An average household in Pakistan's poorest tenth has nearly eight members and spends about Rs 33,000 a month. Applying tax rates to the purchases and bills recorded in the survey, we estimate that around Rs 2,200 goes in tax, as shown in Figure 1. It is collected through the price of cooking oil and soap, the electricity bill and the mobile top-up. Paying it does not require the household to file an income-tax return.
The average household in the richest tenth has four or five members and spends about Rs 152,000 a month. Its estimated tax bill is Rs 17,900, or 11.8 per cent of spending. By this measure, the tax burden is mildly progressive: richer households pay a larger share.
Most of the increase comes from income tax, particularly on salaries. Taxes on purchases and bills take 6.5 per cent of spending in the poorest tenth and 8.1 per cent in the richest; nearly all of the rest of the difference is income tax.
The estimates come from a model built on the 30,123 households in the Household Integrated Economic Survey (HIES) 2024–25. We apply that year's tax rates to recorded purchases and calculate salary tax from reported earnings. Business and farm taxes require a different approach: we assume national collection totals and allocate them across households. Cash assistance is what households say they received; electricity relief is estimated against a regulated tariff benchmark.
This is a selected household account. It excludes public health, education, pensions and other state spending. Beyond the taxes assigned to final purchases and bills, it does not separately estimate taxes paid earlier in production or importing that may be passed on through prices, such as customs duties, corporate taxes and unrecovered taxes on business inputs. The figures describe the part of the system this method can estimate, rather than a household's complete balance with the state.
Figure 1 offers two views. In the first, households are ranked by consumption, adjusted for household size and children. Each tenth contains a tenth of the population. Taxes are then expressed as a share of expenditure, which excludes the notional rent of housing provided without a cash payment. In the second view, households are ranked by income, and taxes are expressed as a share of that income. Throughout the article, income means reported household income plus the income tax we estimate was deducted from monthly-paid employees' earnings before they received their pay. Adding this tax back lets us estimate income before salary withholding; it does not reconstruct every tax and other deduction.
Tax through the shopping basket
As Figure 1 shows, indirect taxes account for almost all the estimated burden outside the richest fifth of the population. The poorest tenth pays 6.5 per cent of expenditure through taxes on purchases and bills, while the richest pays 8.1 per cent.
Exemptions help explain why the rates are fairly close. Wheat flour, fresh milk, vegetables, meat and school fees carry no general sales tax. But many manufactured necessities are taxed, including vegetable ghee, cooking oil, tea and washing powder. Packaged milk has carried the standard 18 per cent rate since July 2024. At that rate, about Rs 15 of every Rs 100 spent at the counter is tax.
Poor households buy taxed necessities even when little of their income is visible to the income tax system. Among richer households, taxes on fuel and utilities take a larger share. General sales tax (GST) on goods moves slightly the other way, from 4.1 per cent of expenditure at the bottom to 3.8 per cent at the top.
Income tax contributes 3.6 percentage points to the rise in the total tax rate between the poorest and richest tenths. Much of it is salary withholding. HIES asks employees for pay after tax and employee contributions, so we work back to the pay that would leave the reported amount after income tax. This calculation produces about Rs 376bn nationally, compared with the Rs 606bn in salary withholding recorded by the Federal Board of Revenue (FBR).
That shortfall does not have a single established explanation. The survey may capture high salaries incompletely; reported pay may already be net of deductions other than income tax; and the model's withholding assumptions may not match every worker's circumstances. Agreement with an administrative total would be useful, but would not by itself show that tax had been assigned to the right households.
What households receive
The Benazir Income Support Programme (BISP) reaches four in ten households in the poorest tenth and fewer than two in every hundred in the richest. Its reported payments are worth about 4 per cent of the poorest group's expenditure.
To value electricity relief, we compare the tariff a household is estimated to pay with the electricity regulator NEPRA's national average tariff of Rs 35.50 per unit for the financial year 2024–25 (FY25). This is a regulated revenue benchmark, not the measured cost of supplying that household. We infer consumption from the reported bill and assume the cheapest tariff category consistent with it, although HIES does not record the billing history needed to establish eligibility.
Under those assumptions, electricity support averages about Rs 1,300 a month in the poorest tenth, rises to around Rs 1,700 in the middle and falls to Rs 940 in the richest. Poor households use fewer units, and four in ten in the bottom tenth report no bill. Richer households use more electricity, but higher applicable tariffs reduce their estimated relief. The richest tenth also pays about Rs 990 a month at rates above the benchmark, and Figure 1 includes that amount as a separate debit in its net balance.
Protected tariffs depend on consumption history and connection characteristics rather than a direct assessment of poverty. For residential connections with a sanctioned load below 5 kilowatts and no time-of-use pricing, the consumption threshold is 200 units in each of the preceding six months. That threshold applies per connection, while household needs depend partly on how many people share it. The survey can illustrate this concern, but cannot measure it fully because it does not identify every household's meters or tariff history.
The government has committed under its IMF programme to replace tariff-differential and cross-subsidy arrangements with targeted support disbursed through BISP by the end of January 2027. If implemented, the change would link electricity assistance more directly to a poverty register.
In the current model, the poorest tenth receives slightly more in cash and tariff support than it pays in the selected taxes and electricity debit. Its net receipt is about Rs 580 a month. The second tenth is close to balance, and the other eight are net payers. These small surpluses at the bottom depend on the electricity assumptions. Valued against a Rs 30 benchmark instead of Rs 35.50, the poorest tenth's margin falls below 1 per cent of expenditure; excluding tariff relief altogether would make it a net payer of about 2 per cent. The calculation shows how much this form of support matters; it does not establish the poorest household's full fiscal position.
The same households ranked by income
Consumption and income do not identify the same people as rich or poor. Consumption reflects what a household uses, including food it grows and housing it occupies. Reported income can fluctuate or cover a different period. Comparing the two gives a fuller picture, provided their differences are not mistaken for evidence of a single cause.
Figure 2 first ranks households separately by consumption and income. Income is more spread out: its ninetieth percentile is 5.5 times its tenth, compared with 4.1 times for consumption. The households at each point on the two curves need not be the same, so the gap between them cannot be read as saving or borrowing.
The third curve keeps the consumption ranking fixed. Within each of the 99 groups shown, median income is below median consumption. Around the middle, the figures are Rs 9,100 and Rs 11,100 per adult equivalent a month. Yet about a quarter of households have reconstructed income above consumption. Borrowing, savings, timing, reporting and differences in definition may all contribute. The survey does not tell us how much each explains.
Figure 3 shows how far households move when ranked differently. One person in three stays in the same tenth; two in three stay in the same or an adjacent tenth. The rankings agree better at the extremes than in the middle. Eight in ten people in the poorest consumption tenth fall within the bottom two income tenths. Six in ten in the richest consumption tenth are also in the richest income tenth.
The bottom income tenth needs particular care. It appears to pay about 20 per cent of income in tax, but its average monthly income is only Rs 13,300 against consumption of Rs 48,800. More than half its members sit outside the poorest consumption tenth. Its high ratio describes a group with unusually low recorded income relative to spending; it should not be treated as the tax rate of a consistently poor group.
Above that bottom tenth, total tax takes between 8.1 and 9.5 per cent of income, with the richest tenth at 9.5 per cent. Rising income tax broadly offsets falling indirect tax as a share of income. The mildly progressive spending pattern becomes broadly flat across most of the income distribution.
Where the money comes from
Figure 4 groups households by their largest source among wages, farming, business and property, remittances, and pensions. Wages lead in every consumption tenth, but their share falls from 78 per cent of households at the bottom to 44 per cent at the top. Business and property rise from 8 to 20 per cent; remittances from 4 to 18 per cent. Farming accounts for between 9 and 13 per cent throughout.
Foreign remittances reach about 1 per cent of households in the poorest tenth and 18 per cent in the richest. Among recipients, average monthly receipts rise from Rs 30,000 to Rs 114,000. Money sent from elsewhere in Pakistan is more evenly distributed. These are patterns of receipt; the survey does not establish what determines who can obtain an overseas job.
Counting six broad sources, including other transfers, 58 per cent of households obtain at least nine-tenths of their positive income from one category. Several wage earners still count as a single category. The share is lower in the poorest tenth, at 54 per cent, than in the richest, at 60 per cent. Cash assistance and private help can give poorer households another source of income alongside earnings.
The payslip and the harvest
Figure 5 compares estimated tax burdens by income source within the richest consumption fifth. Households supported mainly by wages spend about Rs 117,000 a month on average; those supported mainly by farming spend Rs 129,000. They also differ in size and location, so the comparison is between groups rather than otherwise identical households.
The wage-led group has estimated income tax of about Rs 5,900 a month. The agricultural group is assigned about Rs 250; the business-led group about Rs 4,600. Across all the taxes included here, the groups pay 13.0, 7.3 and 10.1 per cent of income respectively. The wage-led group's ratio is nearly twice the agricultural group's under the collection assumptions used.
The assumed business and agricultural tax collections shape the comparison in Figure 5. We assign Rs 150bn of national tax collection to individuals with business and property income, and Rs 3bn to provincial agricultural income tax. Neither is a published, scope-matched subtotal. Both are distributed across households in proportion to simulated statutory liability. Salary tax, by contrast, is calculated directly from each worker's recorded earnings. The resulting comparison is therefore partly determined by the assumed collection totals.
Doubling the business total to Rs 300bn raises the business-led group's rate to 12.8 per cent, close to the wage-led group's 13.0 per cent. Their gap nearly disappears, although the ordering remains. The agricultural group's rate changes little in this exercise because its collection total is held fixed. This sensitivity does not establish how much tax either group actually pays.
The wage–agriculture gap also appears within narrower spending bands. Among households in the top fifth spending Rs 120,000 to Rs 160,000 a month, the estimated rates are 11.0 and 7.3 per cent of income. In the Rs 160,000 to Rs 250,000 band, they are 14.5 and 7.1 per cent. These are useful descriptive checks, but the groups remain unmatched and the collection assumptions remain the same.
We also ask what a common tax schedule would imply. Applying the FY25 federal non-salaried schedule, including its surcharge, to reconstructed business and property income produces about Rs 867bn of liability. Applying it to net farm income and farm-land rent produces about Rs 409bn, with roughly four-fifths of the farm liability in the top income tenth. Only one farming household in five reports enough income to owe anything at all, because the Rs 600,000 exemption protects the ordinary smallholding.
These are illustrative calculations. Agricultural income is taxed by the provinces under different legal bases, and the farm aggregate includes livestock income that does not fall within every province's agricultural tax. The survey's sheets record household shares, so the earning member is only a proxy for the taxable owner. Household surveys may also miss some of the largest incomes. The calculation shows how liability would be distributed under the assumed schedule; it does not measure unpaid provincial tax.
All four provinces enacted agricultural tax reforms during the survey year, intended to align rates with the federal schedule. The first collections were scheduled after that year ended. Sindh subsequently restored its older rates for January–June 2025 and moved the new rates to July. These timing and legal differences make it especially difficult to use one survey year to judge what reform collected in practice.
What the two views tell us
Within this selected account, richer consumers pay a larger share of spending in tax, largely because income tax rises towards the top. Measured against income, the burden is broadly flat across most of the distribution.
The source comparison adds another dimension. On the assumptions used here, wage-led households in the richest fifth face a higher tax-to-income ratio than households supported mainly by farming; the comparison with business income is more sensitive to the collection total assigned to it. The first result is what a tax system built on purchases and payrolls produces by design. The second is the cost of that design: within the richest fifth, on these assumptions, the state asks nearly twice the share of income from a salaried household that it asks from a farming one.
Ranking and expenditure. The analysis covers all 30,123 HIES households. Population weights place a tenth of people in each group. Consumption is divided by adult equivalents, counting children under 18 as 0.8 of an adult. Expenditure excludes the notional rent of owner-occupied, employer-provided and rent-free housing, but retains own-produced food at market value. This is our stated convention rather than an international standard. The Kakwani index is about +0.10 on expenditure and +0.09 on total consumption, using tax and the corresponding welfare measure per adult equivalent, ranked consistently and weighted by people.
Income. Employee earnings include main and second jobs. Detailed farm and enterprise accounts replace reported earnings for members with matching sheets; other members' reported receipts are retained. Farm returns deduct recorded crop and livestock costs, cash land rent and animal purchases. Enterprise income deducts recorded operating costs. Enterprise inventories and changes in the value of animals owned are excluded. Property income uses reported totals, with details as fallback. Using only the separate entries for rent from land and buildings in Section 9A of the HIES questionnaire instead reduces annual rent from about Rs 1.4tn to Rs 0.7tn and the top consumption tenth's mean income by about 4 per cent. This discrepancy is unresolved. No artificial income floor is imposed; 195 households with zero or negative reported income remain in group means.
Tax on earnings. Monthly-paid employees' net cash pay is grossed up under the FY25 salary schedule, including the 10 per cent surcharge above Rs 10mn. This reconstructs salary withholding, but not every deduction from income. Daily-paid employees are excluded from the main withholding calculation; including their pay alongside any monthly earnings of the same person adds about Rs 32bn. Business and agricultural taxes are allocated from the assumed Rs 150bn and Rs 3bn totals. The statutory scenarios apply the non-salaried schedule to each earning member, with household rent assigned to the largest earner in the relevant category. Those assignments are proxies for ownership.
Taxes on purchases. Tax shares are applied to cash purchases across 285 consumption categories, with explicit assumptions about exemptions, product mixtures and informal purchases. An 18 per cent tax on the price before tax is 18/118 of the price paid. The combined petrol-and-diesel category uses an approximate petrol levy share, based on Rs 64.6 a litre over the year and an assumed average pump price of Rs 257. Own-produced and in-kind consumption is assigned no direct tax. Customs duties, corporate taxes passed through prices, and unrecovered taxes on business inputs are not separately estimated.
Cash and electricity support. Cash support includes reported BISP receipts, other government assistance and public zakat. Electricity use is inferred from bills divided by 1.20 to allow for taxes and adjustments, using the FY25 tariff schedule and a favourable eligibility assumption where history is unavailable. Unprotected consumers face the applicable rate on all units; protected consumers retain the previous slab benefit. Bills that fall inside a jump in a tariff schedule, which an annual average of months on either side of a slab boundary can produce, are flagged (15 per cent of households), their units interpolated across the jump and the observed bill taken as the amount paid. Relief and above-benchmark payments are calculated against Rs 35.50 per unit and entered symmetrically in the net account. They are modelled tariff differences, not observed fiscal transfers or measured household supply costs.
Checks and limits. The consumption calculation reproduces published national, rural and urban means closely. Administrative tax totals and survey transfers provide further comparisons but do not validate the estimated distribution. Tax rates are ratios of weighted group means. The results depend on income accounting, tariff eligibility and assumed collections; no confidence intervals are reported. A check restricted to monthly-paid employees who reported last month's net cash pay finds a median of Rs 30,000 in HIES against Rs 40,000 in the Labour Force Survey (LFS); the two surveys count 26mn and 20mn such employees on their own weights. Bonus treatment differs, so this is a partly harmonised descriptive comparison rather than a measure of concealed income.