimate PROB{MM_i = 1 | X} via probit.
  - Step 2: Nearest Neighbor matching with Caliper radius 0.005, without replacement.
  - Step 3: Check matched-sample covariate balance and estimate Equations (1) and (2) on matched sample.
- Falsification (placebo) tests reported in Section 5 to reduce concerns about omitted unobservables.

### Sample balance and matching diagnostics (selected)
- Age difference:
  - Raw sample difference: -3.59 years (p-value 0.00).
  - Matched sample difference: -0.55 years (p-value 0.46).
- Log of income difference:
  - Raw: -350.97 (S.E. 52.11; p-value 0.00; Percent bias -26.9).
  - Matched: -52.79 (S.E. 63.24; p-value 0.40; Percent bias -4.0).
- Access to bank: raw percent bias 57.2 reduced to 0.0 percent bias in matched sample (p-value 1.00).
- Bias diagnostics:
  - Mean Bias reduced from 35.7 (raw) to 2.0 (matched).
  - B statistic reduced from 87.6* (raw) to 7.1 (matched).
  - R statistic reduced from 4.12* (raw) to 0.92 (matched).
- Propensity score distributions: unequal originally; similar in matched sample. Common support condition met; bad matches discarded.

### Determinants of mobile money adoption (select probit marginal effects; Observations: 2467)
- Age: 0.023** (standard error 0.009)
- Agesq: -0.0003*** (standard error 0.000)
- Secondary educ.: 0.968*** (standard error 0.163)
- Female: -0.176*** (standard error 0.057)
- Rural: -0.577*** (standard error 0.068)
- Married: 0.080 (standard error 0.059)
- log of income: 0.098*** (standard error 0.015)
- Access to bank: 0.940*** (standard error 0.130)
- Constant: -1.121*** (standard error 0.281)
- Pseudo R2: 0.1381
- Note: ***, ** and * denote significance at 1, 5 and 10 percent levels.

### Impact of mobile money usage on perceptions about cash (matched sample probit)
- Mobile money users are about 0.3 percentage points more likely than non-users to perceive that “cash is risky” (column A).
- Mobile money users are about 0.3 percentage points more likely than non-users to “dislike carrying cash” (column B).
- Coefficients are highly statistically significant, but effect sizes are small.

### Impact on likelihoods of remittances, saving, and borrowing (marginal/probability impacts; Observations: 1582)
- Mobile money users compared to non-users:
  - 2.4 percentage points more likely to receive remittances (column A).
  - 1.8 percentage points more likely to send remittances (column B).
  - 0.3 percentage points more likely to have savings (column C).
  - 0.3 percentage points more likely to borrow (column D).
- Selected regression coefficients (standard errors in brackets):
  - "I received money": Used Mobile Money = 2.370*** (0.104).
  - "I sent money": Used Mobile Money = 1.795*** (0.095).
  - "I saved money": Used Mobile Money = 0.347*** (0.071).
  - "I borrowed money": Used Mobile Money = 0.314*** (0.075).

### Impact on magnitudes (logged amounts; observations and R2 as reported)
- Amount saved (log): Used Mobile Money = 0.431*** (0.146).
  - Text interpretation: annual savings is about 43 percent larger for mobile money users than non-users.
- Amount borrowed (log): Used Mobile Money = 0.655*** (0.095).
  - Text interpretation: annual borrowing is about 66 percent larger for mobile money users than non-users.
- Remittances received (log): Used Mobile Money = 0.094 (0.164) — not statistically significant.
- Remittances sent (log): Used Mobile Money = -0.106 (0.200) — not statistically significant.
- Observations: amount saved and borrowed regressions use 744 observations; remittances received 586; remittances sent 461.
- R2 (Table 8): 0.064, 0.174, 0.123, 0.072 for columns A–D respectively.

### Robustness and endogeneity checks (selected)
- PSM used to balance users and non-users to reduce endogeneity.
- Reverse causality: regressors capturing adoption motivated by wanting to receive or send remittances added; coefficients on Used Mobile Money remain highly statistically significant and only a little smaller than baseline.
- Falsification (placebo) tests:
  - Sub-sample of non-users with estimated propensity score above average (threshold 0.55) falsely assigned to treatment.
  - Coefficients on the 'false' treatment are in almost all cases statistically insignificant compared to baseline; exception: Table 10 column D shows a smaller and less statistically significant coefficient.
- Additional check: baseline regressions run on randomly selected smaller sub-samples; coefficients remain statistically significant.
- Example smaller random-sample estimates (Observations: 801 each):
  - "I received money": Used Mobile Money = 2.346*** (0.141)
  - "I sent money": Used Mobile Money = 1.926*** (0.145)
  - "I saved money": Used Mobile Money = 0.370*** (0.101)
  - "I borrowed money": Used Mobile Money = 0.393*** (0.101)

### Main conclusions and policy implications
- Main empirical findings:
  - Mobile money users are more likely to perceive cash as risky and less willing to carry cash, though the negative impact on willingness to carry cash is not large.
  - Mobile money usage is strongly associated with higher likelihoods of receiving and sending remittances, saving, and borrowing.
  - Mobile money users save and borrow larger amounts; no statistically significant difference found for amounts of remittances received and sent.
- Policy implications:
  - Rapid expansion of fintech technologies in Africa is likely to reduce, although slowly, the demand for and usage of cash.
  - Differential roles of cash may persist: the "means of payment" function could be gradually overtaken by digital alternatives, while the "store of value" function could remain little impacted due to network failure, limited interoperability, and lack of reliable electricity.
  - Expansion of mobile money could promote financial inclusion (remittances, savings, borrowing), especially if expanded in rural areas where penetration is relatively low.
  - Public sector role recommended: further public investments in foundational infrastructure—such as electricity and mobile networks—would aid private-sector expansion of mobile money given economies of scale.
- Areas for further research:
  - Follow-up research using more recent FinScope data on remittance amounts.
  - Separate assessments of impacts on monetary policy, financial stability, consumer protection, and cybersecurity.

*Source: wpiea2023238-print-pdf - Sections 2.1 and 2.2 (FinScope Uganda 2018; IMF Working Paper).*

### 2.1 Descriptive statistics .............................................................................................

### wpiea2023238-print-pdf - 2.1 Descriptive statistics

### Summary of key findings
- Mobile money penetration and activity
  - Value of mobile money transactions accounted for 94 percent of GDP in 2021.
  - About 53.2 percent of individuals reported having used mobile money in the past 12 months (Observations: 2,830).
  - Frequency of mobile money use among users:
    - Daily: 3.2 percent
    - Once a week: 17.4 percent
    - Several times a month but not weekly: 26.3 percent
    - Less than once a month: 43.5 percent
- Perceptions and behavior toward cash
  - 68.8 percent of respondents perceive cash as riskier than cards and machines (Observations: 2,392).
  - 73.2 percent dislike carrying cash (Observations: 2,766).
  - 82.4 percent prefer to pay in cash because everyone uses cash (Observations: 2,716).
- Financial behavior and inclusion indicators (sample-level and subgroup contrasts)
  - Owns mobile phone: 51.6 percent (Observations: 3,002).
  - Sent remittances in the past 12 months: 38.1 percent (Observations: 2,835).
  - Received remittances in the past 12 months: 43.9 percent (Observations: 2,835).
  - Female respondents: 53.8 percent (Observations: 2,999).
  - Age (years, average): 36.0 (Observations: 2,999).
  - Completed primary education: 42.4 percent (Observations: 2,994).
  - Completed secondary education: 7.9 percent (Observations: 2,994).
  - Completed university: 2.9 percent (Observations: 2,994).
  - Married: 65.0 percent (Observations: 3,002).
  - Rural: 76.1 percent (Observations: 3,002).
  - Has access to a bank account: 10.4 percent (Observations: 3,002).

### Data and sample design
- Data source and representativeness
  - Data: 2018 Uganda FinScope Survey commissioned by Financial Sector Deepening Uganda (FSD Uganda).
  - Sample: nationally representative of individuals aged 16 years and older; estimated adult population 18.6 million at the time.
  - Sampling: three-stage stratified sampling — 320 enumeration areas (EAs) selected probability proportional to size; 10 households randomly selected per EA; one adult randomly selected per household.
  - Survey weighting and analysis: authors use individual survey weights and svyset in Stata to account for survey design.
- Variable construction
  - Mobile money user: "have you used mobile money in the past 12 months?" — dummy = 1 if "yes".
  - Perception about cash: constructed from two yes/no survey statements:
    - "cash is risky" — dummy = 1 if "yes".
    - "I dislike carrying large amounts of cash" — dummy = 1 if "yes".
  - Other background covariates extracted: age, gender, marital status, phone ownership, educational attainment, urban/rural location, access to bank account, annual income (reported elsewhere).

### Contrast between mobile money users and non-users (selected indicators)
- Ownership and demographics (MM Users vs Non-Users; N=1516 vs N=1314)
  - Owns mobile phone: 81.8 vs 18.4 (p-value of diff = 0.000***)
  - Age (years, average): 33.7 vs 37.6 (p-value of diff = 0.000***)
  - Completed primary education: 62.3 vs 22.0 (p-value of diff = 0.000***)
  - Completed secondary education: 15.1 vs 1.1 (p-value of diff = 0.000***)
  - Completed university: 5.8 vs 0.0 (p-value of diff = 0.000***)
  - Female: 62.4 vs 69.0 (p-value of diff = 0.001***)
  - Rural: 60.4 vs 85.5 (p-value of diff = 0.001***)
  - Access to a bank account: 17.7 vs 1.7 (p-value of diff = 0.001***)
  - Annual income (in million Ugandan Shilling): 5.5 vs 1.8 (p-value of diff = 0.000***)
- Financial behavior (MM Users vs Non-Users)
  - Borrows money: 48.8 vs 35.8 (p-value of diff = 0.000***)
  - Saves money: 60.2 vs 39.5 (p-value of diff = 0.000***)
  - Sends money: 67.8 vs 5.1 (p-value of diff = 0.000***)
  - Receives money: 78.6 vs 6.9 (p-value of diff = 0.000***)
- Behavior towards cash (MM Users vs Non-Users)
  - Prefers to pay in cash: 81.8 vs 86.1 (p-value of diff = 0.001***)
    - Note: N=1474 for MM users and N=1237 for non-users for this item.
  - Perceives cash as risky: 73.7 vs 63.2 (p-value of diff = 0.000***)
    - Note: N=1363 for MM users and N=1024 for non-users for this item.
  - Dislikes cash: 77.9 vs 68.5 (p-value of diff = 0.000***)
    - Note: N=1499 for MM users and N=1262 for non-users for this item.
  - Prefers face-to-face banking: 75.7 vs 82.3 (p-value of diff = 0.000***)
    - Note: N=1450 for MM users and N=1177 for non-users for this item.

### Empirical approach (overview)
- Methodological steps (as summarized in paper structure)
  - 3. Empirical Strategy
    - 3.1 Model specification
    - 3.2 Addressing endogeneity
      - Step 1: PSM design
      - Step 2: Matching
      - Step 3: Balance
- Robustness and falsification
  - The paper reports multiple robustness checks and falsification tests (Tables 8–11, Tables 7 and 6 listed) aimed at reducing concerns about selection bias and reverse causality, including tests that control for reverse causality and falsification outcomes.

### Main conclusions reported in the text
- Mobile money users, compared to non-users, are:
  - More likely to perceive cash as risky.
  - Less likely to prefer carrying large amounts of cash.
  - More likely to remit and receive money.
  - More likely to save and borrow, and to save and borrow larger values (results reported in subsequent tables).
- Contribution to literature
  - This paper fills gaps by providing micro-level survey evidence from Sub-Saharan Africa (Uganda) on the relationship between mobile money adoption and perceptions about cash, complementing existing macro-level and non-SSA studies.
  - The authors emphasize that survey-based microdata allow study of preferences and behavior with a large sample (over 3,000 individuals in this case) and hence more reliable empirical estimations.

*Source: wpiea2023238-print-pdf - 2.1 Descriptive statistics (FinScope Uganda 2018; IMF Working Paper).*

### 2.2 Mobile money, financial inclusion, and retail payment facts in Uganda

### 2.2 Mobile money, financial inclusion, and retail payment facts in Uganda

### Mobile money in Uganda
- Mobile money was introduced in Uganda in 2009 and is regulated by the central bank. Mobile money services are fully provided by private sector operators.
- As of September 2023, there were at least seven such operators.
- According to the IMF Survey on CBDC and Digital Payments, mobile money in Uganda is fully backed by bank deposits, and it can be exchanged for legal tender.
- As of December 2022, there were around 25 million registered mobile accounts in the country.
- Finscope (2018) user perceptions:
  - 81% had been using mobile money for more than a year.
  - Main reasons to start using mobile money: send money to others 25.7%; receive money from others 58.4%; other reasons 15.8%.
  - 63.9% experienced network failures in the previous 12 months; 36.1% did not.
  - 9.0% experienced loss of money in the last 12 months when using mobile money; 91.0% did not.
  - Can receive money from someone who uses another network: Yes 44.2%; No 24.4%; Don't know 31.4%.
  - It is cheaper to send money to someone in the same network as you: Yes 58.4%; No 9.8%; Don't know 31.8%.
- Summary interpretation from survey: users are familiar with the technology, primarily use it to transfer remittances, report network failures but low incidence of money loss, and report interoperability and cross-network cost challenges.

### Financial inclusion and retail payments in Uganda
- Diffusion comparisons (2018–2022):
  - Number of debit cards increased by 20 percent.
  - Number of credit cards increased by 25 percent.
  - Number of active mobile money accounts increased by 73 percent.
- By end-2022 counts:
  - Total number of debit cards: about 3.1 million.
  - Total number of credit cards: about 10 thousand.
  - Active mobile money accounts: over 25 million.
- Savings and lending via channels (FinScope):
  - 23 percent of Ugandan savers save through mobile money.
  - 11 percent save through traditional banks.
  - Over 60 percent save through informal mechanisms or by keeping cash at home.
  - Those who save through mobile money are most likely to save once than once a month.
  - Lending through mobile money accounts for only 2 percent of savers.
- Payments:
  - Although most adults still use cash for goods and services including groceries and school fees, 28 percent use digital payments including mostly mobile money.
- Urban/rural contrasts:
  - Access to bank accounts: Rural 6.6 percent; Urban 22.6 percent.
  - Mobile money usage: Rural 46.8 percent; Urban 73.4 percent.
  - Interpretation: mobile money acts as an alternative to bank accounts, especially significant for rural households.

### Macro-level trends
- Outstanding balances on active mobile money accounts (Percent of GDP):
  - Grew from 0.07 percent of GDP in 2011 to 0.8 percent of GDP in 2022.
- Currency in circulation (Percent of M2), trends:
  - Declined from about 30 percent of M2 in 2001 to around 24 percent in 2010.
  - Following introduction of mobile money in 2009, currency in circulation was initially volatile; broadly declining path since 2015.
- Remittances (US$ million), 1999–2022:
  - Grew from about USD 450 million in 2007 to USD 1.4 billion in 2019 before temporarily declining amid the pandemic.
  - Note: remittances received from abroad include, but are not limited to, amounts received through mobile money.
- Outstanding loans from commercial banks (Percent of GDP), 2010–2022:
  - Both total credit and household credit from commercial banks have increased over time.
- Caution stated in source: these macro trends do not necessarily imply causality between mobile money expansion and changes in currency in circulation, remittances, or borrowing; causal assessment is addressed in micro-level analysis.

### Empirical strategy (overview)
- Two main equations:
  - Equation (1): Probit model for binary outcomes:
    - PROB{Y_i = 1} = Φ(α_0 + α_1 MM_i + α_2 X_i)
    - Y_i is a dummy taking value 1 if individual i: (i) perceives cash as risky, (ii) dislikes carrying cash, (iii) receives remittances, (iv) sends remittances, (v) saves money, or (vi) borrows money; 0 otherwise.
    - MM is a dummy = 1 if individual used mobile money in last 12 months; 0 otherwise.
    - X includes age, educational attainment, gender, marital status, urban/rural location, access to a bank account, and income.
    - Parameter of interest: α_1 (average treatment effect of mobile money usage on binary outcomes).
  - Equation (2): OLS for logged monetary amounts:
    - Z_i = β_0 + β_1 MM_i + β_2 X_i + ε_i
    - Z_i is natural log of the actual amount received, sent, saved, or borrowed.
    - Parameter of interest: β_1 (average treatment effect on logged monetary amounts).
- Endogeneity concern: selection into mobile money usage is non-random and correlates with observed covariates (younger, more educated, more men).
- Solution: Propensity Score Matching (PSM) with three steps:
  - Step 1 (PSM design): Estimate PROB{MM_i = 1 | X} = Φ(γ_i + ρ X_i) via probit (Equation (3)).
  - Step 2 (Matching): Nearest Neighbor matching with Caliper radius 0.005, without replacement.
  - Step 3 (Balance): Check matched-sample covariate balance and use matched sample to estimate Equations (1) and (2).
- Falsification test is conducted in Section 5 to reduce concerns about omitted unobservables.

### Sample balance and matching diagnostics
- Matching results (examples highlighting improvements):
  - Age difference between mobile money users and non-users:
    - Raw sample difference: -3.59 years (p-value 0.00).
    - Matched sample difference: -0.55 years (p-value 0.46).
  - Log of income difference:
    - Raw: -350.97 (S.E. 52.11; p-value 0.00; Percent bias -26.9).
    - Matched: -52.79 (S.E. 63.24; p-value 0.40; Percent bias -4.0).
  - Access to bank (raw difference percent bias 57.2) reduced in matched sample to 0.0 percent bias (p-value 1.00).
- Bias diagnostics:
  - Mean Bias reduced from 35.7 in the raw sample to 2.0 in the matched sample (below the 5 percent critical threshold).
  - B statistic reduced from 87.6* (raw) to 7.1 (matched) — B > 25% flagged in raw.
  - R statistic reduced from 4.12* (raw; outside [0.5; 2]) to 0.92 (matched; within range).
- Distributions of propensity scores: unequal in original sample; similar in matched sample. Common support condition met; bad matches discarded.

### Determinants of mobile money adoption (select probit results)
- Reported marginal effects (Dependent variable: Dummy on Mobile Money User):
  - Age: 0.023** (standard error 0.009)
  - Agesq: -0.0003*** (standard error 0.000)
  - Secondary educ.: 0.968*** (standard error 0.163)
  - Female: -0.176*** (standard error 0.057)
  - Rural: -0.577*** (standard error 0.068)
  - Married: 0.080 (standard error 0.059)
  - log of income: 0.098*** (standard error 0.015)
  - Access to bank: 0.940*** (standard error 0.130)
  - Constant: -1.121*** (standard error 0.281)
  - Pseudo R2: 0.1381
  - Observations: 2467
  - Note: ***, ** and * denote statistical significance at 1, 5 and 10 percent level, respectively. Standard errors in brackets.
- Interpretation:
  - Non-linearity in age: adoption increases with age up to a point, then declines (negative agesq).
  - Completing at least upper secondary education increases probability of use by about 1 percentage point.
  - Women are 0.2 percentage points less likely to use mobile money than men.
  - Rural residents are less likely to use mobile money than urban residents.
  - Higher income and access to a bank account increase probability of using mobile money.

### Impact of mobile money usage on perceptions about cash
- Probit regression results (matched sample, controlling for covariates):
  - Mobile money users are about 0.3 percentage points more likely than non-users to perceive that “cash is risky” (column A).
  - Mobile money users are about 0.3 percentage points more likely than non-users to “dislike carrying cash” (column B).
  - Coefficients are highly statistically significant, but effect sizes are small.
- Interpretation:
  - Mobile money usage increases, albeit modestly, users’ awareness of risks associated with cash transactions, possibly reflecting higher perceived safety and affordability of mobile transactions. The data do not directly test this mechanism.

*IMF Working Paper content unit: 2.2 Mobile money, financial inclusion, and retail payment facts in Uganda*

### 2.2 suggests that, while individuals acknowledge mobile money risks related to limited interoperability and

### Mobile Money, Perceptions about Cash, and Financial Inclusion: Learning from Uganda’s Micro-Level Data

### Perceptions about Cash and Mobile Money
- Mobile money users are more likely to perceive cash as risky and less likely to prefer carrying large amounts of cash.
- Regression results (behavior towards cash):
  - Dependent variable "Cash is risky": Used Mobile Money = 0.286*** (standard error 0.083).
  - Dependent variable "I dislike carrying cash": Used Mobile Money = 0.255*** (standard error 0.079).
  - Control coefficients shown (selected):
    - Age = 0.004 (0.015) and -0.001 (0.013) for the two specifications.
    - log of income = -0.074*** (0.024) and 0.001 (0.022).
    - Access to bank = 0.414* (0.253) and 0.328 (0.244).
  - Observations: 1329 and 1557.
  - Note: Pseudo R2 not available from Stata's "svy: probit" command.

### Impact on Financial Inclusion: Likelihood of Remittances, Saving, and Borrowing
- Marginal/probability impacts (Table 6, compared to non-users):
  - Mobile money users are 2.4 percentage points more likely to receive remittances (column A).
  - Mobile money users are 1.8 percentage points more likely to send remittances (column B).
  - Mobile money users are 0.3 percentage points more likely to have savings (column C).
  - Mobile money users are 0.3 percentage points more likely to borrow (column D).
- Regression coefficients reported (selected from Table 6, dependent variable dummies):
  - "I received money": Used Mobile Money = 2.370*** (0.104).
  - "I sent money": Used Mobile Money = 1.795*** (0.095).
  - "I saved money": Used Mobile Money = 0.347*** (0.071).
  - "I borrowed money": Used Mobile Money = 0.314*** (0.075).
  - Observations for each: 1582.
  - Selected controls (examples with significance):
    - Female = 0.245*** (0.085) for "I received money"; Female = -0.146* (0.080) for "I sent money".
    - Rural = -0.325** (0.134) for "I received money".
    - Married = 0.226*** (0.082) for "I borrowed money".
    - log of income = 0.087** (0.034) for "I sent money".
  - Standard errors in brackets. Pseudo R2 not available from Stata's "svy: probit" command.

### Impact on Magnitudes: Value of Savings, Borrowing, and Remittances
- Logged-value impacts (Table 7 / Table 8):
  - Amount saved (log): Used Mobile Money = 0.431*** (0.146).
    - Interpreted in text: annual savings is about 43 percent larger for mobile money users than non-users.
  - Amount borrowed (log): Used Mobile Money = 0.655*** (0.095).
    - Interpreted in text: annual borrowing is about 66 percent larger for mobile money users than non-users.
  - Remittances received (log): Used Mobile Money = 0.094 (0.164) — not statistically significant.
  - Remittances sent (log): Used Mobile Money = -0.106 (0.200) — not statistically significant.
  - Observations: Amount saved and borrowed regressions use 744 observations; remittances received 586; remittances sent 461 (Table 8 presentation).
  - R2 shown in Table 8: 0.064, 0.174, 0.123, 0.072 for columns A–D respectively.

### Robustness and Endogeneity Checks
- Propensity score matching (PSM) used to balance users and non-users to reduce endogeneity.
- Reverse causality addressed by adding regressors capturing adoption motivated by wanting to receive or send remittances:
  - Coefficients on Used Mobile Money remain highly statistically significant and only a little smaller than baseline.
- Falsification (placebo) tests on sub-sample of non-users:
  - Individuals with estimated propensity score above average (propensity score threshold 0.55) were falsely assigned to treatment.
  - Coefficients on the 'false' treatment are in almost all cases statistically insignificant compared to baseline.
  - Exception: Table 10 column D shows a smaller and less statistically significant coefficient.
- Additional check: Baseline regressions run on randomly selected smaller sub-samples of comparable size; coefficients of interest remain statistically significant.
- Placebo and smaller-sample regression examples (selected figures):
  - Smaller random-sample estimates for "I received money", "I sent money", "I saved money", "I borrowed money": Used Mobile Money = 2.346*** (0.141), 1.926*** (0.145), 0.370*** (0.101), 0.393*** (0.101); Observations: 801 each (Annex 1).

### Conclusions and Policy Implications
- Main empirical findings:
  - Mobile money users are more likely to perceive cash as risky and less willing to carry cash, but the negative impact on willingness to carry cash is not large.
  - Mobile money usage is strongly associated with higher likelihoods of receiving and sending remittances, saving, and borrowing.
  - Mobile money users save and borrow larger amounts, but no statistically significant difference is found for the amounts of remittances received and sent in this study.
- Policy implications:
  - Rapid expansion of fintech technologies in Africa is likely to reduce, although slowly, the demand for and usage of cash.
  - Differential roles of cash may persist: the "mean of payment" function could be gradually overtaken by digital alternatives, while the "store of value" function could remain little impacted due to network failure, limited interoperability, and lack of reliable electricity.
  - Expansion of mobile money could promote financial inclusion (remittances, savings, borrowing), especially if expanded in rural areas where penetration is relatively low.
  - Public sector role recommended: further public investments in foundational infrastructure—such as electricity and mobile networks—would aid private-sector expansion of mobile money given economies of scale.
- Areas noted for further research:
  - Follow-up research using more recent FinScope data on remittance amounts.
  - Separate assessments of impacts on monetary policy, financial stability, consumer protection, and cybersecurity.

*IMF Working Paper: Mobile Money, Perceptions about Cash, and Financial Inclusion — Working Paper No. WP/2023/238*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2023/english/wpiea2023238-print-pdf.pdf_