From the World Economic Forum (WEF), Report on the Gender Gap (2023) has been announced.
The "gender gap index" used in the report is a numerical representation of the gender gap in each country. It takes a value from 0 to 1, and the closer the score is to 1, the smaller the gender gap. The report covers 146 countries, but Japan ranks 125th, down 9 places from the previous year.
This time, I learned for the first time that there is such an index in the news. I took this opportunity to study the Gender Gap Index, visualize/summarize the index, and analyze whether the index is related to national indicators (GDP, male-to-female ratio, etc.).
Visualize gender gap index by region and country
First, I colored the map to show the (comprehensive) gender gap index (score) of the 146 target countries.
The range of scores is (0.405 - 0.912). Since both the average and median values are 0.711, there are no countries with extremely low index values (close to 0), and the average value is closer to 1 than 0, so does it mean that the gender gap is small overall?
Looking at the map, it seems that there are large and small values between regions, although it is not clear.
Therefore, here is a plot of each country's score broken down by region*. The blue vertical bar represents the average score for that region, and the country of interest is labeled (rank, country name).
*The territories classified here are divided into the groups defined in the "Gender Gap Report" above.
From this figure, we can see the following.
- Europe, to which many countries belong, has relatively high scores, with Iceland, Norway, and Finland in Northern Europe ranked 1st, 2nd, and 3rd.
- Afghanistan, at the bottom, has an extremely low score compared to other countries.
- On average, the scores for the Middle East/North Africa and South Asia are low.
- South Korea (105th) and China (107th), both of which are close to Japan, also have low scores.
There seems to be a considerable difference between regions and countries in the index.
introduced above The Gender Gap Index is calculated by scoring the gender gap between 0 and 1 for four subcategories (economy, education, health, and politics) and averaging these scores. In keeping with the report, the following notation is used:
List of subcategories
economic : Economic Participation and Opportunity
education : Educational Attainment
health :Health and Survival
politics : Political Empowerment
Shows the distribution of scores for the four subcategories, including the Gender Gap Index.
We can see that the distributions are very different between the categories. For education and health, most countries are distributed near 1, but for politics, the tail tails to the right around 0.1.
Regarding the target countries, it seems that the gender gap is small in education and health, and the gender gap is large in politics.
In order to see the difference between the categories of each country, I drew a "parallel plot" that connects the values of each country with a line. Lines are colored by region.
Looking at this graph It can be read that there is a large variation in scores for "politics," and in the end it seems that the scores for "politics" contribute greatly to the gender gap index.
Japan ranks 138th with a score of 0.057 for “politics”. Political score factors include the ratio of female members of parliament and whether or not there have been female prime ministers in the past.
What factors are associated with the Gender Gap Index?
So, what factors can account for the variation in the gender gap index across countries?
The report lists the following indicators as general indicators for each country.
General metrics
Gender ratio (number of women/number of men) : Population Sex Ratio (%, female/male)
Population growth rate (2021/2020) : Population growth rate (%)
GDP per capita : GDP per capita ($1000)
population : Total Population
These data are mainly obtained from the World Bank. In the analysis that follows, we will use common logarithmic values for "GDP per capita" and "population."
Let's examine the correlation between these four indicators and the Gender Gap Index (including subcategories). However, as for the latter, as mentioned above, some distributions are biased, so we will use the logit transformation*.
In other words, we will look at the correlation for the following combinations.
20 total combinations:
"Y (gender gap index, subcategory) 5" × "X (general index) 4"
*Some original values are 0 or 1, but when performing logit transformation, they are converted to values slightly larger than 0 and slightly smaller than 1.
In JMP, if you specify Y and X above in the Response Screening platform, you can view a report that summarizes statistics for all combinations.
The report below lists all the combinations of Y and X in descending order of R-square.
Looking at the R-squared, there are no combinations that are significantly larger. The largest combination of education and GDP per capita has an R-square of about 0.41 and a correlation coefficient of about 0.64.
Select the combination report you want to examine in more detail, and from the red triangle button to the left of "Response Screening" Bivariate Relationships for Selected Items Select to display the Bivariate Relationships report (a straight line fit to a scatterplot).
The higher the GDP per capita, the higher the score related to education can be read. However, it should be noted that there are countries with low GDP but high scores, and the regression line fails to explain their relationship.
I will pick up only the relationship between the gender gap index and general indicators. None of the relationships are very high when looking at the R-squared.
Considering that these indicators may be explained by multiple factors, we will conduct a multiple regression analysis using the gender gap index as the objective variable* and indicators (GDP, male-female ratio, etc.) as explanatory variables. At the same time, we will also apply and compare models that include region as a factor.
* Earlier, the gender gap index was logit transformed, but the distribution of this index does not change much even if it is left as it is, so for the sake of clarity, we will use the values as they are without transformation.
Here are the results of an analysis using JMP's Fit Model.
- The 4-parameter model on the bottom left is a report when 4 indicators are used as explanatory variables. When the significance level of the test is set to 5%, GDP, gender ratio, and population are significant.
- The 5-parameter model on the bottom right is a report when "region" is used as an explanatory variable for four indicators. "Region" is highly significant, and only GDP is significant here.
Including "regions" increases the degrees of freedom adjusted R-square from 0.258 to 0.410.
Further simplifying the 5-parameter model by removing insignificant factors yields the following 2-parameter model: In this case, the degree of freedom adjusted contribution ratio is 0.407, which is almost the same as the 5-parameter model.
In this model, "region" and "GDP per capita" explain about 44% of the gender gap index. There are other social factors and country-specific factors that can be considered as factors of the gender gap, so if the two factors account for about 44%, I think it can be reasonably explained. .
How to raise Japan's gender gap index
Finally, for the two-parameter model created this time, I expressed the predicted value of Japan's gender gap index as a "prediction profile".
The predicted score would be 0.73 (95% confidence interval 0.704 to 0.756). On the other hand, the actual Japanese score is 0.647.
If you belong to the East Asia-Pacific region and your per capita GDP (logarithm) is 1.61 Although the average score is 0.73 and the 95% confidence interval is (0.704-0.756), Japan's score (0.647) is lower than the lower confidence interval. So it's not very good. "The Doing More to Close the Gender Gap Please work on it."
As we have seen in the analyzes so far, increasing the percentage of women participating in politics seems to be an immediate issue in order to increase the gender gap index. The current ruling party, the Liberal Democratic Party, has set a goal of raising the percentage of female lawmakers to 30% in the next 10 years, but what will happen?
by Naohiro Masukawa (JMP Japan)
