國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
43
Chapter Four: Statistical and Qualitative Analyses
4.1 Empirical Results
In Table 4 we can observe the empirical models’ results which predict women’s representation in the lower chambers in the 18 Latin American countries. For the Models 1 and 2, I use OLS regressions with robust standard errors grouped by country. In the case of Models 3 and, I employ the TSCS analyses along with panel-corrected standard errors.
In the Models 1 and 3, as expected, the egalitarianism of political institutions’
coefficient obtains statistical significance. In addition, the results in Models 1 and 3 show that the gender quota law’s coefficient also has a positive effect on the level of women’s representation in lower chambers. More importantly, the full models’ results, Models 2 and 4, provide notable support for this research’s hypothesis. The coefficient is
statistically significant and positive for the interaction term of egalitarianism of political institutions * gender quota laws. This indicates that, in a country that adopts a gender quota law, the level of women's political representation tends to be much higher when the political institutions of this country are more egalitarian. Each of Model 2 and Model 4 explains 47.8% of the variance in women’s political representation.
Among the control variables, the results show that the type of party list in the PR electoral systems are relevant for female political representation. The results show that when a country adopts a closed party-list PR, this country is more likely to hold a higher level of female political representation. In the case of the degree of religiosity variable, it exhibits mixed results for different models. Preview research argue that countries where citizens are particularly religious, tend to be conservative and less likely to endorse women’s representation in politics (Burns, Schlozman, and Verba 2001; Inglehart and Norris 2003; Morgan and Buice 2013; Mcculloch 2012; Plutzer 1991). However, the
‧
variable of the degree of religiosity attains statistical significance in Models 1, 3 and 4, in contrast, it does not attain statistical significance in Model 2.Table 4. Effects of Egalitarianism of Political Institutions and Gender Quota Laws on Women’s Representation in Lower Chambers in Latin America, 1984-2018.
Explanatory Variables Model 1
(Ordinary Least
Institutions 11.67184** 2.465 11.672*** 2.465
(5.152) (5.226) (3.862) (3.348)
Gender Quota Laws 8.0331116*** -0.203 8.033*** -0.203
(1.467) (3.605) (1.546) (3.738)
Egalitarianism of Political Institutions X Gender Quota Law
15.436** 15.436***
(6.495) (5.751)
Control Variables
Type of Party List 8.124*** 7.997*** 8.124*** 7.997***
(1.795) (1.772) (0.895) (0.870)
Lag of Average District Magnitude -0.549 -0.747 -0.549 -0.747
(0.776) (0.786) (0.692) (0.745)
Lag GDP per Capita (current US$) 3.586*** 3.512*** 3.586*** 3.512***
(1.229) (1.232) (1.288) (1.261)
Degree of Religiosity 4.367* 3.653 4.367*** 3.653***
(2.270) (2.270) (1.034) (1.187)
Lag Polity Score -0.123 -0.022 -0.123 -0.022
(0.184) (0.179 (0.259) (0.262)
President's Ideology 4.63** 4.373* 4.630*** 4.373***
(1.785) (1.790) (1.586) (1.538)
Constant -36.442*** -30.043** -36.442*** -30.043***
(13.128) (13.431) (9.991) (11.153)
Observations (N) 146 146 146 146
R-Squared 0.463 0.478 0.463 0.478
Adjusted R-Squared 0.432 0.444
F-Test 16.77 15.73
Wald Test 328.46 375.11
Notes: standard errors are in parentheses.
*p<0.1; **p< 0.05; ***p< 0.01.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
45 The coefficient of the president’s ideology attains statistical significance in all models.
This finding supports the argument of previous research that presidents from leftist parties tend to appoint more women in the legislature (Escobar‐Lemmon and Taylor‐
Robinson 2005). In contrast, the district magnitude bears little relation to women’s political representation in my sample, suggesting that, even though district magnitudes are relevant due to their leading role in party strategy when choosing candidates (Matland 1995), it might not be an important explanatory factor for analyzing women’s
representation at the country-elections level. In all four models, the effect of the GDP per capita is positive and statistically significant, which is similar to the findings of previous studies (Duflo 2012; Matland 1998; Stockemer 2015). This finding suggests that
economic development has a positive impact on the percentage of female representatives in the lower chamber. Last, the results show that level of democracy does not appear to be a significant predictor of women’s political representation in my sample.
Figure 4: Predicted Percentage of Women Legislators in Latin America.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
46 The hypothesis of this research is supported by the empirical results showed in the previous lines. I demonstrate that the variation of women’s representation in national parliaments in Latin America is driven by the combined effects of the egalitarianism of political institutions and the presence of a gender quota law.
Figure 4 presents the predicted percentage of women legislators in Latin America for countries that have a gender quota law and those that do not at different levels of
institutional egalitarianism based on Model 2. As we can observe, the value for both types of countries increments as the egalitarianism of the political institutions increments, which clearly suggests that countries with gender quota laws and a higher degree of egalitarian institutions are more likely to lead to greater participation of women in the national parliament. Figure 5 shows a difference test in women’s political representation between both types of countries at distinct levels of institutional egalitarianism. It
presents that the difference is not statistically significant when the motivation for encouraging egalitarian institutions is low. However, this difference changes into statistically significant when a threshold of egalitarianism of political institutions is exceled, which is roughly a value of 3.
Figure 5: Difference in Predicted Percentage of Women’s legislators.
‧
Table 5: Additional OLS Models.Explanatory Variables Model 5
(OLS)
Model 6 (OLS)
Model 7 (OLS)
Egalitarianism of Political Institutions 1.922 2.932 6.286
(6.222) (6.823) (6.477)
Gender Quota Laws -1.622 -3.013 0.779
(3.623) (3.623) (3.849)
Egalitarianism of Political Institutions X Gender Quota Law
Lag of Average District Magnitude 0.039 0.565 -0.613
(0.790) (1.143) (0.922)
Human Development Index 83.722***
(15.975)
Lag GDP per Capita (current US$) -0.130 3.506***
(1.806) (1.326)
Degree of Religiosity 8.677*** 5.725* 4.371
(2.647) (3.403) (2.715)
Lag Polity Score -0.159 -0.070 0.013
(0.332) (0.369) (0.313)
President's Ideology 3.700** 3.743* 4.274**
(1.755) (2.021) (1.832)
Women’s Secondary Education (% net) 0.321***
(0.094)
Female Labor Force Participation (% of total
labor force) 0.147
Notes: standard errors are in parentheses.
*p<0.1; **p< 0.05; ***p< 0.01.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
48 In Table 5, I provide the results of additional OLS models that takes into account variables that are considered important explanatory factors by previous women’s political representation literature. The reason that I do not include these variables in the four models in Table 3 is that the number of observations will be reduced because of the missing data for these variables. In Model 5, I replace the lag GDP per capital with the Human Development Index (HDI). We can observe that the level of HDI is still positive and statistically significant, suggesting that the higher the overall education level and societal development of the citizens of a country, the more women tend to be elected for the legislature. Moreover, in Model 6, I include the percentage of women enrolled in secondary education as an additional control variable. The coefficient of women’s secondary education has a positive and significant effect on women’s representation. In Model 7, I include the percentage of female labor force participation as an additional control variable. However, the female labor force participation variable does not seem to be a significant predictor of women’s representation in the sample of this research.
Overall, the coefficient for the interaction term of gender quota laws and egalitarianism of political institutions is statistically significant across these additional models.
4.1.1 Robustness Check
In Table 6, I conduct two tests for checking the robustness of my main finding. To ensure that the result of the interaction term is not driven by particular control variables, I estimate a trimmed model (Model 8) that drops all control variables which do not fall within a p<0.1 level in Model 2. The finding shows that the coefficient for the interaction term is still statistically significant, and the coefficients for closed-list system, GDP per capita, and president’s ideology reach statistical significance and have expected signs. In addition, to make sure that the empirical finding of this research is not driven by a particular way of operationalization of the independent variables, I use an alternative measurement for the gender quota laws to carry out a robustness test replicating the analysis. In Model 9, instead of using a dummy variable, I use the actual percentage of seats required by the gender quota laws. As can be seen, the re-estimated results in Model 9 are consistent with those reported earlier.
‧
Table 6: OLS Models for Robustness Check.Explanatory Variables Model 8
(Trimmed) Model 9
(OLS)
Egalitarianism of Political Institutions -3.563 -8.469
(3.665) (9.059)
Gender Quota Laws 0.206
(3.722)
Actual Gender Quota (%) 5.452
(12.332) Egalitarianism of Political Institutions X Gender Quota
Law
16.380** 34.994*
(6.572) (20.596)
Control Variables
Closed-Party List System 7.385*** 7.076***
(1.081) (2.062)
Lag of Average District Magnitude 0.730
(1.067)
President's Ideology 4.871*** 4.850**
(1.726) (1.995)
Women's Secondary Education (% net) 0.410***
(0.124)
Female Labor Force Participation (% of total labor
force) -0.745**
(0.347)
Constant -14.514* 32.108
(7.907) (24.324)
Observations (N) 146 96
R-Squared 0.460 0.524
F-Test 18.57 11.67
Notes: standard errors are in parentheses.
*p<0.1; **p< 0.05; ***p< 0.01.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
50