In the last post, I compared the US with the EU-28, as if it was one homogeneous state. But is this true? How much do the values and beliefs on gender equity differ between the member states? To approach this question I did another analysis, this time using a boxplot to describe the distribution of the per country means in the EU-28.
What is a boxplot? A Boxplot is a graphical representation on how a variable is distributed over mutltiple observations (in this case countries). It shows the median (the observation whose variable value lies in the middle of the distribution) and the 25% of the observations that lie above and below this value as a box. From there it is calculated, which observations are near enough to the box to fit in the “whiskers” and which are outside of that range, the “outside values”.
From the Stata help entry on boxplots:
Basically, the size and shape of the box and the whiskers tell you, if the data is more homogenoeus or rather heterogeneous. Also you can see if there are outside values that could influence the overall mean strongly in one direction, despite representing just a few “extreme cases”.
1. “When jobs are scarce, men should have more right to a job than women”
Here we see, that the country means in 1993 range between 8% and 70%. The box describing the 50% of countries around the median almost spans from 30% to 50% agreement. Not a small range. The boxplot gets smaller in the other two waves and some more “extreme” outliers emerge. Still the EU-28 do not show a clear and universially shared stance on the statement.
2. “Both the husband and wife should contribute to household income”
The second statement initially shows a more homogenous distribution, than the first one. Still, the whiskers in 2001 range from about 60% to 95% percent. The developement over time isn’t as clear as with the first statement. Actually the size of the box grows in 2001 and then shrinks again. One possible explanation could be, that in 2001 Croatia, Greece and Luxembourg enter the dataset (Cyprus in 2010). While Croatia, Greece and Cyprus generally show agreement rates above the median, Luxembourg has the second lowest value in 2001. This could at least partially be responsible for the picture we see here. I actually did the same analysis while excluding Luxembourg and the boxplots became more similar to the EU-28 means from the last post.
What also becomes clear, is that the stance of the Netherlands on the issue differs strongly from the rest of the EU-28. In 1993 the distance from the Netherlands to the other countries is even greater than the span of the whiskers. This strong outlier position persists through all waves, alltough the agreement rate in the Netherlands rose from 29% in 1993 to 45% in 2010.
3. “Having a job is the best way for a woman to be an independent person”
Here the boxplot becomes significantly smaller after 1993 (but grows again in 2010). When the box (describing the 50% countries around the median) gets smaller, less “extreme” values are included in the whisker range and so in 2001 and 2010 several states are defined as outliers. In 2001 it seems, that the majority of the EU tends to agree on the issue, while there is a small number of countries who differ from that (here the advantage of using a boxplot becomes clear; if I had just compared the range from minimum to maximum values, this developement would have been invisible). In 2010 the outliers get less “extreme” but the variance in the boxplot rises again.
After combining all three results, I believe I can safely say, that the agreement rates in the EU-28 are getting more similar overall. Still I wouldn’t go as far as calling them homogeneous. Even in 2010 the range of the country means is rather large and unevenly distributed, alltough this is partly due to a small number of countries who differ from the median.