**Appendix B. Sample Weight Calculations**


**Table A2.** Share of income groups in Cambodia: our sample vs. population.

**Table A3.** Share of income groups in rural and urban Viet Nam: our sample vs. population.


**Table A4.** Share of rural and urban population in six regions in Viet Nam: our sample vs. population.


For the Cambodian sample, the weights are constructed based on income groups (Table A2). We calculate the weights for the Cambodian sample as follows:

$$\text{Weight}^{KHM} = \frac{IG\_{ip}}{IG\_{i\varepsilon}}$$

where *IGis* is the share of our sample in income group *i* (5 income groups as above); and *IGip* is the share of the population (2017 Media index) in income group *i*.

For the Vietnamese sample, we construct the weights based on: (i) income group in rural and urban areas (Table A3) and (ii) the share of rural and urban population in each region (Table A4). More specially, our weights for the Vietnamese sample are calculated as follows:

$$\% \text{weight}^{VNM} = \frac{IG\_{ip}^{u}}{IG\_{is}^{u}} \ast \frac{Pop\_{rp}^{u}}{Pop\_{rs}^{u}}$$

where *IGuis* is the share of our sample in income group *i* (5 income groups as above) and area *u* (*u* is either rural or urban); *IGuip* is the share of the population (Nielsen Monitoring data) in income group *i* and area *u*; *Popurs* is the share of our sample in each region *r* (6 regions as above) and area *u*; and *Popurp* is the share of the population (following GSO) in each region *r* and area *u*.

### **Appendix C. Estimates Based on Broader Definition of Saving**

Table A5 reports our estimation results for a broader definition of savings that includes not only those who hold savings products (i.e., formal savings) but also those who save in other forms such as keeping money at home, asking some family members to keep money for them, etc. (i.e., informal savings). The dependent variable takes the value one if an individual has any types of savings and zero otherwise. Columns (1)–(3) are the results using the Cambodian sample, while the remaining columns display the results using the Vietnamese sample. We use both the OLS estimator (columns (1) and (3)) and the probit estimator (columns (2) and (4)). Columns (3) and (6) are estimated using the GMM estimator with our conventional instrumental variables. For both countries, we report only the 2nd stage since the 1st stage is similar to the 1st stage reported in Table 8. The estimation results show an increase in the magnitude of the effect of the financial literacy score on the savings decision. A one standard deviation increase in the financial literacy score raises the likelihood of saving by about 12 percentage points among Cambodian respondents and 16 percentage points among Vietnamese respondents, which is twice as large as the effects on formal savings products alone. Similar patterns are also observed when we use the instrumental variable to address the endogeneity of the financial literacy score. Moreover, while the financial literacy score does not have a significant effect on formal savings behavior among Vietnamese respondents, it becomes a significant factor when informal savings are taken into account.


**Table A5.** Effects of financial literacy on savings behavior (broad definition), OLS estimators and IV.

Note: Figures in brackets are standard deviations. \*\*\*, \*\*, and \* denote coefficients significant at the 1%, 5%, and 10% statistical levels, respectively. The dependent variable is whether the respondent saved or not (either in formal or informal ways). Weighted samples are used for all estimations.

### **Appendix D. Estimates Based on Combined Samples of Cambodia and Viet Nam**

Tables A6 and A7 present our estimation results for the combined weighted Vietnamese and Cambodian samples, using the OLS estimator and the GMM estimator, respectively. The dependent variables in Appendix D are: financial literacy score (1), financial knowledge score (2), financial behavior score (3), financial attitude score (4), financial inclusion score (5), and savings behavior (6) and (7). The estimation results show that household income, education, and occupational status are the major determinants of the financial literacy score and its components (especially financial knowledge score and financial behavior score). The financial knowledge score is positively and significantly associated with the financial behavior score, but not with the financial attitude score. Males tend to have a higher financial knowledge score but lower financial attitude score than females.

The OLS results in Table A6 show that financial literacy is positively correlated with financial inclusion and saving behavior. A standard deviation increase in the financial literacy score is associated with an increase in the financial inclusion score of 39 percentage points and in the likelihood of savings of 9 percentage points. Household income, educational level, and, to some extent, occupational statuses are also positively correlated with financial inclusion and savings behavior. While age does not show much correlation with financial literacy score and its components, individuals either under 30 years old or from 30 to 60 years old have somewhat higher financial inclusion and more savings.


**Table A6.** Determinants of financial literacy and savings behavior (combined sample), OLS estimator.

Note: Figures in brackets are standard deviations. \*\*\*, \*\*, and \* denote coefficients significant at the 1%, 5%, and 10% statistical level, respectively. In all estimations, province dummies are controlled for. The weighted sample is used for all estimations.

As before, we attempt to control for endogeneity of the financial literacy score by using the mean financial literacy score at the provincial level. As shown in Table A7, the financial literacy score still has a statistically significant effect on financial inclusion. While the effect of the financial literacy score on narrowly defined savings behavior (i.e., whether the respondents hold any formal savings product) loses its significance after the endogeneity is controlled, the financial literacy score still has a positive

effect on our broader definition of savings (i.e., including those who have savings in informal forms). The latter relationship is significant at the 1% level.

The estimation results shown in Tables A6 and A7 indicate that, after controlling for household income, education, age, occupational status, and other covariates, the coefficients on the "Viet Nam" dummy variable are negative and statistically significant at the 1% level for most equations. This could be attributed to the fact that the financial literacy gap between Cambodia and Viet Nam is rather small, although Viet Nam seems to have higher values in all covariates that determine financial literacy, financial inclusion, and the saving decision. The reasons for this need to be investigated further.


**Table A7.** Effects of financial literacy on savings behavior and financial inclusion (combined sample), IV.

Note: a: Savings is defined based on whether an individual holds any saving product (i.e., formal savings form); b: Savings is defined based on whether an individual has any savings (either in formal savings forms or informal savings forms). Figures in brackets are standard deviations. \*\*\*, \*\*, and \* denote coefficients significant at the 1%, 5%, and 10% statistical level, respectively. The weighted sample is used for all estimations.
