Pakistan, for the last many years, has been grappling with the problem of raising revenue to survive as a viable economic entity. The huge fiscal deficit, coupled with mounting debt servicing, is posing a serious threat to the country’s economic revival. The economic managers have failed miserably to substantially increase revenue according to potential.
Paradoxically, their preoccupation with more and more revenue collection has made them neglect the infrastructure required to administer these very taxes. They are caught in a dilemma: on the one hand, there is mounting pressure from the International Monetary Fund (IMF) and other foreign institutions to lower the fiscal deficit, and on the other, all attempts to increase revenue from the existing taxpayers are proving detrimental to the already ailing economy.
Pakistan needs to learn from the experience of many developing countries that managed to raise revenue by improving their tax administrations. In 1992, Richard M. Bird and Milka Casanegra de Jantscher presented a remarkable book Improving Tax Administrations in Developing Countries (interestingly, this was an IMF publication based on a conference held in Spain in 1991). Since 1992, there has been growing awareness that more efforts are required to improve existing administration if a developing country is keen to explore new sources of revenue.
The old saying “tax policy is only as good as its administration” is outdated. Today’s consensus is that “tax administration is tax policy” (Stanley S. Surrey, “Tax Administration in Underdeveloped Countries”, University of Miami Law Review, xii (Winter 1958) at 158–88. Stanley S. Surrey was, at that time, Professor of Law and International Programme in Taxation at Harvard Law School).
Anyone who has worked in a tax administration of a developing country (like me in Pakistan from 1984 to 1996) can vouch for this. Many well-intentioned laws have been laid to rest by inefficient (which also includes indifferent, corrupt and incompetent) tax administrations. Pakistan is a classical example of this. The Federal Board of Revenue (FBR), the apex administrative body for federal taxes, is the most apt example of an indifferent (though highly oppressive and inefficient) tax administration.
Taxation requires pragmatic thinking and is most effective when developed from a practical and possible agenda. For building a sound tax administration, it is necessary to start its foundation with a Tax Intelligence System (TIS). The widest possible taxpayer base must be identified for any tax to be equitably spread across the whole taxpayer population. Even a small tax at a lower rate, spread over a wide taxpayer base, will invariably yield more revenue than a higher tax on a narrow base.
The levy of General Sales Tax (GST) at 18% has failed to bring the desired results, as it is a higher tax on a narrow base. Had it been an 8% levy across the board, it could have been acceptable as well as successful in terms of yielding more revenue, being a low-rate tax spread over a wide taxpayer base.
How can Pakistan succeed in improving revenue collection when it has no information/intelligence system/unit to maintain taxpayers’ individual profiles? The efforts, through software developed by Pakistan Revenue Automation Limited (PRAL), a 100% state-owned company of the FBR, are neither theoretically sound nor scientifically implemented in conformity with the ground realities of Pakistan.
To exploit the capabilities of the computer network fully, a completely fresh conceptualisation is required in the FBR – preferably not by tax officials, but by system analysts
The FBR, responsible for the collection of federal taxes, has miserably failed to introduce any automated tax system using Artificial Intelligence (AI), despite the fact that it has its own market-wage-oriented company, PRAL, at its disposal, to monitor the economic activities of corporate and business sectors. FBR stalwarts have, to this day, failed to develop a TIS, a computer-based automated system guided by AI. In many countries, such a system monitors “large taxpayers”, identifies the most productive centres of information from which significant data can be extracted, uses simple AI tools that can be set up quickly for basic tax compliance formats, avoids being tax-specific, and can be adapted to any type of tax.
Artificial Intelligence Tax Intelligence System (AITIS) is the idea in vogue these days. Earlier, TIS was implemented even in countries like Botswana in the 1980s and helped in its rapid increase of diamond revenue, as well as proving extremely beneficial for other areas of the economy to expand simultaneously (K.L. De Silva, “Botswana’s New Corporate Tax Intelligence System”, Bulletin, Official Journal of the International Fiscal Association, Vol. 53, No. 7, 1999, p. 302).
The Tax Intelligence System concentrates on third-party information that continuously originates from different areas of the government and quasi-governmental institutions to the tax department. The Tax Department of Botswana, in 1983, on the advice of the IMF, revived its investigation division and the Intelligence Unit, which specialised in gathering information on corporate activity. Previously, all information received was maintained manually in the form of registers.
The problem faced by the Tax Department of Botswana was the same as is now faced by the FBR: the flood of information it was trying to process was too great, and the system became too slow and prone to errors. They found an efficient way to handle it through computerisation. Did the IMF forget its advice to Botswana in the 1980s while dealing with Pakistan in 2025? We have better human resources in Information Technology (IT), and yet we have not achieved what a small African state managed as early as 1985! It is indeed shameful and an eye-opener for the FBR stalwarts and wizards sitting in the Ministry of Finance.
There is an urgent need to set up a TIS in Pakistan, guided by AI, to maximise the scope of revenue collection. Its salient features should include a computer network for intelligence work. It should be able to record and process a large volume of information, increase dramatically the number of new persons registered as active taxpayers, ensure they start filing tax returns and paying taxes, and expose registered taxpayers who need to be investigated, reflected in the increased amount of taxes collected.
To exploit the capabilities of the computer network fully, a completely fresh conceptualisation is required in the FBR – preferably not by tax officials, but by system analysts. The knowledge of tax officials regarding user requirements must be successfully fused with the skills of computer programmers. Independent professional programmers and data scientists should be hired and given the task of building a relational database, i.e. a database in which data relationships could be established electronically.
Tax officials should be left with the task of defining the scope of the project and the method of selecting, collecting and processing the data. The main purpose of the project should be to create a database that could record and process significant information regarding taxpayers, both existing and prospective. It is necessary to determine what is significant and to devise a selective basis for gathering data. This approach will make the work of the FBR proactive rather than reactive.
The problem encountered by the FBR is not how to gather information or the lack of it, but its abundance. Information can be gathered from many sources, e.g. mobile users, the list of electricity and gas consumers, and endless other sources including the Internet and AI. The question is: how useful are these sources? The mere fact that a person has a mobile phone or pays electricity does not in itself determine whether that person may eventually become a taxpayer. Information should be processed on the basis of its usefulness.
The availability of current up-to-date information can assist with the seizure of assets or income in extreme cases of default
The information collected in certain areas of selected cities can be useful if processed from this perspective. Otherwise, it will remain bulk trash without any useful purpose. Therefore, several factors are to be considered in order to deem it significant:
- Information is most useful if it relates directly to expenditure or income.
- If it relates to the ownership of commercial property, such as farms or commercial vehicles, there must be an expectation that these assets will eventually produce assessable income.
- Information is significant if the time lag is short between its receipt and the consequential registration of the new taxpayer. Statutory provisions place time limits on the power of assessing/tax officials to impose additional liabilities on taxpayers, meaning that old information soon loses its usefulness.
Information collected is also needed to have at least the following components:
a. a name (and if possible, the address) of a person;
b. a description of the type of transaction, i.e. whether purchase or sale, or a record of the ownership of property;
c. the date of the transaction;
d. a description of the property or service transacted, e.g. house property or contract payments; and
e. a monetary value above a stipulated threshold, depending on the type of property or service.
The objectives in setting up the database should be to record the “significant” financial transactions entered into by companies during a tax year; to consolidate into one record all the information from different sources relating to one taxpayer/person according to tax years; to send a consolidated report to the field officer in the tax division at the end of the tax year or during the relevant period; to provide quarterly and annual reports to senior managers (e.g. commissioners/collectors) of the data recorded and processed; and to be flexible enough to provide special reports on an ad hoc basis.
A concept of expenditure-income (“exincome”) flows should be developed to create a system that can collect and process information needed by field formations rather than work passively with what they receive. The concept of flows of income, capital, goods, services, etc., within an economy is common in economic theory. It is the basis of the value-added tax system, whereas in Pakistan we are implementing it without the support of a reliable Tax Intelligence System. Goods and services are monitored as they flow from one person to another, and one person’s expenditure becomes another’s income. This concept is at the core of building an AI-based Tax Intelligence System (TIS).
In Pakistan, the major flows are relatively easy to map, as its main source of economic activity is “imports”. The flow of imports can be monitored through the computerisation of all points of customs where imports are handled. Once the exincome stream reaches the contractors, it becomes more difficult to trace, spreading through many channels in a wide delta of economic activity.
The Tax Intelligence System should be able to track some sections of this flow by examining the records of government departments and other large institutions, for which statutory amendments are required in various laws, especially banking laws, even those protecting criminal financial transactions. The TIS can be used by the FBR to alleviate some of the problems associated with revenue shortfalls in Pakistan. Unregistered taxpayers can easily be located, and those most likely to become regular taxpayers can be selected and followed up. Stop-filers can be encouraged or forced to file returns by issuing fairly accurate estimated assessments of their income based on reliable information available in the database.
Estimated assessments can often be wild “top of the hat” guesses that are not taken seriously by taxpayers. If the estimates are too high, the taxpayer is too overwhelmed to respond. If they are too low, the taxpayer would rather pay the tax than file returns.
Near-accurate estimates send a clear message to taxpayers that the Tax Department has reliable information on their activities and could take sterner action if the default continues. In Pakistan, our tax officials are doing just the opposite, and the entire tax system is discredited.
Tax evaders can be quickly detected if the investigation division is able to assess and collect large amounts of additional taxes by detecting companies and other persons engaged in land sales, and companies that are underpricing or overpricing imports and exports. TIS can also be useful when delinquent taxpayers do not pay their taxes. The availability of current up-to-date information can assist with the seizure of assets or income in extreme cases of default.