Organizations using mass media for behavior change

    Our Report

    About this page

    GiveWell aims to find the best giving opportunities we can and recommend them to donors. We tend to put a lot of investigation into the organizations we find most promising, and de-prioritize others based on limited information. When we decide not to prioritize an organization, we try to create a brief writeup of our thoughts on that charity because we want to be as transparent as possible about our reasoning.

    The following write-up should be viewed in this context: it explains why we determined that (for the time being), we won't be prioritizing the organizations in question as potential top charities. This write-up should not be taken as a "negative rating" of the charities. Rather, it is our attempt to be as clear as possible about the process by which we came to our top recommendations.

    Published: October 2012

    Background

    We have considered organizations which aim to use mass media (e.g., radio or television programming, often soap operas) to promote and disseminate messages on contraceptive use, family planning, environmental conservation and other topics in developing countries.

    The three organizations we have considered are:

    In June 2011, we prioritized work on these organizations because of (a) the possibility that by reaching large numbers of people at relatively low cost their programs could be very cost-effective and (b) the existence of a large body of reports which evaluate the effectiveness of this approach.

    After conducting some initial research (details below), in August 2012, we deprioritized work on this cause, though we hope to return to it in the future.

    Details of our process

    We spoke with each of the three organizations and reviewed what we perceive to be the evaluations most likely to support the case for effectiveness of these types of programs. In order to locate studies, we searched the organizations' websites for evaluation and study citations and searched JSTOR and Google for search terms such as "soap operas for social change." We reviewed all evaluation that we found with the exception of a few sources1 which may contain relevant discussion, which we did not have access to at the time that we conducted our review.

    We found more than 20 studies on radio and television soap operas in developing countries. Most of these studies were non-experimental studies, which analyzed listeners vs. non-listeners or pre-broadcast to post-broadcast attitudes by surveying the target populations. We focused our efforts on two studies: one quasi-experimental study and one randomized controlled trial, which we discussed in depth below.

    Based on our review, we did not find the results from the two studies discussed below to be compelling evidence that the interventions in question are an effective method for changing behavior.

    Quasi-experimental study

    Twende Na Wakati was broadcast in Tanzania from 1993-1997. There was one area of Tanzania (called Dodoma) where the program was not broadcast from 1993-1995: the program administrators reached an agreement with Radio Tanzania broadcasters to not broadcast the program there, so as to use it for a comparison group. The choice of Dodoma as a comparison group was not made by random selection; it was the only area where the transmitter broadcast its own programs for part of the day.2 The program was broadcast in Dodoma from 1995-1997. There are two peer-reviewed papers on the Tanzania program.3 We present the results of one paper (on family planning) in the table below. We don’t present the second paper's results (on HIV prevention) but our assessment of it is the same as our assessment of the first paper.

    The results from the study on family planning are presented in the table below. The values represent changes in the treatment and control group areas from the baseline when the program began to follow-up two years later. There are also measurements recorded for the control group for two years following the program. One star indicates statistical significance with a p-value of less than .05; no stars indicates a lack of statistical significance at p less than .05, except where indicated by footnotes.4

    Metric Treatment group (1993-1995) Control group, without program (1993-1995) Control group, with program (1995-1997)
    Ability to name at least one family planning method unprompted5 7% 7% 7%
    Respondent's belief in ability to determine family size6 11%* 6%* 19%*
    Perceived ideal number of children7 -.3 -.4 -.18
    Perceived ideal age at marriage (for women)9 .9 years* .1 years .2 years10
    Self-reported approval of family planning methods11 3%* -6% 5%12
    Talking with spouse about family planning (self-reported)13 4%* 2%* 5%14
    Married womens’ self-reported family planning use (“always” or “sometimes”)15 10%* -11%16 16%*
    Currently pregnant 17 -3% 1% -7%18

    The control group experienced fewer positive changes than the treatment group in 1993-1995. When the program was later broadcast in the control group's area in 1995-1997, the control group experienced a level of change similar to (or in some cases, greater than) that of the treatment group in 1993-1995.

    The study also presents data from health centers in the treatment and comparison area showing a greater rate of new contraception adopters during the time period when the program was broadcast.19

    Assessment of the study:
    The study provides only weak evidence of a causal relation between the program and the observed changes, due to several potential sources of confounding factors:

    • Significant differences between the treatment and control groups: The comparison group was better off than the treatment area in a number of ways (both at baseline and at follow-up). For example, the baseline radio and electricity ownership and access to family planning were higher in the comparison area than in the treatment area.20 The study controls for some differences between the two groups. 21 However, we are not confident in the analysis; there may be confounding factors that are unmentioned in the study and uncontrolled for. This reduces our confidence that the greater level of changes in the treatment area compared to the control area in 1993-1995 was caused by the program rather than other factors.
    • Other, similar radio programs in both treatment and control areas: Examples of programs that were not controlled for in the study were two other similar radio programs (called Zinduka! and Afya ya Jamii) broadcasting nationally, including in Dodoma, during the same time period as Twende Na Wakati.22
    • National changes in family planning access during treatment period: In both the treatment and the control areas, the reported “access to family planning services in village of residence” increased significantly between 1993 and 1997 (from 45% to 78% in the treatment area and from 73% to 87% in the comparison area).23 This expansion of access to family planning services suggests that there could have been additional uncontrolled-for programs taking place. 24 Without further information on whether there were other programs beginning in the comparison and treatment areas within the relevant time periods (1993-1995 for the treatment group and 1995-1997 for the comparison area) it is not possible to be confident that the radio program caused the observed changes.

    Randomized controlled trial (RCT)

    We found one RCT25 involving family planning and HIV prevention entertainment broadcasts. The program is very different from radio-alone programs such as those run by Population Media Center, PCI-Media Impact and Development Media International. In addition to radio programs, it involved written materials, videos, small group discussions, home visits, individual counseling, and a free supply of condoms26 and also the radio programs were not entertainment but instead were informational broadcasts.27 The study found a statistically significant change both in the participants' report of condom use during the last episode of sexual intercourse (from 9% at baseline to 14% at follow-up), and in reports of condoms as a main contraceptive method (from 4% to 10%). There was no statistically significant change in these reports from control subjects.28

    Future analysis

    Were we to continue our analysis, we would look more closely at all of the relevant evaluations, focusing on those which survey the target population about their health behaviors (and related knowledge) as well as evaluation that includes metrics on service use statistics (such as new contraceptive users as reported by health clinics) before and after the programs.

    Note: Development Media International (DMI) is currently running a randomized controlled trial to evaluate its program in Burkina Faso. Midline results are expected in 2013, final results in 2015.29

    Sources

    • Development Media International. How do we prove impact? http://developmentmedia.net/how-do-we-prove-impact (accessed August 20, 2012). Archived by WebCite® at http://www.webcitation.org/6A3xQbBef.
    • GiveWell. Notes from conversation with Roy Head, CEO of Development Media International (DMI), Will Snell, Head of Public Engagement & Development (DMI), and Cathryn Wood, Public Engagement and Innovation Manager (DMI) (April 17, 2012). (DOC).
    • Nariman, H.N. 1993. Soap Operas for Social Change: Toward a Methodology for Entertainment-Education Television. Connecticut: Praeger.
    • Rider, Bill and Paul Ugalde. Finance Director and Director of Development of Population Media Center. Phone conversation with GiveWell (DOC), July 19, 2011.
    • Rogers et al. 1999. Effects of an Entertainment-Education Radio Soap Opera on Family Planning Behavior in Tanzania. Studies in Family Planning 30: 193-211.
    • Sabido, Miguel. 1981. Towards the social use of commercial television: Mexico’s experience with the reinforcement of social values through TV soap operas. Paper presented at the annual conference of the International Institute of Communications (Strasbourg, France). Mexico City, Mexico: Institute for Communications Research.
    • Singhal, A. and E.M. Rogers. 1999. Entertainment-Education: A Communication Strategy for Social Change. Mahwah, New Jersey: Lawrence Erlbaum Associates.
    • Singhal, A., M. Cody, E. M. Rogers, and M. Sabido. 2003. Entertainment-Education and Social Change: History, Research and Practice. Mahwah, New Jersey: Lawrence Erlbaum Associates.
    • Southey, Sean and Lindsey Wahlstrom. Executive Director and Communications Officer of PCI-Media Impact. Phone conversation with GiveWell (DOC), July 28, 2011.
    • Vaughan, P., et al. 2000. Entertainment-education and HIV/AIDS prevention: A field experiment in Tanzania. Journal of Health Communication 5: 81-100.
    • Vaughan, P., and Everett M. Rogers. 2000. A staged model of communication effects: evidence from an entertainment–education radio soap opera in Tanzania. Journal of Health Communication 5: 203-227.
    • Westoff, C. & A. Bankole 1997. Mass Media and Reproductive Behavior in Africa. Demographic and Health Surveys Analytical Reports No.2. Calverton, Maryland: Macro International Inc.
    • Xiaoming, S. et al. 2000. Integrating HIV prevention education into existing family planning services: results of a controlled trial of a community-level intervention for young adults in rural China. AIDS and Behavior 4 (1): 103-110. Summary available at http://www.springerlink.com/content/x541813l1075224u/ (accessed May 4, 2012). Archived by WebCite® at http://www.webcitation.org/67PqIU5ei.
    • 1Singhal et al. 2003, Singhal and Rogers 1999, Westoff and Bankole 1997, Nariman 1993, Sabido 1981.
    • 2 “By prior agreement between the MOH, RTD, and the University of New Mexico (UNM), Twende na Wakati was not broadcast by the Dodoma regional transmitter from 1993 to 1995 to create a comparison area for our effects research [footnote 6]…. The Dodoma broadcast area was chosen opportunistically, rather than randomly, because it was the only transmitter that broadcast its own programs for part of the day. Therefore, our design is a quasi-experiment." Vaughan et al., “Entertainment-education and HIV/AIDS prevention: A field experiment in Tanzania.“ Pg 87.
    • 3 See Rogers et al, “Effects of an entertainment-education radio soap opera on family planning behavior in Tanzania.” and Vaughan et al., “Entertainment-education and HIV/AIDS prevention: A field experiment in Tanzania.”
    • 4The paper presents statistical significance calculated according to 5 different models. We have noted where the results were statistically significant according to at least one of them. Several of the models control for a variety of factors (including access to family planning, income, etc) as noted in the paper. Rogers et al, “Effects of an entertainment-education radio soap opera on family planning behavior in Tanzania,” Pg 200-201.
    • 5“Knowledge of family planning (measured as the respondents' ability to name, unprompted, at least one family planning method) was already high prior to the soap-opera broadcasts. An increase of seven percentage points occurred in both areas from 1993-95 and also from 1995-97 (as shown in Table 3). The three statistical tests for an effect of "Twende na Wakati" were not significant (see Table 4). Exposure to "Twende na Wakati" appears to have had no effect on awareness of or knowledge about family planning methods, but such awareness and knowledge apparently were increasing generally throughout the study area during 1993-97.” Rogers et al, “Effects of an entertainment-education radio soap opera on family planning behavior in Tanzania,” Pg 201-202.
    • 6“A measure of self-efficacy, as indicated by the respondent's belief in his or her ability to determine his or her family size, increased 11 percentage points in the treatment area versus six percentage points in the comparison area from 1993 to 1995 (see Table 3). All three statistical tests support a significant "Twende na Wakati" effect on this self-efficacy variable (as shown in Table 4). Furthermore, this effect was replicated in the Dodoma comparison area where belief in the ability to determine one's family size increased by 19 percentage points from 1995 to 1997. In the Model 5 analyses, exposure to "Zinduka" was also significant (Beta = 0.23, p less than 0.05), suggesting that this radio program contributed to the six percentage point increase in the comparison area from 1993 to 1995.” Rogers et al, “Effects of an entertainment-education radio soap opera on family planning behavior in Tanzania,” Pg 201-202.
    • 7 “The respondents' ideal number of children decreased by 0.3 children in the treatment area and by 0.4 children in the comparison area from 1993 to 1995 (indicated in Table 3). As Table 4 shows, none of the statistical tests indicate a significant "Twende na Wakati" effect on ideal family size between 1993 and 1995. Thus, the declining ideal family size in the study area was not related to the "Twende na Wakati" broadcasts.” Rogers et al, “Effects of an entertainment-education radio soap opera on family planning behavior in Tanzania,” Pg 201-202.
    • 8Calculated using the figures provided in Table 3, Rogers et al, “Effects of an entertainment-education radio soap opera on family planning behavior in Tanzania,” Pg 201-202. We have not calculated whether this change is statistically significant.
    • 9Ideal age at marriage for women increased 0.9 years in the treatment area versus 0.1 years in the comparison area from 1993 to 1995 (see Table 3). All three statistical tests support a significant "Twende na Wakati" effect from 1993 to 1995 (shown in Table 4). However, the effect was not replicated in the Dodoma comparison area from 1995 to 1997, perhaps because of the higher base level of this variable in the Dodoma area in 1995.” Rogers et al, “Effects of an entertainment-education radio soap opera on family planning behavior in Tanzania,” Pg 201-202.
    • 10Calculated using the figures provided in Table 3, Rogers et al, “Effects of an entertainment-education radio soap opera on family planning behavior in Tanzania,” Pg 201-202. We have not calculated whether this change is statistically significant.
    • 11 Approval of family planning increased by three percentage points in the treatment area and decreased by six percentage points in the comparison area from 1993 to 1995 (indicated in Table 3). All three statistical tests support a significant "Twende na Wakati" effect from 1993 to 1995 on approval of family planning (shown in Table 4). Finally, the effect of "Twende na Wakati" is replicated in the Dodoma area from 1995 to 1997, where approval of family planning increased by five percent-age points.” Rogers et al, “Effects of an entertainment-education radio soap opera on family planning behavior in Tanzania,” Pg 201-202.
    • 12The paper does not note whether this change was statistically significant.
    • 13Contraceptive practice was measured with three variables: (1) As Table 3 shows, the proportion of married respondents who reported that they talked with their spouses about family planning increased by four percentage points in the treatment and two percentage points in the comparison area from 1993 to 1995. The effects in Model 1 are not significant, but both the multiple logistic regression and multiple linear regression tests support a significant effect of "Twende na Wakati" between 1993 and 1995 on this variable. This effect was replicated in the Dodoma comparison area from 1995 to 1997; spousal discussion of family planning increased by five percentage points over this period." Rogers et al, “Effects of an entertainment-education radio soap opera on family planning behavior in Tanzania,” Pg 201-202.
    • 14The paper does not note whether this change was statistically significant.
    • 15 "The proportion of married women who practice family planning ("always use" and "sometimes use" are combined) increased by ten percentage points in the treatment area and decreased by 11 percentage points in the comparison area from 1993 to 1995, as Table 3 indicates. All three statistical tests support a significant effect of exposure to "Twende na Wakati" from 1993 to 1995 on married women's use of contraceptives (shown in Table 4). This effect was replicated in the Dodoma comparison area from 1995 to 1997, where the proportion of respondents reporting such use increased by 16 percentage points.” Rogers et al, “Effects of an entertainment-education radio soap opera on family planning behavior in Tanzania,” Pg 201-202.
    • 16Calculated using the figures provided in Table 3, Rogers et al, “Effects of an entertainment-education radio soap opera on family planning behavior in Tanzania,” Pg 201-202. We have not calculated whether this change is statistically significant.
    • 17 “The proportion of married women who were currently pregnant decreased by three percentage points in the treatment area and increased by one percentage point in the Dodoma comparison area from 1993 to 1995 (see Table 3). None of the statistical tests supports a significant effect of exposure to "Twende na Wakati" on current pregnancy.” Rogers et al, “Effects of an entertainment-education radio soap opera on family planning behavior in Tanzania,” Pg 201-202.
    • 18Calculated using the figures provided in Table 3, Rogers et al, “Effects of an entertainment-education radio soap opera on family planning behavior in Tanzania,” Pg 201-202. We have not calculated whether this change is statistically significant.
    • 19” Figure 3 shows the mean number of new and continuing contraceptive adopters at 43 Ministry of Health clinics in the treatment area and at 27 clinics in the comparison area from January 1990 through December 1996.20 The trend lines (computed in the linear regression analysis) for the treatment and comparison areas increased at approximately the same rate during the pre-broadcast period from January 1990 to June 1993. The treatment slope is 0.7 (standard error [S.E.] = 0.05), and the comparison area slope is 0.5 (S.E. = 0.30); the 95 percent confidence intervals for the two trend lines nearly always overlap during this preintervention period. From July 1993 through September 1995, when the radio soap opera was broadcast in the treatment area only, the slope in the treatment area increased to 1.6 (S.E. = 0.21), whereas the slope in the comparison area in-creased only slightly, to 0.6 (S.E. = 0.11). During this field experimental period, the 95 percent confidence-interval bars do not overlap for most months. After September 1995, the trend line in the treatment area reaches a plateau, whereas the slope in the former comparison area increases to 1.2 (S.E. = 0.45), and the 95 percent confidence-interval bars overlap for most months, as the rate of adoption in the former comparison area catches up with that of the treatment area. This pattern supports the hypothesis that "Twende na Wakati" stimulated an increase in the rate of family planning adoption when it was broadcast first in the treatment area and later in the comparison area after 1995.” Rogers et al, “Effects of an entertainment-education radio soap opera on family planning behavior in Tanzania,” Pg 203-204.
    • 20"The comparison-area respondents were more likely to have central-station electricity, were slightly more educated, and were much more likely to live in a village in which family planning services were available. They were also more likely to own a radio and be exposed to at least one other radio program with family planning content (see Table 2). Independent variables including electricity, income, family planning availability, radio ownership, and exposure to family planning radio programs all increased during the period of study in both the treatment and comparison areas. The respondents in the comparison and treatment areas differed initially in terms of several of the dependent variables (see discussion of Table 3 below), with the former having higher means and frequencies for these variables." Rogers et al, “Effects of an entertainment-education radio soap opera on family planning behavior in Tanzania,” Pg 198.
    • 21
      • "Table 4 indicates (1) the results of the logit loglinear (Model 1) or ANOVA (Model 2) tests of significance for the treatment effect in bivariate analysis (treatment versus comparison analysis); (2) the results of the logistic regression (Model 3) or multiway test (Model 4) of significance for the treatment effect in multi-factor ANOVA, controlling for the ten independent variables shown in Table 1 [Percent male, mean age, Catholic, urban resident, number of children, electricity at home, education, mean income, access to family planning services in village of residence, owns a radio]; and (3) the results of the multiple linear regression tests (Model 5) of significance for a "Twende na Wakati" exposure effect, controlling for 20 independent variables (at the ward level of analysis)." Rogers et al, “Effects of an entertainment-education radio soap opera on family planning behavior in Tanzania,” Pg 200-201.
      • For Model 5, "These control variables include number of respondents who listened to "Zinduka"in 1995; number who listened to a radio talk show about health, "Afya ya Jamii" (Health and the Public), in 1995; number who listened to other family planning radio programs in 1995; respondents' access to local family planning services in 1993 and the change in such access from 1993 to 1995 (documented separately by POFLEP researchers in each village of study); the change in respondents' ages from 1993 to 1995; respondents' religion in 1993 and the change in respondents' religion from 1993 to 1995; radio ownership in 1993 and the change in radio ownership from 1993 to 1995; the change in the number of respondents of each gender between the 1993 and 1995 samples; the change in respondents' marital status between the 1993 and 1995 samples; the change in respondents' parity between the 1993 and 1995 samples; respondents' income in 1993 and the change in their income from 1993 to 1995; number of respondents' households with electricity in 1993 and the change in number of households with electricity from 1993 to 1995; respondents' formal education in 1993 and the change in their formal education from 1993 to 1995; and respondents' urban-rural residence in 1993. Both the absolute levels of these control variables in 1993 (or 1995) and the change in these control variables from 1993 to 1995 were used to control for ward-level differences in development at baseline, differences in development rates over the period of study, and sampling error." Rogers et al, “Effects of an entertainment-education radio soap opera on family planning behavior in Tanzania,” footnote 7.

    • 22
      • "Zinduka," another radio soap-opera broadcast from October 1993 to 1996 using the entertainment-education strategy, promoted similar content, including spousal communication and the economic benefits of using family planning.” Rogers et al, “Effects of an entertainment-education radio soap opera on family planning behavior in Tanzania,” footnote 11.
      • “Several other radio programs with HIV/AIDS educational content were broadcast during part or all of our five-year study period. These programs, Afya ya Jamii (Health Facts for Life) and Zinduka (Awake!) were broadcast nationwide, and listenership to both radio programs was somewhat higher in our comparison area than in the treatment area, consistent with the relatively higher radio ownership in the comparison area.” Vaughan, "Entertainment-education and HIV/AIDS prevention: A field experiment in Tanzania." Pg 89.
      • "These tests [Models 1 and 2] do not control for other independent variables that might influence results... [Models 3 and 4] do not control...for exposure to other family planning radio programs..." Rogers et al, “Effects of an entertainment-education radio soap opera on family planning behavior in Tanzania,” Pg 197.

    • 23 Rogers et al, “Effects of an entertainment-education radio soap opera on family planning behavior in Tanzania,” Pg 199.
    • 24 “Could the increased use of family planning methods in Tanzania have been due to the implementation of the 1992 population policy and the expansion of women's access to family planning clinics that took place in the early 1990s? These events occurred nationally and should have had nationwide effects, whereas greater increases in contraceptive use were observed in the study's treatment area from 1993 to 1995.” Rogers et al, “Effects of an entertainment-education radio soap opera on family planning behavior in Tanzania,” Pg 206.
    • 25

      Xiaoming et al, “Integrating HIV Prevention Education into Existing Family Planning Services: Results of a Controlled Trial of a Community-Level Intervention for Young Adults in Rural China."

    • 26 “Young adults in China are at increasing risk for HIV infection, yet there have been few prevention programs to address this growing public health problem. In 1996–1997, we conducted a community-based AIDS prevention intervention in semirural Kunshan county near Shanghai, which was integrated into an existing family planning network. We recruited 748 study participants aged 18–30 years from two townships similar in their socioeconomic characteristics. The experimental township received a multifaceted 12-month intervention that included written materials, videos, radio programs, small group discussions, home visits, individual counseling, and a free supply of condoms. Data from both townships were collected at baseline and at 12-month follow-up. Among intervention participants, AIDS-related knowledge and attitudes improved significantly. Intervention participants also reported an increase in condom use during the last episode of sexual intercourse from 9% at baseline to 14% at follow-up, while those using condoms as their main contraceptive method increased from 4% to 10% (both with a p-value less than .05); no such increases occurred among control subjects. This study demonstrates the feasibility of integrating an HIV prevention intervention into an existing family planning services network. Integrating services in this manner may be an approach to a sustainable AIDS intervention in similar settings where reproductive health services are already established.” Xiaoming et al, “Integrating HIV Prevention Education into Existing Family Planning Services: Results of a Controlled Trial of a Community-Level Intervention for Young Adults in Rural China,” abstract.
    • 27 “Thirty-minute audiotapes on HIV/STD prevention were broadcast once a week, usually during suppertime, through the community cable-speaker network.” Xiaoming et al, “Integrating HIV Prevention Education into Existing Family Planning Services: Results of a Controlled Trial of a Community-Level Intervention for Young Adults in Rural China,” Pg 106.
    • 28 “Among intervention participants, AIDS-related knowledge and attitudes improved significantly. Intervention participants also reported an increase in condom use during the last episode of sexual intercourse from 9% at baseline to 14% at follow-up, while those using condoms as their main contraceptive method increased from 4% to 10%; no such increases occurred among control subjects." Xiaoming et al, “Integrating HIV Prevention Education into Existing Family Planning Services: Results of a Controlled Trial of a Community-Level Intervention for Young Adults in Rural China,” abstract.
    • 29 Development Media International, "How Do We Prove Impact?"

    Two responses below:

    From Peter W. Vaughan, Ph.D., Arvind Singhal, Ph.D., and Scott Connolly, Ed.D.

    Elie Hassenfeld
    Co-Executive Director

    Stephanie Wykstra, Ph.D.
    Research Analyst

    October 23, 2012

    Hi Elie and Stephanie,

    We appreciate this opportunity to respond to GiveWell’s review of research on the impacts of entertainment-education (EE) programs. We would like to use this response to make two basic points. First, we would like to clarify what the logistical and ethical challenges are to implementing a randomized controlled trial (RCT) for public health communication programs to help you understand why no such studies have been published to date. Second, we wish to make sure that GiveWell fully appreciates the strengths of the research on EE that has been conducted in addition to understanding any weaknesses or limitations that inevitably exist in any one individual study.

    By way of disclosure, we all have professional relationships with nonprofits that utilize EE. Peter is a research consultant, coauthor of the papers you reviewed from Tanzania, and has had paid consulting relationships since 1993 with Population Communications International (now PCI-Media Impact) and Rare, another NGO that uses EE. Arvind was a coauthor on one of the Tanzania papers, has studied several other PCI-MI programs, and is also currently a volunteer board member for PCI-MI. Scott is currently the Research Director at Population Media Center (PMC).

    We applaud GiveWell for asking the question about EE efficacy and expending time and resources to undertake a literature review. It has not always been the case that funders have required solid evidence of program impact, and we agree with your premise that non-profit organizations should be held responsible for demonstrating the effectiveness of their programs to make sure each program is meeting its pro-social goals. Such research also enables them to be transparent to their donors and supporters about what exactly they can accomplish with the all-to-scarce human and capital resources that they have. We are very proud to have worked with PCI-MI, PMC and Rare over many years because they have, in our opinion, been the “early adopters” of impact research. Collectively, we have more than 60 years of experience in working with and evaluating EE programs and/or public health.

    We are concerned that the GiveWell review may have missed some of the strengths of the studies you reviewed and not reviewed other studies that we feel bear on this question, and therefore reached an overly negative assessment of the current state of EE impact research. To draw an analogy from statistics, there are two types of statistical error. Type 1 error is the conclusion that an intervention has an effect when it in fact does not; detecting and guarding against a Type 1 error seems to be the primary focus of the GiveWell review. However, it is equally important to consider the consequences of making a Type 2 error, or drawing the conclusion that a program has no effect when in fact it does. No single study can eliminate both types of error, and judgment must be used to balance between these two error risks because attempts to minimize one type of error can cause an increased risk of making the other type of error. We feel this may be the jeopardy of GiveWell’s review as it stands now.

    As your review found, there has been quite a lot of research on EE, but none of it published to date meets the “gold-standard” of a randomized controlled trial (RCT). GiveWell is correct to identify the RCT as the gold-standard for demonstrating impact and especially for attribution of causation (http://blog.givewell.org/2012/08/23/how-we-evaluate-a-study/. No respectable researcher would argue this point. However, we encourage you to recognize that some public health interventions are more amenable to implementation of the RCT approach than others. Your website acknowledges that RCT is not always possible for political, financial or logistical reasons. We feel that you should understand how these issues play out for EE evaluations, and also fully appreciate the ethical considerations for any intervention that deals with public health measures.

    The power of RCT, as you know, derives from the randomization of treatment and control groups. In the EE context, this would mean that the NGO would give up control over which regions received its potentially life-saving programming and which received no program (or better a placebo program). In our experience, NGOs pay careful attention to selecting the areas where they work so that they know they have the best chance of having a large impact. Sites are selected for reasons of need of the population, capacity of local partners, infrastructure to support the programming, and many other factors depending on the group and the program. Working in priority areas is in and of itself a sign of fiscal responsibility that funders do and should demand of the NGOs they support. To do otherwise would be an irresponsible use of resources and pose serious concerns about the ethics of such an opportunity cost. And yet, this is precisely what implementing an RCT study demands. Apparently, DMI has randomized treatment and control sites in their Burkina Faso study of their communication intervention’s impact on child mortality (http://developmentmedia.net/how-do-we-prove-impact), but perhaps they were convinced that all 14 areas in their study were equivalent on key factors. It is, however, our experience that there is often much regional variation in need, demographics, SES indicators, partner capacity, and the other factors.

    There are further serious ethical considerations of using RCT to evaluate public health campaigns. The World Medical Association’s Helsinki Declaration (http://www.wma.net/en/30publications/10policies/b3/) provides guidance to ensure ethical medical testing, and we feel that there are logical extensions for public health research. Explicit in experimental designs is the fact that some people get the potential life-saving treatment/information and some people don’t. In a typical drug trial, these ethical considerations are ameliorated by:

    • Ignorance of whether the new drug is effective or not; it is entirely possible that the new drug is ineffective, or even negatively effective. At the time of the experiment nobody, including the researchers, knows for sure. Hence the reason to do the study and why it is ethical to give someone a placebo or no treatment.
    • Individual patients sign an informed consent form indicating they have been told what the study entails and that they might get the placebo rather than the experimental drug. Study participants are told and understand that they might suffer serious negative effects because of participating in the study.
    • Lack of an existing, proven alternative treatment that could be used instead of the experimental treatment.

    Now, consider the first two issues from the EE perspective:

    • Ignorance: The crux of GiveWell’s current review of EE is that in 2012 we remain in a state of ignorance about the demonstrated impacts of EE. At one time, all three of us would have agreed with this assessment, and it has been a compelling reason for why we have engaged in the EE research that we have done over the years. By way of example, at the time of the Tanzania study you reviewed, while there had been a number of published studies of EE impact, none had been able to attribute causation. This “ignorance” allowed a rationale for using a “comparison area” where the EE radio drama treatment was withheld. Because the Twende na Wakati (TnW) radio program dealt with important reproductive health issues, including HIV prevention and family planning, this withholding of the program’s information may have resulted in negative outcomes for some people living in the comparison area. The research team was acutely aware of this and held a review meeting after 1 year of TnW broadcasts to review results measured to date to determine whether there was sufficient evidence that we should close the experiment down early. This review included study investigators from the University of New Mexico, PCI representatives, as well as outside experts, including the well-known demographer, Charlie Westoff of Princeton University. However, the results at that time were not yet conclusive (not many dependent variables were statistically significant after one year of TnW broadcasts), so we continued with the original design of broadcasting for two years in the treatment area and withholding broadcasts for two years in the comparison area. After year two, the amount of change for many dependent variables was large enough to be statistically significant and this situation changed. The program was then broadcast in its entirety in the former comparison area for both (1) ethical and (2) program impact measurement reasons (elaborated below). It is possible that not all researchers would reach the same conclusion of causal impact that we did from the Tanzania study, as in fact your review does not, but they face a high ethical bar in proposing additional RCT of EE health programs, because we feel we are no longer in a state of ignorance. DMI is apparently aware of this ethical concern, as they make clear on their website that they are attempting to demonstrate something new and are therefore in a state of ignorance “The impact of well-designed media campaigns on attitudes and behavior has been conclusively demonstrated. But there has never been any attempt to measure or even to model the impact of campaigns on mortality.” (http://developmentmedia.net/how-we-calculate-impact).
    • Informed consent: TnW (as many EE programs are) was broadcast at a national level. It was clearly impossible for the research team to receive informed consent from all possible individuals in the broadcast and comparison regions of Tanzania. We did obtain consent for the study from the relevant government ministries, which was the best we could do under the circumstances. Such governmental consent is always an option, but of course, some governments are more representative than others.

    In a good RCT, a large group of subjects who might benefit from the treatment are identified, and then they are randomly assigned to treatment and control groups. Those in the treatment group will be given a carefully proscribed dose of the treatment. The underlying assumption of this is that the samples are large enough so that “uncontrolled” differences among participants will be equalized between the two treatment groups by chance; this eliminates (or reduces) so-called “sampling error”. This is relatively easy to do when the treatment is applied at the level of an individual, but EE is applied at the level of a geographic region. In Tanzania, this was especially hard because each radio transmitter covered quite a large region (risking contamination if the treatment/comparison regions were too close to each other), so we ended up with a small sample of treatment areas (N = 6) and only 1 comparison area. The DMI strategy seems promising, in which they are broadcasting their radio programs on low-power FM transmitters that cover only small regions and have sample sizes of 7 for both treatment and control; hopefully those sample sizes will prove to be adequate (http://developmentmedia.net/how-do-we-prove-impact). They will, however, still be faced with the challenge of self-recruitment to the EE program, which results in individuals in the treatment area with many different levels of exposure to their program, some with none, some with intermediate, and some with complete exposure; not everyone in the treatment area receives equal treatment. We feel this is an unavoidable complication that will generally result in a conservative estimate of program impact because some people in the treatment area never, in fact, receive the treatment.

    What we were able to implement in Tanzania would best be described as a “nonequivalent control group, field quasi-experiment.” How commonly is this accomplished? Three years ago, one of the present authors was part of a group that applied for a grant to study a cohort of Rare’s Pride programs in Latin America. We proposed to the Global Environment Facility (GEF) of the United Nations a design similar to what we used in Tanzania. Here is a portion of the Scientific Technical Advisory Panel’s (STAP) review of our initial proposal.

    2. The STAP welcomes this innovative proposal by UNEP and partners. In particular, STAP congratulates the project proponents for taking seriously the need to test carefully the effectiveness of its interventions. The use of clear theories of change and control groups are critical project components that are virtually absent in the GEF investment portfolio. STAP strongly supports this project on scientific grounds, but has a few comments that we would like to see addressed in the PIF before CEO endorsement.

    What the STAP is saying is that among the GEF’s hundreds, if not thousands, of funded environmental programs (all types of environmental interventions, not just EE programs), our proposal, nearly identical in its rigor to the TnW study except we had 11 treatment and 11 comparison sites, was one of the only they had ever received that included both a comparison area and a theoretical model. We spent several months working with GEF and Rare’s implementing partners to improve the design to include a randomization of sites, and again, for the reasons outlined above, found it to not be possible. Most importantly, we found (1) the implementing partners were unwilling to expend their resources implementing Pride campaigns in non-prioritized regions, and (2) the number of replicate areas required (GEF felt we needed more than the 11 Pride campaigns and 11 comparison sites we had funded) to get large sample sizes to overcome sampling error was simply prohibitive. This implies to us that GiveWell may be insisting on a standard that simply cannot be met in many development/environmental programs. GiveWell runs a serious risk of rejecting, or failing to recommend, many programs unless you are willing to accept the results from less than gold-standard studies.

    In addition to research design, as the STAP points out, another critical factor in attributing causation is for the intervention to have an explicit and coherent theory of how the intervention is expected to cause the change that can be tested. This is a strength of all EE programs, but especially of the long-running drama programs, which are built on a foundation of social science theory, most importantly social cognitive theory, which posits that people learn new behaviors by observing others (including mass media characters) to use as role models of alternative behaviors and to understand their consequences (Bandura, 1994; Vaughan and Rogers, 2000).

    We would like to respond to some of the GiveWell critiques of the Tanzania study and point out what we see as some of the study’s strengths that may have been underappreciated. Also, we would also like to highlight a number of other studies that we feel were well done and lend credibility to the hypothesis of EE impact that we would encourage you to consider in your overall evaluation of the effectiveness of EE.

    GiveWell’s review points out that Dodoma was not randomly selected to be the comparison area and that it differed substantially in demographic, SES, and other variables from the treatment areas. This is true, but was unavoidable due to the structure of Radio Tanzania’s broadcasting. As your review notes, we did use multivariate statistics to control for measured variation in these variables. This is a widely accepted strategy that increases confidence that the results are valid, but cannot completely eliminate the possibility that unmeasured variation is the cause of differences in the amount of change between treatment and control. This “nonequivalence” of the comparison area does pose a threat to the internal validity of the study’s findings in that perhaps something was different in Dodoma that precluded our measuring change there from 1993 to 1995. However, this validity threat is mitigated by the fact that for many of our dependent variables of study, in both survey and clinic data, we did measure changes similar in direction and magnitude in Dodoma once TnW was broadcast there from 1995 to 1997. We see this as strong evidence of a TnW causal impact on the changes simply because if there were unmeasured factors in Dodoma that precluded dependent variable change from 1993 to 1995, those other factors would also have to have changed in 1995, which we find implausible. This design is analogous to a placebo replacement, and we believe allows a causal inference to be drawn.

    GiveWell’s review correctly points out that there were two other nationally-broadcast radio programs with reproductive health content that were being broadcast during at least part of the time of the study (Zinduka produced by John’s Hopkins Center for Communication Programs and Afya ya Jamii produced by the Ministry of Health). It was not possible for the TnW study to change the broadcast patterns of these other radio programs in order to control for them. However, because these other radio programs were broadcast nationally, one would reasonable expect them to have had equivalent impacts in our treatment and comparison areas. Further, listenership to them was measured in some of the surveys and these two listenership variables were entered into the multivariate analyses for Model 5 in the paper (footnote 7) and in our cross-sectional analysis of independent survey data conducted by the Tanzania Demographic and Health Survey (TDHS) that asked about all three radio programs in 1996. Where these radio program listenership variables remained in the model, it is pointed out in the text and footnotes. GiveWell’s review is aware of this (see for example your footnote 6), but seems to discount the widely accepted strategy for statistically controlling for such variables. We anticipate that the DMI study will have similar issues, as there must be many local, regional, and national health initiatives ongoing in Burkina Faso at any given moment over which DMI has no control and for which a sample size of 7 may not be large enough to normalize the differences among treatment groups.1

    GiveWell’s review correctly notes that there was an expansion of village-level access to family planning during the study of about 33 percentage points (pp) in the treatment areas and 14 pp in the comparison area, indicating that access was improving. However, this is precisely the reason to use a comparison area, to control for such factors. Ideally, we would have been able to match sites better, but both sites ended with similar, quite high access (78% in treatment and 87% in comparison). Again, if this were causing the observed changes, we would not expect to see changes in the Dodoma comparison area from 1995 to 1997.

    GiveWell notes, but only in passing, that the Tanzania paper presents an entirely independent set of metrics from reproductive health clinics; 43 clinics in the treatment area and 27 clinics in the comparison area) that shows statistically significant increases in the number of new family planning adopters/month in the treatment area (1993-95) which is replicated almost identically in the Dodoma area when TnW is broadcast there (1995-97). These clinic intake data are substantiated by interviews conducted by clinic staff with new adopters who reported TnW as their source of referral to the clinic. These data are very compelling because (1) they document behavior change at the point and time of behavioral adoption, (2) trends in treatment and comparison areas are identical over the 3-year period prior to broadcast, but clearly and statistically diverge in 1993 when broadcasts of TNW began in the treatment area, and (3) again, any concerns over differences between treatment and comparison areas are mitigated by seeing similar change in family planning numbers in the comparison area after 1995. Further, the growth in family planning access, noted above, is controlled because the same clinics are used for the entire analysis; only clinics that were in existence and had data for all months were included in the study. Source of referral information was also collected for the two other radio programs on the same data sheet, and both had some, but lower levels, of clients reporting them as their source of referral, which addresses GiveWell’s concerns about other radio programs.

    One of the findings of the Tanzania study was that the geographic variation in self-reported listenership rates to TnW was very high at the ward (similar to a US county) level. One of the statistical models used to evaluate the Tanzania survey data was a “dose-response” linear regression model (Model 5), which made use of this natural experiment. That is, it answered the question of whether we measured more dependent variable change in wards with high levels of listenership than in wards with low levels of listenership. For several key dependent variables, the dose response model was significant, consistent with what one would expect if TnW was the cause of the change. While this analysis is a correlation, and therefore can’t be used for causal attribution, it obviates concerns about differences between treatment and comparison areas and is consistent with the findings from the other analyses that utilize the treatment and comparison areas.

    One of the strengths of the Tanzania study, that is not mentioned in the GiveWell review, is that multiple (5) independent data sets were used to triangulate the findings, and all were consistent in showing essentially the same thing, that TnW impacted family planning behaviors (and HIV prevention behavior in the paper that was not included in the review). Given the strong historical and cultural beliefs that govern such behaviors in any society, causing and detecting even small changes over a relatively brief period, is an accomplishment for any intervention.

    A final strength of the Tanzania study was that our research team was largely, but not completely, independent of the PCI team. Research grants were awarded to the University of New Mexico, not to PCI, and the Principal Investigator, Everett Rogers was completely independent of PCI. Vaughan did receive payment for other work from PCI at the time of the study.

    We would like to call your attention to a few other strong studies of EE by other scholars. A recent review of the use of mass media campaigns to change health behavior (Wakefield and others, 2010) in the Lancet assesses all types of communication, but concluded about EE "Effective family planning communication strategies have included the embedding of pro-family-planning messages in entertainment programmes, particularly in a soap opera format, social marketing with expanded distribution of family planning devices, and focused promotional advertising."

    Mai and Kincaid (2006) used a sophisticated new statistical technique, propensity score matching, to assess the impact of the EE TV drama, Shabuj Chaya in Bangladesh on reproductive health outcomes. Propensity score matching approximates the results of what would be expected from an RCT design, and overcomes concerns about self-recruitment to viewership, or the fact that all EE programs are voluntary to watch or listen to, and audience members often differ from non-audience members on many traits. This study found that exposure to the TV drama was positively associated with HIV/AIDS and other health knowledge, and with two health-related behaviors, including (1) an 11 percentage point difference between viewers and non-viewers for current use of modern family planning method, and (2) a 7 percentage point difference for having visited any health clinic in the previous 3 months.

    Piotrow et al. (1990) used an interrupted time-series design in their Nigeria evaluation study on the effects of the family planning television series, In a Lighter Mood. Data about the effects of an entertainment-education project on the number of new adopters of family planning were gathered for a 19 months: (1) 6 months prior to the intervention, and (2) 13 months during the intervention (see Figure 8.1 below). The number of family planning adopters per month increased from a baseline of around 50/month to around 150/month during the intervention. Between 33 and 55 percent of clients cited the TV program as their source of referral.

    Essentially, the interrupted time-series design uses the over-time effects measurements as its own control, the result of which is a within-group comparison of effects over time. The weakness of the interrupted time-series design is also one of its main advantages; there is no control group. This absence of a control solves the ethical problem of withholding the possible benefits of the entertainment-education intervention to individuals in the control area. However, the effects of contemporaneous changes cannot be removed with the certainty, as in randomized control experiments.

    Valente et al. (1994) used a before-after evaluation research design to study the effects of Jakube Farra (Wise Man), an EE radio drama in The Gambia, an Islamic nation in West Africa. Prior to the daily broadcasting of 39 episodes of this radio show in 1991, 19 percent of survey respondents were adopters of family planning methods. Nine months later, 35 percent of listeners had adopted, an increase of 16 percentage points.

    In conclusion, we feel GiveWell has reasonably accurately reviewed the Tanzania paper in so far as identifying its weaknesses, but have drawn an overly restrictive and negative conclusion from it, perhaps because you underappreciated its strengths. Further, the review did not adequately take into account other EE impact research, which in aggregate, using a variety of research methods in diverse cultural settings, to address a variety of public health and other challenges, generally provides consistent findings that EE can be an effective approach to behavior change. GiveWell appears to have set what may be an impossibly high standard for EE evaluations in order to satisfy its desire to determine causality of impact and to minimize the chance of recommending to donors a strategy that is not, in fact effective. However, the stringency of your review may have caused you to fail to recommend EE, when in fact the prevailing evidence suggests you should decide otherwise. We will all eagerly await the DMI study’s results, due in 2013 and 2015, which hold promise for the rigor provided by a RCT design. However, there are no guarantees that this well-designed study will overcome the various challenges outlined above, but we hope they do. Further, the DMI intervention is substantially different from long-running EE programs, so the study’s conclusions may not be relevant to other communications initiatives.

    Thank you for your consideration of these issues.

    Peter W. Vaughan, Ph.D.
    Research Consultant
    1976 Sheridan Ave. So.
    Minneapolis, MN 55405

    Arvind Singhal, Ph.D.
    Samuel Shirley and Edna Holt
    Marston Endowed Professor of Communication & Director of the Social Justice Initiative
    Department of Communication
    University of Texas El Paso
    202 Cotton Memorial
    El Paso, TX 79968

    Scott Connolly, Ed.D.
    Director of Research
    Population Media Center
    P.O. Box 547

    Shelburne, VT 05482

    References cited in this letter

    • Bandura, A., 1994. Social cognitive theory of mass communication. In Media effects: Advances in theory and research, J. Bryant and D. Zillman (Eds.). Mahwah, NJ: Lawrence Erlbaum Associates.
    • Mai, P. Do and D. L Kincaid, 2006. Impact of an Entertainment-Education television drama on health knowledge and behavior in Bangladesh: An application of propensity-score matching, Journal of Health Communication: International Perspectives, 11(3) 301-325.
    • Piotrow, P.T., J.G. Rimon II, K. Winnard, D.L. Kincaid, D. Huntington, and J. Convisser. 1990. Mass media family planning promotion in three Nigerian cities. Studies in Family Planning, 21(5) 265-273.
    • Valente, T.W., Y.M. Kim, C. Lettenmaier, W. Glass, and Y. Dibba. 1994. Radio promotion of family planning in the Gambia. International Family Planning Perspectives, 20: 96-100.
    • Vaughan, P.W. and E.M. Rogers. 2000. A staged model of communication effects: Evidence from an entertainment-education radio soap opera in Tanzania. Journal of Health Communication, 5: 203-227.
    • Wakefield, M. A., B. Loken, and R. C. Hornik. 1261. Use of mass media campaigns to change health behaviour. Lancet, 376 (9748): 1261-1271.

    Select additional references

    • Nariman, H.N. 1993. Soap Operas for Social Change: Toward a Methodology for Entertainment-Education Television. Connecticut: Praeger.
    • Sabido, Miguel. 1981. Towards the social use of commercial television: Mexico’s experience with the reinforcement of social values through TV soap operas. Paper presented at the annual conference of the International Institute of Communications (Strasbourg, France). Mexico City, Mexico: Institute for Communications Research.
    • Singhal, A., M. Cody, E. M. Rogers, and M. Sabido. 2003. Entertainment-Education and Social Change: History, Research and Practice. Mahwah, New Jersey: Lawrence Erlbaum Associates.
    • Singhal, A. and E.M. Rogers. 1999. Entertainment-Education: A Communication Strategy for Social Chang. Mahwah, New Jersey: Lawrence Erlbaum Associates.
    • Westoff, C. & A. Bankole 1997. Mass Media and Reproductive Behavior in Africa. Demographic and Health Surveys Analytical Reports No.2. Calverton, Maryland: Macro International Inc.

    From Albert Bandura, Ph.D. (Stanford University David Starr Jordan Professor of Social Science in Psychology / Emeritus)

    In this letter I address the scientific foundations on which the PMC serial dramas are founded and their worldwide applications to alleviate some of the most pressing societal problems.

    Vehicle of Change

    Long-running serial dramas serve as the principle vehicle for promoting personal and social changes. These productions bring to life people’s everyday struggles and the consequences of different social practices. The storylines speak to people’s fears, hopes, and aspirations for a better life. They inform, enable, motivate, and guide viewers for personal and social changes that can alter the course of their lives. The dramatic productions are not just fanciful stories. They dramatize the realities of people’s everyday life and provide the strategies and incentives to enable them to take the steps to achieve it.

    Hundreds of episodes over several years allow viewers to form strong emotional bonds to the models, who evolve in their thinking and behavior at a believable pace. Viewers are inspired and enabled by them to improve their own lives. This is a highly flexible format that contributes to its generalizability, versatility and power. By including multiple intersecting plotlines via subplots, one can address different aspects of people’s lives rather than just a single issue. Moreover, one can do so at both the individual and social structural level.

    Component Models for Achieving Society-Wide Changes

    There are three major components to the psychosocial approach to fostering society-wide changes. The first component is a theoretical model based on social cognitive theory (Bandura, 1986). It specifies the determinants of psychosocial change and the mechanisms through which they produce their effects. This knowledge, which is the product of a large program of research that I have been conducting on the power of social modeling, provides the guiding principles. The second component is a translational and implemental model. It converts theoretical principles into an innovative operational model. It specifies the content, strategies of change and their mode of implementation. Miguel Sabido, a distinguished playwright and producer in Televisia in Mexico, creatively translated social cognitive theory into the dramatization format.
    We often do not profit from our theoretical successes because we lack effective means to disseminate proven psychosocial approaches. The third component is a social diffusion model on how to adopt psychosocial programs to diverse cultural milieus. Poindexter and Ryerson developed the diffusion model.

    PRINCIPLES GOVERNING THE ENABLING DRAMATIZATIONS

    There are five basic principles guiding the construction of the dramatic serials. These include social modeling, enhancement of personal and collective efficacy, vicarious motivators, emotional engagement, and enlistment of environmental supports (Bandura, 2002 ).

    Social Modeling

    The first principle enlists the power of social modeling for personal and social change (Bandura, 2006), 2009). The dramas include contrasting models exhibiting beneficial or dysfunctional patterns of behavior. The dramatized alternatives and how they affect the course of life, help people to make informed choices to improve their lives. Viewers are especially prone to draw inspiration from, and identify with, transitional models shown surmounting adverse life circumstances like their own. Seeing people similar to oneself succeed by sustained effort raises observers’ beliefs that they too possess the capabilities to improve their lives. The greater the assumed similarity the more persuasive are the model’s successes and failures. Modeling influences do more than build people’s beliefs in their own capabilities. Through their behavior and expressed ways of thinking, competent models transmit knowledge and teach observers effective skills and strategies for managing environmental demands. Acquisition of better means raises perceived self-efficacy.

    Perceived Personal and Collective Efficacy

    Among the mechanisms of human agency none is more central or pervasive than people's beliefs in their causative capabilities (Bandura, 1997). This core belief, called self-efficacy, is the foundation of human aspiration, motivation and accomplishments. Unless people believe they can produce desired results by their actions, they have little incentive to act or to persevere in the face of difficulties. Whatever other factors serve as guides and motivators, they are rooted in the core belief that one has the power to effect changes.

    This self-belief works through cognitive, motivational, emotional and decisional processes (Bandura, 1997). Efficacy beliefs affect whether individuals think optimistically or pessimistically, in self-enhancing or self-debilitating ways. Such beliefs affect how well people motivate themselves through goals and aspirations, and their perseverance in the face of difficulties and adversity. In addition, efficacy beliefs determine how opportunities and impediments are viewed. People of low efficacy are easily convinced of the futility of effort. In the face of difficulties they quickly give up. Those of high efficacy view impediments as surmountable by developing their competencies and perseverant effort. They stay the course in the face of adversity. Efficacy beliefs also affect the quality of emotional life and vulnerability to stress and depression. And last, but not least, efficacy beliefs determine the choices people make at important decisional points. A factor that influences choice behavior can profoundly affect the courses lives take and what one becomes.

    Many of the challenges of life involve common problems that require people to work together with a collective voice to change their lives for the better. Shared belief in their collective power to realize the futures they seek is a key factor in the exercise of collective agency (Bandura, 2000). The higher the perceived collective efficacy, the stronger the groups motivational investment in their undertakings, the more resilient their staying power in the face of impediments and setbacks, and the greater their accomplishments.

    Enabling social modeling serves as the principle vehicle in the dramatic productions in cultivating efficacy for personal and social change. When social change is attempted, it often challenges power relations and some entrenched societal practices supported by individuals who have a vested interest in preserving them even though they no longer serve the common good. Successes do not come easy. To change their lives for the better, people have to challenge adverse traditions and inequitable constraints. They must also be prepared for the obstacles they are likely to encounter.

    There are several ways of building resilient self-efficacy to impediments through social modeling. Prototypical problem situations and effective ways of overcoming them are modeled. People are taught how to manage setbacks by modeling how to recover from failed attempts. They are shown how to enlist guidance and social support for personal and social change from self-help groups and other agencies in their localities. Seeing others similar to oneself succeed through perseverant effort boosts belief in one’s efficacy in the face of obstacles.

    Enlisting Vicarious Motivators as Incentives for Change

    The third feature of the dramatic productions is the use of the vicarious motivators as incentives for change. Self-regulation of motivation and action is partly governed by whether people expect their actions to produce beneficial outcomes or adverse ones. Unless people see the modeled lifestyle as improving their welfare, they have little incentive to adopt it or to stick with it in the face of difficulties. The personal and social benefits of the favorable practices and the costs of the detrimental ones are vividly portrayed by contrast modeling. Seeing beneficial outcomes instills outcome expectations that serve as positive incentives for change, whereas detrimental outcomes create negative outcome expectations that function as disincentives. Observing modeled outcomes heighten motivation when observers see some likeness in sociodemographic characteristics to the inspiring model and judge the modeled behavior to have functional value (Bandura, 1986).

    Attentional and Emotional Engagement

    To change deeply held beliefs and social practices, people must develop strong emotional bonds to enabling models who exemplify a vision of a better future and realistic paths to it. Plotlines that dramatize viewers’ everyday lives and functional solutions get them deeply involved. They form emotional ties to models who speak to their hopes and aspirations. Unlike brief exposures to media presentations that typically leave most viewers untouched, ongoing engagement in the evolving lives of models provides numerous opportunities to learn from them and to be inspired by them. Viewers care deeply about the characters because they are personally relevant.

    Environmental Supports

    Motivating people to change has limited value if they are not provided with appropriate resources and environmental supports to realize those changes. Enlisting and creating environmental supports is an additional aid designed to expand and sustain the changes promoted by the media. To enlist dual sources of change, the dramatic productions are designed to operate through two pathways as shown in the Figure below.

    In one pathway, the serials foster changes by informing, enabling, motivating and guiding viewers. However, people are socially situated in interpersonal networks (Bandura, 2006b, Rogers & Kinkaid, 1981). Therefore, in the second, socially-mediated pathway, media influences are used to connect viewers to social networks and community settings. These places provide continued personalized guidance, as well as natural incentives and social supports for desired changes. Behavioral and valuational changes are promoted within these social milieus. Epilogues, delivered by culturally admired figures, provide contact information to relevant community services and support groups.

    Does it Work?

    PMC has taken seriously the evaluation of the effects of their various programs. Here is the sample of the results. The plotlines for the figures are described in some detail in the following chapter (Bandura, 2006).

    Enrollment In a National Literacy Program (Mexico)

    Enrollments in the national literacy program in the year prior to, during, and following the televised serial drama drawn from data in Sabido, (1981).

    As people develop a sense of efficacy and competencies that enable them to exercise control over their lives, they serve as models, inspiration, and even tutors for others in the circles in which they travel. This concomitant socially-mediated influence can vastly multiply the impact of televised modeling. In the year following the televised series, another 400,000 people enrolled in the self-study literacy program. Through the socially-mediated path of influence, televised modeling can set in motion an ever-widening, reverberating process of social change.

    Family Planning (Mexico)

    National sales of contraceptives in the two years preceding the serial drama promoting family planning and during the year it was broadcast (drawn from data in Sabido, 1981). Family planning centers revealed a 32% increase in the number of new contraceptive users over the number for the previous pre-broadcast year.

    Tanzania (Family planning)

    Mean number of new family planning adopters per clinic in the ministry of Health Clinics in the broadcast region and those in the control region. The values left of the dotted line are adoption levels prior to the broadcast: The values between the dotted lines are adoption levels when the serial was aired in the broadcast region but not in the control region; the values to the right of the dotted line are the adoption levels when the serial was aired in both the broadcast region and previous control region (drawn from Rogers, et al., 1999).

    The study included a number of informative features. The 7-yr. data points permitted a time series analysis in both the broadcast and control regions. The reversal design showed non-change in the control region in ’93 and ’94 but a significant rise in adoption of contraceptive methods during the broadcast years. The study also confirmed a dose-level effect shown on the next page.

    Impact of degree of involvement in the serial drama on women’s spousal discussion of family planning and use of contraceptive methods (from Rogers, et al., 1999).
    The more often people listened to the broadcasts, the more the married women talked to their spouses about family planning and the higher the rate of adoption of contraceptive methods. These diverse effects remained after multiple controls for other potential determinants, including exposure to other radio programs with family planning and AIDS contents, prebroadcast levels and changes in education, increased access to family planning clinics, radio ownership, and rural-urban differences.

    Adoption of Contraception Methods (Kenya)

    Percentage of women adopting contraceptive methods depending on the amount of exposure to family planning communications in the media. The white bard report the level of contraceptive use after controlling for the women’s demographic and socioeconomic characteristics and a host of other potential determinants (from Westoff & Rodriquez, 1995). These controls include life-cycle status, number of wives and children, and a host of socioeconomic factors such as ethnicity, religion, education, occupation and urban-rural resident.

    References

    I am sending as attachments a number of chapters presenting the theory on which the serial dramas are founded; review the global applications of the serial dramas and summarize the need for converging evidence from diverse methods to verify causation in the social science. I also include a brief summary of an invited address I delivered to the British Psychological Society on the global applications of the serial dramas.

    In addition, I include an invited article on selective moral disengagement as a major impediment to environmental sustainability. A much expanded discussion of this issue appears in a book I am writing in a book on Moral Disengagement: How Good People Do Harmful Things and Feel Good About Themselves.

    Bandura, A. (1997). Self-efficacy: The exercise of control. New York: Freeman.

    Bandura, A. (2006). Going global with social cognitive theory: From prospect to paydirt (PDF). In S. I. Donaldson, D. E. Berger, & K. Pezdek (Eds.). Applied psychology: New frontiers and rewarding careers (pp. 53-79). Mahwah, NJ: Lawrence Erlbaum.

    Bandura, A. (2006). On integrating social cognitive and social diffusion theories (PDF). In A Singhal & J. Dearing (Eds.). Communication of innovations: A journey with Ev Rogers, (111-135). Beverley Hills; Sage Publications.

    Bandura, A. (2007). Impeding ecological sustainability through selective moral disengagement (PDF). The International Journal of Innovation and Sustainable Development, 2, 8-35.

    Bandura, A. (2009). Social cognitive theory of mass communications (PDF). In J. Bryant & M.B. Oliver (Eds.). Media Effects: Advances in theory and research (2nd Ed. pp. 94-124). Mahwah, NJ; Lawrence Erlbaum. (see the section on the need for converging evidence from diverse methodologies).

    Concluding Remark

    In short, if I was flush with the coin of the realm this is the type of creative effort that takes on the population taboo that I would support.

    • 1

      In Burkina Faso, DMI requested PMC to refrain from broadcasting two soap operas in their control areas, but approved broadcasting in their treatment areas. PMC was the only health communication initiative to which they made such a request, and DMI stated that their reason was their belief that PMC’s programs might have substantial effects on mortality in the control areas. PMC complied with their request, but DMI’s research design will make it difficult to distinguish between DMI’s public service announcements (which we understand is their primary intervention) and PMC’s soap operas as causal factors in any differentials in behaviors that develop between their treatment and control areas. PMC argued that a better design would be to have PMC’s programs run on a nationwide basis (as Afya ya Jami and Zinduka did in Tanzania), so the only major difference between treatment and control areas would be the DMI intervention, but they exerted substantial pressure for PMC to comply with their request. PMC intends to broadcast the two soap operas in their control areas once their research project is completed.