Sweden: The Law Stopped Asking Whether She Fought, and the Number Went Up

This chamber already holds one essay with Sweden’s name on it, and this is not a second version of it. That piece set the purchase ban of 1999 beside Germany’s legalisation of 2002, asked what a state should do about the people who buy sex, the people who sell it and the businesses between them, and concluded that after two decades neither country could demonstrate the outcome its advocates had promised. Its subject was the sex trade. This one has a different subject entirely, and the difference is worth stating in the first paragraph rather than letting a reader arrive expecting the other argument.

The subject here is a change Sweden made to its ordinary law of rape, nineteen years after the law it is famous for. On 1 July 2018 the threshold moved. Until that date a prosecutor had to show violence, a threat, or that the complainant had been in a particularly vulnerable situation. After it, the question the court asks is whether the other person was participating voluntarily.

Two things make this worth a chamber essay rather than a legal note. The first is that Sweden did not simply pass the law and declare victory: its national crime-prevention agency has now evaluated the change twice, in 2020 and again in 2025, with access to the judgments, the police files and the practitioners. Very few countries that have made this reform have done that, which means the Swedish evaluations are, for the moment, close to the only evidence anybody has about what a consent standard does once it is inside a courtroom.

The second is the number. Sweden appears at or near the top of every European count of sexual offending, and that figure is thrown at any argument in its favour within about two sentences. It deserves a real answer, and the second half of this essay gives it one — including the part that does not go Sweden’s way.

It is worth being exact, because the shorthand has travelled further than the text. Swedes call it the consent law, and Brå — the Swedish National Council for Crime Prevention — notes in its own evaluation that this is not quite the right name: it is not a new statute but an amendment to Chapter 6 of the Criminal Code, and consent is not the concept the provision is built on. The word the provision uses is voluntariness.

The operative sentence now reads that a person who performs intercourse, or a comparable act, with a person who is not participating voluntarily is guilty of rape. In assessing whether participation was voluntary, particular consideration is given to whether voluntariness was expressed by word or deed or in some other way. And there are three circumstances in which a person can never be considered to be participating voluntarily: where participation results from assault, other violence or certain threats; where the perpetrator improperly exploits a particularly vulnerable situation, including unconsciousness, sleep, grave fear or intoxication; and where the perpetrator improperly exploits a position of dependence.

Read carefully, that is a more modest instrument than the one its foreign admirers and its foreign critics both describe. It is not a rule that sex is a crime unless permission was requested and granted in words. The two old criteria did not disappear; they became two of the three circumstances that settle the question in advance. What changed is what happens in the space outside them, where there was no violence, no threat, no incapacity — and where, before 2018, there was also no offence.

One later amendment matters for reading the figures. On 1 August 2022 the minimum sentence for non-aggravated rape rose from two years to three, which is a different kind of change from the definitional one and moves a different set of numbers.

The same reform created something less discussed abroad, and it is the part of the Swedish experiment a reader should watch most closely. Alongside rape, the amendment inserted a new offence: negligent rape, carrying a maximum of four years, together with a parallel negligent sexual assault.

Its logic is this. Rape requires intent — the defendant knew the other person did not wish to take part. Negligent rape covers the man who did not know, but who realised or should have realised there was a risk of it and went ahead anyway. Swedish law already had a category adjacent to this, called intentional indifference, which covers the man who saw a high risk and did not care. The line between indifference and gross negligence is the line between a sentence of years and a sentence of months, and it is drawn inside another person’s head at a moment nobody recorded.

The Supreme Court reached the question for the first time in July 2019. In a case where both lower courts had convicted a man of rape, it substituted a conviction for negligent rape instead. That is a fact worth sitting with before celebrating the new offence: its first significant appearance in the highest court was as a landing place softer than the charge below it.

Brå’s first follow-up, published in June 2020, reviewed every district court judgment from 2019 concerning completed rape of an adult woman — three hundred and sixty-two of them — together with thirty-seven appeal judgments and one from the Supreme Court. Convictions had risen from a hundred and ninety in 2017 to three hundred and thirty-three in 2019, an increase of seventy-five per cent, and Brå’s reading was that new kinds of assault were now reaching trial and resulting in conviction. Twelve of those convictions were for negligent rape.

The second evaluation, published in 2025, is the more useful document: a longer run of years, all convictions for completed rape of an adult in 2023, a sample of processed police reports from the year to June 2023 set against an equivalent sample from 2016, group interviews with investigators in nine police units, seventeen prosecutors interviewed and fifty defence lawyers surveyed.

Its central structural finding is the one to carry away. Prosecutions for rape had nearly doubled, from two hundred and thirty-six in 2017 to four hundred and fifty-five in 2020, and had then settled. Reports rose from just under four thousand in 2015 to around six and a half thousand in 2021 — but that rise, Brå says, largely continued a trend already under way before the amendment. And the increase in reports since 2018 consists entirely of the new kind of case. Reports of violent rape were if anything slightly lower in 2022 and 2023 than in 2016.

The new cases differ from the old ones in one clear respect: the people in them are younger. The median complainant is twenty rather than twenty-four, which means that close to half are teenagers, and the median suspect is twenty-five rather than thirty-one. In nearly half, the parties are acquaintances; in a similar share they are in a close relationship; a stranger is rare.

The most interesting movement between 2019 and 2023 is not in the totals but in what a court was willing to treat as a case at all. In 2019, only just over half of prosecuted new cases ended in conviction. By 2023 that share was seventy-seven per cent — the same as for the old cases. Brå reads this as prosecutors and judges converging on what the provision means and what evidence it requires. Two specific convergences stand out. Cases in which the parties had been intimate before were rarely convicted in 2019 and by 2023 were being convicted at about the same share as cases between people with no such history. And cases in which a person took part willingly, stopped, and the other person continued produced no convictions at all in 2019 and several by 2023.

Alongside that, the evidentiary bar moved. The share of all rape convictions resting on oral evidence alone rose from sixteen per cent in 2017 to thirty per cent in 2023, and in the new cases it reached forty per cent. Swedish appellate law had long permitted this; what changed was that district courts began to act on it, and on a reasonably consistent view of when an account can stand by itself.

What those complainants describe is worth recording precisely, because it is the human content of a definitional change. In about half of the new cases, the person says she refused repeatedly, was not listened to, and eventually gave up and went still. In most of the rest, she describes becoming completely immobile — often on waking to find the act already beginning. Before 2018 those accounts described no offence in Sweden. Neither of them involves the exit this chamber has argued is the whole test of shared and imposed ambiguity, and neither of them involves a fight.

An honest report of an evaluation includes the parts that do not flatter the reform, and this one has several.

The clearance figure has barely moved. The share of reports leading to a prosecution rose from seven per cent in 2017 to eleven per cent in 2023, so around nine in ten reports still end without anybody being charged, and the new cases clear slightly less often than the old because there are no injuries and no toxicology to corroborate them. Convictions for non-aggravated rape have sat at roughly three hundred a year since 2019. The police units handling these investigations had their resources raised by between fifty and ninety per cent since 2018 and conduct more interviews and more tests than before; the proportion cleared did not follow.

The negligent-rape offence remains the reform’s unresolved edge. In 2023 there were twenty-six convictions for it, and every negligent-rape charge that reached judgment produced one — for the revealing reason that prosecutors almost never charge it. It becomes available only after a court has decided that rape is not made out, which makes it less a distinct wrong than a place for a court to land. Brå’s assessment of those twenty-six is blunt: a few appear well justified and address exactly the cases the legislature had in mind, but in the majority of them either an acquittal or a rape conviction would seem more reasonable. Fifteen of the twenty-six rested on oral evidence alone. Brå asks for judicial training so that the provision is not overused as a compromise in cases where what actually happened is unclear, which is a polite way of saying that is what is happening.

The criticism arrives from both directions, and the essay reports both. Three in four of the defence lawyers surveyed think legal certainty has been weakened, and in the autumn of 2024 mothers of convicted sons organised a petition against the way the law is applied. Against them stand feminist scholars who think the provision too narrow rather than too loose — a test built on voluntary participation assumes two free people making free choices, and so fails to reach the difficulty a woman has in expressing her will where the power between the parties is unequal.

And the largest absence is the one Brå states itself. More people now know Sweden has such a law; more people report; that is a real result. But, it writes, it is natural to hope the signal has a concrete effect on how people behave rather than principally producing more reports — and on whether it has, we know very little. The agency that ran the evaluation declines to claim the thing everybody most wants the law to have done.

Now the counter-argument, which arrives in every conversation about Sweden and which deserves to be taken at its strongest. Sweden reports more rapes per head of population than almost anywhere in Europe. If the Swedish settlement works, why does the Swedish number look like that?

The first part of the answer is arithmetic, and it comes from Brå itself rather than from anybody defending Sweden. In 2020 it published a comparison of reported and cleared rapes across thirty European countries and concluded that differences in how rape is defined and how offences are counted make the figures not comparable. In roughly a third of the countries studied, some form of violence or threat was still required. In seventeen of the thirty, only intercourse counted. Sweden’s definition is wider than both, and was already wider before 2018.

Then there is counting. Swedish police register an offence for each occasion a complainant describes, so a relationship in which the same thing happened forty times enters the statistics as forty offences rather than one case; and a report is recorded when it is made, not after an investigation has concluded that it happened. Brå’s own illustration is that in one Swedish city, more than a third of a year’s rape reports traced back to a single prosecution. Across the years 2013 to 2017 Sweden averaged sixty-four reported rapes per hundred thousand inhabitants. Recalculated under German definitions and German recording rules, the same conduct produced about fifteen. Brå’s conclusion is that under comparable rules Sweden would sit somewhere in the middle of the European table rather than at the top of it.

And here is the sharpest form of the point, which is about this essay’s own subject. The 2018 amendment moved a class of conduct from outside the offence to inside it. Brå’s finding that the entire increase in reports consists of the new kind of case is, read as arithmetic rather than as policy, the statement that the reform raised Sweden’s rape figure by definition. A country that protects more people by widening the category will, mechanically and immediately, look worse in any count that treats the category as fixed. Every country now considering this reform is buying that headline along with the law.

It would be convenient to stop there, and it would be dishonest, because there is a second Swedish number that the recording argument does not touch at all.

When the European Union Agency for Fundamental Rights surveyed women directly in 2012 and published in 2014, it did not ask what the police had recorded; it asked women what had happened to them. Lifetime experience of physical or sexual violence ranged from around a fifth of women in Poland to around half in Sweden, Finland and Denmark. The EU-wide survey of gender-based violence run by Eurostat, the Fundamental Rights Agency and the European Institute for Gender Equality, published in 2024 on more than a hundred thousand interviews, again put Sweden at the top of the range for unwanted sexual behaviour at work. Legal definitions and police counting rules explain none of that, because no lawyer and no police officer is involved.

This is what researchers call the Nordic paradox: the countries with the strongest equality records report the most violence. The obvious reply — that women there simply feel freer to say so — is available and genuinely contested. A psychometric study of the survey instrument found the responses comparable across countries, which argues against measurement artefact; and a measure derived from the same survey found Swedish women who had been assaulted slightly more likely than the European average to have told nobody at all, the opposite of what the freer-to-speak explanation predicts. On the other side, a multilevel analysis published in 2021 found that once individual circumstances, survey mode and above all whether a woman knew others subjected to violence were taken into account, the Nordic countries’ position fell sharply and the paradox largely dissolved.

The responsible statement is therefore narrower than either camp wants. The police-recorded comparison is close to meaningless and Sweden’s own agency says so. The survey comparison is a real finding about something, and what that something is — exposure, awareness, disclosure, or some combination moving together — is an open research question. The 2024 survey’s own framing is the most honest available: it reports what it calls disclosed violence, and prints a diagram showing reported violence inside disclosed violence inside actual prevalence, with no claim that any of the three equals another. A number that is explicitly a measure of what women were willing to say is not a league table, and anybody using it as one, in either direction, has stopped doing arithmetic.

Japan made the equivalent move five years later and from a very different starting position, and this chamber has already written that story rather than summarising it here. The essay on the movement made of disclosure records what the 2023 revision did and, more importantly, what it did not do: it found that Japan changed its sexual offence law substantially and its social permissions almost not at all, which is the reverse of the usual sequence, and that the people who forced the change paid for it individually.

The legal architecture is worth setting beside the Swedish one, because the two solved the same problem with opposite drafting instincts. The revision in force from July 2023 replaced the old forcible and quasi-forcible offences with non-consensual sexual intercourse and non-consensual indecency, and defined the wrong as an act committed where the other person was in a state in which it was difficult to form, express or carry through an intention not to consent. Rather than a general standard with three exclusions, it enumerates eight circumstances capable of producing that state — violence or threat, incapacity, intoxication, unclear consciousness, being given no time to react, fear or astonishment, the psychological effects of abuse, and anxiety about disadvantage from a difference in economic or social position. Sweden wrote a principle and let the courts populate it. Japan wrote the list.

Both choices have a cost. Sweden’s is visible in the evaluations: five years of courts working out by trial and error where the boundary sits, with the negligence offence absorbing the uncertainty. Japan’s is the mirror image. A list is easier to apply and harder to stretch, and the situation that fits none of the eight items is the situation the list does not reach.

And Japan has now walked into the same statistical trap. Recorded offences of non-consensual sexual intercourse rose to two thousand seven hundred and eleven in 2023 and three thousand nine hundred and thirty-six in 2024, an increase of forty-five per cent in the first full year under the new definition. The Japanese crime white paper that publishes those figures attaches its own warning: the elements of the offence changed, and the series should not be read across the change as though it were continuous. That is precisely the caution the European argument about Sweden refuses to apply. The same trap, the same escape route, printed by the government that the number embarrasses.

Set the two countries side by side and the comparative finding is narrow, which is the only reason to trust it. What writing consent into the definition reaches is the case that previously was not a case. That is not a small thing and it should not be described as one. Sweden can now count the woman who said no repeatedly until she stopped saying anything, the woman who woke to find something already happening, and the woman who agreed to one thing and then did not agree to the next — and can count them because a court is now allowed to hear them. The chamber’s essay on France records the same arrival there in 2025, after a public argument about ambiguity that lasted seven years and a trial that ended it. The direction of travel is real and it is not a fashion.

What such a law does not reach is the room. Nine reports in ten in Sweden end without a charge, and the reason is evidentiary rather than definitional: the new cases are precisely the ones with no injury, no toxicology, no third party. A standard that asks whether participation was voluntary still has to be proved by somebody describing what happened, to a stranger, months later, against a person who describes it differently. That is not a flaw in the drafting. It is the limit of the instrument, and the chamber found the same limit in South Africa, where the clause came first and the room it could not reach stayed unreached.

This is where the Library’s own smaller essays meet the statute. The Swedish provision says voluntariness is assessed by whether it was expressed by word or deed or in some other way, and that silence and passivity settle nothing — which is, in the flattened language of a criminal code, the same finding as the piece in this corpus on why silence is a multiplier rather than a message, and the reason another holds that a few words before touch create the mood rather than ruin it. The law had to learn what those essays assume: that the absence of a refusal is not information. But a court can apply that only after the evening. A person has to apply it during one, and no legislature in Stockholm or Tokyo has found a way to put a provision in a room.

This house sells evenings and therefore has an interest adjacent to the subject, which should be stated against itself before anything else. An essay arguing that law cannot reach the room is an essay whose conclusion happens to be the thing the house sells: a room where the terms are set in advance by the person receiving, where a word ends anything without explanation, and where the person providing the evening is bound by that and has no standing to reinterpret it. A reader who notices that the argument lands conveniently is noticing something real.

What the house can honestly say is only this. The Swedish evaluations describe, in the flat register of an official document, a specific human situation appearing in courtrooms by the hundred: a person who refused until refusing stopped working, and a person who went still. Both are descriptions of an exit that was not there. An evening in which the exit is built beforehand is not a remedy for that, is not treatment, and is not a substitute for anything the law does badly.

And there is a temptation here the house should name and refuse. It would be easy to sell a room as the place where a woman practises saying the sentence she could not say. That framing makes the house an instrument of somebody’s recovery, which it has no standing to be. What an evening can offer is smaller and more literal: an occasion on which the sentence is available, is used, and changes nothing about how the rest of the evening goes.

It does not claim that the Swedish reform works, in the sense that anybody most wants that word to mean. Brå states plainly that more people know about the law and more people report, and that whether behaviour in sexual situations has changed is not something the evidence shows. Nothing here goes further than Brå went.

It does not claim that Sweden is safe or unsafe relative to anywhere else. The police-recorded comparison is unusable and the essay says why; the survey comparison is a real finding whose interpretation is unsettled, and the essay says that too rather than choosing the convenient half. No rate in this essay is offered as a description of how often anything happens to anybody.

It does not claim that the negligent-rape offence is a good idea or a bad one. It reports that Brå, having read the judgments, thinks most of the convictions under it would have been better as acquittals or as rape convictions, and that the country’s highest court used it once to reduce a conviction the courts below had recorded as rape. Those are findings about application, not about the provision in principle. Nor does it claim that Japan should have drafted as Sweden did, or the reverse: the two drafting choices have symmetrical costs and neither country has run long enough to show which cost is larger.

And it makes no claim about any individual. Every judgment described here is described in aggregate from an official evaluation; no complainant and no defendant in Sweden or Japan is characterised; and the accounts of what complainants describe are Brå’s summary of a file review, not testimony this Library has read. The essay is written by an outsider to both legal systems, from published sources in translation, and a reader who wants the provisions should read them rather than this.