<span style="font-size: 1em">Methods</span>
Estimation Technique
In this study, a linear probability model (LPM) is employed because a logistics regression will not be appropriate due to the nature of the variables and data obtained for the study. That said, in the frst part of regression analysis, the study employs a binary-dependent regression to estimate the efect of violence of all forms on involvement in the political discourse in Sierra Leone. To start with, the study formulates a simple linear probability model of the form
Parameterizing the above model, it could be rewritten as (see Cameron and Trevedi 2005) as where F (·) is a specified function. To ensure that 0 ≤ p ≤ 1 it is natural to specify thus: In a linear form, Equation (2) is rewritten to include variables that would be used for the estimation by employing a maximum likelihood estimation technique. Here, the study assumes that the required assumptions of normality andI.D.D are satisfied. Hence, the estimable equation is then written as
Model three is a binary-dependent variable model. To this, the study estimates the probability of women participation in two mutually exclusive electoral contests: the presidential and parliamentary elections. This is because the two variables are captured in the binary form of 0 and 1 responses. In Equation (3), y represents women participation in presidential election on the one hand; and the second estimation from Equation (3) uses women parliamentary election partici pation (women MP candidates) as the dependent variables. EV represents a set of electoral violence (sexual violence, electoral violence in general, and/or armed confict), which is expected to have an adverse efect on women participation in all forms of the electoral process. Z captures other control variables such as men in active politics, school enrolment, labor forces among others.
Cointegration Analysis
There are several methods for cointegrating analysis to determine the long-run rela tionship between time-series variables. This includes the famous Søren Johansen (1991, 1995) cointegration methods, which deal with multivariate cointegration; there are several single equation models, which include autoregressive distributed lag model by Pesaran and Shin (1999), Stock and Watson’s (1993) dynamic ordinary least squares (OLS), and the canonical cointegration regression (CCR) by Park (1992). Due to the properties of the variables and the limited observation, the study employs the CCR by Park (1992). Hence, it is assumed here that there is one cointegrating equation, which is presented in Equation (4):
where D′ it is the deterministic trend regressor, u is the error and X is a vector of explanatory variables. Following Hanson (2002), it is assumed here that the innovations are strictly stationary and ergodic with zero mean. Park’s (1992) CCR is strictly associated with FMOLS. However, instead, the study employs stationary transformations of the data to obtain least squares estimates and to remove long run dependence between the cointegrating equation and stochastic regressor innovations.
Once that is done, the Hanse instability test is used to test the cointegrating relationship between the variables presented in the model. A confrmatory test using Engel–Granger cointegrating test was done to confrm the already existing results from the Hanson’s instability test.
Variables
1
2
3
4
5
Electoral Violence
0.309** (0.121)
0.329** (0.135)
0.410*** (0.150)
0.790* (0.398)
0.663*** (0.212)
Men active
–0.0570**
–0.0521**
–0.0479**
–0.0957
0.0743*
participatio n in politics
(0.0211)
(0.0236)
(0.0232)
(0.0672)
(0.0359)
Electoral sexual violence
–0.0128
–0.0541
–0.168
0
-
Party ideology: Base (Liberal)
Conservati ve
0.932*** (0.257)
0.539*** (0.137)
Communis t
0.900*** (0.237)
0.488*** (0.126)
Rigging
-0.638 (0.471)
-0.748*** (0.251)
Targeted conflict
0.366** (0.124)
0.262* (0.137)
0.293** (0.135)
0 (0)
Armed conflict
0.361** (0.168)
0.352* (0.177)
Law mandates non-Discrimina tion
0.570* (0.284)
0.575*** (0.151)
Constant
0.832*** (0.212)
0.723** (0.269)
0.903*** (0.311)
–0.296 (0.400)
Table 5.1 Regression result for women presidential and PM candidature and electoral violence
Notes: Robust standard errors in parentheses.
Results from the LPM Model
The results from the study suggest that the presence of election violence of all forms is inimical to the participation of women in politics. The results indicate that electoral violence has the tendency of reducing women participation across all the board. From models 1 to 4, the probability of having no women parliamentary candidate increases by 0.30, 0.33, 0.4, and 0.79, respectively. Except for model 4, all the results were signifcant at or below the 5% signifcant levels. Similarly, the fndings on sexual violence prove that in the period where that are no sexual violence, the probability of having women MP may increase. Armed confict (that does not discriminate) and targeted confict (targeting members on the basis of their political afliation and gender identity) deter women participation in active politics. In two of the models, target confict turns out signifcant at
5% and 10% levels with probability of 0.33 and 0.26 and 0.29. While armed conficts increase the chance of a reduction in women in active politics by 36 and 35%, respectively. These results lend support to the existing argument that confict, in all its forms, deters women from participating in the democratic process.
Existing studies confrm these analyses and results. They corroborate the works of Krook and Sanín (2016) and Shepherd (2014), which found that targeted violence against women in politics and widespread electoral violence to have injurious efects on the rates of women’s political participation. This may even have a future efect and infuence the young female’s decision about future participation in politics. This is in line with Shepherd’s (2014) fnding that 60% of women aged between 18 and 21 and 80% aged over 31 were less likely to run for political ofce and have witnessed how a woman prime minister was mistreated in partisan electoral politics. Thus, violence, as Krook and Sanín (2016) submit, is demoralizing and devastating. Electoral violence presents pol itics in an unattractive form to women to stand for re-election. This eventually makes even existing women politicians bow out of politics after a few months or years of participating.
Time-Series Analysis
That said, all the models passed the basic diagnostic tests that were performed; and, hence, presenting a robust result for interpretation.
Unit Root TestThe study examines the time-series properties of the variables used in the cointegration model. As a requirement, the variables must be I(1) variables for any possible cointegrating relationship to be established. Thus, the variables in the model must be frst diference stationary. The study then employs the DF–GLS unit root testing to test for stationarity properties of the variables. The results indicate that time-series variables are not stationary at levels. They, however, became stationary at frst diference. This indicates that the variables are integrated of order 1 (see Table 5.2).
Cointegration Results (Canonical Cointegration Regression ofWomen Political Participation and Electoral Violence)The results from the cointegrating tests suggest that there is a long-run rela tionship among the variables presented in the model. That is, in the long run, the relationship exists among the men’s engagement in politics, electoral violence, and women’s political engagement. The R-square indicates that about 75% of the variation in the involvement of women in politics is explained by the variables in the models. Serial correlation is not a problem as detected by the Durbin–Watson serial correlation test. From the results (in Table 5.3 above), the support men give to the political activities of women plays a role in improving women’s participation in politics. This was found to be statistically signifcant at less than 1 percentage point. As the support of men in politics increases by 1%, it also increases women’s engagement in politics by 0.67 percent point. This implies that there is a direct relationship between the gender-aware political practices and activities of men and women’s political participation. Where men-in-politics are aware of the need for a violence free, empowering political interaction, this, in turn, favors women; thus, cre ating the enabling environment for greater political participation among women.
Table 5.2 Result of unit root test
Variable
Tau statistics at level
Tau at first difference
Lmenactivepolitics
–0.0551
–7.1801***
Lmensupportpolitics
–1.0261
–2.1936**
lwoactivepartymembers
–02224
–3.5737***
RFM
–7.0295***
Notes: Critical values: –2.6569, –1.9544, and –1.6093, for 1%, 5%, and 10% signifcant levels respectively.
p<0.01, **
p<0.05, *
p<0.1.
Coefficient
Std. Error
t-Statistic
Prob.
0.141302
0.177537
0.795903
0.4446
LMENSUPP ORTPOLITI CS
0.689141
0.202649
3.400670
0.0068
LMENACTIV EPOLITICS
–0.245689
0.113718
–2.160504
0.0561
C
–13.79648
19.11472
–0.721772
0.4870
ARMEDCON FLICTYEAR
0.493204
0.141581
3.483536
0.0059
RIGGING
0.206652
0.110974
1.862166
0.0922
R-squared
0.752468
Mean dependent var
7.966221
Adjusted
0.628702
S.D. dependent var
0.301062
Standard error of regression
0.488119
Sum squared resid
2.382606
Durbin–Wats on stat 1.647676 Long-run variance 0.033586 Cointegration test
Engel–Granger
Engel–Granger Z-statistics
Value
Prob
Hansen instability test
-6.4909
0.0044
19.527
0.0409
0.537
0.2000
Table 5.3 Canonical cointegration regression of women political participation and electoral violence
Variable Coefcient Std. Error t-Statistic Prob.Note: Dependent variable: women active participation in politics.
On the contrary, the higher the number of men in active politics, the lesser the women participation in politics, in the long run. The result shows that a percentage increase in men in active politics will result in a decrease in women participation by 0.25%. What explains this fnding is the natural tendency for men to either compete with women for political positions or present themselves (their masculinity) as the best means through which to get things done in a volatile political environ ment where it is challenging to achieve electoral politics-induced policies without violence—which makes women shy away from active politics. This supports the assumption that politically active men will use strategies which discourages women and entrench their marginalized roles in politics. Through this, women are made to think that politics is a male occupation. This, in turn, feeds into the hegemonic masculine thinking that reinforces the Machiavellian conclusions that the quickest means to destroy a postcolonial African nation is to have women at the helm of political and security decision-making platforms (see Kumah-Abiwu 2017).
The study also found that election rigging had an adverse efect on women’s political participation rate. That election rigging is a disincentive to women political participation rate is reafrmed in the analyses. This is so because of the correlations between poll rigging, electoral violence, and women’s political participation (see Olufunke 2013). Therefore, the high incidence of electoral rigging may lead to post-electoral violence that afects women in disproportionate ways. It is expected, therefore, that the low incidences of alleged or actual election rigging will lead to increased participation of women in politics in Sierra Leone.
Interestingly, it was found that armed confict leads to an increase in the participation rate of women in politics. This result corroborates Hughes’ (2009) fnding based on case studies. Hughes’ study of women, war, and social change points to the fact that the structural and cultural devices may work together with political opportunities after an armed confict to favor women. That the lived experiences of war provide the agency for women to come to the realization that, to prevent a relapse to war is to prevent the men from monopolizing the institutions responsible for war-making, peacemaking, and peacebuilding, they (the women) will increase their political participation in the democratization process of the post-confict era. War is tragic, yes! Yet it provides women access to new political environments and entry points into politics. Thus, the relationship between armed conficts and women’s political participation calls for focusing on the gender-responsiveness of peace accords in the discourses of gender-responsive governance.
Politics in this country, especially in the 1970s and 1980s, was not for women. Joining the men on street rallies was not a simple decision. Because you can be a victim of sexual violence at any time. For me, whenever I go out for rallies. I will ensure that I go with my brothers. On two occasions I escaped being raped: in the 1978 and 1982 elections. I cannot say why men will rape women whenever there was an election. But what I can tell you is that, the women who were victims were never the same again. As the rapists, they were celebrated for the temerity for sexual violence. These rapists were protected by the politicians; they paid them monthly salaries for the protection they were ofering them, and for the service of eliminating their political opponents. [Asked if women were considered political opponents, Mabinty responded that] Women stood for something very diferent from what the men stood for. Even within the same political party, it was not uncommon for women to oppose the men on issues relating to social welfare. We wanted our children to be educated, healthy, and safe. They [the men] wanted power. They got the power they wanted. But in the end, this same power was not used to better the lives of the people of this country. What power did to those male politicians was to transform them into unrepentant rapists and thieves. [Asked what would women have done difer ently if they had this power, Mabinty responded, reading from the manifesto of the SLPP women’s wing zonal elections of 1978]: "we will ask for free edu cation for women and children;" "we will support our fellow women facing economic and social crises;" "we will call for violence-free elections;" "we will ensure that those wo violate the human rights of others are prosecuted." [Looking at me with tear-flled eyes, she continued]: You see, the majority, if not all, of the women of the SLPP and the APC were not in support of what they men were doing. We wanted an electoral system that all Sierra Leoneans, especially the women, can say was peaceful; we wanted to see a politics where the political diferences between the women were settled through non-violent means. Because by the end of the day, people will gain power and lose it, but the idea of power will remain. This idea is not just about violence. Power can also be about progress, togetherness, and peace. This is the type of power we wanted. We did not get it. We, the women, came together into groups within and outside of political parties. The goal was to look for a way to fnd this "new" kind of power that is built on progress, togetherness, and peace. The men, or at least most of them, knew this kind of power was not going to help promote the divisive politics and politics of greed they preferred. So they used acts of violence against women, discrimination against women, and poverty to distract us. All that their violence was aimed at was the politics of distraction. They wanted that space, we wanted to change it. So, our paths were meant to cross; and there can be only one winner: the male politicians or the people of Sierra Leone, whom the key women’s groups represented.