by Samantha McFarland
Although women account for half the
population of the world, they are underrepresented in the political sphere in almost
every country. But what exactly is keeping women from entering the political
realm? This blog posts analyzes the effect of the percent of enrollment in
higher education with the number of women in national parliament. If education
does have a big effect on the number of women in parliament, then this could be
a key way for developing states to develop policy that makes higher education—or
education in general—more accessible for women.
Why is it important for women to
have higher representation in politics? Hunt argues that the presence of women
in politics could lead to very beneficial things within a state. Social science
has proven women are more prone to compromise, collaboration and nurture.
Studies have also shown that higher numbers of women in parliaments enjoy lower
levels of corruption. There is also a correlation between women holding
political office and the overall economic competitiveness of a nation.[1]
A study done Anne Price focuses on
the considerable gains made in both higher education and politics for women. However,
she notes, considerably more attention has been paid to help women make gains
in education around the world than encouraging more elected women officials. Her
research shows that in Muslim countries, the rate of women in politics is directly
proportional to trends in women enrolled in higher education, and after women
in education began to increase, the number of women in national parliaments
went from 2% to 4%.[2]
This study seeks to find the
correlation between enrollment of both men and women in higher education and the
number of women in both houses of national parliaments. As seen in the
histogram, most countries average less than 20% women representation. The
Organization for the Economic Cooperation and Development—the 34 wealthiest
countries in the world—only average about 23%. The Nordic countries, the most
famous group for gender equity, average about 35%. [3] However, non-OECD countries
like Rwanda, Cuba and Argentina boast almost 50% female political
representation.
Table 1. Summary Statistics
Mean
|
Std. Dev.
|
Min.
|
Max.
| ||
Women in Parliament (%)
|
16.17
|
10.41
|
0.0
|
56.3
| |
Tertiary Education Enrollment, Total
|
24.99
|
20.07
|
0.7
|
80.6
| |
GDP (US$ billions)
|
225.26
|
872.29
|
1.0
|
7298.1
| |
Maternal Mortality Ratio (per 100,000 Live Births)
|
264.26
|
298.94
|
8.1
|
1140.1
| |
Women in labor force, ratio to men
|
0.71
|
0.22
|
0.2
|
1.1
| |
Observations
|
77
|
My primary variables are the
percentage of women in both the upper and lower houses of parliaments—the independent
variable—and the total enrollment of the population in tertiary education. In
order to try to isolate the education variable, I also control for several
other variables. First, is GDP.[4] Logically, wealthier countries
will have more infrastructure and policy in place, such as gender
quotas, paid family leave, and child care. It can be difficult to quantify the
cultural and religious issues that may be hindering women from entering
politics. A high percentage of women in the labor force[5] shows that cultural
barriers have been broken down, and women are possibly not the sole parent
raising children at home. There is some potential for slippage here, as the
variable does not control for part-time and full-time workers. Many women work
part-time jobs and are also homemakers. The last control variable is maternal mortality rates
per 100,000 live births. This is an attempt to measure a country’s healthcare
system, and see its effect on women in politics.[6]
One can see a “u” shape effect
between women in parliaments and enrollment in tertiary education in Figure 1.[7] At first, there is
a positive correlation between education and politics, but as the number of
women in education gets higher, the number of women in parliament dips down,
and then rises up again. Once we see the regression in Table 1, the numbers
fall apart. When only education and women in politics are regressed together,
there is statistical significance. For every 1 unit increase in tertiary
enrollment, the percentage of women in parliament goes up by .03 units. However,
in the last Module, when all the variables are accounted for, tertiary
education is no longer significant. I cannot reject the null hypothesis.
However, there is still information
to be gleaned from Table 1. Namely, that the most important and significant
variable here appears to be percentage women in the labor force in Module 4,
which accounts for 14% of the variation of women in politics, and remains
statistically significant, suggesting that these results are not simply chance. GDP remains insignificant, which
may initially seem surprising, but several poor to middle income countries
experience high levels of female representation, such as Rwanda and Argentina. Maternal
mortality is also insignificant, and perhaps for future research a better
variable for quality of healthcare could be found.
Table 1. Effect of Dependent Variables on Women in
Parliament
IV: % Women in Parliaments
|
(1)
|
(2)
|
(3)
|
(4)
|
(5)
|
Tertiary Education Enrollment, Total
|
0.0305
(0.63)
|
0.0061
(0.08)
|
|||
Ln(GDP)
|
-1.0851
|
-0.2513
|
|||
(-1.57)
|
(-0.32)
|
||||
Maternal Mortality Ratio (per 100,000 Live
Births)
|
0.0027
(0.88)
|
0.0000
(0.00)
|
|||
Women in labor force (ratio to men)
|
20.0729***
(4.13)
|
19.5167***
(3.64)
|
|||
Constant
|
15.9603***
|
20.0398***
|
15.8599***
|
1.9784
|
3.1093
|
(9.87)
|
(7.34)
|
(10.89)
|
(0.55)
|
(0.54)
|
|
Observations
|
93
|
77
|
93
|
77
|
77
|
R²
|
0.004
|
0.032
|
0.008
|
0.186
|
0.187
|
Note: OLS Estimates with t-stats in parentheses. *
p < 0.05, ** p < 0.01, *** p
< 0.001
Source: Quality of Governance Data
In conclusion, there is not enough
evidence to show correlation or causation between tertiary education and the
percentage of women in national parliaments. However, there is evidence of an
empirical relationship between women in the labor force and women elected to
national parliaments. Further research and possible policy recommendations
should instead focus on getting women into the labor force. Another potential area for future research is
gender quotas, which many national parliaments have instated. While it is
almost certainly useful to have a highly educated population, there appears to
be other barriers keeping women from entering into politics.
[1]
Hunt, S. (2007). Let Women Rule. Foreign Affairs, 86(3), 109-120.
[2]
Anne Price.
"Differential Support for Women in Higher Education and Politics
Cross-Nationally." Comparative
Sociology. 13, no. 3 (2014): 346-82.
[3]
OECD. "Women and Men in OEDC Countries: Political and Economic
Power." 2014. http://www.oecd.org/gender/38172505.pdf
[4]
GDP is logged in the final regression, as the numbers are so large this makes
it easier to understand.
[5]
Specifically the ratio of women to men in the labor force
[6]
I did test for multicollinearity, a chart is located in the appendix.
[7]
There is a potentially parabolic effect here. In the OLS graph (Table 1) I do
not include this, however. As will be shown and discussed, there is no actual
statistical significance between the two primary variables. Therefore, there is
simply no effect.
Appendix
Variable
|
VIF
|
1/VIF
|
Tertiary enrollment
|
2.01
|
0.496918
|
GDP
|
1.88
|
0.532517
|
Labor force
|
1.44
|
0.692730
|
Maternal mortality
|
1.17
|
0.853382
|
Mean VIF
|
1.63
|
Hunt, S. (2007). Let Women Rule. Foreign Affairs, 86(3), 109-120.
OECD. "Women and Men in OEDC Countries: Political and Economic Power." 2014. http://www.oecd.org/gender/38172505.pdf
OECD. "Women and Men in OEDC Countries: Political and Economic Power." 2014. http://www.oecd.org/gender/38172505.pdf
BIG PICTURE:
ReplyDeleteResearch question, and IV and DV variables are very clear. You do a good job of explaining why you chose the variables that you chose, including the control variables but I don’t easily see where you got your data from (e.g., World Bank Development Indicators, etc.) I think how you measure the variable important too. I think your conclusion is fair, especially given the lack of significance for your IV variable.
NUTS AND BOLTS:
It may be clearer if you label your figures. (Instead of saying as the histogram shows, you can say as Figure 2 shows…the average reader may not know exactly what a histogram is).
MODELLING AND INFERENCE:
I wonder if the relationship in figure 1 is a parabola. If you tested for this and it wasn’t significant, I think you should mention this in the footnote.
Solving your reverse causality problem: I wonder if you can look at when women first begin to enter parliament and the average education level of women then and see if the average education level increases after women are in parliament (no regression, just a simple graph of education over time) and use this to help show that women in parliament help female education.
OTHER:
I’m a little confused by the last sentence in your first paragraph. I’m not sure what is a key way for developing states to create policy that makes higher education more accessible for women. Otherwise a good introduction: clear and gets my attention.
The links for your footnotes don’t seem to work. (I clicked on 4 in the text and it didn’t take me anywhere).
In the paragraph where you are discussing the relationship, you say the number of women in higher education; do you mean percent of women? Also you say that the percent of women in parliament changes by .5 (I think you forgot your unit).
You have a great intro and you clearly identify what you are investigating and what the critical variables are. I do think you may have spent a little too much time in the lit review and background. Its great stuff to have but it takes away the focus from your analysis and results. The last couple of sentences about where certain groups of nations stand with political representation seems slightly out of place and doesn't flow well into your own analysis that follows.
ReplyDeleteYou mention that you take a log of the the women in parliament variable in order to correct a positive skew situation in the data but the summary statistics table explains that the variable was a percentage to begin with. Is that the case? Are you regressing on the percentage change in a percent or was the variable a whole number that you then evaluated percent changes in? How does this decision affect the intepretation of the results?
Your figures and tables aren't always clearly labeled and I think that detracts from the flow of the post. Your regression table is straightforward but the layout makes it a little difficult to read. I would also like to see some more detail about why you ran the 4 modules the way that you did and perhaps a bit of comparison and explanation between them.
Your appendices includes a table about colinearity and I don't see any mention or concern of colinearity in post which may not be necessary but if you are to include the table, I think you should include an explanation of what that table is demonstrating. This is a small style issue, but in the summary statistics table, I am not sure I would leave the GDP variable represented in scientific notation.
Like the other reviewer, I also have a question about the potential for a parabolic relationship that seems to be possible based on Figure 1 and am wondering if you explored that option. If you have, it might be prudent to defend your decision not to use such a model because the figure must be included and suggests the possibility.
Overall, I think you set up the question well but I would to see less time spent in the background and the literature review and more time spent diving in and defending the actual analysis and interpretting those results.
Overall I think you have a great post. Your research question, dependent variable, and independent variable are all presented clearly. I appreciate the graphs and the visuals as they are both informative and add to your post’s aesthetics. I do have a few comments/suggestions though.
ReplyDeleteFirst, I think that your analysis is generally good, but it may be too brief. I know this is difficult because it essentially means taking content away from a fairly strong introduction, but your post could still benefit from it. I am curious, what was the p-value for the women’s education variable on the final module? I’m curious because the .05 value is somewhat arbitrary and you can still make a case if the p-value is .051 or a little bit higher. Some studies use a critical threshold of .1. A p-value of .056 in my mind is not meaningfully different from one that is .049, even though the .049 has an asterisk. The quiet and esttab regressions in stata only show significance in the p-value if it is less than .05, so I believe it is worth investigating.
Also, you may also want to create a footnote regarding your multi-collinearity test because it is included in the appendix but not referenced in the article. I think this is a valuable measure in your study because I initially was wondering whether women’s education and % of women in the labor force would face that problem. So it is a useful figure to have, but it needs to be referenced in some way.
Second, I think that there are few spots where you can be more precise in your word choice. In the sentence “Social Science has proven…” I would advise changing the word “proven.” This might be my own personal caution coming in, but even for relationships where causality is nearly a given, using words like “proven” or “proved” generally establish a level of certitude that social science cannot establish. Similarly, I would recommend changing “effect to “association” in the sentence, “Women in the labor force, seems to have the strongest effect on women in parliament…” You briefly and clearly explain why you included it as a control variable, but I do not think you established enough of a causal mechanism to use “effect.” Also, like a previous commenter mentioned, clarifying the final sentence of the first paragraph will improve your introduction.
Finally, I advise rewording, footnoting, or sourcing one statement in the introduction. The statement, “Higher education was chosen because a well educated society is going to be more open-minded,” has a few assumptions. Even though this statement makes sense, you may want to source it to something because the relationship between education and open-mindedness (however that is defined) could be a project in and of itself and is not necessarily a given without supporting data. That might be nitpicking, but a reference will make that line and the reasoning for your IV selection stronger.
Again, great post. I like your use of graphs, your argument is clear, and you do a really good job avoiding jargon. I think your post can get even better if you shift more of the focus to the analysis section, increase your precision when referring to certitude or causal claims, and ensure that all assertions are referenced or defended. Good luck.