new measures of gender inequality: - the social institutions and
TRANSCRIPT
The Constructionof the Social
Institutions andGender Index
(SIGI)
Branisa, Klasen,and Ziegler
(2009)
Introduction
The OECDGender,Institutions andDevelopmentDatabase
Construction ofthe Subindices
The SocialInstitutions andGender Index(SIGI)
Results
Discussion andFurther Research
References
New Measures of Gender Inequality:The Social Institutions and Gender Index (SIGI)
and its subindices
Boris Branisa, Stephan Klasen and Maria Ziegler
Georg-August-University of Goettingen
Far East and South AsiaMeeting of the Econometric Society
Tokyo, August 2009
1 / 27
The Constructionof the Social
Institutions andGender Index
(SIGI)
Branisa, Klasen,and Ziegler
(2009)
Introduction
The OECDGender,Institutions andDevelopmentDatabase
Construction ofthe Subindices
The SocialInstitutions andGender Index(SIGI)
Results
Discussion andFurther Research
References
Contents
1 Introduction
2 The OECD Gender, Institutions and Development Database
3 Construction of the Subindices
4 The Social Institutions and Gender Index (SIGI)
5 Results
6 Discussion and Further Research
2 / 27
The Constructionof the Social
Institutions andGender Index
(SIGI)
Branisa, Klasen,and Ziegler
(2009)
Introduction
The OECDGender,Institutions andDevelopmentDatabase
Construction ofthe Subindices
The SocialInstitutions andGender Index(SIGI)
Results
Discussion andFurther Research
References
Contents
1 Introduction
2 The OECD Gender, Institutions and Development Database
3 Construction of the Subindices
4 The Social Institutions and Gender Index (SIGI)
5 Results
6 Discussion and Further Research
3 / 27
The Constructionof the Social
Institutions andGender Index
(SIGI)
Branisa, Klasen,and Ziegler
(2009)
Introduction
The OECDGender,Institutions andDevelopmentDatabase
Construction ofthe Subindices
The SocialInstitutions andGender Index(SIGI)
Results
Discussion andFurther Research
References
Motivation
Lack of measures that capture underlying causes of outcome genderinequalities
• Most of the existing measures focus only on gender inequalities inoutcomes like education, health, economic or political participation(e.g., GDI, GEM, GGG)
Problems in constructing composite indices:
• In general most important issues to tackle:Which variables to use? How to combine them?
• In particular: What is a reasonable methodology to capture thedeprivation of women caused by social institutions?
4 / 27
The Constructionof the Social
Institutions andGender Index
(SIGI)
Branisa, Klasen,and Ziegler
(2009)
Introduction
The OECDGender,Institutions andDevelopmentDatabase
Construction ofthe Subindices
The SocialInstitutions andGender Index(SIGI)
Results
Discussion andFurther Research
References
Contribution
First, provide summary measures of the underlying causes of outcomeinequalities for over 100 developing countries:
Social institutions, i.e. long-lasting codes of conduct, norms, traditions,informal and formal laws that spur discrimination of women.
• Five subindices that measure each one dimension
• The Social Institutions and Gender Index (SIGI) that combinesthese different dimensions
• The SIGI and its subindices were officially presented by the OECDon the International Women’s Day 2009 (www.genderindex.org.)
5 / 27
The Constructionof the Social
Institutions andGender Index
(SIGI)
Branisa, Klasen,and Ziegler
(2009)
Introduction
The OECDGender,Institutions andDevelopmentDatabase
Construction ofthe Subindices
The SocialInstitutions andGender Index(SIGI)
Results
Discussion andFurther Research
References
Contribution
Second, method of construction allows to capture the deprivation ofwomen caused by social institutions related to gender inequality inseveral dimensions
• It is inspired by the FGT poverty measures(Foster, Greer, and Thorbecke, 1984),
• makes it possible to penalize high inequality in each dimension,
• and allows only for partial compensation between dimensions.
6 / 27
The Constructionof the Social
Institutions andGender Index
(SIGI)
Branisa, Klasen,and Ziegler
(2009)
Introduction
The OECDGender,Institutions andDevelopmentDatabase
Construction ofthe Subindices
The SocialInstitutions andGender Index(SIGI)
Results
Discussion andFurther Research
References
Contents
1 Introduction
2 The OECD Gender, Institutions and Development Database
3 Construction of the Subindices
4 The Social Institutions and Gender Index (SIGI)
5 Results
6 Discussion and Further Research
7 / 27
The Constructionof the Social
Institutions andGender Index
(SIGI)
Branisa, Klasen,and Ziegler
(2009)
Introduction
The OECDGender,Institutions andDevelopmentDatabase
Construction ofthe Subindices
The SocialInstitutions andGender Index(SIGI)
Results
Discussion andFurther Research
References
Data
OECD GID Database is a cross-country database covering about 130countries with about 20 variables measuring social institutions related togender inequality
Morrison and Jütting (2005),Jütting, Morrison, Dayton-Johnson, and Drechsler (2008),www.wikigender.org.
From the database we choose 12 variables that:
• cover about 120 non-OECD countries from all regions in the world(except North America)
• are between 0 and 1, 0 meaning "no or very low inequality", and 1indicating "high inequality"
• are prevalence rates, legal indicators, indicators of practices.
8 / 27
The Constructionof the Social
Institutions andGender Index
(SIGI)
Branisa, Klasen,and Ziegler
(2009)
Introduction
The OECDGender,Institutions andDevelopmentDatabase
Construction ofthe Subindices
The SocialInstitutions andGender Index(SIGI)
Results
Discussion andFurther Research
References
Data
These variables are assumed to measure one of the four dimensionsoriginally proposed by the OECD:
• Family Code - decision-making power of women in the household.
(parental authority, inheritance, early marriage, polygamy)
• Civil liberties - freedom of social participation.
(freedom of movement, freedom of dress)
• Physical Integrity - different indicators on violence against women.
(violence against women, female genital mutilation, missing women)
• Ownership rights - access of women to several types of property.
(women´s access to land, women´s access to loans, women´saccess to property other than land)
9 / 27
The Constructionof the Social
Institutions andGender Index
(SIGI)
Branisa, Klasen,and Ziegler
(2009)
Introduction
The OECDGender,Institutions andDevelopmentDatabase
Construction ofthe Subindices
The SocialInstitutions andGender Index(SIGI)
Results
Discussion andFurther Research
References
Contents
1 Introduction
2 The OECD Gender, Institutions and Development Database
3 Construction of the Subindices
4 The Social Institutions and Gender Index (SIGI)
5 Results
6 Discussion and Further Research
10 / 27
The Constructionof the Social
Institutions andGender Index
(SIGI)
Branisa, Klasen,and Ziegler
(2009)
Introduction
The OECDGender,Institutions andDevelopmentDatabase
Construction ofthe Subindices
The SocialInstitutions andGender Index(SIGI)
Results
Discussion andFurther Research
References
Association of variables
First , check that variables assumed to belong to one dimension measurethe same underlying concept
• Methods: Kendall’s Tau b, Multiple Correspondence Analysis (MCA)
• Results: positive association of variables in each of the dimensions,except missing women variable in the dimension physical integrity.
• Decision: build 5 subindices using the "missing women" variablefor a separate subindex.
• Missing women seems to be measuring an important and differentaspect of social institutions, namely "son preference" (Sen, 1992;Klasen and Wink, 2003).
11 / 27
The Constructionof the Social
Institutions andGender Index
(SIGI)
Branisa, Klasen,and Ziegler
(2009)
Introduction
The OECDGender,Institutions andDevelopmentDatabase
Construction ofthe Subindices
The SocialInstitutions andGender Index(SIGI)
Results
Discussion andFurther Research
References
Aggregation of variables to a subindex
Second , we choose a weighting scheme and a method of aggregation:
• Using Polychoric PCA (Kolenikov and Angeles, 2009) we extractthe common information of the variables corresponding to asubindex.
• The Subindex is the rescaled First Principal Component (FPC) sothat it is between 0 (no or very low ineq.) and 1 (high ineq.)
• The FPC is a weighted sum of the standardized original variables.
Subindex(Country X) =FPC(Country X)
FPC(Country Worst) − FPC(Country Best)
−
FPC(Country Best)
FPC(Country Worst) − FPC(Country Best)(1)
Exception: Subindex Son Preference takes the value of the variablemissing women.
12 / 27
The Constructionof the Social
Institutions andGender Index
(SIGI)
Branisa, Klasen,and Ziegler
(2009)
Introduction
The OECDGender,Institutions andDevelopmentDatabase
Construction ofthe Subindices
The SocialInstitutions andGender Index(SIGI)
Results
Discussion andFurther Research
References
Contents
1 Introduction
2 The OECD Gender, Institutions and Development Database
3 Construction of the Subindices
4 The Social Institutions and Gender Index (SIGI)
5 Results
6 Discussion and Further Research
13 / 27
The Constructionof the Social
Institutions andGender Index
(SIGI)
Branisa, Klasen,and Ziegler
(2009)
Introduction
The OECDGender,Institutions andDevelopmentDatabase
Construction ofthe Subindices
The SocialInstitutions andGender Index(SIGI)
Results
Discussion andFurther Research
References
Main ideas behind the index
SIGI should
• be used to compare and rank countries .
• be transparent and easy to understand(0 meaning no or very low inequality, 1 meaning high inequality).
• combine the five subindices in a way to reflect the deprivation ofaffected women caused by inequality in social institutions.
• be empirically non-redundant to other existing gender-relatedmeasures.
14 / 27
The Constructionof the Social
Institutions andGender Index
(SIGI)
Branisa, Klasen,and Ziegler
(2009)
Introduction
The OECDGender,Institutions andDevelopmentDatabase
Construction ofthe Subindices
The SocialInstitutions andGender Index(SIGI)
Results
Discussion andFurther Research
References
Main properties of the SIGI
Therefore, the SIGI is an unweighted average of a non-linear functionof the subindices :
• Equal weighting of the subindices as there is no reason for valuingone of the dimensions more or less than the others.
• Non-linearity because we assume that deprivation increases morethan proportionally when inequality increases.
The non-linearity implies two useful properties:
• Penalization of high inequality in every dimension.
• The SIGI allows only for partial compensation : High inequality inone dimension (subindex) can only be partially compensated withlow inequality on another dimension.
15 / 27
The Constructionof the Social
Institutions andGender Index
(SIGI)
Branisa, Klasen,and Ziegler
(2009)
Introduction
The OECDGender,Institutions andDevelopmentDatabase
Construction ofthe Subindices
The SocialInstitutions andGender Index(SIGI)
Results
Discussion andFurther Research
References
The Formula of the IndexFor our specific five subindices, the value of the index SIGI is defined as:
SIGI =1
5(Subindex Family Code)2 +
1
5(Subindex Civil Liberties)2
+1
5(Subindex Physical Integrity)2 +
1
5(Subindex Son preference)2
+1
5(Subindex Ownership Rights)2
Using a more general notation, the SIGI I(X) is computed as:
SIGI = I(X) =1
n
n∑
i=1
φ(xi , 0) =1
n
n∑
i=1
(xi − 0)2 =1
n
n∑
i=1
(xi )2,
X is the vector of the subindices xi , i = 1, ..., n
φ(xi , 0) is the deprivation function with φ(xi , 0) > 0 if xi > 0and φ(xi , 0) = 0 if xi = 0
xi represents the distance to 0, the goal of no inequality to be reached on everydimension.
16 / 27
The Constructionof the Social
Institutions andGender Index
(SIGI)
Branisa, Klasen,and Ziegler
(2009)
Introduction
The OECDGender,Institutions andDevelopmentDatabase
Construction ofthe Subindices
The SocialInstitutions andGender Index(SIGI)
Results
Discussion andFurther Research
References
The Formula of the Index
The SIGI is inpired by the FGT poverty measures (Foster et al., 1984)that are defined for yi ≤ z as:
FGT (Y , α, z) =1n
n∑
i=1
( z − yi
z
)α
, (2)
where yi is the income of individual i , z is the poverty line, α > 0 is apoverty penalization parameter.
In our case:
• 0 is the analogous of z, and 2 is the selected value for α.
• the normalization achieved when dividing by z is not necessary, as0 ≤ xi ≤ 1, i = 1, . . . , n.
• a lower value of xi is preferred contrary to the income case
• we aggregate across dimensions (i).
17 / 27
The Constructionof the Social
Institutions andGender Index
(SIGI)
Branisa, Klasen,and Ziegler
(2009)
Introduction
The OECDGender,Institutions andDevelopmentDatabase
Construction ofthe Subindices
The SocialInstitutions andGender Index(SIGI)
Results
Discussion andFurther Research
References
Contents
1 Introduction
2 The OECD Gender, Institutions and Development Database
3 Construction of the Subindices
4 The Social Institutions and Gender Index (SIGI)
5 Results
6 Discussion and Further Research
18 / 27
The Constructionof the Social
Institutions andGender Index
(SIGI)
Branisa, Klasen,and Ziegler
(2009)
Introduction
The OECDGender,Institutions andDevelopmentDatabase
Construction ofthe Subindices
The SocialInstitutions andGender Index(SIGI)
Results
Discussion andFurther Research
References
SIGI: Which countries score worst and best?
SIGI Family Civil Physical Son OwnershipCountry Ranking Ranking Ranking Ranking Ranking Ranking RankingParaguay 1 .0024832 19 15 6 67 21Croatia 2 .00333 3 29 11 58 17Kazakhstan 3 .0034778 5 77 9 26 14Argentina 4 .0037899 14 76 12 61 7Costa Rica 5 .0070934 23 8 17 34 42Russian Federation 6 .0072524 36 75 10 57 41Philippines 7 .0078831 8 14 4 87 55El Salvador 8 .0082581 17 21 3 22 49Ecuador 9 .0091447 24 20 7 43 54Ukraine 10 .00969 10 45 28 19 3
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.Iraq 93 .2752427 77 108 98 106 87Pakistan 94 .2832434 64 116 47 120 101Iran, Islamic Rep. 95 .3043608 91 120 91 91 94India 96 .318112 100 104 18 118 97Chad 97 .3225771 111 102 84 83 120Yemen 98 .3270495 97 119 64 111 79Mali 99 .339493 112 59 114 36 64Sierra Leone 100 .3424468 98 58 110 39 121Afghanistan 101 .5823044 110 121 93 122 109Sudan 102 .6778067 106 123 111 107 122
19 / 27
The Constructionof the Social
Institutions andGender Index
(SIGI)
Branisa, Klasen,and Ziegler
(2009)
Introduction
The OECDGender,Institutions andDevelopmentDatabase
Construction ofthe Subindices
The SocialInstitutions andGender Index(SIGI)
Results
Discussion andFurther Research
References
SIGI: Quintiles by region
ECA LAC EAP SA SSA MENA TotalSIGIQuintile 1 6 10 4 0 1 0 21Quintile 2 6 8 5 0 0 1 20Quintile 3 1 1 2 1 14 2 21Quintile 4 0 0 1 2 13 4 20Quintile 5 0 0 1 4 10 5 20Total 13 19 13 7 38 12 102
20 / 27
The Constructionof the Social
Institutions andGender Index
(SIGI)
Branisa, Klasen,and Ziegler
(2009)
Introduction
The OECDGender,Institutions andDevelopmentDatabase
Construction ofthe Subindices
The SocialInstitutions andGender Index(SIGI)
Results
Discussion andFurther Research
References
Contents
1 Introduction
2 The OECD Gender, Institutions and Development Database
3 Construction of the Subindices
4 The Social Institutions and Gender Index (SIGI)
5 Results
6 Discussion and Further Research
21 / 27
The Constructionof the Social
Institutions andGender Index
(SIGI)
Branisa, Klasen,and Ziegler
(2009)
Introduction
The OECDGender,Institutions andDevelopmentDatabase
Construction ofthe Subindices
The SocialInstitutions andGender Index(SIGI)
Results
Discussion andFurther Research
References
Discussion
We have proposed several new composite indices for measuring socialinstitutions related to gender inequality.
Strengths:
• focus on a new aspect of gender inequality (roots and notoutcomes)/ empirically non-redundant
• over 100 developing (non-OECD) countries can be ranked andcompared
• transparent and easy to understand
• SIGI: penalization of high inequality, partial compensation
• 5 subindices give a variety of information
22 / 27
The Constructionof the Social
Institutions andGender Index
(SIGI)
Branisa, Klasen,and Ziegler
(2009)
Introduction
The OECDGender,Institutions andDevelopmentDatabase
Construction ofthe Subindices
The SocialInstitutions andGender Index(SIGI)
Results
Discussion andFurther Research
References
Discussion
Limitations of any composite measure:
• The data used as input influence the quality of the composite index.
• By aggregating variables some information is inevitably lost.
• The index does not replace neither a careful investigation of all thevariables nor a case analysis.
• Always some arbitrary decisions regarding weighting andaggregation.
Main limitation of the SIGI and its subindices:
• OECD countries cannot be ranked as variables do not capturesocial institutions in these countries.
23 / 27
The Constructionof the Social
Institutions andGender Index
(SIGI)
Branisa, Klasen,and Ziegler
(2009)
Introduction
The OECDGender,Institutions andDevelopmentDatabase
Construction ofthe Subindices
The SocialInstitutions andGender Index(SIGI)
Results
Discussion andFurther Research
References
Discussion
Nevertheless:
• SIGI can be useful as countries with severe problems arehighlighted, and
• subindices can be useful as problematic issues (e.g., family law)within countries can be highlighted, so that
• policy-makers and the international community can exertpressure.
• the public in the respective countries can be informed.
• Moreover, especially the subindices have explanatory value fordevelopment outcomes (health and education).
24 / 27
The Constructionof the Social
Institutions andGender Index
(SIGI)
Branisa, Klasen,and Ziegler
(2009)
Introduction
The OECDGender,Institutions andDevelopmentDatabase
Construction ofthe Subindices
The SocialInstitutions andGender Index(SIGI)
Results
Discussion andFurther Research
References
Further Research
Work in progress: "Why we should all care about socialinstitutions related to gender inequality?"
Need to invest in the measurement of social institutions related togender inequality in OECD countries as the variables that arenow available in the GID database are not appropriate.
25 / 27
The Constructionof the Social
Institutions andGender Index
(SIGI)
Branisa, Klasen,and Ziegler
(2009)
Introduction
The OECDGender,Institutions andDevelopmentDatabase
Construction ofthe Subindices
The SocialInstitutions andGender Index(SIGI)
Results
Discussion andFurther Research
References
Thank you
26 / 27
The Constructionof the Social
Institutions andGender Index
(SIGI)
Branisa, Klasen,and Ziegler
(2009)
Introduction
The OECDGender,Institutions andDevelopmentDatabase
Construction ofthe Subindices
The SocialInstitutions andGender Index(SIGI)
Results
Discussion andFurther Research
References
Main references
Foster, J. E., J. Greer, and E. Thorbecke (1984). A class of decomposable poverty measures.Econometrica 52, 761–766.
Jütting, J., C. Morrison, J. Dayton-Johnson, and D. Drechsler (2008). Measuring gender(In)Equality: The OECD gender, institutions and development data base. Journal ofHuman Development 9(1), 65–86.
Klasen, S. and C. Wink (2003). Missing women: Revisiting the debate. Feminist Economics 9,263–300.
Kolenikov, S. and G. Angeles (2009). Socioeconomic status measurement with discrete proxyvariables: Is principal component analysis a reliable answer? Review of Income andWealth 55(1), 128–165.
Morrison, C. and J. P. Jütting (2005). Women’s discrimination in developing countries: A newdata set for better policies. World Development 33(7), 1065–1081.
Sen, A. (1992). Missing women. British Medical Journal 304, 586–7.
27 / 27