The BRAC Institute of Governance and Development (BIGD), a Bangladesh-based research institute, took an initiative to track citizens’ perceptions through a Pulse Survey amid the fast-evolving political context following the July 2024 uprising, which led to the fall of the 15-year-long Awami League regime and the emergence of youth as a promising political force.
The recently published findings of the third round of the Pulse Survey have sparked a wide range of reactions. From politicians to policymakers, from academics to citizens, many have weighed in on the merit of the survey findings.
Voting preference was just one of the four focal areas of the Pulse Survey, yet it became the main bone of contention. The other three focal areas are perceptions of the economic and political scenario, assessments of the interim government, and views on youth politics.
In his op-ed, titled “Bangladesh’s BIGD Poll Risks Becoming a ‘Literary Digest’ Moment,” published in The Diplomat on August 23, Mujahed Islam raised serious questions about the methodology used in the survey and opines that the data generated are misleading. This article responds to the concerns raised by Islam.
The Question of Sample Representativeness
Ensuring a truly representative sample in a political poll is extremely challenging. Given the sensitivity of the data it generates, the sampling method of such polls often comes under scrutiny. However, Islam’s comparison of the BIGD Pulse Survey to the Literary Digest poll of 1936 is misplaced.
BIGD did not use a convenience sampling method, i.e., selecting the “easiest-to-reach” citizens, which appears to have been the case in the 1936 poll. Instead, our sample was drawn from the nationally representative sample of 10,240 respondents created through the 2022 Survey of the Bangladeshi People, which we conducted in collaboration with The Asia Foundation.
The 2022 survey adopted a systematic random sampling method to ensure that participants were neither purposively selected nor excluded. The sample was designed to be representative at the divisional level. The standard sampling formula – assuming a 5 percent margin of error, a 95 percent confidence interval and a 1.5 design effect – provided the minimum requirement of 578 respondents per division. The survey reached at least 640 respondents in each division, totaling 10,240, far exceeding the minimum sampling requirement.
The samples were scientifically selected using stratified random sampling across all 64 districts, with a deliberate balance between urban and rural areas, as well as representation by gender and age groups. The survey selected 16 primary sampling units (PSUs) and 10 households per PSU using systematic random sampling. One respondent per household was selected based on gender and age criteria. Respondents’ phone numbers were collected in the survey with explicit consent for future communication.
In the third round of the Pulse Survey, 9,203 respondents from the 2022 survey database were approached, since phone numbers for nearly 10 percent of the original respondents were unavailable. We secured a complete response from 5,489 individuals, resulting in a 60 percent response rate. Although the attrition rate was 40 percent, it did not bias the results or compromise representativeness: baseline characteristics, such as age, gender, education, income, household size, religion and region, showed no significant differences between the surveyed and non-surveyed groups. While the attrition is likely to increase the margin of error, it remains within reasonable bounds.
In his op-ed, Islam also raised concerns about the rural bias of our sample. However, the Pulse Survey’s rural-urban distribution (73 percent vs. 27 percent) shows no statistically significant difference compared to the national breakdown (68 percent vs. 32 percent). Moreover, before publishing the results, we weighted the data to account for any imbalance and found no significant difference compared to the unweighted data.
The Method of Phone Interviews
The op-ed highlighted a significant demographic divide in mobile phone ownership, suggesting that this may have introduced biases in our survey, as it was conducted via phone. However, the most recent Household Income and Expenditure Survey (HIES) reveals that 98.48 percent of households in Bangladesh have access to a mobile phone, with only a slight difference between urban and rural areas.
While the gender divide in access remains significant, our survey was designed to overcome this barrier. In fact, individual phone ownership was not the basis of our sampling. In the 2022 survey, if the respondent did not own a phone, the phone numbers of other household members, neighbors, friends, or relatives were collected. Interviewers made every effort to connect with the original respondent by calling the recorded numbers, ensuring that, regardless of direct phone access, intended respondents were reached, including men and women.
Questionnaire Designed to Minimize Bias
To avoid introducing bias, interviewers refrained from reading out lists of party names that could influence responses to sensitive questions, such as voting preferences. Instead, they asked openly and neutrally: “Who would you vote for in the next election?” Similarly, for questions such as “When do you want the next national election to be held?” respondents had full freedom to refer to any date or month, which BIGD later coded for analysis. Nearly all questions allowed respondents to select “do not know” or “no comment” options, ensuring they were not forced into a choice.
Responsible Reporting of Survey Findings
The survey results were disseminated to the media in a press conference. In the presentation, the methodology and response rates were clearly explained. Regarding voting preferences, researchers emphasized several times that half of the respondents mentioned their indecisiveness, urging the audience to interpret the data with caution.
Following the presentation, the researchers also discussed at length the possible reasons for such indecisiveness. Fear or the inability to trust the interviewers was an important factor, but disillusionment with the major parties – due to their prior and post-uprising activities – and doubts about the capability of new, youth-led parties were likely equally important. This lack of trust in political parties is frequently reflected in both mainstream and social media. In another ongoing qualitative study, BIGD is also recording a strong sense of public doubts and disillusionment.
Additionally, contrary to the dominant political narrative, 51 percent of respondents suggested that the election should be held after major reforms, and an additional 17 percent emphasized the need for urgent reforms. These respondents also specified the types of reform they expected before the election, with law and order, justice, security and corruption emerging as the top priorities. These findings further highlight the prevailing general lack of trust in political parties.
How the Pulse Survey Contributes to the Current Political Discourse
The purpose of the Pulse Survey is not to predict the election outcome but to give visibility to the hopes and concerns of the citizens. Consideration of all three rounds of the Pulse Survey makes it clear that the initial hope ignited by the uprising is gradually fading, reflected in the increasing disapproval of the interim government, widespread concerns about safety and security, mob violence, gender-based violence, and decreasing explicit preference for all traditional parties. If anything, the BIGD Pulse Survey highlights that a successful uprising does not guarantee lasting change, and political parties need to demonstrate how they would contribute to making that change a reality.