Changing basic gender attitudes may be crucial for alleviating discrimination against women and improving gender outcomes. This column describes a unique measurement tool developed by social psychologists, which was adapted to measure gender attitudes of school children in the state of Haryana. It finds that both boys and girls in grades six and seven display strong same-gender preferences, with boys disproportionately associating females with negative attributes and vice versa.
Between 2001 and 2011, the child sex ratio (which measures the number of girls per 1,000 boys between the ages of 0-6) fell from 927 to 914 in India. This differential treatment of girls in early life is a symptom of a much wider problem that includes lifelong discrimination in health and nutrition, education, mobility and security, and economic and social opportunities through adulthood. Why do women face such persistent discrimination in the Indian society and what can be done to alleviate it?
Why women face discrimination in the Indian society
One popular answer is that differences in rewards in the economic and social marketplace drive gender differences (Qian 2008, Burgess and Zhuang 2008). If men earn more than women in the labour force, for example, then it would make sense for parents to favour their sons over their daughters. This would also point to the role of economic incentives in alleviating gender discrimination. But why did the child sex ratio decline in the last 10 years, a period of rapid economic growth that saw the emergence of significant new opportunities for women in the workforce? Why also, did inheritance law reform1 (Deininger, Goyal and Nagarajan 2013; Bhalotra and Chakravarty 2013; Jain 2014), punishment for sex-selective screening and schemes such as Haryana’s Devi Rupak2 (Anukriti 2014) and Delhi’s Ladli which offer cash transfers to parents of girls, fail to arrest the worsening sex ratio ?
One possible answer is that long-standing social attitudes play a major role in shaping the differential behaviour towards boys and girls. A boy who rarely sees his mother leave the house might expect the same behaviour from his wife, even when the economic opportunities available to the latter are significantly different. A girl who sees her father make all major household decisions might defer to her husband even when she has more education and capabilities. Insofar as these attitudes are deeply held and difficult to change through the provision of financial incentives, they may represent a significant challenge to erasing discrimination against women. At the same time, there is also an opportunity – reforming basic attitudes through a targeted intervention early in life might produce long-term improvements in outcomes for women even when the intervention itself is withdrawn.
Measuring gender attitudes of boys and girls in middle school
One such intervention is a school-based gender awareness and sensitisation programme being implemented by Breakthrough, a human rights’ NGO (Non-Governmental Organisation) that aims to eliminate discrimination against women and girls through community engagement, multimedia, and leadership training. The programme works with teachers and adolescents in middle school (grades seven and eight) since these students are still at a stage where opinions and attitudes are malleable. They are also at a key juncture in life – on the cusp of marrying and making fertility decisions. Because school enrolment is nearly universal, schools provide a promising avenue to reach out to adolescents and improve gender attitudes and behaviours at an early age.
Breakthrough’s programme, consisting of teacher training, youth clubs, school-based activities inside and outside the classroom and a nifty media and communications campaign, is currently being rolled out across 150 randomly selected government schools in four districts with among the worst sex ratios in Haryana – Sonepat, Panipat, Rohtak and Jhajjar. We are conducting a randomised evaluation3 to understand the success and cost-effectiveness of this innovative school-based sensitisation campaign. As part of the baseline study4, we measured existing gender attitudes among adolescents in school, who form the target group for the programme (Dhar, Jain and Jayachandran 2014a).
In any evaluation that directly studies attitudes, the measurement of attitudes is a major challenge, especially when participants may not be aware of or unwilling to share their true attitudes. For instance, students interviewed during the baseline may report opinions that they think will be viewed favourably or considered socially acceptable, rather than reveal their true attitudes (known as “social desirability bias”). This challenge is compounded in an evaluation of a programme that explicitly aims to change attitudes. When we eventually collect endline data on attitudes from students after they have been a part of Breakthrough’s programme, they might report to surveyors the attitudes that the programme taught them were ‘correct’, even if they remain unconvinced (known as the Hawthorne effect). In such cases, tools such as Implicit Association Tests (IATs) are useful for eliciting gender attitudes, since they are difficult to manipulate.
The IAT is a computer-based psychometric tool designed to detect the strength of automatic association between different ideas and concepts5. It takes into account not only the associations, but also the time taken to make categorisations. The central idea is that a respondent will match faster, items that they associate more strongly. For example, the IAT used in this project requires users to rapidly categorise two target concepts with an attribute (example, the target concepts ‘male’ and ‘female’ with the attribute ‘good’ or ‘bad’), such that easier pairings (faster responses) are interpreted as more strongly associated in memory than more difficult pairings (slower responses). ‘Male’ and ‘Female’ are depicted through appropriate pictures of boys and girls. Specifically, a respondent who is biased in favour of males will more quickly associate a male picture with a ‘good’ attribute such as ‘success’ than with a ‘bad’ attribute such as ‘pain’, and vice versa.
Figure 1. Screenshots from the Implicit Association Test (Classification of words and images)6
Since the IAT is a computer-based tool, it can measure the time taken for respondents to match concepts with great sensitivity. It uses the average response time over several stimuli (words and images) that are matched with different combinations of the paired words ‘Girl/Boy’ with ‘Good/Bad’, to detect the differences in associations of positive/negative attributes with boys and girls. The differences in response times are in milliseconds, and are thus difficult for respondents to consciously manipulate.
While IATs have been a staple of psychology labs in the US for more than a decade, they have rarely been used in other cultural settings. A pioneering use of the IAT in India was to examine impacts of having a female pradhan on attitudes about females in leadership positions (Beaman et al. 2009).
In our application, implementing an IAT with adolescents in middle schools in Haryana required considerable customisation. After multiple rounds of piloting and testing, we selected images of boys and girls that looked similar to those of the respondents (see the screenshots above), developed simple instructions and chose words that were easy to understand and straightforward to categorise as either good or bad. For instance, “success” and “intelligent” were considered as good words in the social milieu of middle schools in Haryana, whereas “unlucky” and “pain” were bad words.
Figure 2. A respondent taking an Implicit Association Test
Another contextual feature that we had to adapt to was that many students had limited experience using a keyboard. Thus, to give respondents practice in using a keyboard and in categorisation, they began with a practice IAT where images of insects and flowers had to be associated with the right words. Customisation to the local conditions through extensive piloting offered confidence that the IAT would indeed yield meaningful results.
The final respondents were more than 7,000 girls and boys in grades six and seven. The metric that the IAT uses to describe the strength of implicit attitudes is called the D-measure, and it represents the implicit gender preference for boys and girls.
The test revealed that boys disproportionately associate males with good attributes and females with bad attributes. Interestingly, the smallest bias against girls is in Panipat, the study district with the most balanced sex ratio. Girls are not gender neutral, but rather disproportionately associate female with good attributes and male with bad attributes. This finding is quite striking considering that for self-reported attitudes on the survey, many girls display pro-boy bias, for example saying that it is more important for parents to send sons than daughters to school (though the extent of girls’ self-reported pro-boy bias is lower than that of their male classmates’ pro-boy bias).
Figure 3. Implicit preference for boys
An optimistic interpretation of the IAT results is that the pro-girl bias of girls and the pro-boy bias of boys counterbalance each other; on average, adolescents’ implicit associations are fairly gender-neutral. However, that interpretation ignores that men have much more decision-making power than women in the Indian society today: When these children become adults, the boys might be especially influential in shaping household and societal decisions. Thus, addressing the pro-male bias seen among boys - and ensuring that girls do not also develop pro-male bias as they grow older7 - is something policymakers should focus attention on. Adolescence might be an effective age window in which to shape gender attitudes. The Breakthrough intervention, along with other independent and government-sponsored efforts across the country such as revisions in the school curriculum, mass-media and advertisement campaigns for the girl child (featuring celebrities8, for example), activity based programmes such as Meena Manch or Magic Bus, might help achieve this.
- There are various personal laws for inheritance and succession for each religious community in India, which have traditionally discriminated against women. Several changes have been made to these laws over the last century, such as the amendments to the Hindu Succession Act in 2005, to make them less discriminatory against women.
- The Devirupak scheme in Haryana seeks to promote a one-child norm and to decrease the sex ratio at birth by providing substantial monthly benefits, for a period of 20 years, for couples who sterilise after having one child (of either gender) or two girls (and no boy).
- A randomised evaluation is a type of impact evaluation in which participants are randomly assigned to a ‘treatment group’ that receives the intervention being evaluated, or a ‘control group’ which does not receive the intervention. In our case, a school is randomly assigned to the treatment group (receives the Breakthrough sensitisation programme) or the comparison group (existing state of affairs - with no Breakthrough programme).
- The baseline study to measure existing gender attitudes was conducted for children in grades six and seven, prior to the roll-out of the programme. The same cohorts of children are now in grades seven and eight, and the programme is working with them.
- Readers can take an Implicit Association Test online on the following website: https://implicit.harvard.edu/implicit/india/
- English translation of these slides is as follows: The upper left side of these slides has the paired words ‘Boy or Good’; the Upper Right Side has the words ‘Girl or Bad’, capturing the pairing of the target concept (Male/Female) with the Attribute (Good/Bad) The first slide has the word ‘Misfortune’, and the last slide has the word ‘Success’ in the Centre. These words and images at the Centre are stimuli to be matched with the pairings on the right or left.
- Our baseline survey measured self-reported attitudes of the adolescents’ parents. We find that mothers’ self-reported attitudes are as pro-boy as fathers’ self-reported attitudes and much more pro-boy than adolescent girls’ self-reported attitudes (Dhar, Jain and Jayachandran 2014b). This suggests that the pro-girl implicit bias we find among girls might erode as they age. We did not conduct the IAT among parents so cannot directly test this conjecture.
- We thank Prof. Debraj Ray for this suggestion.
- Beaman, L, R Chattopadhyay, E Duflo, R Pande and P Topalova (2009), “Powerful women: Does exposure reduce bias?”, Quarterly Journal of Economics 124(4), 1497–1540.
- Burgess, R and J Zhuang (2000), ‘Modernisation and son preference’, LSE Research Online Documents on Economics 2115, London School of Economics and Political Science, LSE Library.
- Deininger, K, A Goyal and H Nagarajan (2013), “Women’s Inheritance Rights and Intergenerational Transmission of Resources in India”, Journal of Human Resources, 48(1), 114-141.
- Dhar, D, T Jain and S Jayachandran (2014a), Evaluating a School-Based Gender Sensitization Program in India: Report on the Development and Use of Implicit Association Tests,” Policy report.
- Dhar, D, T Jain and S Jayachandran (2014b), “Intergenerational Transmission of Gender Attitudes: Evidence From India,” Working Paper.
- Greenwald, A, B Nosek and M Banaji (2003), “Understanding and using the Implicit Association Test: I. An improved scoring algorithm”, Journal of Personality and Social Psychology 85, 197-216.
- Greenwald, A, D McGhee and J Schwartz (1998), “Measuring individual differences in implicit cognition: The Implicit Association Test”, Journal of Personality and Social Psychology 74 (6), 1464.
- Jain, T (2014), “Where there is a will: Fertility behavior and sex bias in large families”, Journal of Human Resources, 49(2), 393-423.
- S Anukriti (2014), ‘The Fertility-Sex Ratio Trade-off: Unintended Consequences of Financial Incentives’, IZA Discussion Paper No. 8044.
- Qian, N (2008), “Missing Women and the Price of Tea in China: The Effect of Sex-Specific Earnings on Sex Imbalance”, Quarterly Journal of Economics, 123(3), 1251-1285.