## Annex I. Data Details

## Source details

**Canonical URL:** [Annex I. Data Details](https://www.imf.org/-/media/files/publications/wp/2023/english/wpiea2023122-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2023/english/wpiea2023122-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2023/english/wpiea2023122-print-pdf.pdf.json)

---

### Data and survey design
- Data source: Afrobarometer’s Round 7 completed in 2019 covering 34 Sub-Saharan African countries.
- Survey design: face-to-face interviews with a randomly selected, nationally representative sample of 1,200-2,400 adult citizens in each country.
- Main survey questions and exact response codings:
  - Q16 (dependent variable):
    - "Which of the following statements is closest to your view? Choose Statement 1 or Statement 2.
      Statement 1: Men make better political leaders than women, and should be elected rather than women.
      Statement 2: Women should have the same chance of being elected to political office as men.
      Respondents are asked whether they (1) Agree very strongly with Statement 1, (2) Agree with Statement 1, (3) Agree with Statement 2, (4) Agree very strongly with Statement 2, or (5) Agree with neither."
    - Dependent variable: indicator for agreement with women as leaders = 1 if respondent selects option (3) or (4); = 0 otherwise.
  - Q91B (internet use frequency; alternatives preserved):
    - Alternatives: (0) Never, (1) Less than once a month, (2) A few times a month, (3) A few times a week, (4) Every day.
    - Internet usage indicator: = 1 if respondent reported some internet access (options (1)-(4)); = 0 if (0) Never.
  - Q12E (social media news access frequency; alternatives same as Q91B).
    - Social media usage indicator constructed analogously to internet usage.
- Note: not all internet users also use social media.

### Sample coverage and country-level sample sizes (Table 6)
- Total observations: 45,823
- Country sample sizes (selected examples, exact values preserved):
  - Benin — 1,200 — 2.62
  - Botswana — 1,198 — 2.61
  - Ghana — 2,400 — 5.24
  - Kenya — 1,599 — 3.49
  - Mozambique — 2,392 — 5.22
  - Nigeria — 1,600 — 3.49
  - South Africa — 1,840 — 4.02
  - Tanzania — 2,400 — 5.24
  - Total — 45,823 — 100

### Demographics of internet and social media usage (Table 7)
- Aggregate usage:
  - 34.6 percent are internet users.
  - 29.7 percent use social media.
  - For men: 45.0 percent (internet) and 39.0 percent (social media).
- Internet usage by age group (Women / Men):
  - 18-25 — 46.5 / 63.4
  - 26-35 — 39.9 / 55.6
  - 36-45 — 28.4 / 40.2
  - 46-55 — 21.7 / 30.8
  - 56-65 — 15.0 / 21.5
  - Over 65 — 9.1 / 11.5
- Internet usage by education level (Women / Men):
  - No formal education — 6.8 / 11.4
  - Primary — 15.4 / 21.2
  - Secondary — 50.2 / 57.0
  - Post-secondary — 85.8 / 84.8
- Social media news usage by age group (Women / Men):
  - 18-25 — 40.9 / 57.4
  - 26-35 — 35.0 / 49.3
  - 36-45 — 23.7 / 34.2
  - 46-55 — 17.4 / 23.4
  - 56-65 — 10.1 / 15.7
  - Over 65 — 7.7 / 9.0
- Social media news by education level (Women / Men):
  - No formal education — 5.5 / 8.5
  - Primary — 12.1 / 16.8
  - Secondary — 42.8 / 49.4
  - Post-secondary — 77.7 / 76.9

### Empirical specification and controls
- Outcome: indicator for agreement that "Women should have the same chance of being elected to political office as men" (Q16, options (3) or (4)).
- Main independent variables:
  - Internet usage indicator (Q91B, options (1)-(4) vs (0)).
  - Social media news access indicator (Q12E, options (1)-(4) vs (0)).
  - Intensive-margin categorical variables capturing frequency (options (0)-(4)) are also considered.
- Controls include: gender, age (6 groups), education (4 levels), religion (3 kinds), geographical region dummies, employment indicator, urban/rural, lived poverty index, radio/television/newspaper news access indicators, satisfaction with democracy (categorical), perception of government efforts to promote equal rights/opportunities for women (categorical), and other explanatory factors.
- Estimation: Linear probability models using OLS with standard errors clustered at the regional level. Logistic regressions also estimated for robustness.

### Main stylized facts and descriptive findings
- Baseline support:
  - Most people in the sample believe women should have the same opportunity as men to lead; all figures are greater than 50 percent regardless of frequency of internet and social media use.
- User vs non-user descriptive comparison:
  - Around 75 percent of users support gender equality in political leadership, while this fraction drops to 69 percent among non-users.
- Gender differences:
  - Women are more likely than men to agree on female leadership across internet and social media usage categories.
- Sensitivity by gender (descriptive):
  - Using the internet increases the proportion of women who agree by approximately 9 percentage points and men by 7 percentage points.
  - Using social media increases the proportion of women who agree by approximately 10 percentage points and men by 6 percentage points.
- Intensity of use (descriptive):
  - Using the internet and social media every day compared with less than once a month adds about 5 to 6 percentage points to the fraction who agree.

### Key regression estimates and sample statistics (Table 2 and Annex II, Table 8)
- Model coverage and fit:
  - Observations: 40,042 (internet regressions) and 40,044 (social media regressions).
  - R-squared: 0.100 (reported across specifications).
- Main OLS coefficients (specification (1) and examples):
  - Internet usage indicator (specification (1)): 0.032*** (standard error 0.007)
  - Social media usage indicator (specification (1)): 0.035*** (standard error 0.007)
  - Female dummy (all specifications): approximately 0.127*** (examples: 0.127*** with SE 0.007; 0.121*** with SE 0.008)
    - Interpretation: having a female identity is linked to an increase of slightly more than 12.5 percentage points in support for equal chances in political leadership.
- Interaction terms (specification (2) examples):
  - Internet interaction (I_i × F_i): 0.013 (SE 0.011)
  - Social media interaction (I_i × F_i): 0.023** (SE 0.010)
- Intensity-of-use coefficients (specification (3) examples):
  - Internet intensity:
    - Less than once a month: 0.020 (SE 0.016)
    - A few times a month: 0.027** (SE 0.012)
    - A few times a week: 0.039*** (SE 0.009)
  - Social media intensity:
    - A few times a month: 0.034*** (SE 0.013)
    - A few times a week: 0.034*** (SE 0.009)
    - Every day: 0.041*** (SE 0.009)

### Interpretation of main quantitative results
- Extensive-margin effects:
  - Using the internet increases agreement that women should have the same chance to be elected as men by 3.2 percentage points.
  - Using social media increases that agreement by 3.5 percentage points.
- Gender baseline:
  - Female identity increases support for equal political leadership chances by slightly more than 12.5 percentage points.
- Heterogeneous sensitivity:
  - Women’s perceptions are more sensitive to internet/social media use than men’s (positive interaction terms, with social media interaction significant).
- Intensive margin:
  - Intensity coefficients are generally positive and increase with intensity, but Wald tests do not reject equality across subsequent intensity categories; results support that the extensive margin (being a user) is most important.

### Tests of intensity and statistical significance (Wald tests, Table 3)
- All pairwise tests and joint tests for equality across intensity categories have significance levels greater than 10 percent (examples preserved exactly):
  - "Less than once a month equals A few times a month": F-statistic 0.14, Significance Level 0.7092 (Internet); F-statistic 2.13, Significance Level 0.1448 (Social Media)
  - "A few times a month equals A few times a week": F-statistic 0.91, Significance Level 0.3420 (Internet); F-statistic 0.00, Significance Level 0.9481 (Social Media)
  - All above simultaneously: F-statistic 0.66, Significance Level 0.5757 (Internet); F-statistic 1.29, Significance Level 0.2773 (Social Media)
- Conclusion: cannot reject equality hypotheses at reasonable significance levels; no robust evidence of an increasing relationship across intensity categories.

### Robustness checks (Annex III and Tables 4–5)
- Logistic regression results confirm OLS findings (selected Logit coefficients preserved):
  - Internet usage indicator (Logit spec (1)): 0.182*** (SE 0.042)
  - Social media usage indicator (Logit spec (1)): 0.199*** (SE 0.041)
  - Female dummy (Logit examples): 0.699*** (SE 0.038)
  - Interaction terms in Logit: 0.206*** (Internet) and 0.277*** (Social Media)
- Intensity categories in Logit are positive and generally increasing; Wald tests (Table 5) have significance levels > 10 percent.
- Leave-one-country-out analysis:
  - 204 additional regressions estimated (two treatments × three specifications × 34 countries).
  - Results are stable regardless of which country is excluded; graphical summaries in Annex III show point estimates and 95 percent confidence intervals remain consistent.
- Robustness figures:
  - Visualizations report point estimates (dots) and 95 percent confidence intervals (bars) with vertical axis tick values preserved exactly in the figures.

### Controls and additional regression patterns (Annex II, Table 8 — selected coefficients)
- Age group (relative to 18-25 base) — examples:
  - 26-35: Internet: -0.010 (0.007); Social Media: -0.008 (0.007)
  - 46-55: Internet: 0.024*** (0.009); Social Media: 0.027*** (0.009)
  - 56-65: Internet: 0.041*** (0.010); Social Media: 0.042*** (0.010)
- Religion (relative to Christian base):
  - Muslim: Internet: -0.027** (0.011); Social Media: -0.029*** (0.011)
- Education (relative to No formal education base):
  - Primary: Internet: 0.040*** (0.009); Social Media: 0.042*** (0.009)
  - Secondary: Internet: 0.089*** (0.010); Social Media: 0.091*** (0.010)
  - Post-secondary: Internet: 0.131*** (0.012); Social Media: 0.133*** / 0.131*** (0.011–0.012)
- Location and media indicators:
  - Rural area indicator: Internet: 0.015** (0.007); Social Media: 0.015** / 0.014** (0.006–0.007)
  - Radio news indicator: Internet: 0.020*** (0.007); Social Media: 0.020*** / 0.019*** (0.007)
  - Newspaper news indicator: Internet: -0.008 (0.006); Social Media: -0.011* (0.006)
- Political attitudes and government perception:
  - Fairly Well (handling promoting equal rights/opportunities for women): Internet: 0.028*** (0.010); Social Media: 0.028*** (0.010)
  - Very Well: Internet: 0.013 (0.013); Social Media: 0.012–0.013 (0.012–0.012)

### Relationship to existing literature
- Consistent with Shamaileh (2016): internet access improves women's, but not men's, perceptions of female political leadership ability (Arab World).
- Extends literature by:
  - Broadening analysis to a wide range of Sub-Saharan African countries.
  - Investigating both internet and social media usage.
  - Considering effects of intensity of internet and social media use.
- Complements Bacarreza et al. (2023) by demonstrating another potential non-trivial benefit of digitalization for closing gender gaps in political participation.

### Policy implications and recommendations
- Digitalization (internet and social media use) is associated with higher public support for women having equal chances as men to be elected political leaders.
- The larger sensitivity of women to digital usage suggests targeted value of digital platforms for empowering women's political representation and advocacy.
- Lack of a significant intensive-margin effect implies expanding basic access (the extensive margin) may be more immediately relevant than increasing usage frequency for shifting perceptions.
- Recommended policy measures:
  - Improve information and communications technology (ITC) infrastructure.
  - Focus on lowering the cost of energy as part of infrastructure improvements.
  - Encourage measures to improve people's education to provide prerequisites for learning and benefiting from modern technologies such as the internet and social media.

*IMF Working Papers — Digitalization and Gender Equality in Political Leadership in Sub-Saharan Africa (Annex I. Data Details).*

### Annex I. Data Details ..................................................................................................

### Annex I. Data Details

### Data
- Data source: Afrobarometer’s Round 7 completed in 2019 covering 34 Sub-Saharan African countries.
- Survey design: face-to-face interviews with a randomly selected, nationally representative sample of 1,200-2,400 adult citizens in each country.
- Main survey questions used:
  - Question Q16 (exact phrasing preserved):
    - "Which of the following statements is closest to your view? Choose Statement 1 or Statement 2.
      Statement 1: Men make better political leaders than women, and should be elected rather than women.
      Statement 2: Women should have the same chance of being elected to political office as men.
      Respondents are asked whether they (1) Agree very strongly with Statement 1, (2) Agree with Statement 1, (3) Agree with Statement 2, (4) Agree very strongly with Statement 2, or (5) Agree with neither."
    - Dependent variable: indicator for agreement with women as leaders = 1 if respondent selects option (3) or (4); = 0 otherwise.
  - Question Q91B (internet use frequency; response alternatives preserved):
    - Alternatives: (0) Never, (1) Less than once a month, (2) A few times a month, (3) A few times a week, (4) Every day.
    - Internet usage indicator: = 1 if respondent reported some internet access (options (1)-(4)); = 0 if (0) Never.
  - Question Q12E (social media news access frequency; response alternatives same as Q91B).
    - Social media usage indicator constructed analogously to internet usage.
- Note from Annex I descriptive statistics: not all internet users also use social media.

### Empirical specification and controls
- Outcome: indicator for agreement that "Women should have the same chance of being elected to political office as men" (Q16, options (3) or (4)).
- Main independent variables:
  - Internet usage indicator (Q91B, options (1)-(4) vs (0)).
  - Social media news access indicator (Q12E, options (1)-(4) vs (0)).
  - Intensive-margin categorical variables capturing frequency (options (0)-(4)) are also considered.
- Controls: gender, age, education, religion, geographical region, exposure to other media, and other explanatory factors.

### Sample coverage note
- Table 6 in Annex I lists the countries included in the Round 7 sample and the number of individuals surveyed in each country (table not reproduced here).

### Stylized facts and overview
- The paper reports five stylized facts based on the variables described (the facts are summarized in the paper; table references are in Annex I).

---

### Main Findings
- Internet usage association:
  - More people using the internet is associated with a 3.2 percentage point increase in the proportion of people who support that women should have equal opportunities in political leadership as men.
- Social media usage association:
  - Social media usage is associated with a 3.5 percentage point increase in the proportion of people who support that women should have equal opportunities in political leadership as men.
- Gender difference in baseline support:
  - Having a female identity is linked to an increase in the likelihood of supporting women having equal opportunities as men in political leadership positions by slightly more than 12.5 percentage points.
- Heterogeneous sensitivity:
  - Women’s perceptions of gender equality in political leadership are more sensitive to internet and social media use than men’s perceptions.
- Intensive-margin result:
  - No compelling evidence that the intensity of internet or social media usage (the intensive margin) is significantly associated with greater support for women having equal opportunities as leaders compared with the extensive margin.

---

### Relationship to literature
- Consistent with Shamaileh (2016): internet access improves women's, but not men's, perceptions of female political leadership ability (Arab World).
- Extends literature by:
  - Broadening analysis to a wide range of Sub-Saharan African countries.
  - Investigating both internet and social media usage.
  - Considering effects of intensity of internet and social media use.
- Related to Bacarreza et al. (2023): complements findings on digitalization and gender employment by demonstrating another potential non-trivial benefit of digitalization for closing gender gaps in political participation.

---

### Implications highlighted in the paper
- Digitalization (internet and social media use) is associated with higher public support for women having equal chances as men to be elected political leaders, suggesting digital access may help change perceptions of women as leaders.
- The larger sensitivity of women to digital usage suggests potential targeted value of digital platforms for empowering women's political representation and advocacy.
- Lack of a significant intensive-margin effect implies that expanding basic access (the extensive margin) may be more immediately relevant than increasing usage frequency for shifting perceptions.

*IMF Working Papers — Digitalization and Gender Equality in Political Leadership in Sub-Saharan Africa (Annex I. Data Details).*

### 34.6 percent are internet users, but only 29.7 percent use social media. For men, the figures are 45 percent (internet) 

### wpiea2023122-print-pdf - 34.6 percent are internet users, but only 29.7 percent use social media. For men, the figures are 45 percent (internet) 

### Digital access and gender gaps
- 34.6 percent are internet users, but only 29.7 percent use social media.
- For men: 45 percent (internet) and 39 percent (social media).
- Across all age and education groups, women are less likely than men to use the internet and social media.
- Use of the internet and social media increases with education level and decreases with age.
- The observed patterns confirm the International Communication Union's (ICU) findings that the digital gender gap persists in low-income countries.

### Survey findings on attitudes to female political leadership (descriptive)
- Most people in the sample believe women should have the same opportunity as men to lead; all figures are greater than 50 percent regardless of frequency of internet and social media use.
- Being a social media or internet user increases the likelihood of agreeing that women should have the same chance to lead as men:
  - Around 75 percent of users support gender equality in political leadership, while this fraction drops to 69 percent among non-users.
- Women are more likely than men to agree on female leadership across internet and social media usage categories.
- Sensitivity to internet/social media use by gender:
  - Using the internet increases the proportion of women who agree by approximately 9 percentage points and men by 7 percentage points.
  - In terms of social media, these figures are 10 and 6 percentage points for women and men, respectively.
- Intensity of use:
  - More frequent internet and social media use is associated with greater likelihood to agree that women should have the same chance as men in political leadership.
  - When compared to using it less than once a month, using the internet and social media every day adds about 5 to 6 percentage points to the fraction.

### Regression analysis and main quantitative results
- Estimation approach:
  - Linear probability models using OLS with standard errors clustered at the regional level.
  - Controls include age (6 groups), education (4 levels), religion (3 kinds), employment indicator, urban/rural, geographical region dummies, lived poverty index, and three indicators of news access via radio, television, and newspaper. Two categorical indexes added: satisfaction with democracy and perception of government efforts to promote equal rights/opportunities for women.
- Key OLS estimates (Table 2):
  - Internet usage indicator (specification (1)): 0.032*** (standard error 0.007)
  - Social media usage indicator (specification (1)): 0.035*** (standard error 0.007)
  - Female dummy (all specifications): approximately 0.127*** (examples: 0.127*** with SE 0.007; 0.121*** with SE 0.008)
    - Interpretation: having a female identity is linked to an increase of slightly more than 12.5 percentage points in support for equal chances in political leadership.
  - Interaction (I_i × F_i):
    - Internet interaction: 0.013 (SE 0.011) in specification (2)
    - Social media interaction: 0.023** (SE 0.010) in specification (2)
  - Intensity of internet use (examples, specification (3)):
    - Less than once a month: 0.020 (SE 0.016)
    - A few times a month: 0.027** (SE 0.012)
    - A few times a week: 0.039*** (SE 0.009)
  - Intensity of social media use (examples, specification (3)):
    - A few times a month: 0.034*** (SE 0.013)
    - A few times a week: 0.034*** (SE 0.009)
    - Every day: 0.041*** (SE 0.009)
  - Observations: 40,042 (internet regressions) and 40,044 (social media regressions)
  - R-squared: 0.100 (reported across specifications)
- Interpretation of main regression facts:
  - Fact 2 validated: Being a social media or internet user increases the likelihood of agreeing that women should have the same chances as men.
    - Using the internet increases agreement by 3.2 percentage points.
    - Using social media increases agreement by 3.5 percentage points.
  - Fact 3 validated: Women are more likely than men to support gender equality in political leadership (~+12.5 percentage points).
  - Fact 4 supported: Women’s perceptions are more sensitive to internet/social media use than men’s (positive interaction terms).
  - Fact 5 examined: Although intensity coefficients are generally positive and increase with intensity, Wald tests do not reject equality across subsequent intensity categories; results support that the extensive margin (being a user) rather than the intensive margin is most important.

### Tests of intensity and statistical significance
- Wald test results for OLS intensity categories (Table 3):
  - All pairwise tests and the joint test have significance levels greater than 10 percent (examples shown):
    - "Less than once a month equals A few times a month": F-statistic 0.14, Significance Level 0.7092 (Internet); F-statistic 2.13, Significance Level 0.1448 (Social Media)
    - "A few times a month equals A few times a week": F-statistic 0.91, Significance Level 0.3420 (Internet); F-statistic 0.00, Significance Level 0.9481 (Social Media)
    - All above simultaneously: F-statistic 0.66, Significance Level 0.5757 (Internet); F-statistic 1.29, Significance Level 0.2773 (Social Media)
  - Conclusion: cannot reject equality hypotheses at reasonable significance levels; no robust evidence of an increasing relationship across intensity categories.

### Robustness checks
- Logistic regression results (Table 4) confirm OLS findings:
  - Internet usage indicator (Logit spec (1)): 0.182*** (SE 0.042)
  - Social media usage indicator (Logit spec (1)): 0.199*** (SE 0.041)
  - Female dummy (Logit examples): 0.699*** (SE 0.038) and similar across specifications
  - Interaction terms in Logit: 0.206*** (Internet) and 0.277*** (Social Media)
  - Intensity categories in Logit are positive and generally increasing; Wald tests (Table 5) have significance levels > 10 percent.
- Leave-one-country-out analysis:
  - 204 additional regressions estimated (two treatments × three specifications × 34 countries).
  - Results are stable regardless of which country is excluded; graphical summaries in Annex III show point estimates and 95 percent confidence intervals remain consistent.

### Controls and further findings
- Control variable patterns (summary from Annex II, Table 8):
  - Age: inverted U-shape relationship with support for gender equality in political leadership; middle-aged group most supportive.
  - Education: more educated respondents show greater support.
  - Media: those who listen to radio news show greater support.
  - Religion: support is lesser among Muslims compared to other religions.
  - Urban/rural: rural dwellers are more supportive of gender equality in political leadership than urban dwellers.
  - Perception of government: support is greater among those who perceive the government as making efforts to promote equal rights/opportunities for women.

### Policy recommendations and conclusion
- Digitalization is associated with increased perceptions of Sub-Saharan African women having equal chances as men to be elected as political leaders.
- Recommended policy measures:
  - Improve information and communications technology (ITC) infrastructure.
  - Focus on lowering the cost of energy as part of infrastructure improvements.
  - Encourage measures to improve people's education to provide prerequisites for learning and benefiting from modern technologies such as the internet and social media.
- Rationale: significant positive spillover from digitalization on perceptions of female political leadership, alongside other known benefits.

*Source: IMF Working Paper excerpt (Afrobarometer-based analysis) — Authors’ calculations as presented in the supplied content.*

### Annex I. Data Details

### Annex I. Data Details

### Sample Description (Table 6)
- Benin — 1,200 — 2.62
- Botswana — 1,198 — 2.61
- Burkina Faso — 1,200 — 2.62
- Cabo Verde — 1,200 — 2.62
- Cameroon — 1,202 — 2.62
- Côte d'Ivoire — 1,200 — 2.62
- Eswatini — 1,200 — 2.62
- Gabon — 1,199 — 2.62
- Gambia — 1,200 — 2.62
- Ghana — 2,400 — 5.24
- Guinea — 1,194 — 2.61
- Kenya — 1,599 — 3.49
- Lesotho — 1,200 — 2.62
- Liberia — 1,200 — 2.62
- Madagascar — 1,200 — 2.62
- Malawi — 1,200 — 2.62
- Mali — 1,200 — 2.62
- Mauritius — 1,200 — 2.62
- Morocco — 1,200 — 2.62
- Mozambique — 2,392 — 5.22
- Namibia — 1,200 — 2.62
- Niger — 1,200 — 2.62
- Nigeria — 1,600 — 3.49
- São Tomé and Príncipe — 1,200 — 2.62
- Senegal — 1,200 — 2.62
- Sierra Leone — 1,200 — 2.62
- South Africa — 1,840 — 4.02
- Sudan — 1,200 — 2.62
- Tanzania — 2,400 — 5.24
- Togo — 1,200 — 2.62
- Tunisia — 1,199 — 2.62
- Uganda — 1,200 — 2.62
- Zambia — 1,200 — 2.62
- Zimbabwe — 1,200 — 2.62
- Total — 45,823 — 100

### Demographics of internet and social media usage (Table 7)
- Internet usage (percent):
  - Women — 34.6
  - Men — 45.0
- Internet usage by age group (Women / Men):
  - 18-25 — 46.5 / 63.4
  - 26-35 — 39.9 / 55.6
  - 36-45 — 28.4 / 40.2
  - 46-55 — 21.7 / 30.8
  - 56-65 — 15.0 / 21.5
  - Over 65 — 9.1 / 11.5
- Internet usage by education level (Women / Men):
  - No formal education — 6.8 / 11.4
  - Primary — 15.4 / 21.2
  - Secondary — 50.2 / 57.0
  - Post-secondary — 85.8 / 84.8
- Social media news usage (percent):
  - Women — 29.7
  - Men — 39.0
- Social media news by age group (Women / Men):
  - 18-25 — 40.9 / 57.4
  - 26-35 — 35.0 / 49.3
  - 36-45 — 23.7 / 34.2
  - 46-55 — 17.4 / 23.4
  - 56-65 — 10.1 / 15.7
  - Over 65 — 7.7 / 9.0
- Social media news by education level (Women / Men):
  - No formal education — 5.5 / 8.5
  - Primary — 12.1 / 16.8
  - Secondary — 42.8 / 49.4
  - Post-secondary — 77.7 / 76.9

Notes: The table shows the percentages of men and women who use the internet and social media. All figures were calculated using the individual sample weights provided by the survey.

### Regression Results (Annex II, Table 8 — selected coefficients)
Notes: Estimates are weighted using the individual sample weights provided by the survey. Standard errors in parentheses are robust and clustered at the regional level. The dummy variables for geographical regions are not displayed. *** p<0.01, ** p<0.05, * p<0.1.

- Model coverage:
  - Observations (Internet models) — 40,042
  - Observations (Social Media models) — 40,044
  - R-squared — 0.100 (for both Internet and Social Media models)

- Age group coefficients (relative to 18-25 base):
  - 26-35:
    - Internet: -0.010 (0.007)
    - Social Media: -0.008 (0.007)
  - 36-45:
    - Internet: 0.008 (0.008)
    - Social Media: 0.009 (0.008)
  - 46-55:
    - Internet: 0.024*** (0.009)
    - Social Media: 0.027*** (0.009)
  - 56-65:
    - Internet: 0.041*** (0.010)
    - Social Media: 0.042*** (0.010)
  - Over 65:
    - Internet: 0.024 (0.015)
    - Social Media: 0.025 (0.015) / 0.026* (0.015) in one specification

- Religion (relative to Christian base):
  - Muslim:
    - Internet: -0.027** (0.011)
    - Social Media: -0.029*** (0.011)
  - Other:
    - Internet: -0.005 (0.008)
    - Social Media: -0.005 (0.008)

- Education (relative to No formal education base):
  - Primary:
    - Internet: 0.040*** (0.009)
    - Social Media: 0.042*** (0.009)
  - Secondary:
    - Internet: 0.089*** (0.010)
    - Social Media: 0.091*** (0.010)
  - Post-secondary:
    - Internet: 0.131*** (0.012)
    - Social Media: 0.133*** / 0.131*** (0.011–0.012 across specifications)

- Employment and poverty indicators:
  - Employed indicator:
    - Internet: 0.000 (0.007)
    - Social Media: -0.000 (0.007)
  - Lived poverty index:
    - Internet: 0.001 (0.004)
    - Social Media: 0.001 (0.004)

- Location and media indicators:
  - Rural area indicator:
    - Internet: 0.015** (0.007)
    - Social Media: 0.015** / 0.014** (0.006–0.007)
  - Radio news indicator:
    - Internet: 0.020*** (0.007)
    - Social Media: 0.020*** / 0.019*** (0.007)
  - Television news indicator:
    - Internet: 0.004 (0.007)
    - Social Media: 0.003 (0.007)
  - Newspaper news indicator:
    - Internet: -0.008 (0.006)
    - Social Media: -0.011* (0.006)

- Political attitudes and gender-equality handling:
  - Satisfaction with democracy (relative to "Not a democracy" base):
    - Not at all satisfied — Internet: 0.019 (0.022); Social Media: 0.018 (0.022)
    - Not very satisfied — Internet: 0.025 (0.022); Social Media: 0.024 (0.022)
    - Fairly satisfied — Internet: 0.011 (0.024); Social Media: 0.011 (0.023)
    - Very satisfied — Internet: -0.019 (0.025); Social Media: -0.020 (0.024)
  - Handling promoting equal rights/opportunities for women (relative to "Very Badly" base):
    - Fairly Badly — Internet: 0.003 (0.010); Social Media: 0.003 (0.010)
    - Fairly Well — Internet: 0.028*** (0.010); Social Media: 0.028*** (0.010)
    - Very Well — Internet: 0.013 (0.013); Social Media: 0.012–0.013 (0.012–0.012)

*Source: Afrobarometer. Authors’ calculations.*

### Annex III. Robustness Results

### Annex III. Robustness Results

### Overview
- Presents robustness checks for the empirical analysis of "Digitalization and Gender Equality in Political Leadership in Sub-Saharan Africa".
- Visualizations report point estimates (dots) and 95 percent confidence intervals (bars) calculated from robust standard errors clustered at the regional level.
- Titles of each figure indicate the respective regression specification used.

### Figure 1 — Usage Indicators
- Types of usage indicators displayed:
  - Internet | Usage Indicator | Regression (1)
  - Internet | Usage Indicator | Regression (2)
  - Social Media | Usage Indicator | Regression (1)
  - Social Media | Usage Indicator | Regression (2)
- Vertical axis tick values preserved exactly: 0.00, 0.01, 0.02, 0.03, 0.04, 0.05, 0.06
- Countries/units shown (preserved ordering and codes): BENBFOBOT, CAM, CDI, CVE, GAB, GAM, GHA, GUI, KEN, LES, LIB, MADMAU, MLI, MLWMOR, MOZNAMNGR, NIG, SAF, SEN, SRLSTP, SUD, SWZ, TAN, TOG, TUN, UGA, ZAM, ZIM

### Figure 2 — Female Dummies
- Specifications displayed:
  - Internet | Female Dummy | Regression (1)
  - Internet | Female Dummy | Regression (2)
  - Internet | Female Dummy | Regression (3)
  - Social Media | Female Dummy | Regression (1)
  - Social Media | Female Dummy | Regression (2)
  - Social Media | Female Dummy | Regression (3)
- Vertical axis tick values preserved exactly: 0.00, 0.05, 0.10, 0.15, 0.20
- Countries/units shown (preserved ordering and codes): BENBFOBOT, CAM, CDI, CVE, GAB, GAM, GHA, GUI, KEN, LES, LIB, MADMAU, MLI, MLWMOR, MOZNAMNGR, NIG, SAF, SEN, SRLSTP, SUD, SWZ, TAN, TOG, TUN, UGA, ZAM, ZIM

### Figure 3 — Interaction Terms
- Interaction terms between usage indicator and female dummy are reported for the following specifications:
  - Social Media | Female Dummy | Regression (2)
  - Social Media | Female Dummy | Regression (3)
  - Internet | Interaction Term | Regression (2)
  - Social Media | Interaction Term | Regression (2)
- Vertical axis tick values preserved exactly where shown: -0.02, 0.00, 0.02, 0.04, 0.06
- Countries/units shown (preserved ordering and codes): BENBFOBOT, CAM, CDI, CVE, GAB, GAM, GHA, GUI, KEN, LES, LIB, MADMAU, MLI, MLWMOR, MOZNAMNGR, NIG, SAF, SEN, SRLSTP, SUD, SWZ, TAN, TOG, TUN, UGA, ZAM, ZIM

### Figure 4 — Intensity of Use
- Intensity categories and regression specification:
  - Internet | Less than once a month | Regression (3)
  - Internet | A few times a month | Regression (3)
  - Internet | A few times a week | Regression (3)
  - Internet | Everyday | Regression (3)
  - Social Media | Less than once a month | Regression (3)
  - Social Media | A few times a month | Regression (3)
  - Social Media | A few times a week | Regression (3)
  - Social Media | Everyday | Regression (3)
- Vertical axis tick values preserved exactly: -0.03, -0.01, 0.01, 0.03, 0.05, 0.07
- Countries/units shown (preserved ordering and codes): BENBFOBOT, CAM, CDI, CVE, GAB, GAM, GHA, GUI, KEN, LES, LIB, MADMAU, MLI, MLWMOR, MOZNAMNGR, NIG, SAF, SEN, SRLSTP, SUD, SWZ, TAN, TOG, TUN, UGA, ZAM, ZIM

### Key methodological notes (explicit from figures)
- Dots denote point estimates for:
  - Internet usage indicators, social media usage indicators, female dummies, interaction terms, and intensity-of-use indicators as specified.
- Bars denote 95 percent confidence intervals calculated from robust standard errors clustered at the regional level.
- Each figure title corresponds to the regression specification number used for that robustness check.

*IMF WORKING PAPERS Digitalization and Gender Equality in Political Leadership in Sub-Saharan Africa — Annex III. Robustness Results*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2023/english/wpiea2023122-print-pdf.pdf_