«c Copyright 2014 The Author(s) Notice: Changes introduced as a result of publishing processes such as copy-editing and formatting may not be ...»
As mentioned above, in the landmark exploratory study in the field, Schmeidl and Piza-Lopez (2002) examined the potential for gender-specific indicators to inform assessment of the underlying risk of genocide and mass atrocities (root cause indicators, i.e. political participation of women) in addition to developing their own framework for understanding the escalation of risk (precipitating indicators, i.e. high volume of female refugees fleeing conflict). A series of UNIFEM-led studies used this model to explore the utility of local information gathering for early warning about escalating violence (i.e. identifying SGBV early to prevent it from becoming widespread and systematic) (Mosser 2007). This study interweaved generalized forms of violence against women (i.e. rates of domestic violence) with instances of organized violence against groups of women (i.e. armed groups attacking villages and mass rape).
In the rest of this section, we explore how contemporary application of these earlier studies relates to existing early warning frameworks for mass atrocities (R2P crimes).
We compare the earlier mentioned and respected early warning risk lists - Genocide Watch and Genocide Prevention Network – and compare their 2012 risk lists against a 2012 risk list developed from gender-specific indicators suggested by Schmeidl and Piza-Lopez (2002) (see Figure 1) as (equally) capable of identifying the risk of mass atrocities (see Table 1-3).
Before moving on to comparing the datasets, it is important to clarify some potential problems in the analysis. First, gender focused data collection is relatively new. The UNDP and the World Bank – primary sources of the data covered above – have only been collecting data with a ‘gender perspective’ in mind since early 1990s (i.e. 1994 for UNDP). This means that data collection still relies heavily on self-reported surveys from Member States (the primary source for data collection on domestic and sexual violence). Such data is unlikely to be fine-tuned to the gendered implications of data collection in countries with high levels of inequality in their political, judicial and security sectors. For example, significant grievances may not be recorded when women lack access to the formal legal system (Ni Aolain, Haynes and Cahn 2011: 79).
Second, there are methodological problems limiting the accuracy of data. For example, data on female infanticide relies on estimates, often from case studies which are then extrapolated countrywide. The final challenge is the reliance on data produced over different timeframes and the problem with data collection irregularities.
For example, it is difficult to compare data on literacy rates with female attendance rates at primary or secondary school, because many states do not provide regular data on the latter. If such data is to be used for early warning purposes, it must be collected regularly and as accurately as possible (Hudson et al 2012: 152-153).
The definition of violence described by Secretary-General Annan in his landmark 2006 report (see above) focuses on the systematic inequality and lack of empowerment that legitimates violence against women in the home and in society, and the possible overlap of this violence with conflict-related sexual violence. Such ‘very early’ early warning knowledge may be vital for understanding countries at risk of widespread and systematic SGBV but, equally, it may not be. We attempted to incorporate these relevant claims by using a ‘wider’ set of gender-specific indicators in addition to the more limited set of indicators suggested by Schmeidl and PizaLopez (2002).
Our starting point was to look at the countries deemed at high risk of mass atrocities (R2P crimes) by Genocide Prevention and Genocide Watch lists. We then compiled a set of gender-specific root cause indicators to generate a gendered ‘preconditions’ list based on the indicators identified by Schmeidl and Piza-Lopez (see Figure 1).
Figure 1: Root causes for genocide early warning – gendered indicators **INSERT FIGURE 1 HERE ** We soon encountered a number of problems in preparing data that replicated the Schmeidl and Piza-Lopez (2002: 9-11) framework. First, their suggested UNDP Gender Empowerment and Gender Development Index measures have evolved since 2011 into the Gender Inequality Indices (GII) in the UNDP annual report on human development indicators (HDI) (UNDP 2011). It is difficult to find data for some of the economic indicators, such as wage inequality, for many of the countries identified as being at high-risk according to the Genocide Watch and Genocide Prevention lists.
The GII comprises most of the substantive political, economic and social measures that Schmeidl and Piza-Lopez (Figure 1 above) recommend. But, additional economic data and information about female suffrage and Convention on the Elimination of Discrimination against Women (CEDAW) coverage are not included in the GII. We examined separately the relationship between GII rank, CEDAW adherence and female suffrage, as recommended by Schmeidl and Piza-Lopez (2002: 9-11). None of the (small number of) countries that have failed to ratify CEDAW correlated to countries with high GII scores (a high score for a country means their female citizens experience high levels of inequality). Likewise, given that female suffrage is now nearly universal – in principle at least – except in Qatar and Saudi Arabia, there appeared to be little use in applying a suffrage indicator (i.e. Htun and Weldon 2012).
In other words, states that have signed to CEDAW but not ratified the instrument (Iran, Palau, Somalia, Sudan, South Sudan, Tonga and US) are small in number. This, combined with the absence of female suffrage (only Qatar and Saudi Arabia), indicates that such gender-specific data limited will not provide strong early warning indicators of mass atrocities and widespread and systematic sexual violence.
This limited the socio-economic aspects of gender-specific indicators to the GII dataset (Klasen and Wink 2003). We identified three potential alternative composite gender indices: the World Economic Forum Global Gender Gap Index (GGI), the Economist Intelligence Unit’s Economic Opportunity Index (WEOI), and the OECD’s Social Institutions and Gender Index (SIGI). The GGI is a comprehensive annual economic survey but does not consider gender-specific indicators such as reproductive health, political empowerment and (unpaid) labour market participation.
The Economist Intelligence Unit’s Women’s Economic Opportunity Index (WEOI) would have been a useful alternative dataset, but it is not open source (authors did not have available funds to subscribe). The OECD SIGI released its second dataset (first was 2009) in May 2012. This dataset covers non-OECD and non-EU countries that have populations over one million. The focus is on developing countries undergoing rapid economic, social and political development. There are 102 countries surveyed in SIGI, with full data available for 86 non-EU and non-OECD countries. The values range from 0-1 with Argentina scoring the best SIGI at 0.0069 and Mali the worst at 0.6011 (OECD 2012). The SIGI indicator range is much more extensive than GII, with fourteen variables in five categories (see below). Indeed, SIGI attempts to cover the gender knowledge gaps identified by Schmeidl and Piza-Lopez (2002) concerning the inclusion of domestic/‘private life’ variables that affect women’s empowerment (Klasen and Wink 2003). As such, it includes variables that are often overlooked in other composite gender equality indices, including discriminatory family codes, restricted physical integrity, son bias (fertility preference towards boys and suspected practices of infant femicide), restricted resources and unique entitlements, and restricted civil liberties. The inclusion of SIGI indicators was an attempt to respond to critiques about the systematic (economic) bias against women’s lived experience in extant frameworks.
As Table 1 shows, all datasets identify a common group of states, though the levels of risk they identify in each country are significantly different. If we were to predict those countries at highest risk of mass atrocities just by virtue of their gender inequality indices, we see that the GII and SIGI lists produce quite different rankings on this performance. In examining Table 1, compare GII (greatest inequality to those that sit just above the ‘world average’ shaded according to 70% range; 60% range;
50% range and 40% range), with the SIGI range (SIGI range is 0-1, all countries listed that performed worse than median performer – Myanmar, ranked 43rd with SIGI value 0.2405). In comparing the GII and SIGI worst ten performing states, in terms of gender inequality, the two lists agreed on four states’ performance in this area, although it should be observed that GII did not report data for four of SIGI’s worst ten (Guinea, Nigeria, Somalia) and SIGI did not profile two of GII’s worst ten (Central African Republic, Papua New Guinea). Most significantly, in terms of data coverage – and ramifications for early warning – neither GII nor SIGI posted data (due to lack of consistent data) for six of the countries identified by the Secretary-General’s reports to the Security Council (2012) as having a high risk of SGBV (Ban 2012b).
*** INSERT TABLE 1 HERE *** Given disparities in the number of countries identified as ‘high risk’ in the frameworks, cognizant of the fact that policy-makers need to set priorities and allocate resources, we then reduced each list to (a maximum of) twenty-two countries.
We compared situations of ‘high risk’ in the lists to see whether gender inequality provides early warning of countries at risk of atrocities, particularly sexual violence.
*** INSERT TABLE 2 HERE *** *** INSERT TABLE 3 HERE *** This analysis produced mixed results, open, of course, to competing interpretations (see Table 3). The central finding is that in relation to the specific risk of SGBV, none of the non-UN lists correlated closely with the places of concern identified by the Secretary-General in the same year (2012). Of course, the counter argument could be that these lists are not expected to correlate – but that is our point – they are not expected to correlate and this should be a concern for those interested in addressing high-risk situations of SGBV.
The Genocide Watch list produced the highest replication (11 out of 22) with the UN Secretary-General’s list of countries at high risk of widespread SGBV, but this was still relatively low. The GII and SIGI fell short of the Genocide Watch list by three cases, though gaps between the datasets may explain the failure to list these cases.
For example, GII did not provide data for Egypt, Guinea and Somalia, unlike Genocide Watch and SIGI. If we presume that the high SIGI rating for Egypt, Somalia and Guinea would have been replicated in the GII scores, then if we make the same presumption for the GII cases that SIGI did not include (Central African Republic, Cote D’Ivoire, Liberia and Kenya), then the GII list would be equal to Stanton and the SIGI list would be most correctly aligned with the Secretary-General 2012 list. However, the assumption that GII would produce the same scores as SIGI for these countries could be queried given that GII and SIGI scores did not always correlate for the same country.
Findings Do gender inequality lists demonstrate their early warning capacity for countries at risk of atrocities, particularly sexual violence, and do they have greater capacity to highlight risk than early warning lists that exclude gender indicators? As shown above, the results here are mixed. GII and SIGI were able to produce lists that were comparable to the (most expansive) Genocide Watch early warning list and the UN Secretary General list on countries at risk of widespread sexual violence - even though the GII and SIGI use only socio-economic gender inequality indicators and no armed conflict indicators. However, none of these lists was accurate in identifying over 50% of the countries described as being at high risk of widespread and systematic SGBV by the Secretary-General. But, the Secretary-General’s list is a political list and may not reflect all cases and all risks – as the significant differences between the 2012 list and recent 2013 list suggests (Ban 2012b; Ban 2013b).
This preliminary research suggests that gender-specific indicators may be as useful as traditional non-gendered measures for early warning of imminent mass atrocities.
Moreover, failure to consider systematic gender inequality and/or the systematic use of sexual violence as high-risk indicators needs to be reconsidered. Indeed, given the gaps in gender data which negatively impacted on the gender indices, it is possible that gendered indicators might outperform general early warning predecessors. In turn, these frameworks might be improved by including gender-specific indicators, particularly when the focus is on predicting widespread and systematic sexual and gender-based violence. This underscores the importance of states implementing their commitment to Security Council Resolution 1325 (2000) and strengthening their collection and publication of gender-focused indicators.
A second finding is that the UN system has a solid understanding of high risk of sexual violence in situations of conflict, post-conflict and civil unrest even prior to the establishment of the 2014 Early Warning Matrix (Ban 2012b: Annex), but the list is political and vulnerable to exclusions that would not occur if weighed by evidence alone (Security Council Report 2013: 5). This poses the question of whether resources should be poured into refining analytical models before producing lists. While there is an urgent need to improve the quality and quantity of data relating to already established gender focused indicators, the evidence here suggests that there are real limits to what can be achieved with large-N risk assessments that are ultimately vulnerable to political decisions regarding who is listed and who is not. Moreover, lists do not mitigate the need for ongoing system-wide collection and assessment of information from the field and its incorporation into policy-making.