No case for women: gender-based violence and the limits of justice in Pakistan

For an attack on a woman to become a case, it first has to be reported to the police. They then have to register the complaint and investigate it. In court, the questions change: whether the prosecution can proceed, whether the evidence supports a conviction, and whether the accused waits in custody or at home. A case that reaches a verdict can spend years on appeal.

Survey data tells us how often attacks reach the police. Police and court data show what happens to the cases those institutions handle. Together, these sources help us examine reporting, registration, trial outcomes and decisions in the higher courts.

They concern different offences, people and years, and carry no common identifier that would let us trace a case across institutions. We cannot follow an individual attack from a survey into a police return or a court judgment, or calculate the chance that it ends in a conviction. Each stage answers a question of its own. The method note explains the limits of the sources.

Reporting: which attacks reach the police?

Between 2017 and 2020, the provincial Multiple Indicator Cluster Surveys (MICS) asked women and men aged 15 to 49 about physical attacks, rather than sexual violence specifically, in the previous three years. Those who said they had been attacked were then asked whether the police had been told about the most recent incident.

Among 1,618 women with an answer to that question, only 18.6 per cent said the police were told. Figure 1 shows how reporting varies across the combined provincial samples.

In Punjab, police were told about 15.5 per cent of attacks on women, compared with 38.5 per cent of attacks on men. The disparity is a marked gender gap, although most attacks on men also went unreported. Distrust of the police and the state may help explain low reporting, but the survey did not ask respondents why incidents went unreported.

The difference between women and men was less pronounced elsewhere. In Sindh, women reported a slightly higher rate of police notification than men, at 30.5 per cent against 28.4 per cent. That estimate, however, rests on only 138 attacked women; KP's rests on 126. Such small samples limit what can be concluded about provincial differences.

Women with more schooling or household wealth were less likely to say they had experienced a physical attack. In the poorest fifth of households, 1.3 per cent said they had experienced an attack in the previous three years, compared with 0.5 per cent in the richest fifth. Women with more schooling were also less likely to say so.

But the pattern was different when it came to informing the police. Among women who said they had been attacked, neither higher schooling nor greater wealth was associated with a consistent increase in police reporting. The police were told in about 19 per cent of attacks in both groups: women with no schooling and those with secondary schooling or more. Across wealth groups, the share fell unevenly, from 22.6 per cent in the poorest fifth to 14.7 per cent in the richest. Women who owned a phone recorded a reporting share around three percentage points higher than those without one.

The clearest difference was where the attack happened. Of the 1,727 attacked women who said where it occurred, 1,115, or 65 per cent, were attacked at home and 612 elsewhere. Police were told about 14.6 per cent of the attacks at home, compared with 26.8 per cent of those elsewhere. Figure 1 shows the two reporting shares; the home and elsewhere rows describe attacked women only, so they have no share in its left panel.

The data cannot establish whether fear, family pressure or expectations of the authorities account for this disparity. What it does show is that attacks at home were considerably less likely to be disclosed to the police.

Registration: what police records reveal

The survey findings leave open a further question: which incidents come to be formally recorded as offences? Police and prosecution returns obtained by the Sustainable Social Development Organization (SSDO) through right-to-information requests record several forms of gender-based violence, including rape, kidnapping, domestic violence and honour killing. Figure 2 begins with rape registrations and the prosecution activity reported for 2024.

The four provinces recorded 5,163 rape registrations in 2024 and reported filing 3,708 charge sheets. At year end, 1,391 cases were under trial and 420 under investigation. The annual outcome columns recorded 683 acquittals, 46 withdrawals and 21 convictions. These are separate counts, not the successive outcomes of the same 5,163 cases.

A charge sheet, or challan, asks for the accused to be tried. Cases may instead remain under investigation, need further work requested by prosecutors, or be referred to a magistrate for cancellation; the returns do not distinguish these routes.

Punjab accounted for 4,641 registrations and 20 convictions. Sindh's return recorded 243 rape registrations but listed no convictions or acquittals. KP recorded 258 registrations and one conviction; Balochistan recorded 21 registrations and no convictions. The departments' outcome totals do not reconcile with the number of charge sheets filed. They describe what was returned to SSDO, rather than a complete account of each prosecution.

Rape registration rates differed sharply between provinces. Punjab recorded 3.63 rape cases per 100,000 residents in 2024, compared with 0.63 in KP, 0.44 in Sindh and 0.14 in Balochistan. Punjab's rate was 26 times Balochistan's.

Other measures of violence do not follow the same provincial ordering. In MICS, women in Balochistan were slightly more likely than those in Punjab to report having experienced a physical attack. The 2017–18 Pakistan Demographic and Health Survey recorded spousal violence among 49 per cent of ever-married women in Balochistan and 52 per cent in KP, compared with 32 per cent in Punjab and 18 per cent in Sindh. These surveys measure different forms of violence from registered rape, and cover earlier periods. They cannot establish comparative rape prevalence, but neither do the police returns support treating lower registration as evidence of greater safety.

District data helps us examine what else registration reflects. Do more cases appear in places where women have more schooling, own phones and live in towns? Figure 3 compares 122 districts. The accompanying analysis ranks districts within each province, so differences between provinces do not drive the correlations.

Within provinces, districts where more women had secondary schooling, owned phones or lived in towns generally registered more rape cases. Districts where more surveyed women considered wife-beating justifiable in at least one of the circumstances presented generally recorded fewer rape cases. These associations were modest.

The gaps in registration are striking. Twenty-seven of the 122 districts recorded no rape case in either year: 20 in Balochistan and seven in KP. Every district in Punjab and Sindh recorded at least one. All ten districts with the highest registered rape rates were in Punjab.

Kidnapping and abduction registrations showed stronger associations with schooling, phone ownership and urbanisation than rape. Registration rates also tended to be lower in districts where more women considered wife-beating justified. But these figures can also reflect disputes over women's choice of marriage: families sometimes register kidnapping complaints against men their daughters have chosen to marry. The register can therefore include disputes over a woman's choice as well as cases of abduction.

Domestic violence registration rates also tended to rise with schooling, phone ownership and urban share, but showed little relationship with attitudes towards wife-beating. Thirty-five districts recorded no domestic violence case in 2024.

Honour killing showed the opposite relationship with schooling and phone ownership. Registrations were higher where fewer women had secondary schooling or a phone, where marriage before 18 was more common, and where more women accepted wife-beating. The concentration shifts towards upper Sindh and the Balochistan border, including Nasirabad, Kashmore, Khairpur, Shikarpur and Jacobabad.

Unlike offences whose reporting may depend on a surviving complainant, a killing can enter police records without one. Its classification as an honour killing, however, still depends on whether the police identify and record an honour-related motive.

Punjab complicates the wider pattern. Within the province, rape registration rates were higher in poorer districts where fewer women had secondary schooling, reversing the relationship found across the combined district analysis. One possibility is that rape allegations are registered more routinely in Punjab, making recorded rates less dependent on women's access to the police. The data cannot establish whether this explains the reversal or whether differences in underlying violence are responsible.

Taken together, the district patterns suggest that access to the police and decisions about registration help shape the statistics. They cannot tell us how much of each difference comes from reporting and how much from violence itself.

Trial: what do the court records show?

A registered case need not proceed to a completed trial. The Law and Justice Commission of Pakistan's figures for the courts designated to hear gender-based violence cases record convictions, and acquittals after completion of trial, as two categories separate from acquittals under sections 249-A and 265-K and four others. In 2023, these courts recorded 48,395 new cases and 30,295 decisions across the four provinces and Islamabad. They hear offences beyond rape, so their figures concern a wider body of cases than the rape returns examined above. Figure 4 shows the Commission's categories for the 30,295 decisions.

The Commission records 1,451 convictions. It records 6,416 acquittals after completion of trial. Together the two categories are 7,867 decisions, around a quarter of the 30,295.

Convictions therefore accounted for 4.8 per cent of all recorded decisions, but 18.4 per cent of those two categories combined. The denominator changes the answer. Neither figure tells us the chance that a complaint registered with the police will end in a conviction.

The largest category was acquittal under sections 249-A and 265-K of the Code of Criminal Procedure: 12,151 decisions, or 40.1 per cent. These sections allow a court to acquit at any stage when it finds no probability of conviction, and the Commission reports them separately from acquittals after completion of trial. Such an acquittal is a decision on the prosecution's case. A further 936 cases were stopped under section 249 without a verdict.

A further 5,590 cases, or 18.5 per cent, fall in the Commission's category of proceedings initiated under section 512. That section allows a court, subject to conditions, to record the prosecution's evidence when the accused has absconded, so that it can be used if he is later arrested. These are not concluded cases. Another 2,583 cases, or 8.5 per cent, ended in compromise. A further 701 were adjourned without a date.

Punjab accounted for 25,598 of the 30,295 decisions and therefore dominates the combined figures. The other jurisdictions differed: in Sindh and Islamabad, a larger share of decisions fell in the conviction category or in acquittal after completion of trial.

These figures record the category each decision fell into, but not why. They do not explain why the prosecution's evidence was insufficient, why an accused could not be produced, or whether a woman agreed freely to a compromise. A compromise does not, by itself, establish that an allegation was untrue. Proceedings under section 512 preserve the prosecution's evidence but leave the allegation untried against the accused.

Appeal: what changes after a verdict?

A trial verdict can be followed by years of argument over whether it should stand. Sindh's register contains 11,296 criminal appeal and death-sentence confirmation documents. Among the 11,277 documents with usable dates, 36.5 per cent carried an order at least three calendar years after the filing year, and 5.4 per cent an order ten or more years later. These figures cover all criminal offences; the register does not identify violence against women separately.

What happens during that further contest? To examine how rape verdicts were reconsidered, we read the judgments in 268 distinct reported decisions concerning rape or attempted rape against women and girls from Pakistan's seven higher courts. Of these, 251 reviewed at least one existing rape conviction and 24 reviewed an earlier acquittal, with seven decisions doing both.

Half of the 251 conviction reviews, or 126 decisions, set aside every rape conviction under review. Another 106 retained them all. An acquittal on the rape charge did not always mean release: convictions for murder, abduction or other offences could remain. Corrected on 21 October 2026. As published, these figures were 269 decisions, 252 conviction reviews, "exactly half" set aside and 107 retained. One decision has been removed from the coded set; the method note and the corrections log explain why.

Nine decisions replaced rape convictions with convictions for other offences, including six for consensual intercourse outside marriage. Five produced different outcomes across defendants, while five sent cases back without a final verdict. Among the 24 decisions reviewing earlier acquittals, 20 left them standing, three sent the matter back, and only one replaced an acquittal with a rape conviction (Figure 6a).

That one decision, a Federal Shariat Court judgment reported in 2014, is chosen here as an illustration rather than as a typical history. The complainant's FIR was registered in November 1988. The trial court acquitted the accused in July 1999, and the Federal Shariat Court reversed that acquittal in March 2014, convicted him and imposed a four-year sentence, treating the twenty years that had passed as a mitigating factor. It says nothing about usual waiting times; it shows an appellate court changing the legal outcome more than a quarter of a century after the incident.

These decisions come from the Supreme Court, all five High Courts and the Federal Shariat Court. Their composition matters. The Lahore High Court set aside all rape convictions under review in 60 of 100 decisions, Sindh in 26 of 48, the Federal Shariat Court in 18 of 44 and the Supreme Court in 11 of 27. These are shares in the reported record, shaped by which disputes reached each court and which judgments were published. They are not national appeal success rates or a comparison of the courts' fairness.

Even where the rape convictions survived, the punishment could change. In 25 of those 106 decisions—nearly one in four—the court reduced the custodial penalty. It increased it in three and left it unchanged in 75. One replaced a fixed term with life imprisonment; one left a sentence question pending and one did not make its direction clear (Figure 6b). A conviction being upheld was therefore not enough to tell us what happened to its punishment.

In Mehboob, the Sindh High Court upheld the rape conviction and increased the prison term from ten to fourteen years. In a 2018 Lahore High Court decision, the court upheld a father's conviction for raping his eleven-year-old daughter but commuted his death sentence to life imprisonment, reasoning that his execution would further harm her. Both courts accepted that rape had been proved. What changed was the punishment that followed.

Reclassifying the offence changes what the law recognises as having happened. In Muhammad Imran, a Supreme Court majority replaced a rape conviction with one under section 496-B for consensual intercourse outside marriage, reducing the prison term from ten to five years. The majority relied on the interval before the FIR was registered and the absence of injuries. The dissent recorded that the woman had approached the police that evening and undergone a medical examination the next morning. Seeking help and having the police register an FIR were different events, with different implications for how her account was judged. The reclassification also changed the legal character of what had happened to her. The majority itself observed that the complainant could face liability for the reclassified offence, and declined to punish her only because she had not been charged or given an opportunity to defend herself.

The reported record also contains rulings that challenge assumptions about women's credibility. In Atif Zareef, the Supreme Court rejected the use of a woman's sexual history to infer consent or discredit her testimony, and said that the two-finger test must not be used to establish rape. The outcome counts cannot tell us how often such reasoning was followed. They do show why examining the verdict, offence and punishment separately is necessary: a judgment's label alone conceals what the court has decided.

The Sindh register measures the interval within the appeal stage. Individual judgments also permit a longer history to be reconstructed, from the original complaint or FIR to a later court order. Figure 7 traces two criminal rape cases and a separate workplace-harassment dispute, illustrating the time taken across different routes to a legal remedy.

In Asif Masih, 11.5 years passed between the FIR and the Supreme Court decision. The court replaced one petitioner's rape conviction with a conviction under section 496-B and acquitted two others of abduction. In Muhammad Imran, the interval from FIR to Supreme Court decision was eight years.

Nadia Sarwar's workplace harassment complaint began in September 2012. The Ombudswoman upheld it in 2014, but an appeal to the President reduced the penalty to a censure. The Islamabad High Court restored the original order on 31 December 2024, 12.3 years after the first complaint.

Sarwar's case concerned an employment remedy, rather than a criminal conviction. The judgment ordered implementation; it does not establish that the remedy was delivered. These three histories cannot tell us the average wait, but they show why reaching a verdict and obtaining a lasting remedy are different milestones.

What the records cannot tell us about justice

The findings show the obstacles women face at each stage, and how long a decision can remain contested once it is made. Most attacks in the survey sample never reached the police, particularly those at home. In the designated GBV courts, acquittals under sections 249-A and 265-K and proceedings initiated under section 512, where the accused had absconded, together accounted for more recorded decisions than convictions did, or acquittals after completion of trial, or the two combined. Appeals could then undo a rape conviction or reduce its punishment. Reaching a trial verdict was not necessarily the end of the dispute.

These patterns matter for how progress is judged. A fall in registered cases does not establish that violence has fallen. Police figures need to be considered alongside survey evidence, rather than used on their own to rank women's safety.

Improving prosecutions also requires understanding what leads to acquittals under sections 249-A and 265-K, and why accused people cannot be produced. The published totals identify those problems but cannot explain them. Nor can they tell us whether compromises were freely agreed or remedies were eventually enforced.

What remains missing is a record that follows a woman's complaint across these stages. We can see how many registrations, decisions and orders each institution records, but cannot establish how often a woman who reports an attack eventually obtains a remedy. Whether she received that remedy, and whether it was enforced, should be part of how the justice system accounts for its work.

Sana Zaman Butt is a policy researcher and writer whose work spans human rights, international development and public policy. A recent Fulbright Scholar with an MPA from Columbia University, she is interested in data-driven public-interest research on government decision-making, regulation and state capacity, and their effects on inequality and access.

Acknowledgements

Data from Pakistan Case Law (pakistancaselaw.com), provided by Hamadullah Shah, Advocate, its founder. We are grateful to him for access to the reported judgments and for his care over how they are used.