Pakistan’s recently released Economic Census, undertaken as part of the 7th Population & Housing Census exercise, is being described by government officials as a landmark digitisation effort that finally gives the country a comprehensive frame of establishments, structures, and household economic activities. It is indeed the largest mapping exercise of its kind in the country’s recent history, but the report and subsequent coverage also reveal important blind spots, methodological tensions, and policy contradictions that must be confronted before the data can be relied on for major reform or for assessing the economy’s structural problems. Below I summarise what the census achieved, where the evidence shows problems, and why the results matter with direct references to the Pakistan Bureau of Statistics (PBS) report and independent South Asian and international coverage and commentary.
What the census produced the main, verifiable gains
A very large, geo-tagged inventory of structures and establishments. The census documented roughly 38.3 million structures and recorded about 7 million economic establishments during the population listing stage providing province-level breakdowns for housing, economic structures and mixed-use properties. This gives policymakers the first modern digital frame for targeting programmes, sampling and administrative reform.
Digital-first collection, algorithmic cleaning, and rapid scaling. PBS describes an intensive cleaning pipeline that combined a 6,000-entry dictionary, rule-based matching, machine-learning imputation and sample-based field verification (Lahore, Karachi, Faisalabad, Gujranwala, Peshawar and Quetta). PBS reports algorithmic assignments achieved high coverage in trials (e.g., 80–85% automatic updates in some cities) and nation-wide implementation in a compressed time frame. Those technical capacities are real advances for statistical production in Pakistan.
A strong signal that micro and small enterprises dominate employment. Multiple outlets summarising PBS findings report that the economy is overwhelmingly composed of micro-units: roughly 95% of establishments employ fewer than 10 people, and wholesale/retail trade alone accounts for several million units. These findings reshape debates about job creation they imply that employment policy must address millions of very small firms rather than rely only on medium/large manufacturing expansion.
Clear strengths in the PBS Documentation
Transparency about data cleaning and limitations. The PBS report explicitly sets out challenges (a massive “Others” category, Roman-Urdu entries, spelling variants, wrong PSIC codes) and documents its multi-stage verification and correction procedures, including sample field checks and manual corrections where algorithms struggled. That degree of methodological disclosure is good practice and helps external researchers evaluate reliability.
Granularity (structures + unit types + household economic activities). The census did not only count businesses but also mapped household economic activities and structure types (residential, economic, mixed, informal shelters), which is useful for linking livelihoods to infrastructure and service delivery planning. The PBS tables published in the report supply region and sector breakdowns that researchers can use.
The blind spots, contradictions and quality risks
Large “Others”/classification problem: how much of the dataset is ambiguous?
PBS itself documents that ~52% of textual activity entries initially fell into an “Others” category or were poorly described; much of the cleaning required mapping Roman-Urdu, spelling variants and ambiguous descriptions into PSIC codes. While the report explains corrective steps, that starting point is a real warning: large, automated recoding from noisy text can introduce systematic misclassification especially for emerging services, gig activities and mixed household enterprises that do not fit neatly into standard categories. Any policy built on fine sectoral shares (e.g., precise manufacturing vs services employment splits by district) should therefore be used cautiously until independent validation is conducted.
Algorithmic accuracy vs ground truth promising trials, but limited national ground-truthing
PBS reports strong algorithmic performance in pilots (70–85% prediction confidence and high dictionary success in selected cities) and reports national automation rolled out rapidly. But the documented field verification samples (tens of thousands across a few districts) are still small compared with a 7-million-establishment frame. That creates a risk of regional bias: algorithms tuned on urban Urdu/Roman-Urdu patterns may perform worse in rural/linguistically diverse districts, remote areas or among household micro-activities (e.g., home-produced food, seasonal activities). PBS notes portions of data still missing (e.g., 6% with missing employment info), which can distort aggregate workforce totals and employment category shares.
Informality vs formal registration a major policy contradiction
The census puts the informal economy in stark relief. Media summaries (and PBS-derived reporting) highlight that millions of establishments are not formally registered with corporates/SECP one widely reported figure is roughly only ~250,000 establishments formally registered, compared with over 7 million counted units (figures reported in multiple coverage pieces summarising PBS outputs). That gap shows a contradiction between the state’s desire to tax, regulate and grow formal enterprise and the reality that most livelihoods operate informally. Bringing millions of micro units into the formal system requires phased policy instruments (simplified registration, tax thresholds, digital UANs, micro-enterprise support) none of which will be solved by releasing numbers alone.
Evidence notes: press coverage (Tribune, Business Recorder, The Print and other outlets) consistently cites the “7.1 million establishments” headline and the small number of registered firms. These are the core, load-bearing claims about scale and informality.
Shocking structural counts (mosques, madrassas, factories) headline numbers need context
Several news outlets emphasised striking counts such as hundreds of thousands of mosques and only a few tens of thousands of factories, using those counts to argue economic neglect of industry versus social infrastructure. Those comparisons are attention-getting and useful to provoke debate, but they require careful contextual interpretation: (a) a mosque is a different institutional unit than a factory, (b) factories are larger units employing more people on average, and (c) counting buildings does not measure economic output or productivity. The census provides building counts; economic policy must combine these counts with output and value-added data (national accounts, firm surveys) to chart industrial policy.
Why these contradictions matter for policy, donors and researchers
Macroeconomic policy and IMF/World Bank programmes rely on reliable frames. Pakistan is operating under large IMF and World Bank programmes and needs accurate enterprise frames to estimate tax potential, employment elasticities and sectoral productivity. The census is an essential input but the documented limitations mean international partners and domestic policymakers should use the data as a high-quality starting point, not a finished product.
Social and industrial policy will be misdirected if micro-classification errors remain. For example, programmes targeting manufacturing jobs vs wholesale retail will produce different multiplier outcomes. If a share of small manufacturing got misclassified as “retail” due to questionnaire text, interventions may miss small producers who need credit, skills or power subsidies. PBS’s openness about coding problems flags that follow-up enterprise surveys are urgently needed.
The census exposes the scale of informality, but formalisation must be sequenced. The headline that tens of millions are employed across millions of unregistered units suggests that broad formalisation drives (registration + taxation) without support (digital payments, social protection, simplified compliance) would be punitive. The policy contradiction between the desire to broaden the tax base and the practicalities of millions of tiny units is now visible and unavoidable. Media summaries (Tribune, Dawn, BR) echo this point.
Actionable recommendations (evidence-based and modest)
Publish microdata and a reproducible codebook for independent validation. PBS’s public report is strong on methodology; releasing anonymised microdata and the cleaning scripts will let academic teams and international partners test classification, reweight missing employment and run independent verification studies. (This is standard practice in many countries after major censuses.)
Carry out stratified field validation in linguistically and geographically diverse districts. PBS ran thorough urban pilots; the next step is purposive re-visits in remote rural districts, border regions and areas with different language scripts to measure algorithm bias and update the cleaning dictionary. PBS itself documents the value of such verification.
Link the census frame to a phased formalisation pathway. Use the 7-million-unit frame to pilot voluntary digital registration, micro-VAT thresholds, and simplified social protection enrolment beginning with wholesale/retail clusters where the census shows concentration. The evidence of concentrated sectors provides targeting options.
Commission follow-up enterprise surveys on value-added and productivity. The census counts units; to design industrial policy Pakistan now needs representative enterprise surveys (balanced by size strata) to measure output, capital intensity and barriers to growth. International donors (World Bank, ILO, ADB) can partner on technical design.
Conclusion: Progress, but not a panacea
Pakistan’s Economic Census is a major institutional achievement: a nationwide, geo-referenced frame that documents millions of establishments and tens of millions of structures and one that finally brings the informal economy into a mapping format suitable for policymaking. PBS’s report shows methodological care and advances in digital cleaning; independent coverage corroborates the headline numbers and highlights their policy implications. At the same time, the report exposes important blind spots a large initial “Others” classification, dependency on algorithmic recoding with limited national ground-truthing, and a vast gap between counted establishments and formal registration which together create contradictions between ambition (formalisation, revenue mobilization, industrial policy) and on-the-ground reality.
If Pakistan and its partners treat the census as a first, transparent step rather than a final answer by releasing microdata, conducting stratified validation studies, and designing carefully sequenced formalisation pilots, the exercise will mature into an operational statistical frame that can underpin credible, inclusive economic policy. The evidence in the PBS report and in mainstream South Asian and international reportage makes both the promise and the limits of the census unmistakable.
