Key findings
- Item non-response rates: o Overall rates were within ideal levels (below 5%) for all items except for perceived abortion prevalence (12.2%). o Average rates across the 11 items were also within this range for all countries except Morocco (31.3%), Guinea-Bissau (10.3%), Mauritius (8.2%), Angola (8.2%), and Tunisia (6.5%). o Morocco had extremely high refusal rates, ranging from 9.8% (views on marriage autonomy) to 41.5% (contraceptive access regardless of age).
- Interview conditions and item non-response rates: Disaggregation by gender dyads, the presence of others, and respondent attitudes all indicate modest differences at the aggregate level that are unlikely to have meaningfully impacted respondent behaviour.
- Question ordering: Overall, question order had little effect on whether people responded or how they answered, with only a few small exceptions in specific countries (Senegal, Seychelles, and Sierra Leone).
- Psychometric validation: o To rationalise the validation, countries where missing values on scalar scores exceeded the threshold limit were excluded from the analysis, resulting in a 33- country validation analysis. o Item performance varied significantly across these 33 countries, requiring country specific construction of composite scales for valid and reliable measurement. o Some items (Q85A, Q85B, Q86C, Q86D) were consistently poor across countries, often due to wording or phrasing issues that introduced ambiguity in meaning and interpretation. Based on the initial findings, these items were removed from further analysis for developing final scales. o Even within suitable scales as determined in the final analysis, certain items (Q85C, Q86B, Q86E) outperformed others, making them better candidates for single-item analysis. o While scalar use is generally encouraged to minimise measurement error, single-item analysis is permissible based on specific country recommendations. Recommendations on item use are also provided on scalar and single-item analysis for multi-country analysis and comparisons.
Women’s and girls’ reproductive and sexual health rights (including access to contraception and termination of pregnancy) vary widely in Africa due to differing legal, cultural, and religious contexts. For instance, most African countries inherited restrictive colonial-era abortion laws, and an estimated 92% of women of reproductive age in sub-Saharan Africa live in countries that either prohibit abortion or restrict it to cases in which a woman’s life or health is threatened (Bankole, Remez, Owolabi, Philbin, & Williams, 2020).1 Many African societies also have deep rooted cultural norms or religious beliefs that stigmatise abortion (Makenzius, McKinney, Oguttu, & Romuld, 2019), which may discourage people from seeking or providing services in contexts where it is legal (Jim et al., 2023). Abortion, in many countries a bellwether for the broader sexual and reproductive health rights available and accessible to women and girls, is therefore often politicised on the continent (Williams, 2025).
Keep up with the latest headlines on WhatsApp | LinkedIn
Afrobarometer’s Round 10 survey module on sexual and reproductive health and rights (SRHR) asks a series of questions on abortion, access to contraception, sexuality education, and women’s autonomy. The high degree of social sensitivity of these issues raises challenges for both questionnaire design and data analysis. Asking direct questions about these issues may cause discomfort, leading to higher rates of refusal or non-committal responses among some groups (non-response bias). Or survey respondents may feel compelled to provide what they perceive or know to be socially acceptable answers, rather than truthful ones (social desirability bias). These challenges can impact data accuracy and interpretation, thereby diminishing the validity and reliability of obtained responses and limiting use of these results to inform public policy.
This methods note uses the Round 10 SRHR module as a case study to illustrate challenges in surveying socially sensitive attitudes and to showcase some of the diagnostic tools that can be applied to assess the quality of the questionnaire items and their suitability for application in analysis. The diagnostic tools range from descriptive analyses to more advanced psychometric validation testing. Using data from the 38 countries surveyed between January 2024 and September 2025, item response rates indicate that most countries meet recommended benchmarks, thus supporting reliable sub-national analysis. Further disaggregation by gender, age cohort, and religion is therefore permissible. Caution is required, however, when analysing data from Angola, Guinea-Bissau, Mauritius, Morocco, and Tunisia due to elevated levels of non response across numerous items.
In addition, psychometric validation (tests of how well the questions worked) demonstrates that individual items perform differentially in different countries, thus requiring careful selection of items for scalar construction (combining questions into composite measures, or “scales”) and/or single-item analysis. While scalar use of individual items is recommended, this varies by country, requiring selection of different items across different countries. Overall, some items performed consistently better than others across a majority of countries, and recommendations are provided per country for selection and use of individual items as either components of composite scales or for single-item analysis.
Sign up for free AllAfrica Newsletters
Get the latest in African news delivered straight to your inbox
Success!
Almost finished…
We need to confirm your email address.
To complete the process, please follow the instructions in the email we just sent you.
Error!
There was a problem processing your submission. Please try again later.
Based on these findings, we strongly recommend that researchers begin by thoroughly discussing the specific legal and social context of sexual and reproductive rights in each country, as these factors are likely to have an impact on responses. This is particularly applicable for items on the most sensitive topics, such as abortion. In addition, analysts should report itemresponse rates and flag any questions that fail to meet the recommended benchmarks. They should also examine whether there are systematic differences in non-response levels in their specific country or countries of interest based on factors such as interviewer-respondent gender dyads, the presence of others during an interview, and respondents’ reported demeanour or disposition. Lastly, analysts are encouraged to conduct multivariate analyses and to draw from the qualitative insights available in interviewers’ comments as a means of contextualising the results obtained from the quantitative analysis.
Rorisang Lekalake is a senior analyst/methodologist at Afrobarometer
Rajen Govender Rajen Govender is a senior adviser at Afrobarometer.
