Partition in numbers
Almost every figure about 1947 is contested, and most articles pick one and move on. This page does the opposite: every number here comes with the method that produced it and the name of whoever produced it - because on this subject the disagreement between the sources is the finding.
The death toll of Partition is quoted anywhere between 200,000 and 3.4 million - a seventeen-fold spread. It is not that historians are careless. It is that counting bodies and counting absences are different operations, and no one ever did the first properly.
See the estimates and their methods belowHow many died: every estimate, and whose it is
Each marker is a published figure. The spread across the bar is the honest state of the evidence - and the reason you should never cite a Partition death toll without naming the source.
Derived from contemporary administrative and police reporting for the period March 1947 to January 1948. It counts deaths that some authority recorded - which in a collapsing administration is a small fraction of them.
Widely used in mid-century accounts and still common in reference works. Usually offered without a stated method, which is itself worth knowing when you cite it.
The figure most textbooks and news articles use. It sits in the middle of the conventional range and has become conventional largely by repetition.
The top of the range that scholarship has quoted for decades. As of 2009 the published spread still ran from 200,000 to this figure without narrowing.
A different method entirely: comparing the 1931 and 1951 censuses to find how many people of the targeted minorities were simply missing. Later work by the same researchers estimated 2.3–3.2 million deaths. The caveat matters - "missing" also captures unrecorded migration.
Presented as published. The demographic estimates are not a correction of the lower figures so much as an answer to a different question: the conventional range asks how many deaths were recorded, and the census method asks how many people stopped appearing. Both are useful; conflating them is not.
Who moved, and in which direction
Punjab's exchange was close to total. Within months, communities that had lived in the same districts for centuries had swapped sides almost entirely.
From Lahore, Rawalpindi, Sialkot, Gujranwala, Montgomery and Lyallpur - including the canal colonies that Sikh farmers had built out of desert two generations earlier, and lost in a fortnight.
From Jalandhar, Ludhiana, Amritsar, Hoshiarpur, Ambala and the princely states of Patiala, Nabha and Faridkot, where the violence was among the worst anywhere.
Most often cited as 14 to 18 million. Punjab and Bengal account for nearly all of it, and Punjab for the great majority of the deaths.
About 20,700 Muslim women recovered from India and about 9,000 non-Muslim women from Pakistan, out of an estimated 75,000 abducted. Many were returned to families who refused them.
Eleven weeks
From the announcement of the plan to the abandonment of the boundary force. Almost everything that happened, happened inside a single summer.
Why Punjab, specifically
Bengal was partitioned too, and suffered terribly. But Punjab is where the violence became something closer to total, and there are structural reasons rather than moral ones.
The population was interleaved, not blocked. Punjab's Muslims, Hindus and Sikhs did not live in separable regions; they lived in the same districts, often the same villages, sharing wells and harvests. No line drawn anywhere could have produced clean majorities, which meant every line created enormous minorities on the wrong side.
The province was full of trained soldiers. Punjab had supplied a disproportionate share of the British Indian Army through two world wars, and in 1946-47 those men were being demobilised back into their villages. The organised, disciplined character of much of the Partition violence in Punjab - column against column, ambush and counter-ambush - reflects that.
Nobody knew where the border was. The Radcliffe Award was published on 17 August, after independence. For weeks people made irreversible decisions about staying or fleeing on rumour. The one institution meant to hold the ring, the Punjab Boundary Force, was disbanded on 1 September as the killing peaked.
Punjab did not divide. It was cut, and it bled out over a single summer.
The demographic result was near-total unmixing. West Punjab's substantial Hindu and Sikh population is essentially gone; East Punjab's large Muslim population likewise. Two of the three communities that made Punjabi culture what it is no longer live in the half of Punjab where much of that culture was made - which is why two scripts now carry one language, and why Lahore is a city most of the Sikh diaspora has never seen.
Nine figures from 1947
Each with its method attached, because on this subject a number without a method is not information.
Radcliffe arrived on 8 July 1947 and the award was published on 17 August. He had never visited India before and never returned.
Both countries became independent on 14–15 August. The line was published on 17 August. Millions celebrated without knowing their nationality.
Around 4.5 million east and 5.5 million west, out of 12–20 million across the whole subcontinent.
From 200,000 to 3.4 million. No other event of this scale in the twentieth century has a range this wide still in use.
The commonly cited estimate, on both sides. The real figure is unknown and almost certainly higher.
About 20,700 Muslim women from India and 9,000 non-Muslim women from Pakistan - two states negotiating over people as though over property.
About 15 Indian and 10 Pakistani, under Major General T. W. Rees. Operational 1 August, disbanded 1 September as ineffective.
Punjab's two great cities, separated by the award into different countries. The road between them has been effectively closed for most of eighty years.
The Indus Waters Treaty of 1960 allocated the five rivers themselves - the last part of Punjab to be partitioned.
If you are citing a Partition figure
Three rules, which the rest of this page exists to support.
Name the method, not just the number. "Around a million people died" is a claim about recorded deaths. "Around 3.4 million people are missing from the 1951 census" is a claim about demographic absence. They are both defensible and they are not the same sentence.
Say which Punjab, and whether you mean Punjab or the subcontinent. Displacement figures for undivided India, for Punjab province and for one direction of Punjab's exchange are routinely quoted interchangeably, and differ by an order of magnitude.
Do not round a contested range into a single confident figure. The spread is not a gap in this page's research; it is the current state of the evidence, and eighty years of scholarship has not closed it. A source that gives one number with no range is telling you less than it appears to.
The narrative account of what this meant for people, rather than for statistics, is in our guide to Partition.
Frequently asked questions
How many people died in the Partition of India?
The honest answer is a range with a method attached. Conventional estimates run from about 200,000 to 2 million, and general histories usually settle on one million. A demographic approach - counting who is missing from the 1951 censuses rather than counting recorded deaths - gives much higher figures: the 2008 study by Prashant Bharadwaj, Asim Khwaja and Atif Mian found around 3.4 million people unaccounted for, and later work by that group estimated 2.3 to 3.2 million deaths. The methods measure different things and do not converge.
How many people were displaced?
Between 12 and 20 million across the subcontinent, most commonly given as 14 to 18 million - the largest forced migration in recorded history. Punjab supplied roughly 10 to 12 million: about 4.5 million Hindus and Sikhs moving east, about 5.5 million Muslims moving west, nearly all of it within months of August 1947.
When was the Radcliffe Line announced?
17 August 1947, two days after independence. Cyril Radcliffe arrived in India on 8 July, had never been there before, and was given roughly five weeks to draw borders through Punjab and Bengal. The delay in publishing meant millions made irreversible decisions on rumour.
How many women were abducted during Partition?
Around 75,000 is the figure usually given, across both sides, and it is almost certainly an undercount. Under the Inter-Dominion Central Recovery Operation, roughly 30,000 were recovered between 1948 and 1956 - about 20,700 Muslim women from India and 9,000 non-Muslim women from Pakistan. The operation itself is now heavily criticised: women were frequently returned to families who rejected them, and many were moved a second time without their consent.
Why did Partition hit Punjab hardest?
Four things together. Its communities were interleaved village by village rather than separated into blocks, so no line could produce clean majorities. It had supplied a huge share of the British Indian Army, and those men were being demobilised into their villages at exactly that moment. The border was published after independence, so people fled on rumour. And the Punjab Boundary Force, about 25 battalions for some 30 million people, was disbanded as ineffective on 1 September - just as the violence peaked.
Was Gurdaspur deliberately awarded to India?
It is the most persistent controversy about the award and it remains unresolved. Gurdaspur district had a slight Muslim majority overall, and its division gave India road access to Jammu and Kashmir - which mattered enormously within weeks. Whether that was Radcliffe's reasoning, Mountbatten's influence, or a consequence of following the tehsil-level religious figures as instructed, is still argued by historians with access to the same documents.
Keep going
The rest of the data on this site, all sourced the same way.
What the two halves look like today - land, people, wheat, water and debt.
Where Punjabis went after 1947, and a timeline running from 1499 to today.
One language, two scripts, two censuses - and the split that Partition made permanent.
Population by country, and the turban and kirpan in law jurisdiction by jurisdiction.
The ~250 historical Sikh shrines left in Pakistan in 1947, and how few of them still function.
The border split the land. It did not split the language.
Punjabi is still one language on both sides of the line, in two scripts. Learning it is the one thing that reads across the border 1947 drew.
Figures reviewed 2026. Death-toll estimates are presented as published rather than reconciled: the conventional range of roughly 200,000 to 2 million appears across standard scholarship and was still the published spread as of 2009; the higher figures come from the demographic work of Prashant Bharadwaj, Asim Khwaja and Atif Mian (2008) and subsequent research by the same group, which compares the 1931 and 1951 censuses to identify people unaccounted for, and which therefore measures absence rather than recorded death - a distinction this page keeps deliberately visible. Displacement figures of 12-20 million for the subcontinent, and roughly 4.5 million and 5.5 million for the two directions of Punjab's exchange, are the ranges given across standard sources; contemporary record-keeping collapsed, so none of them is a count. Abduction and recovery figures come from the Inter-Dominion Central Recovery Operation and the standard literature on it. Boundary Commission and Punjab Boundary Force details are from the commission's own record and the force's operational history. Where a single confident number is offered anywhere on this page, it is because the underlying record supports one; where a range is given, the range is the finding.