Analyzing Sexual Harassment in Egypt

HM

Harassment is a problem in many parts of the world, Egypt is not an exception. The problem came to the fore after a series of highly publicized incidents in Egypt a few years ago where not just individual people but groups of people were involved in such behaviors. Given the prevalent nature of this problem in the Egyptian society as well as in many other parts of the world the HarassMap project was started in Egypt in 2010 by Rebecca Chiao, Engy Ghozlan, Amel Fahmy and Sawsan Gad. The main idea behind the HarassMap project is to spread awareness about harassment in Egypt and make it socially unacceptable. Recently I have been working with the HarassMap project and had a chance to analyze their data. Hopefully this will help spread awareness about the problem of sexual harassment and also bring to fore how project like HarassMap can help us reduce sexual harassment.

HarassType
Figure 1: Types of Sexual Harassment vs. Number of Incidents

The dataset spans from November 2010 to February 2016. To start off consider the types of harassment reports which are filed by the victims (Figure 1). What is surprising is the top offense is inappropriate or unwanted touching. There were a total of 609 such incidences which corresponds to one indent every 3 three over the span of the dataset. However this should be prefaced with the fact that the incidents are most likely under-reported. The under-reporting issue is discussed in detail below. To give more context about the comments, consider the following examples from actual reports.

  • A group of people hunted and sexually assault a girl who suffered previously from the same incident
  • All of a sudden a car stops in the street and the driver begins to show me one of his body parts that I should NOT see and keeps indicating to me in case I like what I see.
  • A car passed me and the guy in the drivers seat leered out the window and verbally harassed me.
  • Someone followed me on a bike, then cut me off, hindering me from crossing the street. He made lewd remarks, and told me that he loved me, and wanted to do stuff (in details).
  • I was coming home on the metro in Cairo and at Sadat metro, lots of men were getting on the women’s metro cars. They were all ogling and acting in a sexually intimidating way to the women around them (all the women were looking away).
  • Yesterday near the st. leading 2 the museum, saw 2 girls being dragged away from a mob of men.

Now lets consider (Figure 2) the time of the day at which harassment incidents happen. There is a noticeable spike in incidents from 1pm to 2 pm. This could be correlated with the time that people take breaks from work for lunch or other reasons. The only times where the number of incidents are low are from 2 am to 10 am. This implies that women are subjected to harassment most of their waking times which reflects a sad state of affairs.

HarassHours
Figure 2: Number of incidents by Hour of the Day

Now consider the incidents in terms of days of the week (Figure 3). The numbers are really not far off from one another except in the case of Sunday which looks like an outlier. Thus could also be because more people (men and women) are out on Sundays. However this explanation does not explain why Saturday is more like the rest of the days of the week when in fact one would expect it to be similar to Sunday is that explanation was correct.

HarassDaysOfWeek

Figure 3: Number of incidents by Days of the Week

One can also look at this phenomenon by month. Here we see something peculiar, the number of reported incident are comparatively low in September and October. One could speculate that perhaps it is because Eid al-Adha (Islamic Festival) and Muharram (Islamic New Year) fell in these months in the years beings considered so perhaps there is a religious reason for this. However this is unlikely to be true since one does not see a corresponding fall in June and July which roughly correspond to the holy month of Ramadan. In fact people have noted the time around Eid al-Fitr is characterized by a spike in harassment incidents. One thing that can be ascertained by the individual incident reports is that women are less likely to be harassed if they are traveling in a group with at least one man in that group. Thus there are even incidents where a group of women were harassed by other men. One explanation for the drop in incidents could be that the time-span roughly corresponds to the start of the school year and young women are more likely to be traveling in mixed-gender groups.

HarassMonths
Figure 4: Number incidents by Month of the Year

Now it would also make sense to look at the monthly trends over the course of the span of the dataset (Figure 5) to see how different years stack up against one another. One thing that the over trends from 2012 and 2013 are not too dissimilar and that is the case for 2014 and 2015. The first three data points (November 2010-January 2011) correspond to the launch of HarssMap where initially a large number of women did report incidents because of the media exposure. So while the number of incidents being reported has changed the overall trend remains the same. Later on I will discuss how this observation can be used to get a rough estimate of ‘missing’ incidents.

HarassYearMonths
Figure 5: Number of incidents by months over the course of 5 years

One thing that is clear from Figure 5 is that the number of reported incidents has dramatically declined over the course of time. The picture becomes clearer from Figure 6 where we consider the total number of incidents reported per year. It is evident that especially since 2014 the number of incidents that are reported have declined dramatically. This could either be because of decrease of harassment incidents or decrease in incidents being reported. Anecdotal evidence suggests that the later explanation is more likely to be true even though the incidents of sexual harassment may have declined but not as much as what this data suggests.

HarassYears
Figure 6: Number of incidents by months over the course of 6 years (2010 data is partial)

Thus one major issue with the data is the under-reporting of incidences. The problem of estimating incidents of violence against women in conflicts and even in peacetime has been considered by numerous researchers. One very rough way to estimate the number of under counts would be to take the average of incidents from 2011-2013 time period and extrapolate from there. This would imply that the number of incidents are being under-reported by 40-60%. However let us also notice the surge of reports when HarassMap was launched. The spike in the data that we see in the beginning of the dataset is not because of a sudden increase of incidents at the beginning of the experiment or a subsequent decease after three months but rather it is most likely because of decrease in incidents being reports after the initial media coverage. If we consider the first three months of reported incidents as being more representative then we come to the conclusion that even in the 2011-2013 only a quarter of such incidents are being reported. That said even this number is not representative. Firstly as we have not even addressed the issue of media penetration i.e., what percentage of the population is even aware of the availability of a service like HarassMap. Secondly, filing of such reports assumes a certain level of technological competency with computers which is not prevalent in certain segments of society in Egypt currently. These factors taken together would imply that the number of incidents which are being reported are a small fraction of incidents of harassment in Egypt which makes this project even more pertinent.

Now let us consider the spatial aspects of the data. While the dataset is supposed to cover all of Egypt, the incident reports are disproportionately from Cairo and Alexandria (Figure 8-11) . This makes sense given that the main population centers are in these cities as well as the fact that these cities are also the most technologically literate. The following series of Figure shows the locations of incident reports at different levels of granularity. The last map (Figure 12) shows the prevalence of such incidents at Tahrir Square which is place known to be a locus of harassment incidents from time to time.

HarassMapEgypt
Figure 8: HarassMap of locations throughout Egypt
HarassMapCairoEgypt
Figure 9: HarassMap of locations in Lower Egypt (Cairo and Alexandria are prominent)
HarassMapCairo
Figure 10: HarassMap of locations in Cairo and Neighboring Cities
HMCairo
Figure 11: HarassMap of locations in Downtown Cairo and Neighboring Areas HMTahrir
Figure 12: HarassMap of locations in the famous Tahrir Square

Now consider the word cloud for the incident reports in English. Certain words like follow, touch, grab, stop etc stand out which is reflective of the types of incidents which are prevalent in the incident reports. One can also use the word cloud to look back in the data for certain types of incidents. Thus consider taxis, one disturbing phenomenon that one

HM-wc
Figure 13: Word cloud of incident reports.

Another phenomenon that one observes is the clustering of reports e.g., in the days leading up to the Egyptian Elections in November 2011 there was a series of harassment incident reports filed in and around Tahrir square. This is around the time of demonstrations against the military junta and large number of people had gathered at Tahrir square. This gives the perpetrators opportunity to remain anonymous in the crowds.

Given the observations in this blog post what can be done to reduce such incidents and spread awareness? In terms of recommendation, greater public and media awareness of services like HarassMap is necessary. Greater reporting of such incidents will also mean that we have a better idea of the magnitude of this problem. This should be coupled with the usage and availability of services like Facebook Live where perpetrators of such crimes can be recorded so that it will make it easier for authorities to apprehend them.