Page 105 - IJPS-11-2

International Journal of
Population Studies Do female-headed households have poorer finances?

correct response is assigned a value of 1, and the incorrect households have commercial insurance, with female-
response is assigned a value of 0. Question 3: “assuming the headed households (12.45%) having a significantly higher
bank’s annual interest rate is 5% and the annual inflation percentage compared to male-headed households (8.86%).
rate is 8%, what can you purchase with 100 yuan in the On average, households own 0.61 financial investment
bank after 1 year?” The correct response is assigned a value products (ranging from 0 to 4), with female-headed
of 1, and the incorrect response is assigned a value of 0. households (0.65) owning more compared to male-headed
Financial literacy is then determined by measuring the households (0.60).
total score, ranging from 0 to 3. Among the participants, 50% of the household heads
2.3. Methods of analysis were between 56 and 80 years of age, and 53.42% of
female household heads and 49.39% of male household
Descriptive and bivariate analyses (t and χ tests) were heads were in this age range. The educational attainment
2
used to describe and compare sample characteristics. The of female household heads was generally higher than that
missing completely at random (MCAR) test (χ² = 1106.72, of male household heads. Female-headed households
p = 0.053) suggested that the data were MCAR. Therefore, had significantly higher income (10.43) and more urban
listwise deletion was applied in the analysis. We employed residency (78.27%) and were more likely to reside in first-
a combination of propensity score matching and logistic tier cities (36.27%) compared to male-headed households
regression to mitigate selection bias. Initially, male-headed (10.30, 59.75%, and 24.82%, respectively). Female
and female-headed households were matched based on household heads also had significantly higher financial
comparable propensity scores. Subsequently, logistic literacy (0.83) compared to male household heads (0.79).
regression was utilized to examine the influence of the
gender of the household head on the financial health of the 3.2. Propensity score matching
household and obtain the coefficient of influence.
3.2.1. Propensity score matching estimation
2.3.1. Matching method A logit regression was conducted to predict financial
The study utilized three propensity score matching health, which included covariates such as the household
approaches: nearest neighbor one-to-two matching head’s age, education level, occupation type, financial
(cal = 0.03), kernel matching (normal kernel matching, literacy, urban or rural residence status, and the logarithm
bandwidth = 0.3), and radius matching (cal = 0.03). To of the household’s total income in the last year. Propensity
minimize sample loss, all three matching methods were scores were then calculated for each household, and
conducted with replacement. Ultimately, nearest neighbor three matching methods – nearest neighbor one-to-two
matching yielded 16,464 samples, kernel matching matching, radius matching, and kernel matching – were
yielded 11,782 samples, and radius matching yielded employed to pair samples of households with male and
31,339 samples. female heads. Finally, the balance of covariates in the
matched samples was evaluated to ensure comparability.
3. Results
3.2.2. Balance test
3.1. Descriptive statistics
The results of the balance test presented in Table 3
The study used four methods, including standardized demonstrate that upon utilizing these three methods
residuals, studentized residuals, Cook’s distance, and for matching, the explanatory power of the matching
Welch distance, to identify outliers. Samples flagged as variables concerning differences in the dependent
outliers by any of these methods were excluded. After variable converges toward zero. This suggests a substantial
excluding samples with missing values and outliers, the reduction in distributional disparities in the matching
study retained a total of 31,361 samples. variables between the groups of female-headed and male-
Table 2 presents the overall sample characteristics and headed groups, thereby mitigating the confounding effect
gender differences. Approximately 73% of households introduced by these variables. Specifically, the overall
reported having a balanced income and expenditure, mean standard deviation (MeanBias) between the matched
with female-headed households (74.46%) having a groups falls within 10%. Furthermore, the R of the logit
2
significantly higher percentage compared to male-headed model, re-estimated using the matched sample (Ps R ),
2
households (72.16%). Nearly 17% of households had a approaches zero, indicating a weak explanatory capability
savings account, with female-headed households (18.75%) of the model in determining household financial health.
having a significantly higher percentage compared This implies that the gender of the household head tends
to male-headed households (16.85%). Only 10% of toward conditional randomness. The outcomes from all

Volume 11 Issue 2 (2025) 99 https://doi.org/10.36922/ijps.4403