“You need to be able to stand up for what is right”: MTV Shuga Naija’s transformative impact on youth attitudes towards sexual violence in Nigeria

Intervention

The design and content of MTV Shuga Naija is guided by the Theory of Planned Behavior (Ajzen, 1991), which posits that changes in the precursors to targeted behaviors – attitudes, subjective norms, and perceived behavioral control – lead eventually to changes in behavioral intention and, in turn, to changes in behaviors Shuga programming is designed to include content specific to behavioral beliefs, normative beliefs, and control beliefs.

The MTV Shuga Naija television and radio dramas confront challenging issues by featuring diverse characters who express a variety of attitudes, norms, and behaviors. The show presents new, positive attitudes, norms, and behaviors for young people to emulate. At the same time, it also depicts traditional or harmful attitudes, norms, and behaviors, demonstrating how these can negatively affect youth. On the issue of sexual violence, storylines illustrated how behaviors such as respecting the autonomy of sexual partners, supporting victims of sexual violence, and not engaging in sexual violence led to better outcomes, while also depicting the possible negative consequences of harmful behaviors, both for victims and for perpetrators of sexual violence.

In the 2018-2020 seasons of MTV Shuga Naija, issues of sexual violence were portrayed in two key storylines – the rape of a high school girl at a party thrown by older boys and the rape of a sexually experienced “video vixen” and aspiring singer by an influential music producer. Smaller storylines – such as a teacher soliciting sexual favors from a female student to improve her grades or a male student suggestively touching female students – depicted forms of sexual harassment. Storylines showed the victims struggling with the after-effects of sexual assault and harassment, modeling proactive behaviors to seek justice, and ultimately prevailing over entrenched norms and social structures, thereby offering viewers a path to deal with their own experiences of sexual violence.

The target population for MTV Shuga Naija was youth in the states of Kaduna, Kano, and Lagos, although the content in Lagos was broader than in the northern two states. In Lagos, storylines touched not just on sexual harassment and violence but also on sexual education and reproductive health. From March-April 2018, ten 22-minute episodes of the television serial drama were aired on MTV satellite and NTA2 Lagos (Nigeria Television Authority) and again 18 months later. The episodes on NTA2 were available only for viewers in the Lagos area while those airing on MTV satellite were potentially viewable anywhere in the country, although access was generally considerably lower outside of Lagos. Episodes also aired on the www.mtvshuga.com website and on the MTV Shuga YouTube channel. The episodes were complemented by a radio drama, social media (e.g., mtvshuga.com website, Facebook, Instagram, Twitter), and peer education sessions with youth in Lagos. A second 8-episode television drama series that continued the storylines in the earlier series were shown in October/November 2019. This series aired on MTV satellite, MTV Base, Startimes TV, Pop Central, NTA, ONTV, and EBONYLIFE TV, alongside a more extensive radio drama, peer education, and social media.

Beginning in 2019, MTV Shuga commenced limited program activities in Kaduna and Kano, focused solely on reproductive health and absent content and storylines about sexual violence. Activities included a television magazine, a radio drama, social media, and peer education. Themes of this programming

Data

To assess the effects of Shuga on the target groups, surveys were conducted among a randomly sampled cohort of urban youth. Consistent with the stepwise introduction of the intervention, youth were interviewed at three points in time: (1) prior to the implementation of Sahnuyga Naija activities in any of the states (baseline), (2) after implementation of Shuga Naija activities in Lagos but not elsewhere (midline), and (3) after the implementation of Shuga Naija activities in all three states. We focus only on the baseline and endline waves of data because at midline, many of the critical story arcs regarding sexual violence were still emerging, meaning that key learnings for youth were not fully revealed.

At baseline, the target sample size for the study was computed as 2,000 females and 1,000 males between the ages of 14-24 years. With permission from the Bill and Melinda Gates Institute of the Johns Hopkins University, Baltimore, the Center for Research, Evaluation Resources, and Development (CRERD) used the Performance Monitoring for Accountability 2020 (PMA2020 2017) Nigeria sample frame in Kano, Lagos, and Kaduna states to select urban clusters for the Shuga survey. Utilization of PMA2020 enumeration areas (EAs), which were randomly selected from each state’s census enumeration areas, meant that there was a previously established and recent sampling frame of all households located in PMA2020 EAs.

At baseline, in-person interviews were completed with 1,900 females and 986 males. As with most panel surveys, attrition of respondents occurred across survey waves. Of the baseline participants, 70% agreed to be re-contacted for subsequent interviews, and attempts were made to re-contact these individuals for the endline survey.

Because of contact restrictions due to the COVID-19 pandemic and to comply with COVID-19 regulations, data collection for the endline survey was undertaken through phone interviews. Phone calls were made to all respondents in the baseline and midline database who had valid phone numbers and consented to callbacks.¹ The switch to phone surveys compounded earlier attrition, resulting in an endline sample of 899 respondents. Of these, only 641 in total (399 females and 242 males), participated in all three waves.

Survey Instruments

The MTV Shuga Naija survey instruments were designed to capture information on Shuga exposure and indicators from the Theory of Planned Behavior (Ajzen 1991) that guided the development of MTV Shuga Naija programming. The questionnaire focused on issues that were specifically addressed by the storylines for each of the MTV Shuga Naija’s main characters. For example, questions were asked about: 1) behaviors or experiences (e.g., “Has someone ever touched you in a way that made you feel uncomfortable?” “Have you ever talked with your parents or family members about sexual violence?”); 2) attitudes (e.g., “It’s not rape if a girl doesn’t fight back”); 3) subjective norms (e.g., “Most of my friends think I should use condoms when having sexual intercourse”); 4) perceived behavioral control (e.g., “How easy or difficult would it be for you to use condoms with a sexual partner?”); and 5) behavioral intentions (e.g., “During the next 3 months, how likely is it that you will try to persuade your partner to use condoms every time you have sex?”).

Because the endline survey was conducted by phone, a shorter questionnaire was required to minimize respondent fatigue and drop-off. Questions focused on a smaller set of indicators from the TPB that were deemed essential to cover the theory’s broad domains and behavioral outcomes (e.g., contraceptive use, HIV testing, experiences and discussions of sexual violence).

Human Subjects Research Approval

For the baseline survey, National Health Research Committee (NHREC) approval was obtained on January 24, 2018 with NHREC/01/01/2007-24/01/2018. Institutional Review Board (IRB) approval from the Tulane University Biomedical IRB was granted on February 18, 2018. For the midline survey, IRB approval from the NHREC was obtained, which spans August 23, 2019 through August 22, 2020 (NHREC/01/01/2007-23/08/2019B). Approval from the Tulane University Biomedical IRB was granted on September 20, 2019. For the endline survey, CRERD received approval on July 9, 2020. Approval from the Tulane University Biomedical IRB was granted on August 5, 2020.

Statistical Analysis

Our analytical approach is to estimate two-period, two-group difference-in-differences models, comparing changes over time in key outcomes in the treatment area of Lagos relative to those in the comparison areas of Kaduna and Kano. The primary assumption underlying the DID model is that, in the absence of the treatment (MTV Shuga Naija), average outcomes in treatment and control areas would have followed the same trend from baseline to endline. Consequently, any differences observed in those trends at endline are attributed to the causal influence of MTV Shuga Naija. In the DID models, we include a time dummy variable (baseline wave=0, endline wave=1), a binary variable indicating if a person resides in the treatment area of Lagos versus the control areas of Kaduna and Kano, and an interaction between the two. The effect of the interaction terms is the Average Treatment Effect on the Treated (ATT). More explicitly, this is a measure of the mean impact of MTV Shuga Naija on the population in areas that were exposed to it.

We make use of the panel nature of the data set to estimate a fixed effects regression model for each outcome related to experience of sexual violence and for attitudes towards victims of sexual violence. These models capture differences between treatment and control groups through the group fixed effect and control for any temporal factors that affect both treatment and control groups equally through the time fixed effect. Because these models also contain an individual fixed effect, any time invariant variable for an individual drops out of the model. This means that the binary variable for living in the treatment or comparison area drops out, although the interaction with the binary survey wave variable remains.

Because we are concerned about the validity of the parallel trends assumption, we use a method that combines the fixed-effects model with inverse probability weighting (IPW). This approach adjusts for potential confounding by weighting observations by the inverse of their probability of receiving the treatment based on their surveyed characteristics (Joffe et al., 2004). The resulting estimator, known as the doubly robust (DR) estimator, (Callaway, 2021; Lunceford & Davidian, 2004) is robust to misspecification of either the outcome (fixed effects) regression model or the IPW model. The DR model in effect produces a weighted combination of the estimates from the outcome and IPW models, with the weights reflecting the degree of overlap between the treatment and control groups. The DR estimator consistently estimates the treatment effect parameters provided either the regression or the propensity score is correctly specified.

While the dependent variables in each case are binary, we estimate the models using a linear fixed effects regression model in order to avoid the incidental parameters problem inherent in non-linear fixed effects models (Chamberlain, 1980). This problem arises because nonlinear fixed effects models treat individual fixed effects as parameters to be estimated, and, when the number of time periods is small, these models may produce inconsistent estimates of the individual fixed effects, in turn leading to inconsistent estimates of other parameters in the model. Consequently, estimation of the ATT using standard statistical software like Stata is not straightforward,

By using the linear fixed effects model, we are able to directly identify the ATT, the primary focus of our study, as the coefficient on the interaction term for treatment area and survey wave. However, we are unable to estimate average predicted probabilities for observed values of covariates (e.g., the predicted outcome for those receiving the treatment versus those not receiving the treatment) because the individual-specific effect, the intercept for each individual across all time periods, is differenced out of the model through the within transformation or “demeaning” of the data inherent to fixed effects models. Once these individual-specific effects are differenced out in the fixed effects models, we can ascertain changes in the outcome variable from changes in individual covariates but not predictions of the outcome variable for different values of covariates (Wooldridge, 2002). Consequently, we present only the average treatment effects from the DID models.

All models were estimated in Stata 18.0 using the drdid command developed by Sant’Anna (2020). This estimator implements the doubly robust difference-in-differences estimator for the average treatment effect with the inverse probability of tilting (IPT) and weighted least squares (WLS).

We also sought to address the possible influence of changes in the sampling procedure associated with moving from face-to-face interviews, as was done for the baseline wave, to phone interviews, as was necessitated by COVID-19 in the endline wave. Our strategy involves adjusting our sample weights through an iterative proportional fitting procedure (Bergmann, 2011). Because the endline sample was restricted to respondents who had consented to being re-contacted by phone, it contained respondents with a different socioeconomic profile on average than the samples for the first two waves. The unweighted endline sample, for example, was found to be more educated on average; 27.7% of females and 34.5% of males in the endline had post-secondary education as compared with 11.8% and 12.5% respectively at midline.

Using the iterative proportional fitting procedure (also known as “survey raking”), we adjusted our sample weights for endline respondents so that the demographic profile matched that of the baseline sample in terms of gender, state, education and marital status. Sample weights for the endline sample were iteratively adjusted to minimize the discrepancies with the known population distribution of characteristics of the baseline sample until the differences between the two samples converged to pre-specified tolerance level.² Survey raking was conducted using the ipfraking and ipfweight commands in Stata 17 (Kolenikov, 2014, 2019).

Descriptive Results

The characteristics of the sample changed from baseline to endline, with most of the changes attributable to maturation and transitions to adulthood. All respondents aged by three years from baseline to endline; the share of the female sample that was aged 13-17 years decreased from 51% at baseline to 25% at endline in the treatment area (Lagos) and from 42% to 18% in the comparison area (Kaduna and Kano). Similar patterns were observed in the male sample.

The female sample in the treatment areas had higher educational attainment at both baseline and endline relative to the comparison sample. At both baseline and endline, females in Lagos were approximately 14 percentage points more likely to have completed secondary education relative to females in the comparison areas (Treatment: 85% at baseline, 81% at endline; Comparison: 72% at baseline, 67% at endline). For males, educational attainment was higher in the treatment area at baseline (75% of males with secondary education in the treatment area versus 59% in the comparison area) but by endline greater parity was evident (79% of males with secondary education in the treatment area versus 82% in the comparison areas).

Across all areas, few males, at both baseline and endline, reported that they were currently married, consistent with patterns of considerable later ages at marriage for males relative to females in Nigeria. Females in the comparison areas were approximately three times more likely to be married at endline than females in the treatment area. Relatedly, roughly one in four women in the comparison areas had a child by the endline versus only 8% of women in the treatment area. Less than 3% of men in the comparison areas reported having a child at endline, relative to 75 of men in the treatment areas.

At both baseline and endline, the majority of youth owned a cellphone. Women in the treatment area appear to have gotten a cellphone sooner than women in the comparison areas; 63% of women in Lagos owned a cellphone at baseline versus 51% of women in the comparison areas. By endline, roughly eight in ten women in both areas (treatment: 84%, comparison: 80%) owned a cellphone. Approximately equal percentages of males in treatment and comparison areas owned cellphones at both baseline and at endline.

The majority of youth report exposure to both radio and television, with television being the more commonly used than radio. At endline, women in treatment areas were slightly more likely to watch television than women in comparison areas (treatment: 98%, comparison: 90%) but equally likely to listen to the radio (treatment: 59%, comparison: 58%). Males were equally likely to watch television at endline (treatment: 96%, comparison: 95%), but males in the treatment area were more likely to listen to the radio (treatment: 80%, comparison: 63%).

Sexual Violence

From baseline to endline, there were significant changes in experiences of sexual harassment, attitudes towards victims of sexual violence, and the likelihood of discussing sexual violence with family members for both females and males (Tables 2a and 2b).

Reporting of experiences of sexual harassment increased across all categories for both females and males, potentially reflecting a growing awareness of the issue. Reporting of unwelcome attempts to draw participants into a discussion of sexual matters increased from 16.1% to 33.0% for females and 16.3% to 39.7% for males. Unwanted attempts to establish a sexual relationship also increased from 13.7% to 29.4% for females and 17.2% to 43.7% for males. Touching in a way that made participants feel uncomfortable rose from 14.0% to 37.5% for females and 22.7% to 53.7% for males.

Attitudes towards victims of sexual violence generally improved from baseline to endline. The belief that a woman is at least somewhat responsible for being raped while drunk decreased from 80.5% to 60.3% for females and 82.3% to 64.6% for males. The view that women are asking to be raped when wearing revealing clothing dropped from 86.3% to 64.7% for females and decreased from 86.7% to 81.4% for males. Even with these shifts, there remains a significant gender disparity in some beliefs. For example, the belief that women wearing revealing clothing are asking to be raped remains high among males.

The likelihood of discussing sexual violence with family members increased dramatically. Reporting of conversations with family about sexual violence rose from 31.2% to 70.0% for females and 9.0% to 38.5% for males. Mothers were the most common family members to engage in these discussions, with the percentage increasing from 73.7% to 82.9% for females and 42.1% to 56.4% for males. Discussions with fathers also increased, from 23.6% to 21.3% for females and 25.1% to 49.2% for males. These results coincide with a key MTV Shuga Naija theme: engaging parents, family members, and society at large in open conversations about sexual violence may contribute to improved awareness and understanding and ultimately help reduce the occurrence of sexual harassment and assault.

Difference-in-Differences Models

Overall, the multivariate DID results provide support for the effects of MTV Shuga Naija on attitudes towards sexual violence and disclosure of experiences of sexual harassment and sexual violence, but lesser support for increased communication with family members about sexual violence.

The program had dramatic and demonstrable effects on shifting harmful attitudes about sexual violence and victims, particularly among females, with notable decreases in victim-blaming attitudes in instances of rape under various scenarios, such as when a woman is drunk [ATT=-32.0pp, 95% CI: -43.4pp, -20.7pp], initiates kissing [ATT=-29.3pp, 95% CI: -38.9pp, -19.6pp], or does not fight back [ATT=-28.3pp, 95% CI: -40.2pp, -16.4pp], and attitudes justifying rape such as “when women where clothes that expose their bodies, they are asking to be raped” [ATT=-36.6pp, 95% CI: - 46.8pp, -26.3pp] or “if a guy is drunk, he can rape someone unintentionally” [ATT=-12.0pp, 95% CI: -23.0pp, -1.1pp]. The effects for males were less pronounced but still statistically significant for justification for rape when a woman is wearing revealing clothing [ATT=-31.8pp, 95% CI: -45.6pp, - 18.1pp] and for characterization of rape as requiring that a girl fight back [ATT=-19.3pp, 95% CI: - 35.6pp, -3.0pp]. These latter results may indicate that male attitudes are harder to shift or that different types of messages, sources of messages, or other interventions may be more effective with males.

In Lagos, both males and females reported a significant increase in disclosure to interviewers of experiences of sexual harassment, such as unwelcome attempts to discuss sex and offensive gesturing, relative to youth in the comparison states of Kaduna and Kano. For females, significant effects were observed for “unwanted attempts at a sexual relationship” [ATT=12.3pp, 95% CI: 1.1pp, 23.6pp], and “touched you in an uncomfortable way” [ATT=14.8pp, 95% CI: 3.2pp, 26.5pp]. For males, statistically significant effects were observed for “unwanted attempts at a sexual relationship” [ATT=18.5pp, 95% CI: 0.5pp, 36.4pp] and “gestured in an offensive way” [ATT=24.7pp, 95% CI: 8.1pp, 41.4pp]. These results suggest improved awareness and recognition of sexual harassment due to the program, leading to increased reporting.

A key objective ofMTV Shuga Naija was to remove the stigma around victims of sexual violence and to promote open dialogue so that attitudes could shift, norms surrounding appropriate sexual contact could be improved, and victims would be more willing to seek services post-experience. While dialogue with family members about sexual violence indeed improved in the treatment areas, the DID models found that improvements in communication in the comparison areas actually outpaced those in the treatment areas among females [ATT=-19.9pp, 95% CI: -32.1pp, -7.8pp]. Interpretation of this finding is challenging but may lie in the phrasing of the question (“Have your parents or family members ever talked with you about sexual violence?), which provides little information on who initiated the conversation or what the content of the conversation was. It is possible that such conversations might even have perpetuated or articulated some of the same victim-blaming stereotypes and attitudes as admonishments and warnings toward youth.

Limitations and Future Directions

It is important to note that this evaluation faced several challenges. In addition to selective attrition and the switch from face-to-face interviews to phone interviews, there may also have been issues related to differences in the cultures and values between the treatment and comparison areas. The populations of the northern states of Kaduna and Kano are predominantly Muslim, reflected in much lesser autonomy for females,aptterns of early marriage and early child-bearing, and lesser emphasis on schooling and work outside the home for females, in strong contrast to the more western-focused, cosmopolitan populations of the south, including Lagos. The original study design articulated the provision of identical content across all areas, albeit with staggered implementation between the south and the north. However, local governments in the north objected to the explicit content related to sex and sexual violence in MTV Shuga Naija programming and prohibited the airing of any material except that focused solely on reproductive health within the context of marriage. These differences in the population characteristics for the treatment and comparison groups may have implications for one of the key assumptions of this analysis – that treatment and control groups would have followed parallel trends in sexual violence attitudes in the absence of the program. While the doubly-robust DID estimation strategy sought to ameliorate the possible influence of these differences, it was simply not feasible to collect more than one pre-intervention round of data that would have permitted a more robust assessments of trends prior to the implementation of Shuga Naija activities. Hence an explicit test of the parallel trends assumption was simply not possible.

Second, related to attrition, the sample sizes of respondents, particularly of sexually active males, were quite small at endline. The study was originally powered under the assumption of lower rates of attrition, and as a result, the observed differences required for statistical significance for indicators dependent on sexual activity tended to be quite large. Hence, while Shuga may have shifted indicators in the desired direction, our final sample size may have been too small in some cases to observe statistically significant shifts (i.e., at a p-value of below .05).

Finally, starting in March 2020 – seven months before the endline survey – Lagos was placed under stringent COVID-19 lockdowns that were not implemented in the northern states of Kaduna and Kano. This potentially affected the ability of youth in Lagos to interact with others. Hence, changes in the rates of sexual harassment and sexual violence may partially reflect changes in mobility circumstances.

Conclusion

This evaluation provides evidence of MTV Shuga Naija’s effects on attitudes and behaviors related to sexual violence, stigmatization of victims of sexual violence, and open dialogue about addressing sexual violence. During the study period, more rapid attitudinal shifts towards sexual violence and victims of sexual violence were apparent among both male and female youth in the treatment group relative to the comparison group. These results highlight that MTV Shuga Naija was able to achieve important effects in a relatively short period of time. It is well established that it takes time for campaign viewers to progress from campaign exposure to the development of revised perceptions and, subsequently, from the attitudinal shifts to the development of revised behavioral patterns. By opening and normalizing dialogue about sexual violence among youth, MTV Shuga Naija has set the groundwork for subsequent behavior changes that can help reduce the problem of sexual violence.