Pakistan’s debate on widening the tax base has long confused counting people with collecting revenue. The Federal Board of Revenue (FBR) announces the number of returns filed, celebrates additions to the Active Taxpayers List and imposes fresh restrictions on non-filers. Neither a return nor a computer-generated taxpayer number establishes that a person earns taxable income. A modern fiscal State should know the difference.
The first three parts of this series established the scale of informality, tested the revenue arithmetic and distinguished lawful livelihoods from organised criminal activity. The next question is institutional: how can Pakistan build a National Economic Registry (NER) that identifies genuine capacity to contribute without converting every citizen into a suspect?
The starting point must be statistical honesty. Pakistan Bureau of Statistics’ Labour Force Survey 2024–25 reported 83.1 million persons in the labour force under the updated international definition, against 85.6 million under the earlier definition. The difference arises principally from the revised treatment of subsistence producers. The same survey reports informal-sector employment at 72.5% of non-agricultural employment. Neither figure represents a ready-made pool of income-tax payers.[1]
Our ambition of identifying 30 million or more natural persons with taxable capacity is still a hypothesis to test, not a quota to impose. It should be validated against actual incomes, household dependency, business profits, assets, and exemptions. A daily wager may need social protection. A pensioner may owe nothing. An agricultural labourer is not the same as the owner of a profitable agricultural enterprise. An individual may earn from several sources without becoming several taxpayers.
Registration should be universal in reach, selective in taxation and respectful of the Constitution. A workable registry would begin with a verified identity, but not end there. It would record a person’s economic status—student, dependent, salaried employee, self-employed worker, employer, pensioner or person outside the labour force and link, where lawfully justified, to business registrations, declared income, property, vehicles and other economic indicators. Household relationships should help identify dependency, not create a presumption that one family member owns another’s income.
The registry must distinguish three layers. The first is basic identity and economic status, which can support public services. The second is verified tax-relevant information, accessible only to competent authorities for defined purposes. The third is financial-intelligence information concerning suspected crime, subject to separate statutory controls. Combining all three into an unrestricted administrative database would be a grave mistake.
Artificial intelligence can reconcile inconsistent records, detect duplicate identities, classify businesses and identify discrepancies requiring examination. It can, for example, flag the apparent owner of several commercial properties who declares no rental or business income. It can also prevent harassment by recognising that modest electricity use or a small bank deposit does not establish taxable profit.
A risk score is not an assessment order. No algorithm should determine tax liability, freeze an account, block a passport or initiate prosecution without a lawful human decision, recorded reasons and an effective right of challenge. The person affected must be able to see and correct the relevant data, subject to narrowly justified restrictions for genuine criminal investigations.
Pakistan’s Constitution already provides the essential principles. Article 4 requires treatment in accordance with law; Article 10A protects due process; Article 14 safeguards dignity and privacy; and Article 19A secures access to information on matters of public importance, subject to lawful limits. Digital administration must give these protections practical effect rather than treating them as ceremonial assurances.[2]
A statute establishing the NER should define permissible data fields, retention periods, authorised users, audit trails, penalties for misuse, independent oversight and judicial remedies. Every automated recommendation should be explainable in ordinary language. Models should be tested for errors and discriminatory outcomes across regions, occupations, gender and income groups. An annual public report should disclose false-positive rates, successful appeals and actual additional revenue, without exposing private records.
There is an equally important constitutional dimension. Agricultural income taxation, provincial sales taxes on services and important property-related levies fall within provincial competence. A centralised database cannot be allowed to become a device for taking over provincial fiscal powers. The Eighteenth Amendment and the National Finance Commission settlements cannot be displaced by software procurement or administrative instructions.
The better arrangement is a consent-based national information architecture. Provinces should retain their statutory powers, assessment authority and rate-setting responsibilities. Common technical standards can permit secure exchange of verified information while each participating government remains answerable for its own decisions. Data access must be reciprocal, purpose-limited and auditable. Shared technology is not a constitutional licence for central control.
A genuinely cooperative framework would also distinguish a common taxpayer identity from a single taxing authority. Pakistan needs the former. Whether it needs integrated collection functions in particular fields is a separate political and legal question, requiring provincial consent and carefully drafted safeguards.
Formalisation must offer a bargain worth accepting. Small businesses should receive a single straightforward registration channel, proportionate bookkeeping requirements, access to digital payments at reasonable cost and protection against repetitive visits by multiple inspectors. Home-based workers and microenterprises should gain access to social insurance, training, credit and commercial dispute resolution. Registration without benefits will reproduce the mistrust that defeated earlier documentation drives.
Nor should the registry become a new withholding machine. Pakistan’s dependence on collection at source has obscured the distinction between tax on income and tax on transactions. A bank transfer is not income; turnover is not profit; an electricity bill is not proof of capacity to pay. The objective is to replace presumptions with verified assessment, reduce redundant withholding provisions and reward voluntary compliance.
Implementation should proceed in stages. During the first year, a limited pilot could reconcile existing records in selected districts and business categories, with independent evaluation. A second phase could extend simplified registration and correction facilities across provinces. Only after error rates, legal safeguards and appeal mechanisms are demonstrated should broader risk analytics be deployed. No large-scale compulsory data merger should precede the law authorising it.
Success should be measured against demanding indicators: how many previously unrecorded profitable businesses have been identified; how many low-income people have been correctly excluded from liability; how much net revenue has been recovered after refunds and appeals; how much compliance time has been saved; and how many officials have been sanctioned for misuse of information. Inflated registration totals are not reform.
The Financial Action Task Force’s work on new technologies likewise recognises the promise of advanced analytics alongside the need for data protection and accountable deployment.[3] Pakistan should learn from that balance. Better information can make enforcement more selective and fair. Bad governance can make it more intrusive.
The State has spent decades asking citizens to prove that they are not evaders. It is time to ask the State to prove that its classifications are accurate, its procedures lawful and its taxes proportionate.
Thirty million genuine individual taxpayers will not emerge from a button pressed in Islamabad. They will emerge, if the evidence supports that number, from an economy in which registration brings rights, taxation follows capacity and concealed wealth no longer enjoys a competitive advantage. That is the practical promise of artificial intelligence and the test of our institutions.
[To be concluded]
Endnotes
1. Pakistan Bureau of Statistics, Labour Force Survey 2024–25; Ministry of Planning, press release of 26 November 2025, explaining the 83.1m/85.6m definitional distinction.
2. Constitution of the Islamic Republic of Pakistan, Articles 4, 10A, 14, 19A, 160 and relevant legislative entries; National Assembly, current consolidated text.
3. Financial Action Task Force, Opportunities and Challenges of New Technologies for AML/CFT (2021), and related digital-transformation work.
4. Pakistan Bureau of Statistics, Labour Force Statistics and sectoral breakdowns.
5. Huzaima Bukhari and Ikramul Haq, Towards Broad, Flat, Low-rate and Predictable Taxes, third edition (PRIME, 2024), on broadening bases and reducing distortionary withholding.
6. Ikramul Haq, Tackling Informal Economy With AI, Parts I–III, The Friday Times series (2026).