## Annex I. Additional Figures and Tables (wpiea2022245-print-pdf)

## Source details

**Canonical URL:** [Annex I. Additional Figures and Tables (wpiea2022245-print-pdf)](https://www.imf.org/-/media/files/publications/wp/2022/english/wpiea2022245-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2022/english/wpiea2022245-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2022/english/wpiea2022245-print-pdf.pdf.json)

---

### I. Research question and main contributions
- Research question: causal effects of mother education on child education in Vietnam exploiting a natural experiment from the First Indochina War (1946-1954) and the DRV "Mass Education" (ME) program.
- Identification: compare children in the 1997-1998 VLSS whose mothers were school-age girls exposed to ME in DRV-controlled provinces (ME-mothers) with children of mothers who lived in French-controlled areas during the war.
- Key contributions:
  - One additional year of maternal education boosts child education by up to 0.65 years.
  - ME program raised female education attainment by 1.5 years, or 48 percent, from an average pre-treatment 3.1 years.
  - Mechanisms: evidence points to non-income family factors (healthier lifestyles, more durable assets, higher education spending, less spending on unhealthy items); little evidence of improved labor market incomes for ME-mothers or their husbands.

### II. Country context and the ME program
- Historical facts:
  - France colonized Vietnam from 1887 until 1945; female illiteracy estimated near 100 percent before 1945.
  - DRV launched Mass Education ("Bình dân học vụ") after independence in 1945; Decree 20/SL (issued September 8, 1945) mandated free and mandatory illiteracy-eradicating classes.
  - First Indochina War (September 1946 – May 1954) divided Vietnam into DRV-controlled and French-controlled areas; ME expanded rapidly in DRV-controlled North and Central provinces.
- Program features:
  - Highly inexpensive, community-supported; classes delivered in homes, temples, public places.
  - Reported more than 2.5 million people could read and write by 1946 in North and Central Vietnam.

### III. Empirical design and data
- Exposure measures:
  - Binary: ME_M = 1 if mother born in three Northern provinces under complete DRV control (Nghe An, Thanh Hoa, Ha Tinh) and in the treatment cohort.
  - Continuous: Expose_M = number of years a mother was exposed to ME between ages 6 and 11 (varies between 0 and 5).
  - Treatment cohort: mothers born 1940–1945. Controls: 1924–1935 and 1950–1961. Cohorts 1936–1939 and 1946–1949 removed as "contaminated" in binary approach.
- Econometric specifications:
  - First-stage examples: SM = α + β(ME_M * L_M) + γX_M + bp_M + τ_M + ε and SM = α + βExpose_M + γX_M + bp_M + τ_M + ε.
  - Reduced-form and IV: instrument maternal schooling Ŝ_M with ME exposure in child outcome regressions Y̅_C = θ + δ Ŝ_M + ηX_M + bp_M + τ_M + μ.
  - Robust SEs clustered two-way: mother's birth province and household's current commune. Exclude mothers who migrated before age 15.
- Data:
  - 1997-1998 VLSS (~6,000 households); constructed "years of education" as highest grade completed; post-secondary years = 12 + reported years at junior college/university/post-graduate.
  - Occupation-start dates and DRV control timing compiled from provincial party/army histories.

### IV. First-stage and reduced-form evidence (mother and child education)
- First-stage (ME exposure → mother's education):
  - Panel A (binary treated vs control): ME-mothers higher by between 19 to 26 percentage points in probability of reading literacy, math competency, primary education completion, and secondary education completion (Panel A, Columns 1–4).
  - Panel A, Column 5: Years of education: 1.474 *** (standard error 0.297); N 1619; adj. R2 0.535.
  - Panel B (years of exposure 0–5): an additional year of exposure raises years of education by 0.253 ** (standard error 0.094); N 1976; adj. R2 0.499.
  - Panel B: Reading Literacy 0.028 (0.017); Math competency 0.021 (0.019); Primary completion 0.043 ** (0.016).
- Reduced-form (mother's ME exposure → next generation's education):
  - Panel A (binary treated):
    - Average years of children education: 0.894 *** (0.249); N 1206; adj. R2 0.347.
    - Years of education of the most educated child: 0.808 *** (0.191); N 1206; adj. R2 0.385.
    - Percentage of children completing high school: 0.149 *** (0.039); N 1206; adj. R2 0.125.
  - Panel B (years of exposure):
    - Average years of children education: 0.188 ** (0.076); N 1546; adj. R2 0.322.
    - Years of the most educated child: 0.129 (0.076); N 1546; adj. R2 0.349.
    - Percentage completing high school: 0.031 *** (0.008); N 1546; adj. R2 0.123.

### V. IV estimates of maternal education on child education (intergenerational impact)
- Table 5 (no paternal controls):
  - Panel A (first instrument):
    - OLS: Years of maternal education → Average years of children education: 0.308 *** (0.042); N 1206.
    - IV: 0.583 ** (0.280); F-test excluded instrument 13.04; N 1206.
  - Panel B (second instrument):
    - OLS: 0.350 *** (0.033); N 1545.
    - IV: 0.651 ** (0.312); F-test excluded instrument 7.64; N 1545.
  - Interpretation: IV estimates imply one more year of maternal schooling raises children's average education by about 0.58 to 0.65 years.
- Table 12 (controlling for paternal education — direct maternal effect):
  - First instrument: IV = 0.730 ** (0.342); years of paternal education coefficient 0.093 (0.113); F-test 21.68; N 882.
  - Second instrument: IV = 0.853 * (0.458); years of paternal education coefficient 0.065 (0.152); F-test 10.84; N 1145.
  - Interpretation: direct maternal effects about 0.73 to 0.85 child-years per maternal year when controlling for paternal education.

### VI. Robustness checks
- Removing mixed-occupied provinces (Appendix Table A3) yields quantitatively similar estimates:
  - Panel A3.1 (mothers): Treated years of education 1.379 *** (0.371); N 1376; adj. R2 0.513.
  - Panel A3.2 (children): Treated average years of children education 0.862 *** (0.268); N 1021; adj. R2 0.329.
  - Panel A3.3 (IV, mixed-occupied removed): IV average years of children education 0.569 * (0.302) (first instrument, N 1020) and 0.687 ** (0.316) (second instrument, N 1305).

### VII. Mechanisms and suggestive transmission channels
- Maternal labor market and reproduction outcomes (Table 6):
  - No significant effects of ME exposure on maternal non-farm participation, paid work, white-collar work, health insurance, first birth age, number of children, or hours of household chores.
  - Example coefficients (First Exposure Treated): Non-farm participation 0.035 (0.062); Paid work -0.008 (0.024); First birth age 0.470 (0.604).
- Paternal outcomes:
  - Husband education: modest positive effects on reading literacy 0.075 ** (0.036) and math competency 0.090 ** (0.042) (First Exposure Treated).
  - Husband lifestyle (Table 9): reductions in smoking
    - Husband smoking now: -0.024 ** (0.010); Number of cigarettes a day: -0.142 ** (0.063); N 1150.
    - Years of ME exposure: Husband smoking now -0.005 *** (0.002); Number of cigarettes a day -0.029 * (0.015); N 1420.
- Household income, expenditure, assets, living conditions (Tables 10–11):
  - First Exposure Treated:
    - Log real income per capita: 0.298 (0.191).
    - Log real expenditure per capita: 0.153 ** (0.074).
    - Log durable assets' value per capita: 0.269 ** (0.111).
    - Floor quality: 0.174 *** (0.032).
    - Log (expense on tobacco +1): -0.710 ** (0.295).
    - Educational expense per kid aged 6-18: 0.369 ** (0.177).
  - Second Exposure (per year):
    - Log real expenditure per capita: 0.030 ** (0.013).
    - Log durable assets' value per capita: 0.078 *** (0.026).
    - Floor quality: 0.046 *** (0.009).
    - Log (expense on tobacco +1): -0.122 * (0.068).
    - Educational expense per kid aged 6-18: 0.131 ** (0.049).
  - Text interpretation: households of ME-mothers did not have significantly higher income per capita but had significantly higher expenditure per capita and assets per capita; they also showed reduced cigarette expenditures and increased education expenditures per child aged 6–18.
- Overall interpretation of mechanisms:
  - Limited support for maternal labor-market income as the dominant channel.
  - Evidence favors family environment and parental inputs: healthier paternal lifestyles, higher household education spending per child, reduced tobacco spending, improved living conditions and assets.
  - Assortative mating may contribute.

### VIII. Policy-relevant conclusions
- Program-level: Mandatory, free ME had large, persistent, and multi-generational educational impacts despite low fiscal cost.
- Gender gap: ME contributed to reducing the educational gap between girls and boys in North Vietnam; later schooling expansion coincided with girls’ net secondary enrollment rates catching up with and overtaking boys'.
- Intergenerational spillovers: One additional year of maternal education boosted child education by up to 0.65 years (IV), and direct maternal effects controlling for paternal education estimated at about 0.73 to 0.85 child-years per maternal year.
- Policy implication: Educational policies that expand schooling access at the right ages with broad social and government support and low cost can yield large, persistent, and multi-generational benefits relevant for low-income countries seeking cost-effective female schooling expansion.

*Intergenerational Transmission of Education in a Developing Country: Evidence from A Mass Education Program in Vietnam — Working Paper No. WP/2022/245*

### Annex I. Additional Figures and Tables .................................................................................

### Annex I. Additional Figures and Tables .......................................................................................................... 27

### Contents
- "Annex I. Additional Figures and Tables .......................................................................................................... 27"

*Source: wpiea2022245-print-pdf - Annex I. Additional Figures and Tables ..................................................................................*

### References .............................................................................................................

### wpiea2022245-print-pdf - References .............................................................................................................

### I. Introduction — research question and contributions
- Research question: causal effects of mother education on child education in Vietnam exploiting a natural experiment from the First Indochina War (1946-1954) and the DRV "Mass Education" (ME) program.
- Identification: school-age girls exposed to ME in DRV-controlled provinces became better educated (ME-mothers); compare their children in the 1997-1998 Vietnam Living Standards Survey (VLSS) with children of unexposed mothers who lived in French-controlled areas during the war.
- Main contributions:
  - Strong estimated effect: on average, one additional year of maternal education boosts child education by up to 0.65 years.
  - Developing-country evidence: ME program raised female education attainment by 1.5 years, or 48 percent, from an average pre-treatment 3.1 years.
  - Mechanisms: suggestive evidence that non-income family factors (healthier lifestyles, more durable assets, higher education spending, less spending on unhealthy items) are transmission channels; little evidence of improved labor market incomes for ME-mothers or their husbands.

### II. Country context and the Mass Education (ME) program
- Historical context:
  - Colonial period: France colonized Vietnam from 1887 until 1945; female illiteracy estimated near 100 percent before 1945.
  - DRV launched Mass Education ("Bình dân học vụ") immediately after independence in 1945; Decree 20/SL (issued September 8, 1945) mandated free and mandatory illiteracy-eradicating classes for all.
- First Indochina War (September 1946 – May 1954) effectively divided Vietnam into DRV-controlled and French-controlled areas; ME expanded rapidly in DRV-controlled North and Central provinces.
- ME program features: highly inexpensive, community-supported, classes delivered in homes, temples, public places; reported more than 2.5 million people could read and write by 1946 in North and Central Vietnam.

### III. Empirical design and data
- Exposure measures:
  - Binary exposure: ME_M = 1 if mother was born in the three Northern provinces under complete DRV control (Nghe An, Thanh Hoa, Ha Tinh) and in the treatment cohort; mothers born 1940-1945 are treatment cohort, controls are 1924-1935 and 1950-1961; cohorts 1936-1939 and 1946-1949 removed as "contaminated".
  - Continuous exposure: Expose_M = number of years a mother was exposed to ME between ages 6 and 11 (varies between 0 and 5).
- Econometric specifications:
  - First-stage: SM = α + β(ME_M * L_M) + γX_M + bp_M + τ_M + ε  (equation (1))
  - Alternative first-stage: SM = α + βExpose_M + γX_M + bp_M + τ_M + ε  (equation (2))
  - Reduced-form for child outcomes (average child education Y̅_C): equations (3) and (4), analogous to first-stage.
  - IV: Y̅_C = θ + δ Ŝ_M + ηX_M + bp_M + τ_M + μ  (equation (5)), where Ŝ_M is maternal education instrumented by ME exposure.
- Key estimation choices and assumptions:
  - Exclude mothers who migrated before age 15.
  - Robust standard errors clustered two-way: mother's birth province and household's current commune.
  - Exclusion restriction discussed: ME affects child education only through mother's education; potential confounders (mother's health, economic/political connections, other DRV programs) are addressed in section IV.3 (not shown here).
  - Sensitivity: recognition that mixed-occupied provinces and timing of ME rollout could bias instruments toward understating effects; robustness checks remove partially occupied provinces.
- Data:
  - Primary dataset: 1997-1998 VLSS, nationally representative ~6,000 households; links adult children to parents even if not co-resident; includes date/place of birth, migration history, demographics, education, health, labor markets, consumption, income, assets.
  - Construction of key variable "years of education": highest grade completed; for post-secondary, years = 12 + reported years at junior college/university/post-graduate.
  - Occupation/occupation-start dates of DRV control compiled from provincial communist party/army histories (Dang et al., 2021).

### IV. Multi-generational effects — first-stage and reduced-form evidence
- Graphical evidence:
  - Years of maternal education for cohorts 1924-1935 (control), 1940-1945 (treated), and 1950-1961 (control) show parallel trends pre- and post-treatment, with a clear divergence (higher education) for treated cohorts in DRV-occupied provinces during 1940-1945.
  - Male cohorts show negligible differences between DRV- and French-occupied provinces.
- First-stage regression highlights (Table 1 summary, Panels A and B):
  - ME-mothers were between 19 to 26 percentage points higher in the probability of achieving reading literacy, math competency, primary education completion, and secondary education completion (Columns 1–4, Panel A).
  - ME-mothers accomplished 1.5 more years of schooling than peers in French-controlled areas (Column 5, Panel A) — a 48 percent increase from the average 3.1 years in the pre-treatment period.
  - Panel B (continuous exposure): an additional year of exposure to ME leads to 4 to 9 percent higher probability of reading/math competency and completing primary/secondary education (Columns 1–4, Panel B).
  - Panel B, Column 5: an additional year of ME exposure raises education by 0.25 years.
- Treatment/cohort definitions and exposure window:
  - Treatment cohort: mothers born 1940–1945 (would be of primary school age during the War).
  - Control cohorts: mothers born 1924–1935 and 1950–1961.
  - Contaminated cohorts removed in binary exposure approach: 1936–1939 and 1946–1949.
  - Exposure years counted between ages 6 and 11, bounded 0–5.

### V. Mechanisms and suggestive findings on transmission channels
- Observed household and parental characteristics (VLSS 1998):
  - Husbands of ME-mothers had a healthier lifestyle and better education.
  - ME-mothers' households lived in better conditions and had more durable assets.
  - ME-mothers' households spent significantly more on education and less on unhealthy items.
  - ME-mothers and their husbands did not have significantly higher incomes or dramatically improved labor market outcomes compared to unexposed counterparts.
- Interpretation:
  - Family environment and parental inputs (non-income channels) likely important in transmitting maternal education gains to children's education.
  - Limited labor market improvements in the pre-1990s centrally planned Vietnamese economy may have increased time mothers spent at home, amplifying the transmission to children.

- Key numeric findings (preserved exactly as reported):
  - One additional year of maternal education boosts child education by up to 0.65 years.
  - ME program raised female education attainment by 1.5 years, or 48 percent on average, from an average 3.1 years of education in the pre-treatment period.
  - ME-mothers were between 19 to 26 percentage points higher in the probability of achieving reading, math literacy, and completing primary or secondary education (Panel A, Columns 1–4).
  - An additional year of ME exposure leads to 4 to 9 percent higher probability of reading and math competency and completing primary or secondary education (Panel B, Columns 1–4).
  - An additional year of ME exposure raises education by 0.25 years (Panel B, Column 5).
  - Exposure varies between 0 and 5 years.

*Source: wpiea2022245-print-pdf - References .............................................................................................................*

### 15. This exclusion ensures we know the girls studied in their birthplaces, as we do not have

### wpiea2022245-print-pdf - 15. This exclusion ensures we know the girls studied in their birthplaces, as we do not have

### ME Exposure and Mother's Education
- Panel A (First Exposure Measure) — treated effect on mothers:
  - Reading Literacy: 0.204 ***
  - Math competency: 0.192 **
  - Primary education completion or above: 0.263 ***
  - Secondary education completion or above: 0.187 **
  - Years of education: 1.474 ***
  - Standard errors (in parentheses): (0.073) (0.089) (0.065) (0.084) (0.297)
  - N: 1622, 1622, 1618, 1618, 1619
  - adj. R2: 0.460, 0.460, 0.521, 0.395, 0.535

- Panel B (Second Exposure Measure) — years of exposure to ME:
  - Reading Literacy: 0.028
  - Math competency: 0.021
  - Primary education completion or above: 0.043 **
  - Secondary education completion or above: 0.023
  - Years of education: 0.253 **
  - Standard errors (in parentheses): (0.017) (0.019) (0.016) (0.014) (0.094)
  - N: 1979, 1979, 1975, 1975, 1976
  - adj. R2: 0.414, 0.425, 0.481, 0.371, 0.499

- Robustness checks:
  - Removing mixed-occupied provinces (Appendix Table A3 Panel A3.1) yields very quantitatively similar estimates; sample size drops only slightly.

- Addressing alternative explanations:
  - Dang et al. (2021) provide analysis addressing socioeconomic and selection channels; children in DRV-controlled South Vietnam where ME was not implemented did not show better outcomes than French-controlled areas.
  - No evidence that control group was more likely to receive remittances from overseas (1998 survey).

### ME Exposure and the Next Generation's Education (Reduced-form effects)
- Outcomes used: children’s average years of education; years of education of the most educated child; percentage of children completing high school.
- Panel A (First Exposure Measure) — Treated mothers:
  - Average years of children education: 0.894 ***
  - Years of education of the most educated child: 0.808 ***
  - Percentage of children completing high school: 0.149 ***
  - Standard errors (in parentheses): (0.249) (0.191) (0.039)
  - N: 1206 (all columns)
  - adj. R2: 0.347, 0.385, 0.125
  - Note: ME-mother had 1.5 more years of education on average herself.

- Panel B (Second Exposure Measure) — Years of exposure to ME:
  - Average years of children education: 0.188 **
  - Years of education of the most educated child: 0.129
  - Percentage of children completing high school: 0.031 ***
  - Standard errors (in parentheses): (0.076) (0.076) (0.008)
  - N: 1546 (all columns)
  - adj. R2: 0.322, 0.349, 0.123

- Robustness: Removing mixed-occupied provinces (Appendix Table A3 Panel A3.2) yields similar estimates; one-year exposure impact on the most educated child becomes statistically significant.

### On the Causal Impact of Maternal Education on Children's Education
- Confounding factors considered:
  - Alternative DRV policies, political/economic connections, selective migration.
  - Evidence against these: children of exposed fathers (ME-fathers) do not have significantly higher education than children of non-ME fathers.

- Table 3 (Father’s ME Exposure and Children's Education) — main estimates not statistically significant:
  - Panel A (First Exposure Measure) — Treated:
    - Average years of children education: 0.898 (0.531)
    - Years of education of the most educated child: 0.565 (0.502)
    - Percentage of children completing high school: 0.109 (0.097)
    - N: 974; adj. R2: 0.431, 0.470, 0.187
  - Panel B (Second Exposure Measure) — Years of exposure to ME:
    - Average years of children education: 0.121 (0.102)
    - Years of education of the most educated child: 0.069 (0.114)
    - Percentage of children completing high school: 0.015 (0.016)
    - N: 1256; adj. R2: 0.394, 0.432, 0.150

- South Vietnam falsification:
  - Table 4 shows mothers exposed to DRV in South Vietnam (where ME was not systematically carried out) do not have children with significantly higher education:
    - Panel A Treated: 0.553 (avg years), 0.453 (most educated), 0.033 (high school)
    - Panel B Years of exposure: 0.113, 0.126, 0.007
    - N: 1313–1659; adj. R2: 0.268–0.295 (Panel A) and 0.261–0.289 (Panel B)

- IV estimates to identify intergenerational impact (upper bound interpretation):
  - Table 5 — Panel A (First instrument):
    - OLS: Years of maternal education -> Average years of children education: 0.308 *** (0.042)
    - IV: Years of maternal education -> Average years of children education: 0.583 ** (0.280)
    - F-test of the excluded instrument: 13.04
    - N: 1206
  - Table 5 — Panel B (Second instrument):
    - OLS: 0.350 *** (0.033)
    - IV: 0.651 ** (0.312)
    - F-test of the excluded instrument: 7.64
    - N: 1545
  - Interpretation: IV estimates imply one more year of maternal schooling raises children's average education by about 0.58 (first instrument) to 0.65 (second instrument) additional years.

### Potential Transmission Mechanisms (Section V)
- Overview:
  - Reduced-form regressions for exposure measures; evidence points to family factors (better lifestyle, stronger focus on education) rather than clear labor-market income channels.

- V.1 Maternal labor market and reproduction outcomes (Table 6):
  - No significant effects of ME exposure on maternal non-farm participation, paid work, white-collar work, or health insurance.
  - First birth age and number of children: statistically insignificant impacts.
  - Hours of household chores: ME exposure measures do not change this outcome.
  - Selected coefficients (First Exposure Treated):
    - Non-farm participation in past seven days: 0.035 (0.062)
    - Paid work: -0.008 (0.024)
    - White-collar: 0.064 (0.074)
    - Health insurance: 0.069 (0.096)
    - First birth age: 0.470 (0.604)
    - Number of children: -0.025 (0.214)
    - Hours of household chores: 1.054 (1.446)
    - N varies by outcome; e.g., N 1623 for many outcomes.
  - Second Exposure (Years of ME exposure) coefficients similarly small/non-significant; e.g., Years of ME exposure on non-farm participation: 0.009 (0.014).

- V.2 Paternal education, labor market outcome, and lifestyle:
  - Table 7 (Father’s education):
    - First Exposure Treated shows marginal positive effects on husband's reading literacy: 0.075 ** (0.036) and math competency: 0.090 ** (0.042); other education measures smaller and not significant.
    - Second Exposure Years of ME exposure: reading literacy 0.013 (0.009), math competency 0.016 (0.011).
  - Table 8 (Father’s labor market outcomes): no significant impact of maternal ME exposure on father's non-farm participation, paid work, white-collar work, or health insurance.
  - Table 9 (Father’s lifestyle):
    - First Exposure Treated:
      - Husband ever-smoked: -0.005 (0.034)
      - Husband smoking now: -0.024 ** (0.010)
      - Number of cigarettes a day by husband: -0.142 ** (0.063)
      - N: 1150 for these outcomes
    - Second Exposure Years of ME exposure:
      - Husband ever-smoked: -0.001 (0.005)
      - Husband smoking now: -0.005 *** (0.002)
      - Number of cigarettes a day by husband: -0.029 * (0.015)
      - N: 1420

- V.3 Household income, expenditure, assets, and living conditions (Table 10 and 11):
  - Table 10 (per capita, log terms and living conditions):
    - First Exposure Treated:
      - Log of real income per capita: 0.298 (0.191)
      - Log of real expenditure per capita: 0.153 ** (0.074)
      - Log of real durable assets' value per capita: 0.269 ** (0.111)
      - Log of living area per capita: 0.046 (0.058)
      - House structure: 0.056 (0.041)
      - Floor quality: 0.174 *** (0.032)
      - N ranges: 1494–1623
    - Second Exposure Years of ME exposure:
      - Log of real income per capita: 0.044 (0.030)
      - Log of real expenditure per capita: 0.030 ** (0.013)
      - Log of real durable assets' value per capita: 0.078 *** (0.026)
      - Log of living area per capita: 0.009 (0.014)
      - House structure: 0.024 ** (0.009)
      - Floor quality: 0.046 *** (0.009)
      - N: 1838–1980
  - Table 11 (expenses on tobacco and education):
    - First Exposure Treated:
      - Log (expense on tobacco +1): -0.710 ** (0.295)
      - Educational expense per kid aged 6-18: 0.369 ** (0.177)
      - N: 1622 and 1144 respectively
    - Second Exposure Years of ME exposure:
      - Log (expense on tobacco +1): -0.122 * (0.068)
      - Educational expense per kid aged 6-18: 0.131 ** (0.049)
      - N: 1979 and 1348 respectively
  - Interpretation: households of ME-mothers did not have significantly higher income per capita but had significantly higher expenditure per capita (15 percent) and assets per capita (26 percent) according to text; they also showed reduced cigarette expenditures and increased education expenditures per child aged 6–18.

### IV Inter-generational Estimates Controlling for Paternal Education (Direct maternal effect)
- Table 12 — IV regressions controlling for paternal education:
  - IV regressions — First instrument:
    - Years of maternal education -> Average years of children education: 0.730 ** (0.342)
    - Years of paternal education: 0.093 (0.113)
    - F-test of the excluded instrument: 21.68
    - N: 882
  - IV regressions — Second instrument:
    - Years of maternal education -> Average years of children education: 0.853 * (0.458)
    - Years of paternal education: 0.065 (0.152)
    - F-test of the excluded instrument: 10.84
    - N: 1145
  - Interpretation: IV estimates controlling for paternal education imply direct maternal effects on child education of about 0.73 to 0.85 additional child-years per maternal year.

### Conclusions and Policy-relevant Findings
- Program-level findings:
  - The mandatory, free ME program had large, persistent, and multi-generational educational impacts despite low fiscal cost to the DRV government.
  - The ME program contributed to reducing the educational gap between girls and boys in North Vietnam; similar programs and later educational expansion coincided with girls’ net secondary enrollment rates catching up with and overtaking boys' in recent decades.

- Intergenerational spillovers:
  - One additional year of maternal education boosted child education by up to 0.65 years (IV estimates), larger than many estimates from advanced-country literature.
  - When controlling for paternal education, the direct maternal effect is even larger (about 0.73 to 0.85).

- Suggested transmission channels:
  - Evidence does not strongly support maternal labor-market income as the dominant channel.
  - Evidence points to family pathways: educated mothers, healthier paternal lifestyles (e.g., reduced current smoking and cigarettes per day), higher household expenditure on education per child, reduced household spending on tobacco, and improved household living conditions and assets.
  - Assortative mating may also contribute to observed household patterns.

- Policy implication:
  - Educational policies that achieve broad social and government support and low cost—especially those that expand schooling access at the right ages—can yield large, persistent, and multi-generational benefits.
  - Lessons may be relevant for low-income countries seeking cost-effective ways to expand female schooling and generate long-term intergenerational gains.

*Source: wpiea2022245-print-pdf - 15. This exclusion ensures we know the girls studied in their birthplaces, as we do not have*

### Annex I. Additional Figures and Tables

### Annex I. Additional Figures and Tables

### Figure A1 — Map of Vietnam under DRV and the French Occupation (1946-1954)
- Note: Different colors indicate the number of years under the DRV’s occupation.
- Provinces under complete DRV control throughout the First Indochina War (nine years under DRV control): Nghe An, Thanh Hoa, Ha Tinh, Quang Ngai, Binh Dinh, Phu Yen (darkest-blue color).
- Some mountainous North provinces (Ha Giang, Cao Bang, Tuyen Quang, Bac Can) fell to DRV control early in the war.
- Disclaimer: Country borders or names do not necessarily reflect the IMF’s official position. The dividing line for the North and the South after the First Indochina War was established by the Geneva Accords in 1954.

### Table A1 — Summary Statistics (Mother cohorts, husbands, and children)
- Cohorts covered: 1924-1935 and 1950-1961 mother cohorts; 1940-1945 mother cohorts.
- Significance notation: ***(p<0.01), **(p<0.05), *(p<0.1).

Key maternal education statistics (means, S.D., and DRV–French differences):
- Reading Literacy (1924-1935 & 1950-1961): DRV mean 0.72, S.D. 0.45; French mean 0.69, S.D. 0.46; Difference 0.03.
- Reading Literacy (1940-1945): DRV mean 0.89, S.D. 0.32; French mean 0.65, S.D. 0.48; Difference 0.235.
- Math competency (1924-1935 & 1950-1961): DRV mean 0.64, S.D. 0.48; French mean 0.65, S.D. 0.48; Difference -0.005.
- Math competency (1940-1945): DRV mean 0.78, S.D. 0.42; French mean 0.59, S.D. 0.49; Difference 0.189 **.
- Primary education completion or above (1924-1935 & 1950-1961): DRV mean 0.64, S.D. 0.48; French mean 0.58, S.D. 0.49; Difference 0.025.
- Primary education completion or above (1940-1945): DRV mean 0.70, S.D. 0.46; French mean 0.40, S.D. 0.49; Difference 0.300 ***.
- Secondary education completion or above (1924-1935 & 1950-1961): DRV mean 0.47, S.D. 0.50; French mean 0.42, S.D. 0.49; Difference 0.048.
- Secondary education completion or above (1940-1945): DRV mean 0.46, S.D. 0.50; French mean 0.23, S.D. 0.42; Difference 0.237 ***.
- Years of education (1924-1935 & 1950-1961): DRV mean 6.62, S.D. 3.90; French mean 5.98, S.D. 4.10; Difference 0.643 **.
- Years of education (1940-1945): DRV mean 6.93, S.D. 3.26; French mean 4.86, S.D. 3.35; Difference 2.066 ***.

Key paternal education statistics:
- Reading Literacy (1924-1935 & 1950-1961): DRV mean 0.89, S.D. 0.32; French mean 0.92, S.D. 0.28; Difference -0.03.
- Reading Literacy (1940-1945): DRV mean 0.95, S.D. 0.22; French mean 0.89, S.D. 0.31; Difference 0.062.
- Math competency (1924-1935 & 1950-1961): DRV mean 0.85, S.D. 0.35; French mean 0.89, S.D. 0.31; Difference -0.04.
- Math competency (1940-1945): DRV mean 0.93, S.D. 0.26; French mean 0.87, S.D. 0.34; Difference 0.06.
- Primary education completion or above (1924-1935 & 1950-1961): DRV mean 0.83, S.D. 0.38; French mean 0.77, S.D. 0.42; Difference 0.052.
- Primary education completion or above (1940-1945): DRV mean 0.83, S.D. 0.38; French mean 0.74, S.D. 0.44; Difference 0.094.
- Secondary education completion or above (1924-1935 & 1950-1961): DRV mean 0.65, S.D. 0.48; French mean 0.57, S.D. 0.50; Difference 0.076 **.
- Secondary education completion or above (1940-1945): DRV mean 0.55, S.D. 0.50; French mean 0.52, S.D. 0.50; Difference 0.026.
- Years of education (1924-1935 & 1950-1961): DRV mean 8.49, S.D. 3.59; French mean 8.01, S.D. 3.49; Difference 0.471 *.
- Years of education (1940-1945): DRV mean 8.83, S.D. 3.82; French mean 7.86, S.D. 3.83; Difference 0.978.

Key children education statistics:
- Secondary education completion or above (1924-1935 & 1950-1961): DRV mean 0.50, S.D. 0.44; French mean 0.52, S.D. 0.44; Difference 0.02.
- Secondary education completion or above (1940-1945): DRV mean 0.61, S.D. 0.40; French mean 0.66, S.D. 0.38; Difference 0.052.
- High school education completion or above (1924-1935 & 1950-1961): DRV mean 0.21, S.D. 0.35; French mean 0.22, S.D. 0.36; Difference 0.01.
- High school education completion or above (1940-1945): DRV mean 0.29, S.D. 0.36; French mean 0.41, S.D. 0.39; Difference 0.122 **.
- Average years of education of children (1924-1935 & 1950-1961): DRV mean 7.85, S.D. 3.54; French mean 8.01, S.D. 3.94; Difference 0.16.
- Average years of education of children (1940-1945): DRV mean 9.02, S.D. 3.00; French mean 9.96, S.D. 3.06; Difference 0.947 *.
- Education of the most educated children (1924-1935 & 1950-1961): DRV mean 9.30, S.D. 4.09; French mean 9.65, S.D. 4.40; Difference 0.35.
- Education of the most educated children (1940-1945): DRV mean 11.06, S.D. 3.12; French mean 12.06, S.D. 3.00; Difference 0.997 **.

### Table A2 — Number of Years of Potential Exposure to the ME Program (by birth year and occupation timing)
- Individuals assumed to start primary school at age six.
- Maximum number of years of potential exposure restricted to 5 years (number of years required to achieve a primary school degree).
- Individuals assumed to stop going to school if never enrolled at age 11.
- ME program start dates by area:
  - DRV-occupied provinces: ME program started in 1945.
  - French-controlled provinces: ME program started in 1954.
- Example timeline entries (columns: Birth Year | DRV-occupied Provinces (1945) | French-occupied Provinces (1954) | Mixed-occupation Provinces Example 1 (1947) | Example 2 (1951)):
  - 1924: 0 | 0 | 0 | 0
  - 1934: 0 | 0 | 0 | 0
  - 1935: 1 | 0 | 0 | 0
  - 1936: 2 | 0 | 0 | 0
  - 1937: 3 | 0 | 1 | 0
  - 1938: 4 | 0 | 2 | 0
  - 1939: 5 | 0 | 3 | 0
  - 1940: 5 | 0 | 4 | 0
  - 1941: 5 | 0 | 5 | 1
  - 1942: 5 | 0 | 5 | 2
  - 1943: 5 | 0 | 5 | 3
  - 1944: 5 | 1 | 5 | 4
  - 1945: 5 | 2 | 5 | 5
  - 1946: 5 | 3 | 5 | 5
  - 1947: 5 | 4 | 5 | 5
  - 1948 ... 1961: 5 | 5 | 5 | 5

### Table A3 — Robustness: Removing Mixed-Occupied Provinces
Panel A3.1: ME Exposure and Mother’s Education (separate regressions; robust SEs clustered two-way)
- Panel A: First Exposure Measure (Treated coefficients and standard errors)
  - Reading Literacy (1): Treated 0.189 ** (0.078). N 1377. adj. R2 0.442.
  - Math competency (2): Treated 0.177 * (0.094). N 1377. adj. R2 0.441.
  - Primary education completion or above (3): Treated 0.260 *** (0.072). N 1375. adj. R2 0.503.
  - Secondary education completion or above (4): Treated 0.175 * (0.088). N 1375. adj. R2 0.394.
  - Years of education (5): Treated 1.379 *** (0.371). N 1376. adj. R2 0.513.
- Panel B: Second Exposure Measure (Years of exposure to ME)
  - Reading Literacy (1): 0.031 * (0.018). N 1677. adj. R2 0.395.
  - Math competency (2): 0.029 (0.020). N 1677. adj. R2 0.413.
  - Primary education completion or above (3): 0.049 *** (0.016). N 1675. adj. R2 0.470.
  - Secondary education completion or above (4): 0.030 ** (0.014). N 1675. adj. R2 0.370.
  - Years of education (5): 0.276 *** (0.088). N 1676. adj. R2 0.477.

Notes for A3.1: Mixed-occupied provinces removed. Regressions control for gender, religion groups (including Buddhism and Christianity), major ethnic group, birth year fixed effects, birth province fixed effects; samples exclude those who migrated when younger than 15.

Panel A3.2: Mother’s ME Exposure and Children's Education
- Panel A: First Exposure Measure (Treated)
  - Average years of children education (1): Treated 0.862 *** (0.268). N 1021. adj. R2 0.329.
  - Years of education of the most educated child (2): Treated 0.810 *** (0.191). N 1021. adj. R2 0.373.
  - Percentage of children completing high school (3): Treated 0.145 *** (0.037). N 1021. adj. R2 0.120.
- Panel B: Second Exposure Measure (Years of exposure to ME)
  - Average years of children education (1): 0.202 *** (0.048). N 1306. adj. R2 0.304.
  - Years of education of the most educated child (2): 0.158 *** (0.049). N 1306. adj. R2 0.335.
  - Percentage of children completing high school (3): 0.035 *** (0.006). N 1306. adj. R2 0.119.

Notes for A3.2: Mixed-occupied provinces removed. Regressions control for gender, religion groups (including Buddhism and Christianity), major ethnic group, maternal birth year and birth province fixed effects; samples exclude those who migrated when younger than 15.

Panel A3.3: Impact of Maternal Education on Children's Education (OLS and IV; mixed-occupied provinces removed)
- Panel A. First instrument
  - OLS regressions (Years of maternal education)
    - Average years of children education: 0.334 *** (0.042).
    - Years of education of the most educated child: 0.275 *** (0.041).
    - Percentage of children completing high school: 0.039 *** (0.005).
  - IV regressions (Years of maternal education)
    - Average years of children education: 0.569 * (0.302).
    - Years of education of the most educated child: 0.536 ** (0.192).
    - Percentage of children completing high school: 0.095 * (0.047).
  - F-test of the excluded instrument: 9.30 for all three outcomes. N 1020.
- Panel B. Second instrument
  - OLS regressions (Years of maternal education)
    - Average years of children education: 0.355 *** (0.034).
    - Years of education of the most educated child: 0.306 *** (0.034).
    - Percentage of children completing high school: 0.043 *** (0.005).
  - IV regressions (Years of maternal education)
    - Average years of children education: 0.687 ** (0.316).
    - Years of education of the most educated child: 0.536 *** (0.149).
    - Percentage of children completing high school: 0.120 * (0.060).
  - F-test of the excluded instrument: 6.73 for all three outcomes. N 1305.

Notes for A3.3: Mixed-occupied provinces removed. Regressions control for gender, religion groups (including Buddhism and Christianity), major ethnic group, maternal birth year and birth province fixed effects; samples exclude those who migrated when younger than 15. Robust standard errors in parentheses adjusted for two-way clustering (mother's birth province and household's current commune); *** p<0.01, ** p<0.05, * p<0.1.

*Intergenerational Transmission of Education in a Developing Country: Evidence from A Mass Education Program in Vietnam — Working Paper No. WP/2022/245*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2022/english/wpiea2022245-print-pdf.pdf_