“Unfortunately, violence against women is not the only injustice women face globally; it is one of the many inequalities that impede the full development of socially excluded women globally.” ~ Zainab Salbi, Iraqi-American activist and Founder of “Women for Women International”
“Rape is one of the most terrible crimes on earth and it happens every few minutes. The problem with groups who deal with rape is that they try to educate women about how to defend themselves. What really needs to be done is teaching men not to rape. Go to the source and start there.”
― Kurt Cobain, Songwriter and Musician
Conflict Countries’ Rate of Sexual Violence: A/B Testing
Since the Vietnam War, we see how conflicts take many forms and involve different actors that develop within the country, or organize as rebel forces that gain support from other countries–unwittingly, or not. In 2010, Syria was an authoritarian regime with a population of 20.4 million split between urban and rural areas. Since its 2011 political uprisings turned proxy war, Syria’s population has fallen to an estimated 17.3 million because the conflict turned violent, and thereby causing over 5 million to seek refugee. [Source: https://www.worldometers.info/world-population/syria-population/] One could argue that Syria’s conflict has exploded to the extent that warrants the most investigation into war crimes, like assassinations, chemical weapons use, and state implemented sexual violence compared to other similar conflicts. Aside from the Iraq War in 2004, the Syrian conflict represents the worst violence–not just in the region–but globally since the Balkan Crisis of the 1990’s. The complete A/B testing and visualization conducted by PITAPOLICY and Pitaconsumer may be found here.
A/B Testing of Conflict Country Cases
There will be a statistical difference between ‘Group A’, Syria, and ‘Group B’, world, in rates of sexual violence against women during conflict.
There will be no difference in Test A (Syria case study) and Test B (rest of world) in rates of Form or prevalence of sexual violence against women during conflict. This means that the means of both ‘Group A’ and ‘Group B’ will be equal.
For example, the Balkans Crisis resulted from in the dissolution of Yugoslavia where simultaneous independence struggles among six provinces sought their own states. However, this struggle devolved into certain states implementing ethnic cleansing measures to influence which countries succeeded in asserting independence based on population numbers. As a result, certain states carried out human rights abuses and sexual violence crimes, which skyrocketed into war crimes to the point of warranting global intervention by the North Atlantic Treaty Organization (NATO). Even NATO intervention did not curtail the abuses as other countries, like Russia, intervened by supplying military assistance.
Between 1992 to 1995, the conflict erupted into outright genocide resulting in the death of over 200,000 people, 2.3 million fleeing their homes–the largest modern refugee crisis since World War II–and a record number of sexually based offenses, according to the international nonprofit the Borgen Project and the United Nations High Commission for Refugees. [Sources: https://borgenproject.org/10-facts-about-the-bosnian-genocide/ and https://www.unhcr.org/3ae6a0c58.pdf]
Since the dissolution of Yugoslavia, several conflicts erupted globally. In particular, the Syrian conflict has also morphed into Refugee Crisis that mirrors the internal and external dynamics witnessed after Yugoslavia’s dissolution morphed into the Balkans Crisis followed by the Bosnian Genocide. The Syrian situation repeats the 1990s catastrophe to that the extent that the human rights abuses, proxy actors, food shortages, and sexually based offenses carried out by state and non-state actors has produced 5.27 million Syrian refugees. [Source: https://data2.unhcr.org/en/situations/syria] This is more than double the Balkans conflict. Thus, we will test if the Syrian Crisis outpaces other conflict countries in the same time period of 2011 to 2015 regarding the sexual-violence crimes because this is noted as a warcrime, not just a casualty of war.
For a more detailed discussion on the sexual violence statistics in non-conflict countries in neighboring Arab countries, please visit the United Nations report.
Sample during a set of active conflict years. We pulled population data from the UN data site. We selected the year 2013 because this year is the average of years 2011 and 2015.
For a complete discussion of PITAPOLICY’s methodology for the A/B Testing experiment behind the study of Syria’s rate of sexual violence between 2011 to 2015, please refer to the data science study and its detailed analysis here.
The target variable for this study is “Sexual Violence Index”, which was comprised of combining both ‘form’ and ‘type’ of sexual violence, followed by counting the occurrences to determine pattern.
**Note: Following the definition used by the International Criminal Court (ICC), we use a definition of crimes of sexual violence which includes (1) rape,(2) sexual slavery,(3) forced prostitution,(4) forced pregnancy, and (5) forced sterilization/abortion. Following Elisabeth Wood (2009), we also include (6) sexual mutilation, and (7) sexual torture. This definition does not exclude the existence of female perpetrators and male victims. According to the SVAC Dataset
We focus on behaviors that involve direct force and/or physical violence. We exclude acts that do not go beyond verbal sexual harassment and abuse, including sexualized insults or verbal humiliation.~Sexual Violence in Armed Conflict Data Project
**Note: Variables to Consider Countries inhabiting more of this type of actor “Rebel” will have higher rates of Form or prevalence of sexual violence against women during government and territorial conflict. Create a new index: a column that combines both columns (both violence and type). What other possible weak points? Proportion of cases larger than the average. Count of the number of cases. Proportion of cases/population. Focus across 3 years.
The study’s key variables to construct Sexual Violence Index are as follows:
- Number incident reports by Human Rights Watch
- Number incident reports by U.S. State Department
- Number incident reports by Amnesty International
- Form of Sex Crime
- Number of Sex Crimes
For more detail about how featured variables were cleaned, wrangled, and recoded, please contact PITAPOLICY at: qayyum at pitapolicyconsulting.com.
Variable: Actor Type
There are six actor types in conflict: types 1 , 2, 3, 4, 5, and 6.
- Type 1: State or incumbent government (in UCDP dyadic, this actor type is called ‘Side A’).
- Type 2: State A2 (in UCDP dyadic, this actor type is called ‘Side A2nd’). These are states supporting the state (1) involved with conflict on its territory.
- Type 3: Rebel (in UCDP dyadic, the actor type is called ‘Side B’).
- Type 4: State supporting ‘Side B’ in other country.
- Type 5: Second state in interstate conflict (in UCDP dyadic, this actor is called ‘Side B’).
- Type 6: Pro-government militias (PGMs).
Although ‘3’, or ‘Rebel’ was the most prevalent in 3,020 of the conflict country cases, ‘actor_type’ ‘6’ was the second most prevalent and defined as “Pro-government militias” (PGMs) in 2,335 country cases.
Variable: Conflict Type
Type UCDP/PRIO is a nominal variable with three categories:
- Type 2: Interstate Conflict.
- Type3: Intrastate Conflict.
- Type 4: Internationalized Internal Armed Conflict.
Types of conflict range between government and territorial.
Based on the literature review conducted by Mehrunisa Qayyum of PITAPOLICY, this study characterized Syria as both types of conflict: Government & Territory. Similarly, Yemen is both types of conflict: Government & Territory
In our ‘Sexual Violence by Actor’ plot we see that there are six different ‘actor types’ committing different types of sexual violence, measured by ‘form’. “Type 1″ represents ‘State or incumbent government (in UCDP dyadic, this actor type is called ‘Side A” while “Type 3” represents Rebel (in UCDP dyadic, the actor type is called ‘Side B’) “Type 2” represent states supporting the state “Type 1” involved with conflict on its territory, and highlighted as an occurrence after year 2006 and increasingly between 2011 to 2015. Syria represents this situation regarding Russia and Iran.
So we need to parse out Syria subset, or ‘Group A’, aside from the global sample between years 2011-2015. Source: https://www.dunderdata.com/blog/selecting-subsets-of-data-in-pandas-part-1
Summary Statistics of Group A: Syria
Here we see the breakdown of summary statistics for our experiment on Syria, which is ‘Group A’. There are 28 cases between years 2011 and 2015. Unlike the U.S. Three huge entities specifically report rape or other sexual violence related to the conflict. Although the State Department shows in column ‘state_prev’ 18 reports mentioning sexual violent occurrences, both Amnesty International(‘ai_prev’) and Human Rights Watch (‘hrw_prev’) provide a higher, and the same number of reports mentioning sexual violence in Syria conflict: 23.
We need the mean in ‘sex_violence’, our independent variable of interest, for both groups. We used the numpy mean built-in method to calculate; shared immediately below: ‘Group A’: 1.6785714285714286 ‘Group B’: -16.710108073744436
Summary Statistics of Group B: World
‘Group B’ in this experiment represents all countries experiencing conflict between the same time period (our control of years 2011 through 2015) except Syria since Syria is our ‘Group A’. ‘Group B’ includes 1,573 cases for comparison.
Between both ‘Group A’, Syria, and ‘Group B’, world, we will run a t-test to review our hypothesis. We will run an ‘Independent Samples t-test’, which compares the means for two groups: ‘Group A’ (Syria) and ‘Group B’ (world). Our T-test considers that both samples have different means, variance, and sample sizes.
t-test sex_violence Ttest_indResult(statistic=1.1516107098127624, pvalue=0.24966745563067755)
fromscipyimport stats #import researchpy as rp #~ We are using 2 dataframes, Syria and world, so use researchpy. x= df_Syria[‘sex_violence’] y= df_world[‘sex_violence’] ttest=stats.ttest_ind(x,y) print (‘t-test sex_violence’, ttest)
We calculate a t-statistic of ‘1.1516’ with a p-value of ‘0.249’.
After conducting a t-test between ‘Group A’ and ‘Group B’, we calculated a t-statistic of ‘1.1516’. We can reject the null hypothesis of equal means between ‘Group A’ and ‘Group B’ regarding their ‘sex_violence’ level at the 0.25 level. (The p-value represents the probability of getting the data above if the null hypothesis were true in the population.) Specifically: Syria’s average rate of sexual violence compared to the conflict countries’ average rate of sexual violence are not the same. Although the t-statistic is not very low, it is important to note that our sample size for Syria (‘Group A’) is 25, which is much less than the sample size of all the other conflict countries (‘Group B’), which is 1451 country cases.
In conclusion, Syria’s sexual violence index represents a different combination of documented watchdog reports and a higher pattern and form of sexual violence types as represented by our constructed variable ‘crime_counted_another_way’, which was derived from the ‘form’ type and ‘form’ counts, or occurrences of sexual violence. As a result, the Syrian crisis could represent the worst form of violence against women in conflict since the Balkans crisis and the resulting Bosnian Genocide.
Other areas for review include examining the ‘actor_type’ carrying out the sexual violence ‘sex_violence’ as well as conducting an A/A test between Syria and Yemen. We propopse conducting the latter t-test because Syria and Yemen represent two Middle Eastern countries experiencing similar state and non-state actor conflict that are continuing until today and share similar sample sizes. The rates of violence may differ depending on whether there are more types of actors involved. Is there a statistical difference between these two countries with more actors than less actors.
For further study of conflict countries: we will compare/contrast the case study Yemen, which submerged into conflict at the same time as Syria and also resides in the Eastern Mediterranean region.