Ingmar Weber
Although the United Nations Charter guarantees gender equality for all member states, today gender disparities remain a global issue and addressing them is a top priority for international organisations. Reducing and ultimately erasing the gender gap is both a moral obligation but also a crucial ingredient for economic development . By limiting women’s access to education and economic opportunities an immeasurable amount of human resources are lost and a significant part of the population is not able to develop its full potential.
As the digital age transforms the way we go about our daily lives, we can observe such gender disparities transfer into the online world. But are online gender disparities a full reflection of inequality “on the ground”?
Measuring gender inequality
To quantify gender inequality around the globe and to track changes over time, for example in response to policies put in place, the World Economic Forum annually publishes the “Global Gender Gap Report”.
This report ranks countries according to a numerical gender gap score. These scores can be interpreted as the percentage of the inequality between women and men that has been closed. In 2013 the leading country Iceland had an aggregate score of 0.87, whereas Yemen scored lowest with 0.51 (An interactive map visualising the data for 2013 can be found here )
The emphasis of the report is on the relative gender difference for the variables considered, rather than the absolute level achieved by women. So a country where both men and women have extremely limited, but equal access to education and economic opportunities would be ranked higher than a country where both men and women fare better in absolute terms, but where men do disproportionately better.
Its scores are based on “hard data” such as the ratio of female-to-male earned income and the ratio of women-to-men in terms of years in executive office . Indicators related to country-specific policies, such as culture or customs, do not contribute. The final index computed is an aggregation of four categories “Economic Participation and Opportunity”, “Educational Attainment”, “Health and Survival”, and “Political Empowerment”.
In our own research, done by Gabriel Magno and myself at Qatar Computing Research Institute , we augment the Global Gender Gap Report by incorporating a new dimension: activity on online social networks. In other words, we are interested in questions such as whether countries where men have a higher online “status” than women are also countries with large gender inequalities. Knowing the answer could, for example, help develop real-time, global indicators of reduction in gender gaps.
For our study, described in full length here , we gathered public data for tens of millions of Google+ and Twitter users. Whereas in Google+ users explicitly stated their gender and their country, we had to infer both of these for Twitter users. The most likely gender was inferred using a first-name-to-gender dictionary. To infer the most likely country a user is in we used their self-stated location which was mapped to countries using a public geo-coding tool .
Our emphasis was on studying correlations between online and offline indicators of inequality. We did this both for the purpose of validation, to be sure that what we measured online is linked to “real world” phenomena, and for the purpose of devising new indicators, when an important online measure is not in good agreement with existing indicators.
We analysed data for 45 countries for which we have data for at least 1,000 female and 1,000 male users, both for Google+ and Twitter. Using this data we observed the expected behaviour of the user number gap: countries with relatively fewer female users of Google+ or Twitter were countries with more pronounced gender inequalities. In this visualisation, this can be seen by looking at the distribution of red (female Twitter users in data sample) vs blue (male Twitter users in data sample) in a pie chart and observing that in countries with smaller fractions of the gender gap closed, marked in lighter green, male Twitter users tend to outnumber female users.
We were also expecting to see the same pattern with respect to “status” of women online. In countries with large offline gender inequalities women would have a lower status online compared to men. However, we observed the opposite : in a country such as Pakistan with only 55 percent of its gender gap closed, women on Google+ enjoy on average more followers (25 vs 16) and a higher PageRank (22 percent higher) than their male counterparts. Similarly, Pakistani women on Twitter have on average more followers (600 vs 222) and are included in more Twitter lists (3.6 vs 1.3). Here Pakistan just serves as an illustrative example for a wider trend.
One possible explanation for this phenomenon may be selection bias: If a woman in a very unfavourable environment is still very active online then she is likely to be an overachiever and might be representing the so-called “Jackie Robinson Effect”. Jackie Robinson was a highly talented and enormously successful US baseball player who became the first African-American to play in Major League Baseball. If Jackie had been only good, rather than great, it is unlikely that he would have been given a chance to play rather than a slightly less talented white player. Similarly, one might imagine that women who are online in countries where women have more limited access to the internet compared to men must be extraordinary to begin with. This phenomenon has also been observed in politics, where women are often compelled to perform better than men becaue doing “just equally well” would not suffice to “make it” onto the political arena.
Combining just three Twitter indicators of gender inequality relating to (i) per-country differences in the number of lists women and men are included in, (ii) per-country differences in the fraction of their tweets with hashtags, and (iii) per-country differences in their fractions of retweeted, non-original content , we obtained an aggregate index that across 45 countries achieved a correlation of r=.73, typically interpreted as a “strong linear relationships”. The match between the predicted gender gap values and the actual ones can be explored in the figure below.
As more and more economic activity becomes digital and moves online , as more and more education happens online through MOOCs and other initiatives , and as more and more of political participation happens online, we are convinced that quantifying gender inequality also has to take into account online activity. Measuring global digital gender gaps can be a first step towards implementing policies to reduce the gap.–Al Jazeera