15 CHAPTER IV DATA ANALYSIS 4.1 Descriptive Statistics The descriptive statistics in Table 4.1 provide an overview of the variables used in analyzing the relationship between FDI inflows and the gender pay gap in Asian developing countries. The dependent variable, the gender Gross National Income (GNI) gap, comprises 330 observations and exhibits a large standard deviation (6238.281), indicating significant variability in income disparities between genders across countries. On average, women earn 7247.676 USD less in GNI than men in the Asian developing countries included in the dataset. The minimum value of -3747.985 reflects instances of reversed pay gaps where women may earn more than men, while the maximum value of 35797.57 underscores extreme income disparities in some countries. Table 4. 1 Descriptive Statistics Variable Obs Mean Std. Dev. Min Max Countries 330 17 9.536 1 33 Gender GNI Gap 330 7247.676 6238.281 -3747.985 35797.57 FDI 330 2.791 5.116 -37.173 17.199 Paid Maternity Leave Policy 330 216.691 286.679 49 1215 Gender Education Gap 330 .999 .91 -.563 2.773 The independent variable, FDI inflows, shown in Table 4.2, initially exhibited moderate to high skewness (-1.94) and extremely high kurtosis (21.218), prompting a log transformation to normalize the data. After transformation, as shown in Table 4.3, the log of FDI inflows (Ln(FDI)) reduced the number of valid observations to 296, as log transformations are undefined for zero or negative values. The reduced mean (0.757) and standard deviation (1.122) indicate a narrower range of values compared to the raw FDI variable. As shown in Table 4.1, the moderator, paid maternity leave policy, averages 216.691 days, with a standard deviation of 286.679 and values ranging from 49 to 1215 days. This reflects substantial differences in leave policies across Asian developing 16 countries. The control variable, the gender education gap, has a mean of 0.999, indicating that men, on average, have approximately one additional year of schooling compared to women in the dataset. This variable has a moderate standard deviation of 0.91, with values ranging from -0.563 (where women outperform men in education) to 2.773, highlighting educational disparities between genders. Table 4. 2 FDI Detail Descriptive Statistics Variables Obs Mean Std. Dev. Min Max p1 p99 Skew. Kurt. FDI Inflows 330 2.791 5.116 -37.173 17.199 -11.79 16.37 -1.94 21.218 Table 4. 3 Ln(FDI) Descriptive Statistics Variable Obs Mean Std. Dev. Min Max Ln(FDI) 296 .757 1.122 -2.733 2.845 4.2 Correlation Matrix The pairwise correlations table (Table 4.4) reveals that as Ln(FDI) increases, the Gender GNI Gap tends to increase slightly (correlation = 0.037, p = 0.525), although this relationship is not statistically significant. This suggests that higher FDI inflows are linked to a slight widening of gender wage disparities. The Paid Maternity Leave Policy has a weak negative correlation with the Gender GNI Gap (-0.067), indicating that longer paid leave policies are associated with slightly smaller gender income gaps, though the effect is also insignificant. Both FDI inflows and the Paid Maternity Leave Policy show correlations with the dependent variable that align with theoretical expectations. In contrast, the Gender Education Gap has a moderate negative association with the dependent variable (-0.248), which contradicts the expectation of a positive correlation. Instead, the results suggest that reducing the Gender Education Gap tends to narrow income disparities. It is important to note that correlation identifies relationships between variables but cannot infer causality (Pott, 2008). The correlation coefficients between the independent variable, moderator, and control variable are all well below the common threshold of 0.7, indicating no multicollinearity issues in the dataset. This ensures that the independent variables are sufficiently distinct for use in further regression analyses without distorting the results. 17 Table 4. 4 Pairwise correlation Variables (1) (2) (3) (4) (1) Gender GNI Gap 1.000 (2) Ln(FDI) 0.037 1.000 (0.525) (3) Paid Maternity Leave Policy -0.067 0.193 1.000 (0.227) (0.001) (4) Gender Education Gap -0.248 -0.111 -0.222 1.000 (0.000) (0.057) (0.000) 4.3 Multicollinearity Test Although the pairwise correlation results suggest no issues with multicollinearity, the Variance Inflation Factor (VIF) test is necessary to confirm this (Kutner, Nachtsheim, & Neter, 2004). The results in Table 4.5 indicate that multicollinearity among the independent variables in the model is very low. All variables—Ln(FDI) (1.075), Gender Education Gap (1.047), and Paid Maternity Leave (1.045)—have VIF values well below the commonly accepted threshold of 5, with tolerance values (1/VIF) above 0.8. This indicates that there are no multicollinearity issues among the variables. The mean VIF is 1.056, further confirming the absence of significant multicollinearity. Low multicollinearity ensures that the regression coefficients are stable and the standard errors are not inflated, allowing for precise and reliable estimates. These results validate the robustness of the regression model and indicate that no adjustments are needed to address multicollinearity. Table 4. 5 Variance inflation factor VIF 1/VIF Ln(FDI) 1.075 .931 Gender Education Gap 1.047 .955 Paid Maternity Leave Policy 1.045 .957 Mean VIF 1.056 . 4.4 Autocorrelation and Heteroskedasticity Test In panel data, the likelihood of autocorrelation is high because past events often influence current outcomes (Pesaran, 2015). Similarly, heteroskedasticity is common in panel data due to inherent differences across countries or over time (Pesaran, 2015). 18 To detect the presence of first-order autocorrelation in the residuals of the model, the Wooldridge test was used. Based on Table 4.6, the p-value (0.0036) is below 0.05, leading to the rejection of the null hypothesis (H₀). This indicates significant first-order autocorrelation in the panel data residuals. In other words, the error terms are correlated across consecutive time periods for the same entity. Furthermore, to test whether the variance of the error terms differs across Asian developing countries or over time, the Modified Wald test for groupwise heteroskedasticity in a Fixed Effects (FE) regression model was employed. According to Table 4.7, since the p-value is less than 0.05, the null hypothesis (H₀) is rejected. This confirms the presence of significant groupwise heteroskedasticity, indicating that the error variances differ across developing countries in Asia in the panel data analysis. Table 4. 6 Wooldridge test for autocorrelation in panel data H0: no first order autocorrelation F( 1, 30) = 9.961 Prob > F = 0.0036 Table 4. 7 Modified Wald test for groupwise heteroskedasticity in fixed effect regression model H0: sigma(i)^2 = sigma^2 for all i Chi2 ( 11) = 41201.10 Prob > chi2 = 0.0000 The true effects of variables may be masked or misattributed due to improper handling of autocorrelation and heteroskedasticity (Worral & Pratt, 2004). Thus, I applied the cluster-robust standard errors to address heteroskedasticity and autocorrelation within clustered panel data models. 4.5 Empirical Findings 4.5.1 Base Model The results of the base model panel data regressions, shown in Table 4.8, provide key insights into the relationship between FDI inflows and the gender pay gap, with the gender education gap included as a control variable. For Asian developing countries, FDI inflows (Ln(FDI)) have a positive and 19 statistically significant coefficient (569.3, p < 0.05). This suggests that an increase in FDI inflows is associated with a widening of the gender pay gap, supporting the hypothesis (H1) that FDI inflows exacerbate income disparities between men and women. Similarly, in the ASEAN region, the results align with this study’s expectations regarding the relationship between FDI inflows and the gender pay gap in this sub- region. The coefficient for Ln(FDI) is even larger than that for all Asian developing countries (982.4, p < 0.05) and statistically significant. This suggests a stronger positive relationship between FDI inflows and the gender pay gap in ASEAN compared to the broader set of Asian developing countries. These findings imply that in ASEAN countries, FDI inflows may benefit men more than women, further widening wage inequalities. Thus, hypothesis (H1) holds true in this sub-region as well. The adjusted R² for the ASEAN region is slightly higher than that for the broader Asian developing countries, suggesting that the model’s variables are somewhat more predictive in the ASEAN sub-region. These findings are consistent with existing literature, which suggests that while FDI tends to reduce gender wage inequality in developed countries, it often retains or even exacerbates gender wage gaps in developing countries (Chaudhuri & Mukhopadhyay, 2014; Oostendorp, 2009). In this case, FDI increases gender wage disparities in both the ASEAN region and the broader group of Asian developing countries. Chen, Lai, and Ge (2011) explain that this issue stems from institutional weaknesses and inadequate gender equality frameworks, which are common in developing economies and contribute to their larger gender pay gaps compared to developed nations. However, the coefficient for the gender education gap is negative but remains insignificant in both the broader Asian developing countries and its sub-region, ASEAN, suggesting a limited influence of education disparities on the gender pay gap. This finding contrasts with Aguayo-Téllez (2012), who argued that limited access to education for women is a key driver of how FDI widens gender pay gaps in developing countries. 20 Overall, despite ASEAN's efforts in regional economic cooperation and its status as the leading recipient of FDI inflows among Asian developing countries, the findings indicate no significant differences between the broader Asian developing region and ASEAN as its sub-region. 4.5.2 Interaction Model The interaction model investigates the moderating effect of paid maternity leave policies, with the results shown in Table 4.8. When the interaction term is introduced, the findings reveal differing dynamics compared to the base model. Previous studies suggest that the interaction between paid maternity leave policies and FDI inflows can modify the direct effects of FDI on the gender pay gap (Oostendorp, 2009; Blanton & Blanton, 2015; Vahter & Masso, 2019). 4.5.2.1 Asian Developing Countries In Asian developing countries, the coefficient for Ln(FDI) decreases from 569.3 in the base model to 485.9 in the interaction model and loses statistical significance, although it remains positive. This aligns with findings by Vahter and Masso (2019), who explain that the interaction between paid maternity leave policies and FDI inflows often reduces the significance of FDI’s direct effects. Such policies establish fair labor standards, limiting the extent to which FDI-driven companies can exploit wage differences, thereby protecting workers and promoting fairer wage practices (Vahter & Masso, 2019). While the positive coefficient for Ln(FDI) indicates that FDI may contribute to increased gender income disparities, this effect is not statistically significant. Braunstein (2006) suggests that FDI-driven wage disparities are often influenced by skill-based premiums, but these do not necessarily translate into significant gender income gaps, as broader socio-economic factors, such as urban-rural income inequalities, may play a larger role. The interaction term between paid maternity leave and FDI inflows has a small positive coefficient (0.358). Although not statistically significant, this suggests that paid maternity leave could strengthen the relationship between FDI and the gender income gap. Amin and Islam (2022) provide a possible explanation: financially supporting women during maternity leave may increase short-term costs for employers, particularly in competitive, FDI-driven industries. This can lead to a 21 perception that hiring women is costlier than hiring men, resulting in wage reductions or fewer opportunities for women, thereby reinforcing the link between FDI and gender income disparities (Amin & Islam, 2022). The direct coefficient for paid maternity leave policy (-1.214) is negative but not statistically significant, suggesting that such policies may help reduce gender wage disparities regardless of FDI inflows. Oostendorp (2009) supports this notion, arguing that regulations like paid maternity leave reduce occupational gender gaps by requiring multinational corporations to adhere to fair labor standards, thereby weakening FDI’s direct impact on wage disparities. Overall, while the results are not statistically significant, they highlight the potential of strong labor policies to mitigate gender income inequality in FDI-influenced economies. 4.5.2.2 The ASEAN region For the ASEAN region, the coefficient for Ln(FDI) remains positive and statistically significant (4747.4, p < 0.05), indicating that the relationship between FDI inflows and the gender pay gap persists strongly in this sub-region even when moderation is considered. This finding can be explained by indirect evidence from prior research. Studies suggest that FDI influences gender wage disparities differently in developed and developing countries (Chaudhuri & Mukhopadhyay, 2014; Oostendorp, 2009), with traits of developing economies, such as institutional weaknesses and insufficient gender equality frameworks, playing a significant role (Chen, Torsin, & Tsang, 2022). Most ASEAN member states, apart from Singapore, exhibit these characteristics, which are associated with wider gender pay gaps compared to developed nations (Endo & Ikeda, 2022). The interaction term (Ln(FDI) * Paid Maternity Leave Policy) is negative (-4345) and marginally significant (0.05 < p < 0.1), suggesting that paid maternity leave may have a slight mitigating effect on the gender pay gap exacerbated by FDI inflows, partially supporting hypothesis 2 (H2). Previous studies (e.g., Herr, Roy, & Klerman, 2020) explain that maternity leave policies can weaken the relationship between FDI inflows and the gender wage gap by reducing career interruptions, improving women’s labor 22 market attachment, and encouraging shared parental responsibilities. These measures address wage disparities in FDI-intensive sectors. Additionally, government-funded maternity leave reduces employer costs, fostering workforce stability and enabling women to participate more equitably in FDI-driven industries (Amin & Islam, 2022). However, the positive direct effect of paid maternity leave in ASEAN (17.81), though insignificant, indicates that on its own, the policy can widen the gender pay gap by increasing employer costs. This, in turn, may lead to lower wages or fewer opportunities for women in higher-paying roles (Amin & Islam, 2022). This effect is further exacerbated in labor-intensive, FDI-driven industries and by societal norms that reinforce caregiving roles, limiting women’s career advancement without complementary policies such as shared parental leave or anti-discrimination measures (Vahter & Masso, 2019; Del Rey, Racionero, & Silvia, 2021). The adjusted R² for the ASEAN region increases slightly in this model, from the base model to 0.109. This small improvement indicates that the interaction term adds explanatory value in the ASEAN context. The overall difference between ASEAN and Asian developing countries in the interaction model highlights the interplay between sectoral composition, policy effectiveness, and cultural norms. While FDI remains a significant driver of gender wage disparities in ASEAN, broader Asian developing countries exhibit a weakened and less significant relationship, likely due to more effective policy moderation and a less concentrated impact of FDI across sectors. Table 4.