2006.the Uses and Misuses of the Genderrelated

Embed Size (px)

DESCRIPTION

use

Citation preview

  • The Uses and Misuses of the Gender-related Development Index and GenderEmpowerment Measure: A Review of theLiterature

    DANA SCHULERDana Schuler is a researcher at the Development Economics Research Groupat the University of Gottingen, Germany

    Abstract The 1995 Human Development Report introduced two newmeasures of well-being: the Gender-Related Development Index (GDI)and the Gender Empowerment Measure (GEM). The two indexes werecreated with the intention of attracting more attention to genderinequality issues. This paper first of all reviews the attention the indexesreceived in the publications of the United Nations DevelopmentProgramme itself, concentrating on their use in national and subnationalHuman Development Reports. It also reviews how the two indexes wereused in academia and the press. The main result of the review is that theGDI in particular seems to be a measure that is not used appropriately. Inmost cases of misuse, the GDI was wrongly interpreted as a measure ofgender inequality. Due to the many misinterpretations, the potentialpolicy impact the GDI and GEM can have seems limited.

    Key words: Gender-related Development Index, Gender EmpowermentMeasure, Gender inequality

    Introduction

    The 1995 Human Development Report introduced two new measures ofwell-being: the Gender-Related Development Index (GDI) and the GenderEmpowerment Measure (GEM). The GDI takes into account genderinequality in its overall assessment of aggregate human development ina country. It measures development in the same dimensions as theHuman Development Index (HDI), discounting them for gender inequal-ity. The GEM is meant to measure whether women and men hold equalpower in the political and economic sphere. The two indexes werecreated with the intention of attracting more attention to genderinequality issues.

    Journal of Human DevelopmentVol. 7, No. 2, July 2006

    ISSN 1464-9888 print/ISSN 1469-9516 online/06/020161-21 # 2006 United Nations Development ProgrammeDOI: 10.1080/14649880600768496

  • This paper is built upon an extensive literature review. First of all,it examines the attention the indexes have received within the publicationsof the United Nations Development Programme (UNDP) itself, concen-trating on their use in national and subnational Human DevelopmentReports.1 It also reviews how the two indexes were used in academicarticles and the press. The main result of the review is that both indices,seem to be rarely and inappropriately used. In most of the cases were amisuse occurred, the GDI was wrongly interpreted as a measure of genderinequality in itself. Therefore, the potential policy impact of the GDI inparticular seems questionable.

    The paper is structured as follows. First of all, the next sectionconsiders the correct interpretation of the GDI and GEM and explainsbriefly how the measures are computed. A literature review of thepublications that have correctly addressed the GDI and/or the GEMfollows, while the fourth section focuses on their adoption in developingcountries. The fifth section summarizes alternative measures designedto improve upon or widen the concepts encapsulated in the GDI andGEM. Then, the subsequent section highlights the incorrect uses ormisuses that have been made of the UNDPs gender-related indices.Finally, the seventh section evaluates the policy impact of the indexes andconcludes.

    Correct use of the GDI and GEM

    This section reviews briefly the correct interpretation of both indexes, theGDI and the GEM, including how to interpret the gap between the GDIand the HDI.2

    One of the criticisms brought up about the HDI was that it does nottake into account inter-group inequality in a society, assuming thateveryone in the society has reached average achievements for the threedifferent components of the index (life expectancy, education andincome). However, given that there are differences in achievements inthe population, such differences should be taken into account if anaversion to inequality between groups in a society exists. From this notion,Anand and Sen (1995) developed the GDI. The idea is to penalize theHDI if gender inequality exists in any of the three dimensions incorporatedin the HDI. The larger the gap between men and women in achievementsof life expectancy, education and income, the more the GDI differs fromthe HDI. The gap between the HDI and GDI therefore depends on theobjective gender inequality between men and women in one of thecomponents of the HDI, as well as on the penalty given to this genderinequality.

    To compute the GDI, first of all indicators of achievement for men andwomen are calculated separately. Secondly, based on Atkinsons methodof incorporating aversion to inequality (Atkinson, 1970), the equally

    D. Schuler

    162

  • distributed index is calculated for each component of the HDI as follows:

    Equally Distributed Index~ female population share female index1{e z male population share male index1{e

    g1=1{eIf e is equal to zero, then the simple arithmetic mean of female and maleachievements is calculated. The Human Development Report assumes an evalue of 2 indicating a social preference for equality. This step of thecalculation of the GDI has been misunderstood by most of the studies thatmisuse the index (see later Common mistakes section).

    The GDI therefore is to be interpreted as the HDI discounted forgender disparities in its components and should not be interpretedindependently of the HDI. The gap between the HDI and the GDI is to beinterpreted as the loss of human development due to gender inequalityand the size of penalty given to gender inequality.

    Finally, it is important to bear in mind that, as the methodologychanged in 1999, the GDI values before and after 1999 are not directlycomparable. The method of calculation changed due to a computationalmisspecification pointed out in Bardhan and Klasen (1999). Only sincethen have the HDI and the GDI been fully comparable. Before 1999 theequally distributed income index was calculated using female and maleproportional income shares. In 1999 the method was changed to use thefemale and male income index calculated in the same way as the incomeindex included in the HDI.

    The GEM consists of three indicators: male and female shares ofparliamentary seats; male and female shares of administrative, professional,technical, and managerial positions; and power over economic resources asmeasured by womens and mens estimated earned income (PurchasingPower parity, PPP US$).3 The GEM is meant to be interpreted as an index ofgender equity in political and economic participation and decision-making aswell as power over economic resources. However, due to a computationalmisspecification, it cannot be interpreted as such. To be such a measure, theincome component of the GEM would have to be based on income shares noton income levels (see Klasens contribution in the current JHD special issue).The income component of the GEM differs from the other components in theway it is calculated, incorporating an adjustment for overall income levels. Ashighlighted by Klasen in his contribution in the current JHD issue, this isproblematic because a poor country can never achieve a high value for theGEM even if equality of earned incomes occurred. Thus, changes in anddifferences between countries in the GEM are only interpretable if the change/difference is decomposed in changes/differences in female and male incomelevel and overall income level.

    Literature review correct use examples

    Not only have the GDI and GEM been used relatively rarely, but one alsohas to search for examples where the GDI or GEM were used correctly.

    Uses and Misuses of the GDI and GEM

    163

  • This section describes publications that use the indexes correctly. As theinterpretation of the GEM is complex due to the aforementionedcomputational shortcoming, all publications interpreting the GEM in theway the UNDP (1995) intended it to be interpretable are also included inthis section.

    Besides the Human Development Reports, the GDI and GEM seem tohave attracted minimal attention in the international press and, even moreimportantly, the introduction of the indexes do not seem to have started apublic debate on the overall issue of gender inequality. In press articles, ifthe GDI and GEM are mentioned, the focus is on the ranking of a specificcountry compared with other countries without an analysis of the gendersituation in the country (The Economist, 2002; Manila Standard, 2003;Korea Times, 2004; Central News Agency Taiwan, 2004; Global NewsBites The Nation, 2005).

    In academia, the GDI and GEM elicited a series of papers thatevaluated their strengths and shortcomings. These papers are notsummarized here, as the focus of this paper is on the use of and not theconceptual or methodological criticisms on the two measures. All otherpapers that used the GDI correctly and interpreted the GEM in the sameway as already mentioned are included.

    While several national and subnational Human Development Reportshave used the indexes, many did not report them at all. Among those thatdid report them, some of them merely presented the values of the GDI orGEM for their country at the national level, while others calculated theindices at the level of the local government. In any case, the interest is inthose studies that analyzed gender-related development more deeply, bynot only reporting but also analyzing the indexes. These are cited below.

    National Human Development Reports examples of good use

    As already mentioned, the GDI needs to be compared with the HDI (eitherusing the difference or the ratio between the two). The GDI is notinterpretable in itself if conclusions about the welfare loss due to genderinequality are to be drawn. This seems obvious. However, the later sectionon misuses shows that it is not.

    The 2004 National Human Development Report (NHDR) of Jordan(UNDP, 2004c) and the 2003 NHDR of Bolivia (UNDP, 2003b) areexamples of reports that compare the GDI with the HDI for analysis ofgender issues. The illustration of the gap between the HDI and GDI in theJordan NHDR shows in which regions of the country the loss of humandevelopment due to gender disparities is most severe. The Bolivian NHDRshows the dynamics of GDI compared with the HDI between two points intime, namely 1992 and 2001. The post-1999 computational method for theGDI is used in both cases.

    To draw conclusions on gender development from the GDI and theHDI in a specific country, a detailed descriptive analysis of the data used

    D. Schuler

    164

  • for computation is necessary. The summary of the UNDP networkdiscussion on GDI and GEM in 2005 highlights that it is important notonly to focus on aggregate gender-sensitive measures but also to analyzeindividual factors contributing to gender inequality. This is what the 2004NHDR of Jordan (UNDP, 2004c) partially does. It calculates the gapbetween male and female achievements in education by governorate. Theincome component of the GDI is not presented disaggregated by gender,but the GDP component of the HDI is compared with that of the GDI bygovernorate. For interpretation of the gap between the two, unemploy-ment statistics by gender and statistics on women labor force participationare presented.

    A good example of an extensive descriptive analysis can be found inthe 1999 NHDR of Cambodia (UNDP, 1999a). Regarding education, 1998data on adult literacy and schooling attainment are disaggregated bygender and income quintile. Age-specific gross enrolment rates show thewidening gap between males and females when age increases. In addition,net enrolment rates are analyzed, school performance between males andfemales is compared, and the reasons for gender disparity in schoolenrolment and in economic returns to schooling are analyzed. Thechapters on employment and health and nutrition are equally at length.

    When interpreting the GEM, a disaggregation of the figure into itscomponents and comparison of the trends in achievements between menand women is equally important. None of the reports shows trends in bothmale and female achievements. For instance, the 2004 NHDR of Indonesia(UNDP, 2004b) presents figures of the GEM by province for the years 1999and 2002. The falls in the GEM in nine provinces were analyzed to reflectchanges in the proportion of women in DPRDs the RegionalParliaments of Indonesia. The 1999 NHDR of Cambodia (UNDP, 1999a)uses the GEM to evaluate the position of the country in Asia, finding thatits score is among the lowest. The report also shows that GDP per capita isnot strongly correlated with the GEM in the Asian region.

    Academic articles

    The analysis of the GDI and GEM in academic articles can be classified inthree categories. Firstly, some papers have analyzed what the GDI andGEM actually measure. The question behind the analysis is whether thetwo measures significantly add new information to what is provided by theHDI or GDP per capita (PPP adjusted). For instance, McGillivray andPillarisetti (2003) analyze whether the GDI and GEM provide insights inthis respect. The authors test, using the Pearson zero-order correlationcoefficient and the Spearman rank-order correlation coefficient, whetherthe extent of redundancy justifies the effort involved in calculating andreporting the new indicator, introducing a threshold to define redun-dancy. The results indicate that the GDI is practically indistinguishableempirically from the HDI. The GEM is a more insightful indicator, when

    Uses and Misuses of the GDI and GEM

    165

  • compared with GDP per capita. For the low human development countrygroup, it is significantly negative correlated to GDP per capita. The paperalso offers a series of more demanding tests, such as the constant andslope coefficient test and the proxy variable test. Here the results aremixed. For the medium development group, the hypothesis that GEM orGDI and HDI are the same variable could not be rejected. The authorsconclude that based on simple measures of statistical association, GDI andGEM provide few new insights into ordinal and cardinal well-beingachievements.

    As UNDPs composite indicators were introduced because ofdiscontent with real GDP per capita as a sole indicator of well-being, itis interesting to compare degrees of inter-country inequality in theseindicators of human development. Ogwang (2000) compares inter-countryinequality in the HDI, GDI, GEM and two measures of GDP per capita. Thecoefficient of variation, income-weighted Theil-1 and population-weightedTheil-2 index is calculated for all five measures of well-being. The numberof countries for which inter-country inequality is computed variesaccording to the availability of data and is lowest for the GEM, with 94countries. Inter-country inequality in real GDP per capita is found to behigher than in all three human development indicators. Moreover, the GDIand GEM present higher degrees of inequality than the HDI, with the GEMexhibiting the highest degree of inter-country inequality out of the threeindices. This supports the argument that both the GDI and the GEM (butmostly the second) add new information to the comparison of humandevelopment between countries.

    McGillivray and Pillarisetti (1998) examine the determinants of theGEM. Using a cross-sectional database of 116 countries, they find thatthere is a significantly negative relationship between fertility rate andgender empowerment. This is in line with UNESCOs (2003) report, whichsimilarly finds evidence of the positive correlation between fertilitydecline and womens empowerment. On the other hand, the relation-ship between empowerment and female labor force participation issignificantly positive. Direction of causality is not analyzed further in eithercase.

    The second group of papers analyze gender differences in a specificthematic field, such as the health sector or the labor market, and try to givean indication of whether the GDI or the GEM tap these observed genderdifferences. However, these papers interpret the GDI or the gap betweenthe GDI and HDI wrongly. For instance, Blackburn and Jarman (2005)create a measure of segregation; that is, a measure of the tendency for menand women to be employed in different occupations. The authors examinethe relationship of this measure with the GDI and GEM. Both indexes arehigher in countries with higher female employment rates in the formalsector. The findings indicate that the GDI and GEM are highly positivelycorrelated with gender segregation in 16 industrialized countries for theyear 2000. The authors highlight that this result is surprising since theory

    D. Schuler

    166

  • would assume that segregation comes at a disadvantage for womensempowerment. Hank and Jurges (2005) find that a higher level of genderempowerment as measured by the GEM is positively associated with amore equal division of labor in the household.

    The third category of papers comprises those that aim to analyze theimpact of certain policies on the improvement of gender empowerment.4

    Carreiras (2005) constructs a measure of gender inclusiveness in themilitary, shedding light on how human development and genderempowerment are correlated with a symmetric situation of women inarmed forces compared with men.

    This section has shown that the use of the GDI and GEM has beenlimited. The reason for this might be their limited informational andempirical value added, as revealed by some of the aforementioned articles.Even in the national and subnational Human Development Reports, theindicators rarely constituted a starting point for an extensive and soundanalysis of gender inequality.

    Adoption of the GDI and GEM in developing countries

    The choice of indicators in the GDI and GEM is not always suited fordeveloping countries. Therefore, when the data are available, itmakes sense to widen the analysis to sectors and political levels thatare perceived as important there, but are not included in the GDI andGEM.

    The Human Development Reports produced in several Indian statesusually use data on representation in regional parliaments for assessingwomens empowerment. The 2004 West Bengal HDR (UNDP, 2004e), forexample, presents figures of percentage participation of women in GramPanchayats, Panchayat Samitis, and Zilla Parishads. Whereas repre-sentation of women in the national parliament in India is minor, thefigures for panchayats indicate a considerable participation of women atlower levels of political bodies (even by 1993, 35.5% of all members inGram Panchayats, 34.6% in Panchayat Samitis, and 36.1% in ZillaParishads were women). In 2003 the Indian government decided toreserve women 33 percent of seats in local municipal bodies (for examplein Panchayats). The 2002 Maharashtra HDR (UNDP, 2002b) presents a lessoptimistic picture of womens empowerment in 2001. The reportconcludes that women panchayati members face male resistance, whichmakes it difficult for them to take independent decisions.

    The GDI and GEM use an estimation of non-agricultural wage for thecomputation of the income component. The rural agricultural sector,however, is an important source of employment for women. The 2004West Bengal HDR (UNDP, 2004e) therefore describes the employmenttrends for men and women in rural as well as urban areas.

    Other subnational Human Development Reports Himachal Pradesh2002 (UNDP, 2002a) and Tamil Nadu 2003 (UNDP, 2003e) also present

    Uses and Misuses of the GDI and GEM

    167

  • figures on women employment and political empowerment, including forexample voting percentages Tamil Nadu.

    The lack of subnational data on gender in the political sphere hindersthe adoption of the GEM in some Indian state Human DevelopmentReports. For the GDI these restrictions do not bind.

    New measures inspired by the GDI and GEM

    There are two types of newly created gender-related measures. On the onehand, there are measures that incorporate new dimensions not includedin the GDI or GEM and that have been tailored to evaluate morerealistically the gender situation in specific regions of the world.5 On theother hand, the need to create a direct measure of gender inequality hasmotivated another range of papers.

    Concerning the first category of measures, several of the national andsubnational reports introduced new indicators, as it was felt that the GDIand GEM did not reflect the specific issues and concerns of a developingeconomy.

    The 2004 Gujarat HDR (UNDP, 2004a) introduces the GenderDevelopment Measure-1 (GDM-1) and the Gender Equity Index (GEI).The GDM-1 measure has the same components as the HumanDevelopment Measure (HDM-1), with the exception of per-capitaexpenditure. GDM-1 consists of the following components: education asmeasured by (a) adult literacy rate and (b) combined enrolment ratio;ability to lead a long and healthy life as measured by (a) life expectancy atbirth, (b) incidence of disability and (c) incidence of morbidity; availabilityof housing and related facilities as measured by (a) availability of a durablehouse and (b) availability of water, electricity and sanitation; andparticipation as measured by (a) the work participation rate of mainworkers, (b) the percentage of electors voting and (c) contestants inelections per lakh population. The GDM-1 is calculated by first computingthe indexes for men and women separately and then calculating theequally distributed index for each component. The GDM-1 is a simpleaverage of the four indexes.

    The GEI, which has the same components as the GDM-1, calculatesthe ratio of malefemale achievements for each indicator and thenaverages the indicators in each component. The first obvious drawback ofthe GDM-1 is that it does not include the same standard of living indicatorsas does the HDM-1 and thus both indicators cannot be compared.Secondly, the measure mixes two dimensions: basic capabilities andwomens empowerment. The GDI and GEM separate these twodimensions. As it is possible that different countries have gender equalityin basic capabilities, but look very different in the dimension ofempowerment, the separation of the two dimensions is preferable.Thirdly, the inclusion of empowerment indicators might exacerbate thefact of a possible substitution of a high achievement in one indicator and

    D. Schuler

    168

  • low in another. The GEI is more useful, as it is a measure of genderinequality. However, the same criticism as for the GDM-1 on mixing twodimensions holds here.

    Dholakia (2005) similarly suggests a widening of the GDI. TheModified Gender-related Development Index (MGDI) adds two variablesto the original index: gender-specific percentages in ministerialpositions, and gender-specific economic participation rates. An equallydistributed equivalent achievement is calculated for each variable. TheMGDI is then a simple average of all indexes. Different to the HDI andGDI, it is assumed that the female and male population is distributedequally, assuming a weight of 0.5. Again the same criticism as for the GEIapplies.

    Another indicator proposed in the same paper is the EquallyDistributed Gender-related Development Index (EDGDI), which addsfertility to the set of variables. As there are six variables for females andonly five for men, the indexes are first calculated for males and femalesseparately for each variable and then averaged to yield an aggregate maleand female index. Using the aggregate male and female index, the equallydistributed equivalent achievement is calculated, which is similar to theEDGI. A low fertility rate is interpreted as favorable for the empowermentof women. The inclusion of this variable poses several problems. First ofall, low fertility rates in developed countries hint also at difficulties forwomen to combine career and children. Secondly, the inclusion of thisvariable rewards countries such as China with its One-Child-Policy, whichcannot be interpreted as an empowerment of women.

    Other papers tried to improve on the existing GEM. Mehta (1996)constructs three measures of gender empowerment for developingcountries. The GEM-1 is the average of five indexes for: politicaldecision-making, relative exercise of the right to vote, relative lifeexpectancy, relative literacy rate, and per-capita income share. Theindicators of political decision-making are: relative representation ofwomen in Lok Sabah, legislature, Gram Panchayat, Panchayat Samiti andZilla Parishad. The GEM-2 takes the average of one index of politicaldecision-making and the four indexes for literacy, life expectancy, incomeand casting of the vote. It gives 20% weight to each of the five indicators.The GEM-3 is based on different weights for each of the nine indicators.6

    The three new GEMs are used to assess the situation of empowerment in16 states of India. Kerala ranks the first for all the three measures, andMaharashtra second. Some states, for instance Assam and Uttar Pradesh,rank differently for all the three measures. The incorporation of relativelife expectancy in a measure of empowerment might be questionable.Furthermore, the weighting of GEM-2 and GEM-3 seems subjective. On theother hand, the inclusion of the variables relative excise of the right tovote and representation of women in regional parliaments might be animprovement over the GEM in order to assess empowerment at thesubnational level in a developing country.

    Uses and Misuses of the GDI and GEM

    169

  • Time use is considered an important dimension of inequality betweenwomen and men in developing as well as developed counties.7 In order toassess the real workload of women compared with men, time-use statisticsare necessary. A large proportion of women are active in the informalsector. If data on activities in the informal sector are available, thedistinction between household and business activity may be unclear. Forthis reason, it is difficult to value womens work. Bhatia (2002) creates theGender Disparity Index by averaging four indexes based on the absolutedifference of weekly average time spent by a man and a woman ondifferent activities (SNA [System of National Accounts], extended SNA,non-SNA activities and unpaid SNA). SNA activities are defined as (I)primary production activities, (II) secondary activities, and (III) trade,business and services. Extended SNA activities are (I) householdmaintenance, management, and shopping for own household; (II) carefor children, the sick, elderly and disabled in the own household; and (III)community services and help of other households. Non-SNA activities are(I) learning activities; (II) social and cultural activities, mass mediaconsumption; and (III) personal care and self-maintenance. The fourindexes are calculated using the formula:

    Iij~xij{xmin

    xmax{xmin

    where i is the state, j is the indicator and xj represents the absolutedifference of time spent by men and women in the three differentcategories. Not surprisingly, the paper finds that women spent a lot moretime than men in unpaid work.

    As mentioned earlier, the GDI and GEM do not adequately reflect thegender situation in developed countries either. Kjeldstad and Kristiansen(2001) develop a regional gender equality index for the communities ofNorway to measure the status of women. Three types of indicators areincluded: a demographic indicator; welfare and economic indicators; andadministrative indicators. The demographic indicator included is definedas the number of women per 100 men aged 2039 years. The rationale isthat adult women move out of communities that fail to offer suitableemployment to find a job elsewhere. The administrative indicators includethe percentage of preschool children enrolled in publicly approveddaycare institutions and the female percentage of municipal councilmembers. Daycare coverage is placed in the responsibility of municipa-lities. Therefore it is assumed that a high coverage implies a good genderequality climate. Secondly, a high coverage permits mothers to combinechildcare and paid work. It is further assumed that a high femalepercentage of municipal council members is accompanied with a relativelyhigh legitimacy of gender equality policies in the community. The otherindicators are sex disaggregated educational, income and labor forceparticipation rates.

    The misinterpretations of the GDI highlight the need for theconstruction of an indicator that directly measures gender inequality.

    D. Schuler

    170

  • There are several papers that already proposed such a gender inequalityindicator.

    Akder (1994) shows that the HDI can be disaggregated by groups in asociety treating each group as if it were a separate country. Some NHDRscalculate the HDI for men and for women separately. For example, the1996 NHDR of Turkey finds a significant lower HDI for women (0.648)than for men (0.824) at the national level. Conversely, the 2003Kazakhstan (UNDP, 2003d) NHDR presents an slightly lower HDI formen (0.768) than for women (0.787) for the year 2002: women seem tohave advantages in life expectancy and education that are not offset by thedisadvantages in income.

    Dijkstra and Hanmer (2000) construct the Relative Status of Women(RSW) index, which uses the same indicators as the GDI. The RSW index iscalculated as follows:

    RSW~1

    3

    EfEmz

    LfLmz

    wfwm

    where Ef and Em are male and female educational attainment indexes, Lfand Lm are the male and female life expectancy index, and wf and wm arethe male and female rate of return to labor time. The indexes for malesand females are calculated in exactly the same way as they are for the GDI.

    Dijkstra (2002) creates the Standardized Index of Gender Equality(SIGE) with the aim to avoid the methodological limitations of GDI andGEM. The SIGE consists of five indicators: education, health, labor marketparticipation, share in higher labor market occupations/positions andshare in parliament. Indicators are defined as the relative achievement offemales to males for the first three indicators and as the female share forthe last two. For each country and indicator the resulting score isstandardized by expressing the score as number of deviations from themean of scores of all countries. The index is a simple arithmetic average ofthe standardized scores.

    Social Watch (2005) developed the GEI as a direct measure of genderequality. The index has three dimensions: education, economic participa-tion and empowerment. Gender equity in the education dimension ismeasured as the female-to-male ratio in literacy rates and in enrolmentrates at the primary, secondary and tertiary level. In the economicparticipation dimension, the percentage of women in total paid jobs(excluding the agricultural sector) and the ratio of female income to maleincome are used. Empowerment is measured by the percentage of womenin high administrative and management positions, in parliament and indecision-making posts at the ministerial level. The GEI is the simpleaverage of the indicators for the three dimensions.

    The latter two measures differ from the previous one, in that they mixdimensions of empowerment and capabilities. Again, the convenience ofdoing so is questionable. As already mentioned, it is possible for a countryto have achieved high levels of gender equality in basic capabilities, while

    Uses and Misuses of the GDI and GEM

    171

  • women remain disempowered. Therefore, the separation of the twodimensions seems preferable.

    The Economic Commission for Africa (2004) introduced the AfricanGender and Development Index (AGDI) in 2004, which is a compositeindex that has two parts: the Gender Status Index (GSI) and the AfricanWomens Progress Scoreboard (AWPS). The former incorporates elementsof basic capabilities, economic power and political power, using severalindicators to measure them:

    N enrolment rates, dropout rates, and literacy rates in the categoryeducation;

    N child health indicators, life expectancy at birth, new HIV infection, timespent out of work in the category of health;

    N wages in formal, agricultural, and informal sector;N income from formal, agricultural, and informal enterprise and from

    remittances;N time use for market, non-market, and leisure activities;N employment;N access to resources including access to management positions; andN political power as measured by positions held in public sector and civil

    society institutions.

    For every variable the female-to-male ratio is calculated. The GSI is aweighted average of the indicators. For details on the weightingprocedure, see Economic Commission for Africa (2004).

    The AWPS measures government policy performance regardingwomens advancement and empowerment in four areas: womens rights,capabilities, economic opportunities and political power. It tracks govern-ment progress in ratifying relevant conventions such as the Convention onthe Elimination of All forms of Discrimination against Women and in theimplementation of policies in line with international charters or policydocuments on such issues as violence against women, health or education.

    The AGDI, which combines the above indexes into one by averagingover the GSI and the AWPS, has been piloted in 12 countries. The value ofthe index resides in its comprehensiveness. Disaggregated scrutiny of thesubcomponents of the GSI is necessary for judging the performance ofcountries.

    The work by Sugarman and Straus (1987) was one of the earlieststudies that developed an empirical measure of gender empowermentsimilar the AGDI. Di Noia (2002) updates this measure using different datasources or substituting some indicators. The so-called Gender EqualityMeasure consists of seven indicators of economic gender equality, fourindicators of political gender equality and 13 indicators of legal genderequality. The legal indicators included are what distinguish this measurefrom the GEM. A value of one was assigned to a state if the appropriatestatute (e.g. providing warrantless arrest based on probable cause indomestic violence cases) was present. Sex discrimination laws were also

    D. Schuler

    172

  • taken into account (e.g. in the area of financing). For the economic andpolitical gender equality indicators the attainment of women relative tothat of men was calculated. The exact procedure of how to compute theindex can be found in the two papers. The Gender Equality Measure ofSugarman and Strauss therefore measures the extent to which womenhave the same access as men to economic resources, legal rights, orpositions of political power.

    There are a range of other papers that also dealt with the question ofwhich dimensions to include in a measure of gender inequality. Forexample, Wieringa (1999) presents the Gender Equality Index, a compre-hensive measure that consists of the following components: gender identity,autonomy of the body, autonomy within the family and the household,political power, social resources, material resources, employment andincome, and time.8 So far, to our knowledge, the index has not beencomputed for a country. See also Malhotra et al. (2002) for a review of empi-rical measures of empowerment used at the micro-level and the macro-level.

    Furthermore, several studies create indicators of empowerment at themicro-level.9 Ackerley (1995), for example, analyzes empowerment in creditprograms. Her measure of empowerment is derived by interviewingborrowers about their knowledge on input costs, product yield, andprofitability of the loan-funded activity. If the interviewee could answer a setof questions she scored one, and zero otherwise. Ackerley finds that creditprograms that encourage and enable women to participate directly in theactivity funded by their loan, are those that are most successful in empo-wering women. See also Oxaal and Baden (1997) for a review of other studies.

    This section has shown that there is a wide range of papers proposingnew measures. These can be classified in two types. Firstly, there arepapers that directly try to widen the GDI or the GEM by the inclusion ofnew dimensions. On this regard, some measures have been proposed toimprove on the GDI and GEM in order to better take into account thegender situation in specific regions of the world. Secondly, there arepapers that propose direct measures of gender inequality by, for example,calculating the female-to-male ratios for the dimensions of the GDI. Thesedirect measures of gender inequality are deemed as highly useful. This isbecause, as highlighted in the next section, there is demand for a directmeasure of gender inequality, rather than a measure that adjusts overalllevels of well being for gender inequality.

    Common mistakes

    The most common mistake made in studies using the GDI was to interpretit as a measure of gender inequality. Furthermore, most studies presentboth correct and incorrect interpretations. From this fact it can begathered that the formula of computation of the GDI is not understoodvery well, while there is a strong demand for a direct measure of genderinequality.

    Uses and Misuses of the GDI and GEM

    173

  • The 1999 NHDR of Kenya states the gender development index(GDI) has been designed to measure gender disparities in the level ofachievement in human development (UNDP, 1999b, p. 62). The 2003NHDR of Kenya (UNDP, 2003c) describes the GDI as a measure thatcaptures the differences between the achievements of women and men.Moreover, the report wrongly states that a value close to one signifiesachievement of equality for women and men. However, the next sentencecontinues with the correct interpretation and states that the GDI is ameasure of human development adjusted for gender inequality. The 2004NHDR of Macedonia also states first, correctly, that the GDI makes anadjustment for the average achievements, but continues wrongly: andshows the inequality between women and men in major humandevelopment areas(UNDP, 2004d, p. 37). The 2002 HDR of Rajasthan(UNDP, 2002c) defines the GDI as an adjusted version of the HDIreflecting gender inequalities, interpreting its individual values rather thancomparing them with the HDI. Other examples include the 2005 NHDR ofAlbania (UNDP, 2005a), the 2004 subnational Human Development Reportof West Bengal (UNDP, 2004e), and the 2002 subnational HumanDevelopment Report of Himachal Pradesh (UNDP, 2002a).

    In academic articles using the indexes the same mistake is common.For instance, Blackburn and Jarman (2005, p. 4) interpret the GDI andGEM as two measures constructed to indicate the extent to whichnational labor markets approach gender equality. This interpretationwould be partially correct for the GEM but definitely incorrect for the GDI.Fernandez et al. (2005) intend to analyze, among other things, therelationship between marital sorting and gender inequality, interpretingthe GDI as a measure of gender inequality and female status. The findingsare that the GDI, ratio of female-to-male enrolment rates, and thepercentage of female labor force affect marital sorting in a significantlynegative way. The frequency of this misinterpretation is further indicatedby a list of papers in which the same mistake is found: Berger (2002),Wieringa (1997), Kabeer (1999), Oudhof (2001), Pollert (2005), Saith andHarris-White (1999), United Nations Research Institute for SocialDevelopment (2005), Yu and Sarri (1997), and Kodoth and Eapen(2005). For instance, the latter affirm that the Indian state of Kerala hasbeen considered relatively free from restrictions against women, and verifythis with the GDI ranking of Kerela on top of other states. Even the WorldBank states that the GDI is a measure of gender disparity (2002, p. 2).

    Interestingly, as White (1997) highlights, the 1995 HumanDevelopment Report, which introduced the measure, was the first tomistakenly interpret the GDI as a measure of gender inequality. Severalstatements in the report are misleading:

    1. No society treats its women as well as its men. This is obvious fromthe GDI value. A value of 1.00 reflects a maximum achievement in basiccapabilities with perfect gender equality. But no society achieves such avalue(UNDP, 1995, p. 75). This statement suggests that the GDI is

    D. Schuler

    174

  • interpretable in itself and the higher its value, the lower is genderinequality. Looking at the GDI value by itself does not tell anythingabout the state of gender inequality in a society. A society with nogender inequality can still have a low GDI due to low level of overallhuman development.

    2. Gender equality does not depend on the income level of a society China is ten ranks above (GDI: comment of the author) Saudi Arabia,although its real per capita income is only a fifth of Saudi Arabias(UNDP, 1995, pp. 7678). Again the GDI is interpreted in itself and as adirect measure of gender inequality.

    3. The GDI values of all countries have improved since 1970. Not asingle country has slipped back in the march towards greater genderinequality (UNDP, 1995, p. 78).

    The 2000 Human Development Report also makes incorrect statements inthis respect: Of the 143 countries, as many as 30 have a GDI value of lessthan 0.50, showing that women in these countries suffer the doubledeprivation of low overall achievement in human development and lowerachievement than men (UNDP, 2000, p. 153).

    Another misleading interpretation firstly found in the 1995 HumanDevelopment Report is the following:

    4. One way of gauging the gender inequality in a country is to compareits GDI value with its HDI value. This can be simply done by taking thepercentage reduction of the GDI from the HDI, or: (HDI-GDI)/HDI(UNDP, 1995, p. 79).

    This percentage gap between the GDI and HDI is to be interpreted as theaggregate welfare loss due to gender inequality, and the penalty given togender inequality in the dimensions included. Other papers that used thisformula also interpreted the percentage gap between the GDI and HDI inwrongly. White (1997) states that the ratio of the GDI to the HDI is ameasure of gender disadvantage. Forsythe et al. (2003) assess the level ofgender inequality in the countries by calculating the aforementionedformula. The Philippine 1997 NHDR (UNDP, 1997) also interprets thepercentage gap as a measure of gender disparity.

    Equally impossible is to interpret a rise in the GDI without looking atthe dynamics of the HDI. The 2004 NHDR of Jordan (UNDP, 2004c), forexample, states that the rise in its GDI indicates that gender equalityappears to have improved. As shown in the third section, few reportsanalyze the dynamics of GDI and HDI in comparison.

    Nowhere is there perfect equality in all three dimensions of the GDI.As the GDI is a 1 e average of female and male achievement indexes, theGDI can only be lower than the HDI or equal to the HDI. Several NHDRsnevertheless present figures where the GDI exceeds the HDI. This is likelyto be due to the use of different data for computation of HDI and GDI.None of these NHDRs points at this issue and questions the data used forcalculation of the GDI. It could also be due to computational error. What is

    Uses and Misuses of the GDI and GEM

    175

  • the reason cannot be explored due to lack of information in the reports.The reports where this mistake was found are the following:

    N The 2005 NHDR of Albania (UNDP, 2005a) calculates the HDI and GDIfor different regions. In Tirana the GDI has a value of 0.864 and an HDIof 0.830 for 2004. The reason might be that the indexes are based ondifferent data sources. The GDI is not compared with the HDI and isinterpreted wrongly as a measure that captures differences in theachievements of men and women.

    N The 2004 Indonesian HDR (UNDP, 2004b) analyzes gender issues bycomparing the HDI and GDI. However, for the district Kota Kediri theHDI is given with a value of 0.661 and the GDI with a value of 0.691 for2002. In the same table the variable inputs for both measures are listedfor both men and women. The district shows highly unequal values inall three variables for men and women.

    N Most striking is this problem in the 2003 Indian subnational HumanDevelopment Report of Assam (UNDP, 2003a). On page 110 a tablewith the values of GDI and HDI for all districts is given. Excludingthe last five districts listed, the GDI exceeds the HDI in every district.In the North Carcha Hills the GDI has a value of 0.877 and the HDIa value of 0.363. The 2002 subnational Human DevelopmentReport of Himachal Pradesh (UNDP, 2002a) includes the sameproblem.

    N The 1998 NHDR of Cambodia (UNDP, 1998) highlights the point thatthe aggregate GDI is slightly higher than the HDI in 1997. In adisaggregation of the GDI for rural and urban areas as well as byquintile, it becomes clear that the GDI is higher than the HDIfor every category. The report continues with presenting exten-sive descriptive statistics on achievements of men compared withwomen for each category included in the GDI. This analysis showsthat there are large gender gaps in literacy and school enrolment ateach level that discriminate against women. Furthermore, femalemonthly earnings are substantially lower than those of males, even ifthey are in the same age group and have the same educationalachievements.10

    Some NHDRs do present a simple correlation between GDP on one handand the GDI, GEM, and HDI on the other 1998 and 1999 CambodiaNHDRs (UNDP, 1998, 1999a) and 2005 Korea NHDR (UNDP, 2005c). Thiscorrelation is then sometimes interpreted as a causal relationship. The2005 NHDR of Korea states that the correlation between the GDP, theHDI, and the GDI turned out to be highly positive, which meant thateconomic affluence is a prerequisite in achieving human development(UNDP 2005c, p. 19). The income index of the GDI for males and femalesis based upon female and male total GDP. The same does apply for theGEM. Therefore it is clear that the correlation between GDP, the HDI, theGDI, and the GEM will not be low.11

    D. Schuler

    176

  • Impact on policy and conclusions

    Due to many misinterpretations of the GDI, and the low attention the GDIand GEM attracted in the international press, the policy impact the twomeasures can have seems questionable. Secondly, the GDI and GEM werecriticized because they do not adequately reflect gender inequalitydimensions neither in developing countries, nor in developed nations.Finally, shortcomings in the analysis of gender issues in national andsubnational reports not only reflected in the lack of adequate data, butalso in the lack of deeper descriptive analysis using the existing data reduce the possibilities for their influence on policy.

    The network discussion on the GDI and GEM (UNDP, 2005d)reviewed experiences of the policy impact of the two indices. Thesummary of responses states that the impact of GDI and GEM seems tohave been limited in Central and Eastern Europe, but also for most of theother regions no impact on policy was reported. This is also the result ofthe review of studies for this paper. In contrast, the summary report of thediscussion (UNDP, 2005d) highlights the example of Korea, where theGEM has been used directly for lobbying purposes. The government hasresponded by taking action against the low representation of women inpolitical and economic sectors.

    In conclusion, it can be said that the GDI is a measure prone tomisinterpretation. Moreover, the misuses highlight the demand for a directmeasure of gender inequality. One measure that could be easily obtainedis calculating the average of the female-to-male ratios for the variablesincluded in the GDI.

    The GEM is also incorrectly interpreted due to the computationalfeature of the income component. There is only one recommendation inthis case. The GEM would be much more consistent if the incomecomponent were based purely on male and female income shares (seeKlasens contribution in this JHD special issue).

    Finally, many papers that proposed direct measures of genderinequality also proposed to mix the dimensions included in the GDI andthe GEM. However, keeping two distinct measures is preferable, since theseparate interpretation of these dimensions is more informative.

    Acknowledgements

    The author would like to thank Stephan Klasen, Susana Franco, Tim Scott,Sarah Burd-Sharps and all the other participants at the workshop to reviewthe GDI and GEM, organized by HDRO in New York, 2021 January, forhelpful comments. Furthermore, the author would like to thank KlausSeipp for assistance in the literature search. The author would also like toacknowledge the Government of The Netherlands for their support for thispaper through UNDPs Gender Thematic Trust Fund.

    Uses and Misuses of the GDI and GEM

    177

  • Notes

    1 The term subnational reports is used here to refer to the reports that have beenproduced in several Indian states.

    2 For a detailed explanation on how to calculate the indices, see the appendix to theguest editors introduction to this Journal of Human Development (JHD) issue, or thetechnical note in UNDP (2005b).

    3 The GEM uses unadjusted incomes: no logarithmic transformation of income is made.For more details on its calculation see the appendix to the Editors Introduction in thecurrent JHD issue.

    4 The third category of papers also comprises those that aim to analyze the impact ofcertain policies on the improvement of gender inequality. These papers interpret theGDI wrongly as a measure of gender inequality. For instance, Schnepf (2004) analyzeswhether reform policies in transition countries lead to an improvement in the GEM,GDI and a measure of the loss in human development due to gender inequality.

    5 A drawback of such a measure is that only similar countries can be compared using themeasure. Hence, the value of global gender indicators that allow comparisons for allcountries is not questioned.

    6 The weights are as follows: 0.05 for the indexes of political decision-making, 0.15 forthe index of life expectancy and that of casting a vote, 0.20 for the index of literacy, and0.25 for the index of income.

    7 Another dimension that is perceived as very important when trying to measurewomens status in society is violence against women. However, it is impossible toinclude this by now due to lack of data.

    8 The ideas for the components of the Gender Equality Index were collectivelyformulated during a workshop held in The Hague in January 1997.

    9 Ackerley (1995), for example, analyzes empowerment in credit programs. His measureof empowerment is derived by interviewing the borrowers about their knowledgeabout input costs, product yield, and profitability of the loan-funded activity. If theinterviewee could answer a set of questions she was scored with a one, and a zerootherwise. Ackerley finds that credit programs that encourage and enable women toparticipate directly in the activity funded by their loan are those that are most successfulin empowering women.

    The briefing written by Oxaal and Baden (1997) reviews some of the studies. Hashemiet al. (1996) used a model based on eight indicators of empowerment in their study ofrural Bangladesh: mobility, economic security, ability to make small purchases, abilityto make larger purchases, involvement in major household decisions, relative freedomfrom domination within the family, political and legal awareness, and involvement inpolitical campaigning and protests. Furthermore, the Canadian InternationalDevelopment Agencys (CIDA) approach of measuring empowerment is introduced.The CIDA measures empowerment in the legal, political, economic and socialdimension. Social empowerment, for, instance is measured by the extent of trainingand networking among local women, as compared with men.

    10 Time worked was not taken into account.11 To evaluate the contribution of GDP on gender inequality, one could analyze whether

    increases in GDP result in a higher share of womens GDP per capita compared with thatof men; that is, whether women disproportionably gain from increased incomecompared with men. Otherwise, there would be the possibility of investigating in whichway changes in the gap between GDI and HDI are related to changes in GDP per capita.For analyzing the relationship between income and gender inequality in economicstatus, the earnings gap between males and females would be one possible indicator.

    References

    Ackerley, B.A. (1995) Testing the tools of development: credit programmes, loaninvolvement and womens empowerment, IDS Bulletin, 26(2), pp. 5667.

    D. Schuler

    178

  • Akder, A.H. (1994) A means to closing gaps: disaggregated Human Development Index,Occasional Paper 18, UNDP, New York.

    Anand, S. and Sen, A. (1995) Gender inequality in human development: theories andmeasurement, Occasional Paper 19, UNDP, New York.

    Atkinson, A.B. (1970) On the measurement of inequality, Journal of Economic Theory, 2,pp. 244263.

    Bardhan, K. and Klasen, S. (1999) UNDPs gender-related indices: a critical review, WorldDevelopment, 27(6), pp. 9851010.

    Berger, L. (2002) Conflict prevention, gender and early warning: a work in progress,Prepared for the 3rd Expert Consultative Meeting on Gender and Early Warning,London 11 February.

    Bhatia, R. (2002) Measuring gender disparity using time use statistics, Economic andPolitical Weekly, 37(33), pp. 34643469.

    Blackburn, R.M. and Jarman, J. (2005) Gendered occupations: exploring the relationshipbetween gender segregation and inequality, GeNet Working Paper No. 5, GenderEquality Network, University of Cambridge.

    Carreiras, H. (2005) Patterns of gender integration in the military: a cross-nationalassessment of womens participation in the Armed Forces, updated version of paperpresented at the ECPR workshop The Future of Gender Equality in the European Union,Granada, April [http://www.essex.ac.uk/ecpr/events/jointsessions/paperarchive/granada/ws5/Carreiras.pdf] (accessed 4 November 2005).

    Central News Agency Taiwan (2004) Taiwan ranks 1st in Asia in terms of womenparticipation in politics, 21 September.

    Dholakia, K. (2005) Human Development Index and status of women, mimeo, TheUniversity of Texas at Dallas [http://www.utdallas.edu/,kruti/HDI_status_%20of_women_072305.pdf] (accessed 5 September 2005).

    Dijkstra, A.G. (2002) Revisiting UNDPs GDI and GEM: towards an alternative, SocialIndicators Research, 57, pp. 301338.

    Dijkstra, A.G. and Hanmer, L.C. (2000) Measuring socio-economic gender inequality:toward an alternative to the UNDP Gender-related Development Index, FeministEconomics, 6(2), pp. 4175.

    Di Noia, J. (2002) Indicators of gender equality for American states and regions: anupdate, Social Indicators Research, 59(1), pp. 3577.

    Economic Commission for Africa (2004) The African Gender and Development Index,ECA, Addis Ababa.

    Fernandez, R., Guner, N. and Knowles, J. (2005) Love and money: a theoretical andempirical analysis of household sorting and inequality, Quarterly Journal of Economics,114(4), pp. 12431284.

    Forsythe, N., Korzeniewicz, R.P., Majid, N., Weathers, G. and Durrant, V. (2003) Genderinequalities, economic growth and economic reform: a preliminary longitudinalevaluation, Employment Paper No. 45, ILO, Geneva.

    Global News Bites The Nation (2005) The Worst Performer, 19 July.Hank, K. and Jurges, H. (2005) Gender and the division of household labor in older

    couples: a European perpective, Working Paper No. 76, Mannheim Research Institutefor the Economics of Aging, Mannheim.

    Hashemi, S.M., Schuler, S.R. and Riley, A.P. (1996) Rural credit programs and womensempowerment in Bangladesh, World Development, 24(4), pp. 635653.

    Kabeer, N. (1999) The conditions and consequences of choice: reflections on measure-ment of womens empowerment, UNRISD Discussion Paper, No. 18, United NationsResearch Institute for Social Development, Geneva.

    Kjeldstad, R. and Kristiansen, J.E. (2001) Constructing a regional gender equality index:reflections on a first experience with Norwegian data, Statistical Journal of the UnitedNations ECE, 18, pp. 4149.

    Kodoth, P. and Eapen, M. (2005) Looking beyond gender parity: gender inequities of somedimensions of well-being in Kerala, Economic and Political Weekly, 40(30),pp. 32783286.

    Uses and Misuses of the GDI and GEM

    179

  • Korea Times (2004) Kora 28th in human development, 16 July.Malhortra, A., Schuler, S.R. and Boender, C. (2002) Measuring womens empowerment as

    a variable in international development, background paper prepared for the WorldBank Workshop on Poverty and Gender: New Perpectives, 28 June.

    Manila Standard (2003) 2003 Human Development Index reveals crisis, Financial TimesInformation Manila Standard, 11 July.

    McGillivray, M. and Pillarisetti, J.R. (1998) Human development and gender empower-ment: methodological and measurement issues, Development Policy Review, 16,pp. 197203.

    McGillivray, M. and Pillarisetti, J.R. (2003) Adjusting human well-being indices for genderdisparity: insightful empirically?, paper presented at the WIDER Conference onInequality, Poverty and Human Well-Being, Helsinki, Finland, 3031 May.

    Mehta, A.K. (1996) Recasting indices for developing countries: a gender empowermentmeasure, Economic and Political Weekly, 26 October.

    Ogwang, T. (2000) Inter-country inequality in human development indicators, AppliedEconomic Letters, 7, pp. 443446.

    Oudhof, K. (2001) The GDI as a measurement instrument on gender aspects ofdevelopment in the ECE region, Statistical Journal of the United Nations ECE, 18,pp. 125.

    Oxaal, Z. and Baden, S. (1997) Gender and empowerment: definitions, approaches andimplications for policy, BRIDGE Report, No. 40, Institute of Development Studies,Brighton.

    Pollert, A. (2005) Gender, transformation and employment in Central Eastern Europe,European Journal of Industrial Relations, 11(2), pp. 213230.

    Saith, R. and Harris-White, B. (1999) Gender-sensitivity of well-being indicators,Development and Change, 30, pp. 465497.

    Schnepf, S.V. (2004) Transformations of gender relations in Central and Eastern Europe:the impact of reform on gender equality, mimeo, United Nations Coordination Unit inTajikistan [http://www.untj.org/files/reports/ifu-paperGenderinTransition.pdf] (accessed10 November 2005).

    Social Watch (2005) Roars and Whispers Gender and Poverty: Promises versus Action,Social Watch, Montevideo.

    Sugarman, D. and Straus, M. (1987) Indicators of gender equality for American states andregions, Social Indicators Research, 20, pp. 229270.

    The Economist (2002) Self-doomed to failure Arab development, 6 July.UNDP (1995) Human Development Report 1995, Oxford University Press, New York.UNDP (1997) Philippine Human Development Report 1997, UNDP, Philippines.UNDP (1998) Cambodia Human Development Report 1998: Womens Contribution to

    Development, UNDP, Phnom Pehn.UNDP (1999a) Cambodia Human Development Report 1999: Village Economy and

    Development, UNDP, Phnom Pehn.UNDP (1999b) Kenya Human Development Report 1999: Gender and Human

    Development, UNDP, Nairobi.UNDP (2000) Human Development Report 2000, Oxford University Press, New York.UNDP (2002a) Himachal Pradesh Human Development Report 2002, UNDP, Shimal.UNDP (2002b) Human Development Report Maharashtra 2002, UNDP, Mumbai.UNDP (2002c) Rajasthan Human Development Report 2002, UNDP, New Delhi.UNDP (2003a) Assam Human Development Report 2003, UNDP, Dispur.UNDP (2003b) Informe de Desarrollo Humano de Genero en Bolivia, UNDP, La Paz.UNDP (2003c) Kenya Human Development Report: Participatory Governance for Human

    Development, UNDP, Nairobi.UNDP (2003d) Kazakhstan Human Development Report 2003: Water as a Key Factor of

    Human Development in Kazakstan, UNDP, Almaty.UNDP (2003e) Tamil Nadu Human Development Report 2003, UNDP, Chennai.UNDP (2004a) Gujarat Human Development Report 2004, UNDP, Ahmedabad.

    D. Schuler

    180

  • UNDP (2004b) Indonesia Human Development Report 2004: The Economics of Demo-cracy: Financing Human Development in Indonesia, UNDP, Jakarta.

    UNDP (2004c) Jordan Human Development Report 2004: Building sustainable liveli-hoods, UNDP, Amman.

    UNDP (2004d) National Human Development Report 2004-FYR Macedonia: Decentra-lization for Human Development, UNDP, Skopje.

    UNDP (2004e) West Bengal Human Development Report 2004, UNDP, Kolkata.UNDP (2005a) Albania Human Development Report 2005: Operationalizing the MDGs in

    Albania, UNDP, Tirana.UNDP (2005b) Human Development Report 2005, Oxford University Press, New York.UNDP (2005c) Korean Human Development Report: Gender, UNDP, Seoul.UNDP (2005d) Discussion: revisiting the GDI and GEM, HDRStats-Net Consolidated

    Reply, UNDP, HDRO [http://hdr.undp.org/docs/nhdr/consolidated_replies/Revisiting_GDI_GEM_31October2005.pdf] (accessed 20 November 2005).

    UNESCO (2003) EFA Global Monitoring Report, UNESCO, Paris.United Nations Research Institute for Social Development (2005) Gender Equality: Striving

    for Justice in an Unequal World, UNRISD, Geneva.White, H. (1997) Patterns of gender discrimination: an examination of the UNDPs Gender

    Development Index, mimeo, Insitute of Social Studies, The Hague.Wieringa, S.E. (1997) A reflection on power and the gender empowerment measure of the

    UNDP, mimeo, Institute of Social Studies, The Hague.Wieringa, S.E. (1999) Summary of workshop discussions and decisions, report from the

    Workshop on GDI/GEM Indicators, The Hague, 1318 January 1997.World Bank (2002) The gender dimension of bank assistance: an evaluation of results.,

    Report No. 23119, World Bank, Washington, DC.Yu, M.Y. and Sarri, R. (1997) Womens health status and gender inequality in China, Social

    Science and Medicine, 45(12), pp. 18851898.

    Uses and Misuses of the GDI and GEM

    181