For the first time in more than two decades, Pakistan has produced a national economic census. It counted over 7.1 million establishments and geo-tagged nearly 40 million structures, making it the largest digitisation exercise in South Asia. Government officials have presented it as a landmark step toward building a reliable statistical frame of the economy. But behind the headline figures, experts say the report also exposes serious blind spots and contradictions that raise questions about how useful the dataset will be for policymaking.
The last economic census in Pakistan was conducted in 2000. Since then, the country’s economy has transformed, but policymakers have had to rely on outdated estimates and patchy surveys.
That is why Bilal Gilani, Executive Director of Gallup Pakistan, sees the new census as historic. “It is wonderful that this has finally happened after 23 years. The world has changed dramatically in that time, yet we were still relying on outdated figures. This provides a digital baseline that Pakistan has never had before,” he said.
The census was integrated with the 7th Population and Housing Census of 2023. By combining the two, the government says it saved Rs7 billion. Enumerators equipped with digital tablets geo-tagged 38.3 million structures, of which 79 per cent were residential and 13.4 per cent were economic. A total of 7,142,941 establishments were documented, employing 25.3 million workers.
Dr Lubna Naz, Professor of Economics & Director of the Centre for Business and Economic Research at IBA, described the scale of the effort as both impressive and unfinished. “It is both a milestone and a work in progress. Integrating seven million establishments into a single digital frame and geo-tagging 40 million structures is unprecedented, but the data needs further strengthening before it can serve as a robust base for policy,” she said.
Yet how the census was conducted, its methods, tools and checks is as important as the numbers themselves.
Methodology, discrepancies and the role of AI
The census marked Pakistan’s first attempt to fully digitise economic enumeration. By integrating the exercise with the population census, enumerators were able to collect data on businesses while also recording household information, using digital tablets for geo-tagging.
However, reliability became a challenge. More than half of all entries, 52 per cent, were initially recorded as “Others” because of vague activity descriptions, spelling mistakes or Roman Urdu. To resolve this, the Pakistan Bureau of Statistics (PBS) turned to artificial intelligence, machine learning and natural language processing to clean and classify the data.
Testing showed uneven results. In Lahore, early trials revealed that the algorithm achieved only 24.68 per cent accuracy in matching establishments with their correct industrial codes. After revisions, the rate improved to 85.23 per cent in Karachi and 86.50 per cent in Faisalabad.
Nationwide, the final model was implemented with an estimated accuracy of around 80 per cent. On a dataset of over seven million establishments, that margin means as many as 1.4 million could be misclassified.
Dr Naz questioned the approach: “The process depended on limited testing in three urban centres, which restricts its representativeness. Large parts of Pakistan’s linguistic and regional diversity may not have been adequately captured,” she explained.
Gilani took a more forgiving view. “What matters is that a baseline has been created. Even with some misclassifications, this is a leap forward compared to having no census at all,” he said.
What the census leaves out
Beyond classification challenges, experts say the bigger issue lies in what the census excluded. The methodology covered establishments in fixed structures and household-based activities such as embroidery, tailoring, poultry, tuition and home food production. But it did not count mobile vendors, street stalls, small beauty parlours, or the rapidly growing sector of freelancers and digital businesses.
Dr Naz warned that this skews the picture: “Many unstructured activities fall into services, retail or small-scale production. By leaving them out, the census shifts the apparent weight of the economy towards more formal sectors. Women’s contributions, in particular, are statistically visible but economically invisible, because no financial values were recorded,” she said.
The report acknowledges this weakness. Household activities were recorded, but 40.49 per cent fell into a catch-all “miscellaneous” category. This means nearly half of the household economy remains unclassified and effectively unknown.
Contradictions and anomalies
A deeper review of the report reveals inconsistencies between different data tables and classifications.
Education mismatch: Adding up the unit types gives a total of 290,729 educational establishments (242,616 schools, 11,568 colleges, 214 universities and 36,331 madrassas). Yet under the industrial classification, education establishments number 326,868. This leaves over 36,000 establishments unexplained.
Health mismatch: Hospitals are counted as 119,789 establishments, but under the industrial codes for human health and social work, the number rises to 123,973. The report does not provide a breakdown to reconcile this gap.
Mosque workforce anomaly: The census counts 600,403 mosques employing 2.06 million people, more than the entire factory sector. But it is unclear if this includes volunteers, caretakers or teachers, making the figure hard to interpret.
Structures vs establishments: The census documents about 6.3 million economic structures but over 7.1 million establishments. This implies an average of 1.13 businesses per structure, confirming that many plazas, markets and multi-storey buildings host multiple enterprises. Yet the report does not analyse this density, missing an opportunity to shed light on commercial clustering in Pakistan’s cities.
These contradictions point to weaknesses in how the data was cleaned, tabulated and presented.
Geo-tagging: a digital first
One of the most widely advertised achievements of the census was geo-tagging. Enumerators recorded the GPS coordinates of nearly 40 million structures across Pakistan, enabling for the first time a geo-referenced mapping of businesses.
Gilani sees this as a digital breakthrough: “It is a major achievement. For the first time, every shop and establishment is literally on the map. This can change how Pakistan plans its urban and commercial growth,” he said.
Dr Naz urged caution: “Geo-tagging holds real promise for urban planning and investment targeting, but it is only as good as the classifications attached to it. If the activity codes are inconsistent, then linking them to coordinates does not solve the reliability problem,” she said.
Missing depth: finance and gender
Another major gap is the absence of financial and gender-disaggregated data. The census records the number of establishments and their workforce, but does not measure income, investment or financing sources. This makes it impossible to assess the economic value of household-based enterprises or their sustainability.
Dr Naz emphasised the implications: “By not monetising household-based work, Pakistan misses crucial insights into home-based enterprises, limiting the ability of policymakers to design effective credit, training, and business support programmes,” she said.
The lack of gender data is also a concern. With no breakdown of male and female workers in establishments, the census cannot accurately capture women’s role in the economy. This makes their contribution both visible in numbers yet invisible in value.
The risk of standing still
Globally, economic censuses are conducted regularly. India and Bangladesh already use them to capture financing, investment, and digital businesses. Pakistan’s exercise, while important, risks being a one-off if not institutionalised.
Dr Naz warned: “If Pakistan does not repeat and improve its economic censuses, it will face policy blind spots and lose competitiveness in emerging sectors like e-commerce and freelancing, where women and youth are most active.”
Gilani agreed that continuity is essential, though he stressed the importance of the progress already made. For him, the census marks the beginning of a digital statistical system that Pakistan must now build upon.
A foundation to build on
The Economic Census 2023 is undeniably historic: the first in more than two decades, the first digital, and the largest of its kind in South Asia. It provides a baseline picture of Pakistan’s economy that did not exist before.
Yet its flaws are equally significant. Contradictions between tables, anomalies in workforce numbers, exclusions of informal and digital sectors, and a heavy reliance on imperfect AI all raise questions about how much weight policymakers should place on its findings.
As Dr Naz summed up, it is best seen as both a breakthrough and a work in progress. The real test will be whether Pakistan can refine, repeat and expand this effort into a reliable system. For now, the country has a digital frame of its economy. Whether that frame is sturdy enough to support sound economic decisions remains uncertain.
