Economic Census 2024 had coverage error below 3%, reliability not undermined: BIDS

The Bangladesh Bureau of Statistics (BBS) missed 3,65,472 units in the 2024 Economic Census, an error which, according to a Post Enumeration Check (PEC), did not undermine the overall reliability of the census.

The PEC shows that the census covered 97.05% of units, meaning 2.95% were left out. The highest number of missed units was found in City Corporation areas, especially in the informal and service sectors.

The PEC, conducted by Bids, found large differences in Net Coverage Error (NCE) depending on location, with City Corporation areas showing the highest rate of undercount.

The PEC report says no intentional bias or motivated error was detected. The mistakes were random and did not undermine the overall reliability of the census. However, it recommends stronger monitoring in urban and informal areas, better training for field staff, and improvements in digital supervision.

According to the review, the national NCE stands at 2.95%, meaning nearly 3% of economic units were missed in the 2024 Economic Census.

But the error varied sharply by area. City Corporation areas recorded an NCE of 5.99%, indicating significant under-coverage in major urban centres. Overall urban areas had an error rate of 3.90%, higher than the national average.

Upazila sadar/growth centres showed an NCE of 2.48%, municipalities slightly lower at 2.33%. Rural areas performed somewhat better, with an NCE of 2.55%, still showing gaps but much lower than big cities.

The report was released at a seminar at the Bids building in Agargaon, Dhaka, where Bids Research Director Mohammad Yunus presented the findings.

The seminar was told that the BBS initially estimated a total of 11,877,364 economic units in the census. But according to the PEC, the actual number is 12,242,836 units — nearly 3 percent higher than the preliminary count.

Dhaka division recorded the highest NCE at 4.19%. Sylhet had the lowest at 2.01%, while Rangpur showed 3.04%, Chattogram 2.59%, Rajshahi 2.51%, Barishal 2.5%, Mymensingh 2.45%, and Khulna 2.04%.

Temporary establishments — such as seasonal shops or short-term project sites — had the highest NCE at 8.29%. Permanent establishments had a much lower rate of 3.03%.

In the industrial and service sectors, the NCE was 2.08% and 3.05% respectively.

For the PEC, 352 sample areas were selected from all eight divisions — 100 in Dhaka, 63 in Chattogram, 48 in Rajshahi, 42 in Khulna, 80 in Rangpur, 24 in Mymensingh, 19 in Barishal, and 16 in Sylhet. Villages, towns, municipalities, Upazila Sadar areas and City Corporations were all included to get an accurate picture of where errors occurred.

According to the report, the undercount was not the same everywhere. City Corporation areas had the highest rate because businesses move frequently, many new ventures start up, and temporary or unregistered establishments are more common. Small service-sector businesses, temporary stalls, footpath shops and informal units were missed most often. In contrast, registered establishments and those with Tax Identification Numbers were almost always counted.

The PEC also reviewed inconsistencies between census data and PEC responses. It found mismatches in establishment type (14.13%), nature of economic activity (8.03%), type of ownership (6.51%), and fire safety information (5.91%). The highest inconsistency was in business registration information — nearly one-third (32.90%) did not match. Bids said these differences mainly arose from misunderstanding by respondents or unclear explanations, not from deliberate misinformation.

Responding to a question, Director Yunus said temporary establishments have much higher "move-in, move-out" rates than permanent ones. Permanent structures take time and cost to change, but temporary ones can disappear instantly, causing coverage problems.

Yunus noted that internationally a Net Coverage Error of up to 5% is considered acceptable. The 2.95% NCE in the Economic Census is within that standard. He added that no country can achieve zero coverage error in a census.

S M Shakil Akhter, former project director of the Economic Census and now planning secretary, pointed out field-level limitations. He said the accuracy of a census depends on the skills, training, access and safety conditions of the data collectors. Learning a large and complex manual in three to five days of training is difficult, yet these enumerators are the most crucial part of the census.

He added that access restrictions to garment factories, industrial sites or sensitive places often hindered data collection. In many cases, security officers, management or owners were reluctant to share information.

BBS Director General Mohammad Mizanur Rahman said the 2024 Economic Census was conducted during an unusual national situation. After the change of government in August 2024, many business owners could not be reached and many establishments were closed. These factors contributed to the NCE, although the rate remains low.

The seminar was chaired by Bids Director General Professor AK Enamul Haque.