The press conferences counted missiles. They counted jets, terror camps, intercepted drones, and the precise number of minutes Operation Sindoor lasted. What no government spokesperson in New Delhi or Islamabad stood up to count, in the four days between the night of May 7 and the ceasefire of May 10, 2025, was the number of families sleeping on the floors of school gymnasiums, the children who lost their parents in the scramble onto a stranger’s truck, and the elderly who could not be moved and were left to shelter in basements while artillery walked across the ridgelines above them. The 2025 India-Pakistan conflict produced two stories. One was the story of weapons systems and deterrence thresholds, told in real time on television. The other was the story of the civilians the weapons displaced, and that story was told almost nowhere.

This is the second story. It is the account of what happened to the people of the Line of Control when the doctrine of immediate military retaliation, the doctrine that produced the strike on nine sites across Pakistani territory, met the geography of one of the most densely militarized civilian landscapes on earth. Roughly fifteen million people live in the contested Kashmir region. A large share of them live within artillery range of a border that, for four days in May, became a live front. They were not combatants. They were farmers in Uri, shopkeepers in Poonch, schoolteachers in the Neelum Valley, day laborers in Chakothi. When the firing began, they did the only thing available to them. They ran. The evacuation of the Kashmir frontier in May 2025 was the conflict’s invisible dimension, and the argument of this analysis is that its invisibility was not an accident. A military doctrine that markets itself on precision and proportionality has a structural incentive to keep the displaced out of frame, because the displaced are the part of the ledger that precision cannot explain away.

What follows is a reconstruction of that displacement, mapped on both sides of the Line of Control. It traces the civil defence machinery that was switched on before the strikes, the evacuation of the Jammu frontier districts, the conditions inside the shelter camps, the parallel and far less documented flight of civilians in Pakistan-administered Kashmir, and the staggered, uneasy return that followed the ceasefire. It engages the central dispute that hangs over the whole episode, the question of whether what happened was an organized civil defence operation or a chaotic exodus that exposed how little either state had prepared its border population for the wars those states were willing to fight. And it argues, finally, that the human cost documented here belongs in the same conversation as the missile counts, because a strategy is not only what it destroys on the other side. It is also what it does to the people it claims to protect.

Background and Triggers

The road to the May evacuations began on April 22, 2025, in the Baisaran meadow above Pahalgam, where militants moved through a crowd of tourists, separated men by religion, and shot twenty-six people at close range. The Pahalgam massacre and its minute-by-minute sequence is the trigger event for everything that followed, and it set in motion an Indian response calibrated for maximum pressure. Within twenty-four hours, New Delhi had suspended the Indus Waters Treaty, closed the Attari-Wagah crossing, expelled Pakistani diplomats, and cancelled visa categories. The treaty suspension in particular signalled that this crisis would not be confined to the diplomatic register. India’s decision to put water itself on the table marked the conflict as one in which civilian systems, not only military ones, were considered legitimate instruments of coercion, a shift examined in detail in the analysis of the Indus Waters Treaty as a strategic weapon.

Along the Line of Control, the practical consequences arrived faster than the diplomacy. Skirmishing began around April 24. By the night of April 29, Pakistani forces had opened fire along the international border in the Jammu sector, and the near-nightly exchanges that border residents describe as a grim background condition of their lives began to intensify in both volume and reach. People who had lived through the pre-2021 years, before the ceasefire understanding that had quieted the LoC, recognized the rhythm. What they did not yet know was how far past that old baseline the coming weeks would go.

The water dimension deserves a specific mention here, because it reached the Pakistani border population before any missile did. After the Indus Waters Treaty suspension, local media in Muzaffarabad reported in the last days of April that water had been released from the Uri facility into the Jhelum, producing unannounced fluctuations downstream. Whether or not every detail of those reports was accurate, the effect on the population of Pakistan-administered Kashmir was immediate and psychological. A community that lives along a river it does not control, downstream of a country that has just announced it considers the river a legitimate pressure point, reads a sudden change in water level as a message. The fear that arrived in the Neelum and Jhelum valleys in late April was not yet fear of artillery. It was fear of the river, and it primed the population for flight before the conventional phase had even begun. The escalation, in other words, was already displacing people psychologically while it was still officially diplomatic.

The first formal signal that the Indian state expected a war reached civilians on May 5, when India’s Ministry of Home Affairs ordered a nationwide civil defence exercise to be conducted on May 7 across hundreds of designated civil defence districts in multiple states and union territories. The order specified the operationalization of air raid warning sirens, crash blackout drills, the training of civilian volunteers, and the rehearsal of evacuation plans. The significance of that directive is hard to overstate. India had not conducted a civil defence mobilization on this scale since 1971, during the war that produced Bangladesh. For more than five decades, the machinery of civilian wartime protection had existed mostly on paper, in the dormant structures of the Civil Defence organization and in district disaster management plans written for floods and earthquakes. The May 5 order was the state telling its own population, in the only language a bureaucracy has, that it considered a conventional exchange with Pakistan probable enough to rehearse for.

For the border population of Jammu and Kashmir, the May 5 announcement landed differently than it did in Delhi or Mumbai, where the drill was an abstraction practised between office hours. In Poonch, Rajouri, Uri, and Akhnoor, the people being asked to rehearse evacuation were the people who would actually have to evacuate. The drill was not a simulation of their future. It was a preview of it, and many of them understood it that way immediately. Some began quietly moving children and elderly relatives toward relatives in safer towns before a single Indian missile had been fired. The civil defence exercise, intended as preparation, functioned for the frontier districts as an early warning, and the population that read it correctly was the population that had learned, across decades of shelling, not to wait for official permission to save itself.

There is a grim logic to how the warning signs accumulated, and the border population read that logic better than most analysts did. The treaty suspension on April 23 said the conflict would not stay diplomatic. The skirmishing from April 24 said the Line of Control was already active. The April 29 firing on the international border said the danger had widened beyond the LoC proper into the Jammu plains. The May 5 civil defence order said the state itself now expected a conventional exchange. Each signal was public, and each one moved the probability of war higher in a way that the people living within artillery range could not afford to ignore. By the time the strikes came, a portion of the frontier population had already pre-positioned itself, and that pre-positioning, invisible in any official count because it happened before any order, was the first wave of a displacement that the headline figures would never fully capture.

Two nights later, the warning became the event. In the early hours of May 7, India launched Operation Sindoor, the twenty-three-minute strike campaign against nine sites in Pakistan’s Punjab and in Pakistan-administered Kashmir. The strikes were the opening of the conventional phase of the crisis, and the complete operational picture of Sindoor shows how rapidly the exchange escalated from there. Pakistan’s army responded within hours, and its response was not aimed at Indian airbases or military formations alone. It fell on the towns. The artillery blitz on Poonch that began that morning, examined separately in the reconstruction of Pakistan’s retaliatory shelling, killed civilians and destroyed homes in the first hours of the conflict. The evacuation was no longer a contingency in a district plan. It was happening, in the dark, under fire, and the question was no longer whether the frontier would empty but how, how fast, and at what cost.

The Civil Defence Drill That Came First

The civil defence exercise of May 7 deserves its own examination, because it is the one part of the displacement story that the Indian state designed, announced, and executed deliberately, and because its design reveals what the state thought civilian protection meant in 2025. The exercise carried the working name of a mock drill, and it was conducted on the same day Operation Sindoor’s strikes were launched, which produced an unsettling overlap. Across hundreds of districts, sirens that had not sounded in earnest for half a century were tested. Volunteers practised guiding people to designated assembly points. In several cities, including Delhi, blackout rehearsals dimmed neighbourhoods at scheduled times. Schools and colleges ran evacuation formations. The footage that circulated showed orderly lines, marshals in vests, and the visual grammar of a state that had its civilian protection apparatus under control.

The reality along the Line of Control complicated that grammar within a single news cycle. The drill was national, but the threat was not. For most of India, the May 7 exercise was a rehearsal whose stakes were theoretical. For the Jammu frontier, it was a rehearsal that the same night’s events turned into the real thing, and the gap between the two experiences is the first crack in the official account. A blackout drill in a city four hundred kilometres from the border is a civics exercise. A blackout in Poonch on the night of May 7, with shells already landing, is a survival measure performed by people who had no time to practise it. The civil defence machinery, switched on simultaneously everywhere, was calibrated for the average district and not for the districts that would actually be shelled.

There is a longer history embedded in the choice to revive civil defence at all. The Civil Defence organization in India dates structurally to the 1960s, built in the shadow of the wars with China and Pakistan, and its mandate was precisely this: the protection of civilian life and property during external aggression. For decades after 1971, that mandate atrophied. The organization was repurposed toward natural disaster response, its wartime functions preserved as legal text rather than living capability. The May 5 order reactivated the wartime function, and in doing so it admitted something the Indian security establishment rarely says aloud. The doctrine of immediate retaliation, the posture that had hardened after Pahalgam and that the analysis of India’s evolving response over the fourteen days before Sindoor traces in detail, carries a domestic cost. If India will strike Pakistan whenever a major attack occurs, then India must also be ready to evacuate its own border whenever it strikes. The civil defence revival was the quiet acknowledgement that the new doctrine is a two-front commitment, one front facing the enemy and one front facing the citizens who live where the enemy can reach.

The fifty-four-year gap since the last comparable mobilization is itself a piece of analysis. Between 1971 and 2025, India fought the Kargil war, weathered the 2001-2002 standoff after the Parliament attack, conducted the 2016 surgical strikes and the 2019 Balakot operation, and lived through the entire post-1989 insurgency in Kashmir. None of those episodes triggered a national civil defence exercise of this scale. The 2025 mobilization therefore signals that the Indian state assessed this crisis as carrying a conventional-war risk it had not assigned to any confrontation in half a century. For the analyst, that assessment is data about how seriously New Delhi rated the escalation ladder in May 2025. For the border resident, the same assessment was a far simpler message: the government thinks this could become a real war, and you live where a real war would be fought.

What the drill could not manufacture in two days was the physical infrastructure that civil defence actually requires. The single most repeated complaint from border residents in the weeks that followed concerned the absence of bunkers. Community shelters, the kind of hardened structures that would let a frontier village ride out an artillery exchange in place rather than fleeing, existed in some sectors and not in others, and where they existed they were often too few for the population. The government had, in the years after the post-2019 security reorganization, funded the construction of individual and community bunkers in parts of the border belt, and the programme had delivered thousands of such structures in the most exposed villages of Poonch, Rajouri, and the international border districts. But coverage was incomplete and uneven, the structures were distributed according to a logic that did not anticipate the May 2025 intensity, and an individual bunker built for a single family does not help the family next door. A drill cannot substitute for a bunker. When the shelling came, the villages with adequate shelter could shelter, and the villages without it had only one option, which was to leave.

The bunker shortfall is the hinge on which the entire organized-versus-chaotic debate turns, so it is worth being precise about it. A bunker programme that reaches some villages and not others does not produce a uniformly protected border. It produces a patchwork, in which a family’s odds of being able to stay safely depend on which sector they happen to live in and whether the limited construction reached their hamlet. In May 2025, that patchwork meant that the evacuation was not a single coordinated movement but a mosaic of different local situations: sheltering in place where infrastructure allowed, organized evacuation where the administration reached people in time, and self-organized flight where neither shelter nor timely transport existed. The civil defence exercise rehearsed the leaving. It did not, because two days is not enough time and because the underlying investment had not been made at the scale the doctrine required, give every frontier family the choice of staying safely. That distinction, between an evacuation people chose for tactical reasons and an evacuation people were forced into because no alternative existed, runs through every account that follows.

The Indian Side: Evacuating the Jammu Frontier

The evacuation of the Indian side of the Line of Control concentrated in the Jammu region, in a belt of districts where the border is not a remote abstraction but the edge of the field behind the house. Poonch, Rajouri, and the Akhnoor sector bore the heaviest displacement, with Uri and parts of Baramulla district in the Kashmir Valley, along with Kupwara and the Samba sector, drawn in as the exchanges spread along the line. The figure that recurs in the reporting from those days is tens of thousands of civilians moved, with thousands sleeping in organized shelters on consecutive nights, though the precise total was never consolidated into a single official number, a gap that itself tells part of the story.

It is worth laying out the geography as a map, because the displacement was not uniform and its shape carries meaning. Think of the affected belt as running from the international border in the Jammu plains, northwest through the LoC districts of Poonch and Rajouri, across the Pir Panjal, and on to the Uri sector and the Kupwara frontier in the Kashmir Valley, with the Samba sector and other international border stretches activated by drone incursions and shelling in the conflict’s later days. Each segment of that arc had its own evacuation character. The Jammu plains and the Samba sector saw blackouts and precautionary movement driven heavily by the drone threat. Poonch and Rajouri saw the most lethal artillery and the most desperate flight. Uri and Kupwara saw overnight shelling that emptied specific villages along identifiable routes. This is the bilateral displacement map’s Indian half, and reading it as a single arc rather than a list of incidents shows that the entire inhabited frontier, from the plains to the high Valley sectors, was simultaneously in motion.

Poonch was the epicentre. The town and its surrounding villages took the most concentrated artillery fire of the conflict, and the civilian death toll there was the highest of any single area on the Indian side. The shelling began in the early hours of May 7 and did not let up for days. Residents described a barrage unlike the routine exchanges they had learned to live with, one that struck inside the town itself rather than along its military periphery. Homes were hit. A water tank exploded. Shrapnel tore through windows and walls. The people of Poonch did not wait for a convoy. They evacuated themselves, in whatever vehicle could be found, and the chaos of that self-organized flight, with neighbours pouring into the few cars available and families briefly unable to locate their own members, is documented in the firsthand accounts that survivors later gave. The district administration mobilized in parallel, opening shelter sites and arranging transport, but the administration was responding to an evacuation that the population had already started, not directing one from the front. In Poonch the sequence ran population first, state second, and that ordering is the single most important fact about the worst-hit district’s experience.

Rajouri, adjacent to Poonch along the same stretch of the Line of Control, experienced a closely related version of the same event. The district lies in the same Pir Panjal foothills, its villages are scaled the same way against the same border, and when the shelling spread it produced the same response: families moving toward the district’s safer interior towns and toward Jammu, schools and community buildings pressed into service as shelters, an administration improvising a reception system for people who were already on the road. The reason Rajouri does not feature in the public memory of May 2025 as prominently as Poonch is not that its experience was categorically different. It is that Poonch’s casualty figures were higher and its imagery more dramatic, and the coverage followed the drama. The displacement in Rajouri was real, substantial, and largely uncounted in the headline narrative, which is a small instance of the larger invisibility this analysis is built around.

The Akhnoor sector, closer to the Jammu plains, sat at the junction where LoC shelling and international border firing overlapped. Residents near the LoC in the Akhnoor sector were advised by the authorities to move to safer places as the countermeasures were being organized, and the population complied with the same mixture of official direction and self-initiated movement seen elsewhere. Akhnoor’s position is analytically useful because it shows the displacement was not confined to the high mountain LoC. It reached down into the more accessible plains country, which means the protective challenge was not a niche problem of a few remote villages but a belt-wide problem spanning very different terrain.

In Uri, on the Kashmir Valley side of the line, the pattern was similar in substance and slightly different in texture. Uri sector villages, including Gingal and the cluster around Salamabad, came under overnight artillery fire that killed civilians in their homes; a woman was among those killed in the Uri sector as the shelling continued. Here the evacuation took on the appearance the official account prefers: villagers gathered at roadsides, boarded trucks, and were moved out in identifiable groups toward shelter further from the line. Photographs from those mornings show Kashmiri village women waiting for transport, families climbing onto truck beds, the visual record of an evacuation in progress. The administration in Baramulla district arranged for the receiving end, with schools and colleges on the outskirts of safer towns converted into temporary shelters. The difference between Poonch and Uri is partly a difference in how concentrated and how sudden the shelling was, and partly a difference in how much lead time the population had, but in both cases the underlying reality was the same. The frontier emptied.

The transport and logistics of moving tens of thousands of people off a live frontier deserve their own note, because they are where the improvised nature of the response is most visible. There was no pre-staged fleet of evacuation vehicles waiting in the border districts. The movement happened on whatever wheels existed: private cars, the trucks that normally carry goods and were redirected to carry people, district vehicles, and buses requisitioned as the administration caught up with the event. Routes mattered enormously, because a road within line of sight of the firing is itself a hazard, and the safest hours to move were not always the hours when movement was most urgent. The Jammu and Kashmir Police and the central paramilitary forces, alongside the army units in the sector, were drawn into managing the movement, guiding traffic, and securing the routes, which meant the same security apparatus conducting the conflict was simultaneously conducting the civilian evacuation. That dual tasking is a feature of fighting a war across an inhabited border, and it is one more reason the protective response was stretched.

The receiving infrastructure clustered around Jammu city and other towns set back from the line. Educational institutions were the default, because a school or a college offers what an emergency needs at scale: large covered halls, toilets, water, a compound that can be secured, and a staff accustomed to managing crowds of people. Colleges on the outskirts of Jammu became shelter camps almost overnight. District administrations, drawing on disaster management protocols originally written for other hazards, set up registration, arranged cooked food, and tried to provide separately for the particular needs of women, of small children, and of the elderly and the sick. The machinery that responded was real, and the officials who ran it worked hard under pressure. What the machinery could not do was pretend that the people arriving at its doors had arrived by choice, or that the infrastructure waiting for them had been built for this rather than improvised from whatever the district happened to own.

Where people fled from matters because it maps almost exactly onto the geography of the casualties. The same Jammu region districts that produced the heaviest evacuation, Poonch above all, produced the bulk of the civilian deaths India recorded from the conflict. India’s own accounting, which placed civilian deaths in the conflict at around twenty-one with the majority caused by Pakistani mortar and artillery fire in the Jammu region, describes a single phenomenon from two angles. The dead are the people the shelling caught. The displaced are the people the shelling did not catch because they left in time, or were moved in time, or simply ran. To treat the evacuation as a logistics story separate from the casualty story is to miss that they are the same story, and the comprehensive assessment of civilian casualties on both sides and the displacement mapped here are two readings of one ledger. There is even a precise causal sense in which the evacuation explains why the casualty figure was not higher. Every family that got out of Poonch before a shell found their house is a casualty that did not happen, which means the evacuation, for all its chaos, was doing exactly the work civilian protection is supposed to do. The problem is not that the evacuation failed. The problem is that the evacuation was necessary at all, on this scale, because the alternative of safe in-place sheltering had not been built.

Inside the Shelter Camps

The shelter camp is where the evacuation stopped being a movement and became a condition. Families who had left Poonch or Uri or Akhnoor with whatever they could carry arrived at a college on the edge of Jammu, or at a community hall in a safer town, and there they waited, because there was nothing else to do and nowhere else to go. The camps were the physical form the displacement took for the days the conflict lasted, and the conditions inside them are the part of the record that the official framing works hardest to smooth over.

The basic provisions were present. Cooked food was distributed, and the images of evacuated families eating meals at temporary shelters, of people standing in queues to receive food at a college converted for the purpose, are part of the documentary record of those days. Local administrations and volunteers organized the distribution. Water was arranged. Floor space was allocated. For an improvised response to a sudden mass movement, the camps functioned, in the narrow sense that people were fed and sheltered and not left entirely to themselves.

That narrow sense is not the whole sense. A college hall is a shelter, but it is not a home, and the families inside the camps were not on a difficult trip. They were people whose houses might at that moment be taking shells, who did not know whether they would have a house to return to, and who had no information about how long the situation would last or what they were supposed to do next. The psychological weight of that uncertainty does not show up in a photograph of a food queue. The camps held a population that was physically safe and emotionally suspended, and the suspension was its own kind of damage. Children were out of school. Adults were out of work, with crops in the fields behind the line and livestock unattended. The economic clock kept running while the families sat in the halls, and for a border population that mostly lives close to the margin, days of displacement translate directly into months of recovery.

The composition of the camp population shaped what the camps needed and where they fell short. A frontier evacuation does not move a representative slice of a city. It moves the specific demographic of the border villages, which skews toward the agricultural and the rural, includes a high share of elderly residents whose adult children may work elsewhere, and brings in livestock owners who in many cases would not leave without their animals or who left their animals behind at real economic cost. Pregnant women, infants, the chronically ill, and people with disabilities were all inside the camps, and each of those categories needs something a generic floor-space-and-food model does not automatically provide. A converted college can shelter a crowd. It cannot, without deliberate planning that two days does not allow, become a facility appropriate for a medically diverse population under acute stress. The gaps that residents reported, the thinness of provision for specific needs, were not failures of effort by the people running the camps. They were the predictable result of asking an improvised general-purpose shelter to do a specialized job.

The camps also revealed the limits of a system designed for the average disaster rather than for this one. Disaster management plans written for floods assume a hazard that passes and a population that returns to rebuild. An artillery front is different. It does not recede on a predictable schedule, it can reactivate without warning, and it makes the return itself dangerous. The shelter camps were stocked and staffed for a shorter, simpler event than the one they actually had to absorb, and the strain showed in the gaps: in the families who found the provision for their specific needs thinner than promised, in the elderly for whom a college floor is a genuine medical risk, in the sheer difficulty of running registration and reunification for a crowd that had arrived in pieces. None of this means the response was a failure in the absolute. It means the response was an improvisation, and an improvisation under fire, however competent, is not the same thing as preparedness, and the official language that described the camps preferred the vocabulary of preparedness because preparedness is the version that reflects well on the doctrine.

Registration and family reunification were quieter problems that the camp system had to solve in real time. When a population arrives in pieces, in whatever vehicles were available, across multiple routes and multiple hours, the receiving system inherits the task of reassembling it. Who is here, who is missing, which child belongs to which family that has not arrived yet, where is the elderly relative who was supposed to be in the second vehicle: these are the questions a shelter has to answer before it can do anything else, and answering them for a crowd that did not arrive as an organized convoy is genuinely hard. The camps did this work, and they did it without the digital pre-registration systems that a prepared civil defence apparatus would have had in place. The reunification largely succeeded, which is to the credit of the staff and volunteers, but the fact that it had to be improvised at all is one more entry in the ledger of what advance investment would have made unnecessary.

There is a particular cruelty in the timing that the camp residents understood better than anyone. The civil defence drill of May 5 had rehearsed exactly this, the movement of people to designated places. But the drill had rehearsed the choreography without building the stage. The people in the Jammu colleges were performing, for real and under shellfire, a sequence the state had practised two days earlier as a demonstration. The demonstration had been smooth. The real thing was not, and the distance between the two is the distance between a state that can stage civilian protection and a state that has actually invested in it.

The Other Side of the Line: Displacement in Pakistan-Administered Kashmir

Everything documented so far has a mirror image across the Line of Control, in Pakistan-administered Kashmir, and the mirror image is harder to see clearly, which is itself a finding. The displacement on the Pakistani side was real, substantial, and considerably less documented than the displacement on the Indian side, and the asymmetry in documentation should not be mistaken for an asymmetry in suffering. Civilians in Pakistan-administered Kashmir fled their homes in May 2025. The record of their flight is simply thinner, because the area is less accessible to independent reporting, because the Pakistani state had its own reasons to control the civilian-impact narrative, and because international attention, having framed the conflict as an Indian operation, paid less sustained attention to the Pakistani civilian cost than the symmetry of the situation warranted.

The strikes of May 7 landed inside or near populated areas in Pakistan-administered Kashmir. A mosque structure near Muzaffarabad, the territory’s main city, was damaged, with civilian deaths reported in its vicinity, and Muzaffarabad itself sits in the Neelum and Jhelum river corridor that runs close to the Line of Control. A second mosque structure in the Kotli area, which India identified as a militant facility and the local population knew as a place of worship, was also damaged, and the dispute over what those buildings were is itself part of the conflict’s contested record. Residents of Pakistan-administered Kashmir described being forced to flee their homes and take shelter after the missile strikes, the same compressed sequence their counterparts on the Indian side experienced, the same night. As the conflict extended into the cross-border artillery exchanges of the following days, the firing fell on the Pakistani side of the line as it fell on the Indian side, and the towns and villages within range responded the way border populations always respond. They moved.

The Neelum Valley, which traces the Line of Control along the Neelum river, was among the most affected areas, and it is the part of the Pakistani side for which a little more of the picture survives. Local administration officials in the Neelum Valley acknowledged that many residents had left and that, after the ceasefire, a significant share were waiting to see how the situation developed before deciding whether to return, an account that maps closely onto the behaviour of displaced populations on the Indian side. Chakothi, the crossing point sector, and the cluster of communities along the Jhelum corridor near Muzaffarabad experienced the firing directly. The Neelum corridor in particular had a specific vulnerability that compounded the danger: it is narrow, it is hemmed by mountains, and its road network offers limited routes out, which means that an evacuation there is slower and more exposed than an evacuation across the more open Jammu plains.

Terrain on the Pakistani side is worth dwelling on, because it shaped the displacement in ways the Indian-side geography did not. Pakistan-administered Kashmir’s population is strung along river valleys, the Neelum, the Jhelum, the corridors that the mountains permit, and those valleys run close and roughly parallel to the Line of Control for long stretches. There is less depth, in many sectors, between the line and the inhabited valley floor than there is on the Indian side, and there are fewer large set-back towns of the kind that Jammu city provided as a receiving area. A family fleeing the Neelum Valley is fleeing along a single constrained corridor toward Muzaffarabad, and Muzaffarabad itself was a strike location. The result is that the Pakistani-side displacement was, if anything, more geographically trapped than the Indian-side displacement, with fewer good options for where to go, and the thinness of the documentation should not be allowed to obscure that the underlying physical situation was at least as difficult.

The civilian cost on the Pakistani side, as Pakistan accounted for it, was not smaller than the Indian cost. Pakistan stated that roughly forty civilians had been killed in the conflict from Indian air strikes and cross-border shelling, with over a hundred wounded, figures that exceed India’s stated civilian toll. Both sets of numbers are contested, both states had incentives in the direction of their own framing, and the verification gaps are real, a problem the analysis returns to below. But the contestation over the exact figures should not obscure the structural point. The displacement was bilateral. The Line of Control does not shelter the population on either side from the population’s own government’s decisions, and in May 2025 the people of the Neelum Valley and the people of Poonch were living the same four days, fleeing the same kind of fire, sleeping in the same kind of improvised shelter, and waiting on the same uncertain ceasefire. The shadow war’s doctrine, examined across this series as a posture that treats the other side’s territory as an operational theatre, has a corollary that the doctrine’s architects rarely state. The theatre is inhabited. On both sides of the line, it is inhabited by people who did not choose the war and could not leave the geography.

The question of what protective machinery existed on the Pakistani side is harder to answer than the equivalent question for India, and the difficulty is itself informative. India at least announced a civil defence exercise, gave it a date, and generated a public record of its existence. The Pakistani-side civilian protection response is far less visible in the accessible record, which could mean it was less organized, or could mean it was simply less documented, or both. What the available accounts do show is the same fundamental pattern: a population near the line moving away from the firing, local administration acknowledging the displacement after the fact, and a return that waited on the population’s own assessment of the ceasefire’s durability. The absence of a visible, announced, large-scale civil defence apparatus on the Pakistani side is a real gap in what can be said, and an honest analysis names the gap rather than filling it with assumptions. What can be said with confidence is that civilians in Pakistan-administered Kashmir were displaced in numbers that mattered, that the river-valley geography made their flight harder, and that their experience has been documented less thoroughly than it deserves.

That thinness of the Pakistani-side record has one further consequence worth naming. When a conflict’s humanitarian cost is documented unevenly, the better-documented side becomes the whole story, and the less-documented side becomes a footnote, and the footnote status then feeds back into the politics. An Indian audience could process the 2025 conflict as an event in which India’s border civilians suffered and Pakistan’s were an abstraction, and a Pakistani audience could do the reverse, and both audiences would be working from a real but partial record. A bilateral displacement map, the kind this analysis attempts, is not a neutral exercise. It is a corrective, because it insists that the civilians of Muzaffarabad and the civilians of Poonch occupy the same row in the same ledger, and that any account which renders one of them invisible has failed as analysis before it has even begun as politics.

There is a deeper reason the bilateral framing matters, and it goes to the logic of the conflict itself. The 2025 crisis, like the longer history of India-Pakistan diplomatic and military confrontation, is conducted by two states that each present their own civilians as victims and the other’s as either combatants or abstractions. That asymmetry of sympathy is not incidental. It is load-bearing, because a public that sees only its own side’s civilian suffering will support escalation more readily than a public that sees the suffering on both sides as the same kind of thing. The displacement record, documented bilaterally, is therefore politically inconvenient to both governments in the same way and for the same reason. It shows that the cost of the doctrine is paid, in identical currency, by the people both states claim to be protecting, and that the line between protected citizen and enemy abstraction is a line the governments draw, not a line the artillery respects.

The Return: What Came After the Ceasefire

The ceasefire of May 10, 2025, arrived after four days of conflict, and the negotiations that produced it through the military hotline and external pressure ended the firing faster than most of the displaced had expected. The guns going quiet, however, did not end the displacement. It changed the displacement into a new and slower problem, the problem of return, and the return is the phase of the story that received the least coverage of all, because by the time it was happening the cameras had largely moved on.

The return was staggered, not immediate, and it was staggered for rational reasons. The ceasefire of May 10 was fragile from the first hour. Reports of violations followed quickly, blackouts were reimposed in border towns including Amritsar and in the Samba sector as suspected drones were detected in the days after the formal ceasefire, and the aftermath of the ceasefire was not a clean return to peace but an uneasy, tested quiet. A border family weighing whether to go home was weighing it against the genuine possibility that the firing would resume, and many of them did the prudent thing and waited. On the Indian side, businesses in Srinagar reopened and a measure of normalcy returned to the valley’s towns within a day or two, but the frontier villages emptied themselves more cautiously. On the Pakistani side, the Neelum Valley administration’s own account of residents waiting before returning describes the same calculation. The ceasefire was a reason to hope. It was not yet a reason to trust.

When the displaced did return, many of them returned to damage. The shelling of Poonch had destroyed or damaged hundreds of homes. Villages in the Uri sector had houses with walls breached and roofs gone. On the Pakistani side, the structures near the Muzaffarabad strike and the communities along the shelled corridors had their own damage to absorb. Return, for these families, did not mean resuming an interrupted life. It meant arriving at a house that needed rebuilding, in a season when the agricultural year had already been disrupted, with savings spent on the days of displacement and compensation processes that, where they existed, moved at administrative speed rather than at the speed of need. The political leadership on the Indian side spoke of rehabilitation for the affected border population, and rehabilitation programmes were initiated, but the gap between a promise of rehabilitation made in the immediate aftermath and the lived experience of a family rebuilding a shelled house months later is a gap that the border population of Kashmir has learned, across many such cycles, to expect.

The mechanics of the return deserve a closer look, because the staggering was not only a matter of individual families weighing risk. It was also shaped by what the administrations on each side signaled and failed to signal. On the Indian side, there was no single moment at which the frontier districts were declared safe and the shelter populations told to go home. The colleges and schools serving as shelters needed to return to their actual function, which created a quiet administrative pressure toward closure, but the formal all-clear that would have given families confidence was not the kind of announcement the system was built to issue. People read the environment instead. They watched whether the shelling resumed, whether the blackouts held, whether neighbours who had ventured back stayed back. Return by observation is slower and more uneven than return by instruction, and it left the most cautious families, often those with the most to lose from a wrong guess, in the shelters longest.

The geography of return mirrored the geography of flight. The sectors that had emptied most violently were the sectors that rebuilt most slowly, because the same shelling that drove the fastest evacuation also did the most damage to return to. Poonch, the epicentre of the civilian toll, faced the hardest version of the return problem, with damaged housing stock, disrupted services, and a population that had every reason to move deliberately. The Uri sector villages, Gingal and Salamabad among them, had their own breached homes to assess. On the Pakistani side, the Neelum Valley’s narrow settlements and the damaged structures around Muzaffarabad and Kotli posed the same question in the same terms. Return was not a single event that the ceasefire triggered. It was a sector-by-sector process metered by damage, by trust, and by the slow return of the services that make a frontier village habitable.

There is a quieter dimension of the return that the infrastructure accounting misses entirely. The 2025 conflict, and the communication restrictions and border tensions that came with it, made the already fragile connections across the Line of Control harder still. Kashmir is a divided society in the most literal sense, with families split across the line by the partition of the region and by the surges of militancy that drove people across it in earlier decades. For those families, the May conflict was not only a physical danger. It was another turn of the screw on ties that were already barely holding, another period in which a relative on the other side became unreachable, another reminder that the line runs through households and not only through territory. The return home, when it came, was a return to a home that the conflict had made lonelier, and that loss does not appear in any damage assessment because no damage assessment is built to measure it.

The return also sorted people by vulnerability in ways the camps had partly masked. Inside a shelter, a family with resources and a family with none received roughly the same cooked meal and the same floor space. Outside it, back in the villages, the differences reasserted themselves immediately. A household with savings, an undamaged house, and working-age members could begin recovery within days. A household that had lost its home, its stored harvest, or an earning member to the shelling faced a recovery measured in seasons, and the compensation machinery was not calibrated to that gap. The return, then, was the point at which a displacement that had been experienced collectively became a recovery experienced individually, and the families least able to absorb the shock were the ones the post-ceasefire accounting was least equipped to see.

By the time the displacement had fully unwound, weeks after the ceasefire, the episode had largely vanished from public attention. The strategic conversation moved on to the questions it found more interesting, the performance of weapons systems, the credit dispute over the ceasefire, the doctrinal lessons for the next crisis. The evacuation, having no clean resolution and no dramatic number attached to it, simply faded. That fading is the final stage of the invisibility this analysis began by naming. The displacement was invisible while it happened because the official framing had no incentive to show it, and it became invisible afterward because a slow, undramatic recovery does not compete for attention with a fast, dramatic war.

Key Figures

A humanitarian event of this scale has no single protagonist, but it has categories of people whose decisions and experiences shaped how the displacement unfolded, and naming them is the way to keep the account from collapsing into an abstraction about populations and percentages.

The District Administrations

The district magistrates and their administrations in Poonch, Rajouri, Baramulla, and the other affected districts were the state as the displaced population actually encountered it. They were not the architects of the doctrine that produced the crisis, and they were not consulted on the timing of Operation Sindoor. They were handed the consequences and given hours to manage them. They opened the shelter sites, requisitioned the schools and colleges, arranged the cooked food and the transport, and ran the registration. Their performance was the difference between an improvised response that functioned and an improvised response that collapsed, and on the Indian side it largely functioned, which is a real if limited achievement. The honest assessment of the district administrations is that they did competent work inside a system that had not given them the tools for the actual scale of the event, and that the credit they deserve for the competence should not be allowed to launder the deficiency in the system above them.

What the district administrations improvised was not just shelter but a whole temporary architecture of civilian life. A requisitioned college becomes, overnight, a site that needs cooked food in quantity, drinking water, sanitation, some medical presence, a registration desk, and a way of keeping families together inside it. None of that exists by default in a college. It has to be assembled, by named officials making phone calls, in the same hours that the shelling is driving more people in. The district administrations of the Jammu frontier did that assembly under conditions that would have justified failure, and the absence of a humanitarian catastrophe inside the Indian-side shelters is the measure of how much unglamorous administrative work was done in those hours by people whose names will not appear in any account of the conflict.

The Civil Defence Apparatus

The Civil Defence organization and the volunteers mobilized under it were the formal answer to the question of who protects civilians during external aggression. The May 5 order reactivated them after decades of dormancy in their wartime role. The volunteers who guided people, ran the drills, and staffed the response did genuine service. But the apparatus as a whole was being asked to perform a function it had not seriously trained for in fifty years, with two days of notice, and the structural truth is that an organization cannot be revived by an order. It can only be revived by sustained investment over years, and that investment had not been made. The civil defence apparatus of May 2025 was a real institution performing above its actual readiness, and that is both a credit to the people in it and an indictment of the priorities that had let it atrophy.

Those who actually filled the civil defence role were, in large part, local volunteers, wardens, and municipal staff pressed into a function most of them had only ever seen described on paper. They learned the choreography of the May 7 drill days before they had to perform it for real. That they performed it at all is a credit to a kind of civic willingness that does not appear in any budget line. But willingness is not the same as a trained, equipped, standing capacity, and the distance between the two is exactly the distance that fifty years of treating civil defence as a ceremonial relic had opened. The volunteers closed that distance in May 2025 with their own effort. They should not have been asked to.

The Border Families

The people who evacuated were not passive recipients of a state operation. They were the first responders to their own crisis. The family in Poonch that loaded itself into a neighbour’s car before any convoy arrived, the woman near Uri taking her three daughters toward relatives because she did not trust that anyone had planned where she should go, the residents of the Neelum Valley who left on their own reading of the danger: these were active decisions made by people exercising the only agency the situation allowed them. The border population of Kashmir has a hard-earned expertise in self-preservation, accumulated across decades of being the part of the country where the country’s conflicts are physically conducted. In May 2025 that expertise saved lives, and it is worth stating plainly that the lives it saved were saved by the civilians themselves as much as by anyone sent to help them.

That expertise is also generational, transmitted the way hard knowledge is transmitted in places that need it. The grandparents in a Poonch household had lived the high-shelling years before the 2003 ceasefire. The parents had lived the crises of 2016 and 2019. The children were now logging their own entry in that inherited ledger. Each generation teaches the next which rooms are safer, which routes lead out, which relatives inland will take a family in, how much warning the firing tends to give. It is a curriculum no one chose to write, and the fact that the border population of Kashmir carries it at all is itself a quiet indictment of how routine their displacement has been allowed to become. A population should not have to be this good at fleeing.

The Officials Who Set the Frame

The political and military leadership on both sides made the decisions that produced the displacement and the decisions that shaped how it was described. The Indian leadership chose the doctrine of immediate retaliation, ordered the strikes, and then controlled a public narrative that foregrounded precision and kept the displaced in the background. The Pakistani leadership chose the form of its retaliation, the artillery on the towns, and ran its own narrative of victimhood and response. Neither leadership stood up to give the displacement the same prominence it gave the missile counts. That was a choice, not an oversight, and the people who made it are as much a part of the displacement story as the people who lived it, because the invisibility of the displaced was authored at the level where frames are set.

The Analysts and the Record

A final category belongs in any honest accounting of the displacement, the small number of analysts and observers who insisted the civilian story be recorded at all. Michael Kugelman’s argument that the humanitarian dimension of the 2025 conflict was systematically underweighted, and Radha Kumar’s longer historical framing of the Line of Control’s civilians as the recurring casualties of crises they do not control, are not neutral background. They are part of the event. An event that is not documented effectively did not fully happen in the public record, and the people who did the documenting against the pull of a strategic conversation that found the missiles more interesting were doing necessary work. The 2025 conflict had no shortage of people willing to count missiles. It had a shortage of people willing to count the displaced, and that shortage is itself one of the findings. The displacement needed its chroniclers as much as it needed its relief workers, because a humanitarian cost that leaves no record generates no pressure, and a cost that generates no pressure is a cost the system has no reason to reduce.

Consequences and Impact

The consequences of the May 2025 displacement extend well past the days the camps were open, and they fall into registers that the conflict’s strategic accounting was not built to capture.

The immediate human cost is the clearest. Tens of thousands displaced on the Indian side, a substantial and less precisely counted displacement on the Pakistani side, civilian deaths concentrated in the same border belt that produced the heaviest evacuation, hundreds of homes destroyed or damaged across both sides of the line. The displacement and the casualties are the same event seen from two distances, and the bilateral civilian casualty picture is incomplete without the displacement that runs alongside it.

Economic damage is larger than the immediate cost and longer-lasting. The border population of Kashmir is overwhelmingly agricultural and small-trade, living close to the margin in a short growing season. Days of displacement at the wrong point in the agricultural calendar cost a season. A shelled house costs savings that were not there to spend and a rebuilding process that competes with the next planting. Compensation and rehabilitation programmes, where they functioned, ran at bureaucratic speed. The net effect is that the economic shock of four days of conflict is paid off by border families over months and years, and that long tail of cost is invisible precisely because it is slow.

The psychological cost is the least measured and arguably the most durable. The families who sheltered in the Jammu colleges, the children who experienced the shelling and the chaotic flight, the elderly moved out of their homes under fire: these are populations carrying trauma that no damage assessment records. For a border population the trauma is not a single event. It is cumulative, layered onto the residue of every previous cycle, and the 2025 conflict added a thick new layer onto people who were already carrying the residue of the 1999 Kargil war, of the long high-intensity shelling years before the 2003 ceasefire, of the renewed exchanges that followed the 2016 and 2019 crises. The communication restrictions and the strain on cross-line family ties compounded it. A child who spent four days in a college hall listening for the next shell does not file that experience away when the ceasefire is announced. It becomes part of how that child understands the place they live, and the place they live is a place their own country has decided it will defend by methods that periodically require them to flee it. The return home was a return to a place the conflict had made materially poorer and emotionally heavier, and the emotional weight is the part of the cost that compounds across a lifetime rather than clearing with a rebuilding grant.

The cost was also distributed unequally inside the displaced population itself. Evacuation under fire is hardest on the people least able to move quickly, the elderly, the disabled, the very young, pregnant women, the sick. A doctrine that relies on rapid civilian flight as its de facto protection mechanism is, in practice, a doctrine that protects the mobile better than the immobile. The shelter camps did what they could to level that difference once people were inside them, but the journey to the camp, the chaotic phase before the administrative response caught up, sorted people by physical capacity in a way that no compensation programme afterward could correct. That is a structural feature of protection-by-evacuation, and it is one more reason that protection-by-infrastructure, the bunkers that let people shelter where they are, is not an interchangeable alternative but a materially different and fairer thing.

There is a consequence for the relationship between the border population and the state that is harder to name but real. Every cycle in which the state’s doctrine produces the displacement and the state’s preparation fails to soften it teaches the border population a lesson about where it stands in the order of national priorities. The woman near Uri who said the government should have planned where people would go was not making a narrow complaint about one evacuation. She was describing a relationship. The frontier population of Kashmir is asked to absorb the physical consequences of national strategy and is given, in return, an improvised response and a rehabilitation promise that moves slowly. That exchange has a cumulative effect on how the border feels about the centre, and a state that needed the border population’s cooperation and steadiness in a crisis would be wise to treat the protection deficit as a political cost and not only a humanitarian one.

There is an institutional consequence as well, and it cuts in a more ambiguous direction. The May 2025 displacement was a real-world stress test of India’s revived civil defence posture, and the test produced data. It showed that the apparatus could be switched on, that district administrations could improvise a functioning response, and also that the underlying investment, the bunkers, the trained capacity, the infrastructure built for an artillery front rather than a flood, was not adequate to the doctrine the state had adopted. Whether that data produces sustained investment or fades with the news cycle is the open question. The honest reading of the institutional history, given how civil defence atrophied for fifty years once the immediate threat receded, is that the political incentive to invest in unglamorous civilian protection is weak, and that the lesson of May 2025 is at real risk of being filed and forgotten.

The strategic consequence is the one this analysis most wants to insist on. The 2025 conflict is being studied for what it taught about deterrence, about the implications of two nuclear powers exchanging fire, about the performance of doctrine under pressure. The displacement belongs in that study and not in a separate humanitarian appendix. A doctrine of immediate retaliation is, among other things, a decision about the border population, because the border population is the population that will be evacuated every time the doctrine is exercised. To evaluate the doctrine honestly is to count that cost as part of the doctrine’s price, and the strategic literature that omits it is not being rigorous. It is being selective.

The displacement also had a consequence for how the conflict was understood beyond the subcontinent, and it is the consequence of an absence. International coverage of the four days followed the strategic drama, the air battle, the missile exchanges, the question of escalation between two nuclear states. The bilateral displacement, the millions moved on both sides, was the part of the story most likely to convey the conflict’s actual human scale to an outside audience, and it was the part least represented. A reader following the 2025 conflict closely from abroad could come away with a vivid sense of what was launched and almost no sense of who fled. That gap in the external record is not cosmetic. It shapes how much pressure, if any, the humanitarian cost of the next crisis will generate, and a cost that was never widely seen is a cost that generates no pressure at all.

Analytical Debate

A central dispute hangs over the May 2025 evacuation, and it can be stated as a single question. Was what happened on the Line of Control an organized civil defence operation, a state competently protecting its border population during a war, or was it a chaotic exodus, a population saving itself because the state had not prepared for the war it chose to fight? The two readings are not merely different emphases. They are different verdicts on the same evidence, and the choice between them matters because it determines whether May 2025 is read as a success to be repeated or a failure to be corrected.

The case for the organized reading is not empty. The Indian state did issue a civil defence order in advance. District administrations did open shelters, arrange transport, and distribute food. The evacuation in sectors like Uri did proceed in identifiable, managed groups. The machinery that the state pointed to as evidence of organization genuinely existed and genuinely functioned at a basic level. An official describing the response as an organized civil defence operation was not inventing the organization. It was there.

The case for the chaotic reading is built on what the organized reading leaves out. The evacuation of Poonch, the area worst hit, was self-organized under fire before the administration’s response arrived, with families pouring into whatever vehicles existed and briefly losing track of their own members. The defining civilian complaint was the absence of bunkers, the absence, that is, of the single piece of infrastructure that would have made organized in-place sheltering possible and that fifty years of civil defence atrophy had failed to provide at adequate scale. The civil defence drill rehearsed the choreography two days before the event but could not build the missing infrastructure in that window. And the testimony of the displaced themselves, the woman near Uri saying the government should have planned where people would go, is not the testimony of a population that experienced a well-run operation.

An honest adjudication is that both readings describe real features of the same event, and that the organized reading is true at the level of the district response and false at the level of the system, while the chaotic reading is true at the level of the system and unfair to the district response. The district administrations and the civil defence volunteers performed an organized response. The state, across the decades that preceded May 2025, had not organized the conditions that would have let that response be anything other than an improvisation. A competent improvisation under fire is a real achievement and it is also, definitionally, evidence of inadequate preparation, because a prepared state does not need to improvise. The verdict, then, is that May 2025 was an organized response to a situation the state had failed to prepare for, and that the official framing’s error is not that it claimed organization but that it used the organization at the bottom of the system to imply preparedness at the top, which the evidence does not support.

It is worth being precise about what would change this verdict, because an adjudication that cannot say what would falsify it is not an adjudication but a preference. The organized reading would earn the stronger claim if the evidence showed an adequate bunker network the population chose not to use, a clear and early all-clear and return process, and a civil defence apparatus that was trained and equipped rather than reactivated by decree. The chaotic reading would earn the stronger claim if the district response had actually collapsed, if the shelters had failed, if the improvisation had not held at all. Neither pure case is what the evidence shows. The evidence shows a functioning bottom and an unprepared top, which is why the split verdict is not a hedge but the most accurate description available. Anyone arguing for a cleaner verdict in either direction is choosing which half of the evidence to set aside.

The debate also looks different depending on which side of the line it is applied to. On the Indian side, there is at least a real argument to have, because there was a formal civil defence order, a drill, and a visible administrative response to assess. On the Pakistan-administered side, the organized reading has even less to work with. The displacement there was, by most available accounts, closer to the pure self-evacuation end of the spectrum, with the protective machinery thinner and the terrain less forgiving. A debate framed only around the Indian-side response can therefore mislead, because it implicitly treats the better-documented half of the displacement as the whole. The bilateral frame this analysis has insisted on is not only a humanitarian corrective. It is an analytical one, because the organized-versus-chaotic question genuinely has different answers on the two sides of the line, and a verdict that does not say so is incomplete.

This adjudication is consistent with how the most careful outside analysts have framed the conflict’s civilian dimension. Michael Kugelman, in his analysis of the humanitarian register of the 2025 crisis, has been among the clearest in arguing that the human cost of the conflict was systematically underweighted relative to its strategic drama, that the civilian story was the conflict’s missing dimension in international coverage, and that the displacement on both sides deserved attention proportional to the missile exchanges it accompanied. Radha Kumar, whose work on the modern history of Kashmir traces the civilian experience of India-Pakistan military crises along the Line of Control across decades, situates May 2025 in a long pattern in which the border population is the constant casualty of escalations it does not control, and in which each crisis is managed rather than prevented, leaving the same communities to be evacuated again and again. Kugelman’s frame emphasizes the documentation deficit, the way the displacement was rendered invisible. Kumar’s frame emphasizes the historical recurrence, the way the displacement was nothing new. The two frames are complementary, and together they support the reading offered here. The May 2025 evacuation was both underdocumented and unsurprising, and a state response that is competent in the moment but predictably necessary every few years, with the underlying protective investment never made, is a response that has confused crisis management for civilian protection.

One further dimension of the debate deserves naming, because it is where the analysis is most contested. A defender of the Indian doctrine would argue that the displacement, however regrettable, is the price of a posture that ultimately protects civilians by deterring the attacks that target them, that the alternative to immediate retaliation is the slow normalization of mass-casualty terrorism, and that a few days of evacuation is a smaller civilian cost than the Pahalgam massacre the doctrine exists to deter. That argument is not frivolous and it deserves a real answer rather than dismissal. The answer is that the argument proves too much if it is used to make the displacement disappear from the ledger. A cost can be justified and still be a cost. If the doctrine is worth its price, then the honest course is to name the price, count it, and invest in reducing it, by building the bunkers and the infrastructure that would let the border population shelter in place rather than flee. The doctrine’s defenders and its critics can disagree about whether the price is worth paying. What neither should accept is a public accounting in which the price is not shown at all, and the persistent invisibility of the displaced suggests that the official preference is for exactly that unshown accounting.

Why It Still Matters

The May 2025 evacuation matters now, well after the camps closed and the displaced returned, for reasons that run directly into the strategic future of the India-Pakistan relationship.

It matters because the doctrine that produced it is still in force. The Indian posture that hardened after Pahalgam, the commitment to immediate and visible military retaliation for major terrorist attacks, was not a one-time decision. It is the standing policy, and the diplomatic and strategic trajectory of the relationship gives little reason to expect the conditions that trigger it to disappear. Every future exercise of that doctrine will produce another evacuation of the same border belt, because the geography does not change and the doctrine’s logic runs through that geography. The displacement of May 2025 is not a closed historical episode. It is a template, and the people of Poonch and Uri and the Neelum Valley know it, which is why their return was so cautious and their trust in the ceasefire so provisional.

It matters because the institutional lesson is still unlearned. The stress test of May 2025 generated a clear finding, that the revived civil defence posture functions at the level of improvisation but lacks the infrastructure for the doctrine it serves. The window in which that finding can be converted into sustained investment, into bunkers and trained capacity and protection built for an artillery front, is the window of attention that follows a crisis, and that window closes fast. The history of Indian civil defence is a history of exactly this lesson being learned in a war and forgotten in the peace that follows. Whether May 2025 breaks that pattern or repeats it is genuinely undecided, and naming the displacement clearly is part of what keeps the pressure on the decision.

There is also the backdrop against which the next evacuation would happen. The 2025 conflict was, at every moment, a confrontation between two nuclear-armed states, and the displacement unfolded inside that fact. A border population that evacuates during a conventional exchange is also a population living with the knowledge of what a worse escalation would mean, and the psychological weight of the flight cannot be separated from that backdrop. The strategic community studies the escalation ladder of the 2025 crisis with care. The people of the frontier live on its first rung, and every analysis of how close the rungs are is also, whether it says so or not, an analysis of how exposed those people are. Keeping the displacement in the strategic conversation is a way of keeping the human stakes attached to the ladder being studied.

It matters because the lesson is bilateral and the correction would have to be too. The displacement was not an Indian problem with a Pakistani footnote. It was a single humanitarian event distributed across both sides of a line, and the protective deficits, the thin bunker provision, the reliance on flight, the slow rehabilitation, are recognizable on both sides. That symmetry is politically inconvenient, because it does not fit a narrative in which one side is purely the aggressor and the other purely the victim. It is also the most important practical fact about the displacement, because it means the populations of Poonch and the Neelum Valley have more in common with each other than either has with the strategic planners who speak in their name. Any serious reduction of the displacement cost would have to be built on recognizing that shared exposure, and the first step toward that recognition is refusing to let the two halves of the displacement be told as separate stories.

It matters because the invisibility is a choice that can be reversed. The displaced were kept out of the frame because the official narrative had no incentive to show them, and because the strategic conversation found the weapons more interesting than the people the weapons moved. Neither of those is a law of nature. An analytical record that puts the displacement back in the frame, that maps it on both sides of the line, that insists the civilians of Muzaffarabad and the civilians of Poonch belong in the same ledger as the missile counts, is a small correction to a large omission. The correction does not change what happened in May 2025. It changes whether what happened is understood, and understanding is the precondition for the displacement being treated, in the next crisis, as a cost to be reduced rather than a fact to be hidden.

The namable claim of this analysis can be stated in a sentence. While missiles flew and jets fought, millions of civilians on both sides of the Line of Control fled their homes, and the shadow war’s doctrine has human costs that its architects have a structural incentive never to acknowledge. The four days of May 2025 proved both halves of that claim. The displacement was real, it was bilateral, and it was very large. And it was, almost everywhere that mattered, kept out of sight. The work of this analysis has been to bring it back into sight, because a strategy that is honest about what it destroys on the other side should be equally honest about what it does to the people it claims to defend, and the evacuation of the Kashmir frontier is the part of the 2025 conflict where that honesty is still owed.

Frequently Asked Questions

Q: How many civilians were evacuated during the 2025 India-Pakistan conflict?

The most consistent reporting from the conflict described tens of thousands of civilians evacuated from villages near the frontier on the Indian side alone, concentrated in the Jammu region districts of Poonch, Rajouri, and the Akhnoor sector, along with Uri and other sectors in the Kashmir Valley. Thousands slept in organized shelters on consecutive nights. A precise consolidated total was never published by either government, which is part of why the displacement is hard to pin down. On the Pakistan-administered Kashmir side, the displacement was also substantial but considerably less documented. The honest answer is that the evacuation ran into the tens of thousands at minimum on the Indian side and was large on the Pakistani side, with the exact figure obscured by the absence of a single official accounting.

Q: Where did evacuated families go?

On the Indian side, most evacuated families were moved to towns set back from the Line of Control, with Jammu city and its outskirts serving as a major receiving area. Schools and colleges were the default shelter sites, converted into temporary camps because they offered large covered halls, sanitation, water, and securable compounds. Some families did not go to official shelters at all but moved in with relatives in safer towns, often having left before any organized convoy arrived. On the Pakistan-administered Kashmir side, displaced residents similarly moved away from the shelled corridors near the Line of Control, with the Neelum Valley among the areas people left. The common pattern on both sides was movement from the immediate border belt toward whatever lay just beyond artillery range. These journeys were often strikingly short. Safety in this geography is not a distant city but the next town inland, which means the displaced rarely traveled far and remained close enough to their homes to worry about them constantly, a proximity that shaped the anxious, provisional quality of the whole displacement.

Q: Were the evacuations organized or chaotic?

Both descriptions are accurate for different parts of the event. The Indian state issued a civil defence order in advance, and district administrations opened shelters, arranged transport, and distributed food, which is real organization. But the evacuation of Poonch, the worst-hit area, was largely self-organized under fire before the administration’s response arrived, with families using whatever vehicles they could find. The defining civilian complaint was the absence of adequate bunkers, meaning many people had no option but to flee. The fair adjudication is that the district-level response was organized while the underlying system was unprepared, so May 2025 is best understood as a competent improvisation rather than a well-resourced operation.

Q: What was the civil defence drill conducted on May 7, 2025?

On May 5, 2025, India’s Ministry of Home Affairs ordered a nationwide civil defence exercise for May 7 across hundreds of designated districts in multiple states and union territories. It included testing air raid sirens, blackout rehearsals, training of civilian volunteers, and rehearsal of evacuation plans. It was the largest such mobilization since 1971, the year of the war that produced Bangladesh. The drill happened to fall on the same day Operation Sindoor’s strikes were launched, so for the Line of Control districts the rehearsal and the real event collapsed into each other within hours.

Q: How long were civilians displaced during the conflict?

The active conflict lasted four days, from the night of May 7 to the ceasefire of May 10, 2025, and the most intense displacement coincided with that window. But the displacement did not end with the ceasefire. The return was staggered over the following days and weeks because the ceasefire was fragile, with reported violations and reimposed blackouts in border areas, and many families chose to wait before going home. For those returning to destroyed or damaged houses, the displacement effectively extended into the months of rebuilding that followed, so the four-day figure describes the acute phase but understates the full duration of disruption.

Q: Were children separated from families during the evacuation?

Firsthand accounts from the evacuation of Poonch describe exactly this kind of brief separation. As shelling hit the town in the early hours of May 7, residents poured into the few available vehicles in a self-organized rush, and for a period it was difficult for families to locate all their members. These were generally short separations resolved within hours rather than permanent ones, but they capture the chaotic texture of an evacuation that the population conducted itself, under fire, faster than any organized system could direct it. The reunification challenge was one of the clearest indicators that the worst-hit area’s evacuation outran the state’s capacity to manage it.

Q: What happened to people who could not evacuate?

The people who could not leave, principally the elderly, the sick, and those caring for them, were the population for whom the absence of community bunkers mattered most. Where hardened shelters existed, they could ride out the shelling in place. Where shelters did not exist or were too few, the immobile population sheltered in basements and the most protected interior spaces of their homes while artillery exchanges continued nearby. This is the group the civil defence system was least equipped to protect, because an evacuation-based response inherently assumes people can move, and in-place protection requires infrastructure that decades of civil defence atrophy had not adequately built along the frontier. Their situation also placed an impossible choice on their families, who had to decide between staying behind with a relative who could not travel and moving the rest of the household to safety. That dilemma, faced in the dark with shells landing, is one of the least visible costs of a protection model that substitutes flight for infrastructure.

Q: How did the evacuation differ on the Indian and Pakistani sides of the Line of Control?

The substance was similar and the documentation was not. On both sides, civilians fled the shelled border belt the same nights, sheltered in improvised accommodation, and returned cautiously after the ceasefire. The Indian-side displacement, concentrated in the Jammu region, was relatively well documented through reporting and photography. The Pakistan-administered Kashmir displacement, including in the Neelum Valley and around Muzaffarabad, was real and substantial but far less covered, because the area was less accessible to independent reporting and because international attention framed the conflict primarily as an Indian operation. The asymmetry in the record should not be mistaken for an asymmetry in suffering.

Q: Have all displaced civilians returned home?

Most displaced civilians on both sides had returned within weeks of the May 10 ceasefire, once the quiet held. But the return was conditional and slow rather than immediate, because the ceasefire was fragile and border families had rational reasons to wait. Those returning to destroyed homes faced a rebuilding process that extended the disruption well beyond the physical return. For families divided across the Line of Control, the conflict also worsened already strained cross-line ties, a form of loss that no return resolves and no damage assessment records.

Q: How many civilians died in the 2025 India-Pakistan conflict?

The figures are contested. India stated that around twenty-one civilians were killed, with the majority caused by Pakistani mortar and artillery fire in the Jammu region, particularly Poonch district. Pakistan stated that roughly forty civilians were killed by Indian air strikes and cross-border shelling, with over a hundred wounded. Independent verification was limited, and both states had incentives shaping their accounting. What is clearer than the exact totals is the geography: the civilian deaths concentrated in the same border belt that produced the heaviest evacuation, which is why the casualty story and the displacement story are best read as one. The fuller picture is examined in the dedicated assessment of civilian casualties on both sides.

Q: Why was the displacement not covered as much as the military operations?

Several factors converged. The official narratives on both sides had no incentive to foreground the displaced, because a doctrine that markets itself on precision and proportionality is undercut by a visible mass of forced evacuees. The strategic conversation found weapons systems and deterrence thresholds more dramatic than a slow humanitarian story. And the displacement had no single dramatic number or clean resolution to anchor coverage, so it faded once the active conflict ended. The invisibility was partly a deliberate framing choice and partly a structural feature of how conflicts get reported, with the fast and dramatic crowding out the slow and undramatic.

Q: What was Operation Sindoor and how did it trigger the evacuations?

Operation Sindoor was India’s strike campaign launched in the early hours of May 7, 2025, hitting nine sites in Pakistan’s Punjab and in Pakistan-administered Kashmir in response to the April 22 Pahalgam massacre. The strikes opened the conventional phase of the crisis. Pakistan’s army retaliated within hours, and crucially its retaliation included heavy artillery shelling of Indian border towns rather than military targets alone. That shelling, falling on places like Poonch, is what made the evacuation of the frontier unavoidable. The complete operational account of Sindoor is covered separately in this series.

Q: Did civilians get any warning before the conflict began?

The clearest formal signal was the May 5 civil defence order announcing the May 7 nationwide drill, which border populations correctly read as the state preparing for a probable conventional exchange. Some families in the frontier districts began moving vulnerable relatives to safer areas on the strength of that signal alone. Beyond that, the warning border residents had was the warning they always have, the intensifying skirmishing along the Line of Control through late April. There was, however, no specific evacuation order issued to the frontier before the strikes, so for most families the actual trigger to leave was the shelling itself rather than any advance instruction.

Q: What were conditions like inside the shelter camps?

The basic provisions functioned. Cooked food was distributed, water was arranged, and floor space in converted schools and colleges was allocated, with district administrations and volunteers running the response. But the camps held a population that was physically safe and emotionally suspended, with no information on how long the crisis would last, whether their homes still stood, or what to do next. Children were out of school and adults out of work, with the economic clock running. The system, built on disaster management protocols designed for hazards like floods, was stocked and staffed for a shorter, simpler event than an artillery front, and the strain showed in the gaps. Privacy was minimal, with families partitioning hall space as best they could. Medical provision was thin beyond basic first aid. And the camps had no real mechanism for the thing the displaced most needed, which was reliable information, leaving rumor to fill the vacuum that an official communication channel should have occupied.

Q: How does the May 2025 displacement compare to earlier India-Pakistan crises?

It belongs to a long and recurring pattern. The border population of Kashmir has been displaced by escalations repeatedly across the decades since the region was partitioned, and scholars of the region’s modern history place May 2025 within that recurrence rather than treating it as unprecedented. What was distinctive in 2025 was the scale and suddenness, driven by India’s doctrine of immediate retaliation and Pakistan’s choice to shell towns rather than only military positions, and the revival of a national civil defence mobilization not seen since 1971. The continuity, the constant being the border population paying for escalations it does not control, is the more important point than the novelty.

Q: What does the evacuation reveal about India’s military doctrine?

It reveals that the doctrine of immediate retaliation is a two-front commitment, one front facing Pakistan and one facing India’s own border citizens, who must be evacuated every time the doctrine is exercised. The May 5 civil defence revival was the quiet official acknowledgement of this. The deeper finding is that the doctrine had been adopted without the matching civilian-protection investment, the bunkers and infrastructure that would let the frontier population shelter in place rather than flee. An honest evaluation of the doctrine has to count the displacement as part of its standing price, which is the analytical point this assessment most insists on, and which connects to the broader question of what it means for two nuclear-armed states to fight a limited war across an inhabited border.

Q: Is the displacement risk likely to recur in future conflicts?

Yes, and the border population understands this clearly, which is why its return in 2025 was so cautious. The doctrine that produced the evacuation remains in force, the geography that channels it does not change, and the conditions that trigger Indian retaliation show no sign of disappearing. Unless sustained investment is made in frontier civilian protection during the window of attention that follows a crisis, a window that the history of Indian civil defence suggests closes quickly, the next major escalation will produce another evacuation of the same border belt. The May 2025 displacement is best understood not as a closed episode but as a template for what the doctrine will keep producing.