How companies avoid inclusion – and even save money by doing it

Large employers in Germany are required to hire people with disabilities – or pay a penalty. But they can avoid the fine with a trick. For the first time, we show exactly how companies save millions of euros – and where the money goes

“I would like to work in a library. Or read to elderly people.” These would be good jobs for Ing Han Ong. But instead, the 45-year-old sits at a white wooden table every day. He folds cardboard boxes and puts a needle, a plaster, and a note for a vitamin D test inside. 200 times a day. For the past four years. Ong works in a workshop for people with disabilities in Hamburg. He packages the tests for the healthcare company Cerascreen. There, he earns 260 euros of pocket money per month, for about 30 hours of work each week. 

Ing Han Ong’s situation is similar to that of many of the more than 300,000 people working in workshops for people with disabilities in Germany: they earn very little money for their work, and many live below the poverty line.

Outside the workshop system, Ong would earn more money, and he would not depend on social benefits. But the general job market is not inclusive. There are very few jobs for people with disabilities. And even those who do find work face additional obstacles in the open job market. Structural barriers, uncertainty, bullying, and harassment at the workplace are part of everyday life for many people with disabilities.

To help more people with disabilities enter the general job market, there is a rule in Germany: companies with 20 or more employees must allocate at least 5 percent of their jobs to people with disabilities. If companies do not hire enough people with disabilities, they must pay a penalty: the compensation levy.

This rule is meant to make it easier for people like Ong to access the German job market. But in 2023, most employers did not hire enough people with disabilities. More than 100,000 employers had to pay the compensation levy. And the law allows them to use a trick that helps them save money – or avoid the levy completely.

This joint investigation by andererseits, Süddeutsche Zeitung, and FragDenStaat shows for the first time how big the problem is: it involves many millions of euros that should actually support people with disabilities. Instead, the money is used to support workshops for people with disabilities – even though, according to the United Nations, these workshops in their current form violate human rights.

The cost-saving trick

The trick is simple: if companies have hired too few people with disabilities, they can give contracts to workshops for people with disabilities and save part of the compensation levy. They can offset up to half of the money they pay for the work done in the workshops.

Many employers in Germany use this trick – and they save a lot of money. In 2022, they saved about 84 million euros. Employers in Berlin are not included in this figure. That means 84 million euros went into supporting the workshop system – and not into measures that promote inclusion in the open job market. With that money, around 20,000 office workplaces could have been made wheelchair-accessible. Or several thousand personal workplace assistants could have been paid for a whole year.

In 2022, employers in Thuringia saved the most. In Bremen, they saved the least through this trick. In federal states with particularly large numbers of companies, millions of euros flow directly into workshops for people with disabilities – in Bavaria, for example, more than 14 million euros in 2022.

If employers give enough contracts to workshops, they do not have to pay the levy at all. In Bavaria, in 2022, almost 2,500 companies were able to reduce their compensation levy to zero. That was almost one in ten companies that would otherwise have had to pay the levy.

This does not only mean that companies save money. It also means that less money flows into the compensation levy. This levy is like a pot of money that is supposed to be used for inclusion. If companies pay less into it, then less money can be spent on inclusion.

Part of the money from the compensation levy goes to the Federal Employment Agency. It uses the funds to provide support for people with disabilities in the job market. Most of the money remains with the integration offices of the individual federal states. They use it, for example, to pay for inclusion projects or workplace assistants. However, about 3 percent of the fund also goes directly to workshops for people with disabilities. In 2022, that was almost 15 million euros. This means that even the money from the compensation levy is used to support non-inclusive structures.

The job market is not inclusive

What companies save on the levy is, for many of them, only a small amount, says Ulrich Scheibner. He used to lead the Federal Association of Workshops. Today, he is one of the strongest critics of the workshop system and the compensation levy. He says the reasons why employers use the trick are not primarily financial. Instead, they often fear that creating inclusive workplaces would require too much effort. Many still have prejudices.

The inclusion researcher Gudrun Wansing confirms this: “Just because someone has a so-called severe disability does not automatically mean they are not qualified or not capable. Unfortunately, this is often claimed.”

Ing Han Ong experienced this himself. For five years, he had an external placement at a supermarket in Hamburg. This means he worked in the open job market, but unlike his colleagues, he was officially still employed by a workshop. It paid him 395 euros a month. He did not receive an employment contract. He stocked shelves and checked whether food was still within its expiry date. Not his dream job, Ong says, but he had managed to make the transition from the workshop for people with disabilities into a regular company. Until his supervisor complained that he worked too slowly. At first, she said this behind his back to other colleagues, Ong recalls; later, she criticised him directly. It simply wasn’t possible to work any faster, Ong said. “She did a bit of bullying, she was kind of against me—against disabled people,” says the 45-year-old. 

After five years at the supermarket, he quit. Not only because of weekend work, but also because he was treated badly. Unfortunately, Ong says, it is often the case that people trample on people with disabilities. “We are all just human, and even people without disabilities make mistakes sometimes.”

Today he is back in the workshop. His pocket money is also financed through contracts with workshops. Ong does not think it is right that companies profit from his daily work: “It does feel a bit like exploitation.”

What Ong experiences at work every day has a big effect on his self-esteem, says Gudrun Wansing: “For people with disabilities, it makes a difference whether they walk through the gate of a company in the morning and are part of a workforce, or whether they are picked up from their home by a transport service and driven to the workshop.”

And yet, less than half of companies in Germany hire enough people with disabilities. Even public-sector companies struggle with inclusion. Public companies work for all people in Germany, yet they, too, save a lot of money through contracts with workshops. In some federal states, public companies save much more than private ones. For example, in Thuringia, public-sector companies saved a quarter of the compensation levy in 2022 through workshop contracts, while private companies saved less than a fifth.

The Federal Association of Workshops does not see it as a problem that companies can save the compensation levy through such contracts. It sees it as an important “compensation for disadvantages” in the market.

Exclusion is sold as inclusion

Packaging, cigarette paper, fuel caps – most people have already handled items made in workshops by people with disabilities. Very few knew this at the time. There is no requirement to label products as such.

The fact that so many companies award contracts to workshops is also due to a misunderstanding: many believe—or act as if—working with workshops is good for inclusion. The healthcare company Cerascreen, for which Ong folds cardboard boxes every day, wrote in response to our inquiry that it wants to contribute to an inclusive working world: “We are proud of this partnership and see it as an example of how business and inclusion can go hand in hand.”

And even at Lufthansa, one of the largest employers in Germany, they advertise a form of inclusion that isn’t really inclusion at all. In 2024, the Lufthansa Group hired too few people with disabilities. “The Lufthansa Group therefore also uses other ways to support people with disabilities, such as awarding contracts to workshops for disabled people,” their press office told us.

Instead of hiring enough people with disabilities, the Lufthansa Group gives contracts to workshops where people only earn pocket money. These contracts benefit the company twice over: in response to our inquiry, Lufthansa wrote that its companies make use of the option to offset these workshop contracts against the compensatory levy.

And so, inclusion in the job market progresses very slowly.

A study by the Institute for Employment Research shows that some employers deliberately hire fewer staff. More employees also means an obligation to hire more people with disabilities: for example, a team of 39 people requires only one such person, but a team of 40 people requires two. In another study, three-quarters of all surveyed employers said they could not find suitable applicants with severe disabilities.

Workshop critic Scheibner doubts this: “This is usually just an excuse.” He says that the supposed lack of suitable applicants is actually due to conditions in the company. These conditions must change. Anyone can work well – if their individual needs are considered. 

For Ong, this means a calm working environment and a clear structure of tasks. Too many tasks at once confuse him, he says. A precise plan would help him complete his work independently.

What politics could change

“We do not have a legal problem,” says Wansing. There are enough government measures to support inclusion, they are just not used enough. For example, making a workplace accessible or providing personal assistants. Or the “Budget for Work.”

With the “Budget for Work,” workshop employees can practice the transition into the general job market. The government pays companies up to three-quarters of the employee’s salary. In addition, workplace assistance is provided. Yet fewer than 1 percent of workshop employees actually move from workshops into the general job market.

Politicians have also recognised that workshop contracts from companies could play a role. The previous German government wanted to abolish the rule that allows companies to save money through workshop contracts. But that did not happen: the government dissolved prematurely at the end of 2024. The new government under Chancellor Friedrich Merz has so far made no new plan on the topic.

In response to an inquiry, a spokesperson from the Ministry of Labor and Social Affairs wrote that it must now be examined whether the measures will be addressed in the future, including the planned abolition of the rule allowing workshop orders to count toward the levy.

For Ing Han Ong, this change would only be a first step in the right direction. Besides changing the system, a change in mindset is also needed. After his bad experiences in the general job market, he has had enough for now. He would only go back if there were people there with knowledge of disabilities and who show understanding so that he is not treated as poorly as he was in the supermarket. Until then, he mainly wishes for a higher salary than the 260 euros per month he earns in the workshop. “Unfortunately, meals are deducted from that,” he says. When asked, his employer Cerascreen describes the work in the workshops as “meaningful, fairly paid employment.”

Further Credits:

Inside the everyday Facebook networks where far-right ideas grow

To experience the interactive, original publication, please click here. You can find the full text without interactive and scrollytelling elements below.

Far-right ideas are gaining ground. They are thriving in online worlds which are not easily visible to large swathes of the public but which spread misinformation that has real-world consequences.

Take last year. In the summer of 2024, riots broke out across parts of the UK, fuelled by misinformation that spread on social media.

The violent disorder was primarily aimed at asylum seekers and Muslims, including an incident of rioters setting fire to a hotel housing asylum seekers.

The rioting was as surprising as it was appalling. It was mostly carried out by local people who were not members of formal far-right organisations. Some rejected the far-right label, carrying banners that read: “We’re not far-right, we’re just right.”

What we found was a community bound together by a deep distrust of government and its institutions, whose members trade in anti-immigrant sentiment, nativism, conspiracy and misinformation.

Experts say their posts include content that is far-right and extremist, and that such online spaces can play a role in radicalisation.

Our starting point was to trace the social media activity of almost two dozen people charged with online offences in connection with the 2024 summer riots. The majority of those charges were brought for posts made on Facebook.

The circles below represent a set of Facebook groups.

We found the public-facing Facebook profiles of those charged and any public Facebook groups of which they were members. We traced five of these individuals to three groups which together had a membership of 267,000.

The groups in the wider network have much in common including titles that celebrate figures on the right. They also have overlapping members and contain similar posts – including hundreds expressing support for those arrested in connection with the summer riots.

Many of the people posting in support of the rioters took issue with people being jailed for online offences. This included support for Lucy Connolly, whose arrest has made her a cause célèbre for the far right.

When we looked into the moderators and admins – who manage group membership, approve requests and have the authority to remove posts or comments among other things – we found that these groups were part of a wider network of at least 16 connected groups.

The groups have a combined membership of close to 600,000 – a figure that is growing week-on-week.

Using a large language model via OpenAI’s API, we analysed 51,000 posts made in three of the largest groups prior to, during and after the 2024 summer riots. We did this to get a sense of what the individuals charged during the riots – and the people defending them – engage with online. The analysis below explores how these conversations overlap with broader far-right ideology.

Membership of one of these groups does not signify any wrongdoing – for example, we found a small number of posts calling out racism – and there will inevitably be users who joined these groups through curiosity or because their friends had.

However, experts to whom we showed our findings said some comments posted in the groups were far-right in their nature. They pointed to a worrying normalisation of far-right ideologies on Facebook, still the world’s largest online social media platform.

Here is an overview of the key themes observed in the groups since their inception through to mid-May 2025, alongside expert analysis of the psychological processes driving this content.

1. Distrust of mainstream institutions

Many of the group members expressed support for Reform UK, with a sizeable proportion of the posts – about two in five of the sample – voicing widespread criticism of government institutions and figures, including anti-establishment and populist rhetoric.

In many of the posts, this included the use of contemptuous, dehumanising or villainising language. Ingrained in these posts was a deep-seated distrust of mainstream politics and various mechanisms of the state: Labour and the Conservatives were “traitors”, “treacherous”, “scum”; the police and the judiciary “two-tier”; the media “controlled” – even the RNLI was villainised as a “taxi service” for “illegals”.

Example post

9 October 2024

WHY IS THIS LABOUR DICTATOR BEING ALLOWED TO GET AWAY WITH SO MUCH, HE LIED TO GET ELECTED, AND HE’S LIEING NOW. I CAN’T BELIEVE HE’S NOT BEING HELD ACCOUNTABLE FOR HIS ACTIONS. TRAITOR COMMITTING TREASON ON A VERY LARGE SCALE. DESTROYING THIS ONCE GREAT COUNTRY.

322 likes 53 comments 34 shares

Analysis: Voters expressing dissatisfaction with the government is as old as democracy itself. In many of the posts seen here, however, that criticism crossed into more radical rhetoric, painting the government and its institutions as illegitimate actors working against the nation. These claims were typically emotionally charged but unsubstantiated, reflecting a broader erosion of institutional trust evident across the sample.

Anki Deo of the advocacy group Hope Not Hate said the main beneficiary of the loss of trust in institutions was Reform. She said the party was part of a wider ecosystem – which included GB News, social media influencers, far-right political groups and others with non-electoral aims – all of which drove anti-establishment rhetoric.

This in turn spilled into these Facebook groups, “filling the gap” for “people who are disillusioned with their usual parties and general politics”.

She said the Guardian’s deep-dive analysis of three of the groups in the network pointed to a worrying online environment: “Public Facebook groups like those used for the data sample contain a mixture of political discussion alongside extremely harmful language.”

Sander van der Linden, a professor of social psychology at Cambridge University, who has researched the impact of conspiracy and misinformation and its connection to extremism, said the undermining of democratic institutions on display in these groups echoed fascist methods used throughout history.

“Generally, fascists and extremists are trying to undermine institutions of truth, facts and education because that is what’s standing in their way – an informed citizenry,” he said.

However, he said the people reading these posts – or even posting themselves – were unlikely to be aware that attacking the credibility of mainstream institutions was a tactic associated with the current far right.

“Regular people interacting with this content often don’t know that they’re part of some playbook or agenda. Political elites and opinion leaders such as Farage, [the far-right party] Homeland, or Tommy Robinson do take a page out of the fascist playbook and are using it to dupe regular people into engaging with their rhetoric. So they come up with a narrative such as ‘the mainstream media is lying to you’, or ‘scientific institutions are looking to censor people’.

“And then they try to get regular people to support, amplify and engage with those topics.”

Dr Julia Ebner, a radicalisation researcher at the Institute for Strategic Dialogue and an expert on online radicalisation, said it was worrying that populist parties, including Reform, were increasingly using a rhetoric that fuelled some of the key radicalisation elements online – whether the undermining of institutions, out-group demonisation, existential threat ideas or the emphasis on a victimised in-group of patriots.

“If Reform UK doesn’t want to be considered far-right it has to make a better effort to distance itself from the most extreme voices supporting its political agenda. Likewise, many Reform UK voters might not like to be associated with far-right ideas but it is important that they are aware of the psychological elements driving all forms of extremism,” she said.

2. The scapegoating of immigrants

Immigration has become one of the most important issues for British voters, with two-thirds of adults polled in May saying the total number of people entering the UK was too high.

Posts in the network about immigration were commonplace, making up about one in seven of the entries captured in our sample.

Our analysis differentiated between criticism of government policy or migration statistics and demeaning, dehumanising, generalised or out-and-out racist posts.

One in 10 fell into this latter category.

Example post

7 April 2025

Can someone explain to me why Britain imports people from a 3rd world cultures that have medieval beliefs of the 17th century. Who put all our women and children at risk of being assaulted?

131 likes 18 comments 8 shares

Analysis: The targeting of groups based on ethnicity, religion or origin echoes harmful historical tropes that remain prevalent across certain online spaces. There were 4,718 anti-immigrant posts within the sample.

The pattern of demonisation in the Facebook groups was noticeable in our analysis, with some posts portraying migrants as dangerous, deceitful, criminal or culturally incompatible.

Others were more subtle, using insinuations and generalisations about “military-aged men” and “grooming gangs” – both commonly used in these forums as coded language for an inherent threat – or the broad-stroke labelling of all immigrants as “illegals” or an “invasion”.

Ebner said participation in this kind of discourse played a role in radicalisation. “Immigrants are often the scapegoats, especially in far-right extremist communities [where they are] systematically demonised and dehumanised.”

The issue, she said, was that the gradual condoning of or justification for extremist ideologies – including the demonising and dehumanising language used about immigrants seen here – could form part of a “toxic cocktail”, resulting in riots or violence against groups or individuals.

Deo, of Hope Not Hate, noted that while many people had legitimate grievances, such as the cost of living crisis or the decline of their local high street, the scapegoating of immigration was not legitimate.

“The way in which they’re engaging [with these issues] is rarely in and of itself – it’s always in contrast to another group. The conversation around the welfare system, for example, which is legitimate, is very often based on this premise of mistruth, which is that people who immigrate to the UK can just instantly access benefits.”

While overt racism was rare in these groups – as is to be expected in a Facebook-moderated space with guidelines about what constitutes hate content – it wasn’t wholly absent: immigrants were variously described as “criminal”, “parasites”, “primitive”, “scum”, “lice” or having “dinosaur” beliefs or values.

Some of these posts were still accessible on the site in mid-May, months or years after being posted, despite Meta’s community standards and the groups’ own stated rules.

3. ‘White British people are fed up’

Alongside anti-immigrant sentiment, there was a strong vein of nativism, with the posters tending to cling to a common in-group identity – that of being “indigenous”, “British”, “white” and “Christian” – which they perceived to be under threat, saying Britons were now “second-class citizens”.

One in 25 posts in our sample fell into this bucket. Deo said the implication in this grouping was that “white, British or English people are the pure group and everyone else is other”.

Example post

13 April 2025

In the 28 years of Blair’s forced policy of mass importation of the third world, we have allowed our politics & judiciary to be corrupted, to the extent that indigenous British people have become second class citizens.

29 likes 9 comments 11 shares

Analysis: Nativism – defined as the belief that a country should be populated only by those considered part of the “native” group – appeared in nearly 2,000 posts in the sample. As here, many portrayed immigration as a form of demographic attack, echoing narratives of victimhood and cultural erasure.

The repetitive nature of this messaging can be particularly harmful, according to Van der Linden. He said that the psychological mechanism at play here, when this kind of post is frequently appearing in someone’s social media feed, was one of “illusory truth”.

“The more often you hear something, the more likely your brain is to think that it’s true. So the more you keep hearing that immigrants are replacing the white population, the more it starts to feel like that is something that must be true. There’s social reinforcement.”

Ebner said that the posts in this group included a perceived “existential threat” to the in-group, as well as some signs of “identity fusion” – feelings of oneness with the in-group.

“When this sense of very strong pro-group commitment is activated in combination with a sense of existential threat from a demonised out-group, that’s when extremist ideologies can become dangerous,” she said.

“This type of identity fusion with an in-group, in this case among people who perceive themselves as a victimised in-group of patriots or white Brits, can then be the final ingredient that tips someone over into the violent spectrum; into being willing to do anything on behalf of the group to save them from a perceived threat from a demonised out-group.”

4: ‘I’m not far-right … I’m just right’

Although there was some condemnation of last year’s riots in the groups analysed, we also found hundreds of posts supporting what the commentators saw as legitimate protest or – in the case of those charged with online offences – freedom of speech.

Many of these posts were contemporaneous to the unrest in July and August 2024, as the riots unfolded and initial sentencing took place.

Example post

02 August 2024

1 last time…WE ARE NOT FAR RIGHT.WE ARE TRUE BRITS.

8.5k likes 814 comments 5.9k shares

Analysis: This section revealed a tension in how users engaged with the “far right” label. Some firmly rejected the term, often framing it as an unfair slur used to silence dissent, particularly around issues such as immigration or political protest. Others appeared to embrace it, equating it with “common sense” or national pride.

They included posts decrying the arrest of Lucy Connolly, a childminder and the wife of a Conservative councillor from Northampton. She was jailed for 31 months in October after calling for hotels housing asylum seekers to be torched.

Deo said this reflected views in other online spaces where people “saw the riots as a watershed moment where their beliefs were finally being recognised” and portrayed the online offenders as “ordinary people expressing legitimate concerns”.

She said this was true of online and offline groupings: “Many of those who participated in the riots do not view themselves as far-right, racist or even anti-immigration. The majority will not be part of any far-right organisation or group.

“The line of what makes someone a member of the far right is increasingly blurred: whereas previously someone might have had to join an organisation, now people can participate or just observe through online groups, dipping in and out.”

One unique and distinguishing element to the patterns of behaviour we found was the online environment in which it was taking place. Ebner said that while the narratives around existential threats, dehumanisation of out-groups and nativism were similar to what had been seen throughout far-right history, what was new here was that the online spaces amplified a lot of these dynamics.

“It’s the algorithmic amplification – the speed at which people can end up in a radicalisation engine – which is different. There are also elements related to new technologies such as fabricated videos or deepfakes, as well as chatbots’ automation.”

She said spaces such as these Facebook groups “definitely play a role in radicalisation of individuals”.

“There is a sense that a lot of this rhetoric has been normalised to an extent that they no longer have to fear legal consequences or societal pushback,” she said, adding that this had only been strengthened by Meta’s announcement this year that it was reversing its takedown policies on certain content.

“[Extremist] rhetoric is welcome again because we’ve seen a reversal of removal policies and takedown guidelines that were in place,” Ebner said. “Not all of the users in the groups analysed by the Guardian show the same degree of radicalisation and extremist ideologies but there are definitely some users who showed clear signs of extremism,” she said of our sample.

Van der Linden said the unprecedented reach of social media meant the dynamics around this rhetoric were playing out differently to how they had in the past.

“You may be able to find two people in your neighbourhood who feel the same way, but now you can connect with thousands of individuals who feel the same way as you in a matter of seconds, which leads to a misperception about what the consensus is in society,” he said.

The way the rhetoric spilled over into offline violence was also distinct to other periods in history, he said. “Obviously, people have been sitting in rooms saying nasty things about other groups of people for a very long time. But whereas before you would have to commit violence in person, now there’s a buildup that happens online, which was just much more difficult without the internet.”

5. ‘Entry points’ for deeper conspiracies

One stark theme we have uncovered in the analysis is the prevalence of conspiracy theories and misinformation, which account for about one in 20 posts in the network.

Denial of the climate crisis featured heavily, as did replacement theory – referred to elsewhere as the “great reset” – and the view that the World Economic Forum or other “shadowy” elite bodies were dictating policy in the UK and elsewhere.

Example post

29 June 2023

There’s something VERY SINISTER going on in Great Britain today .This should concern every critical thinker . This should concern EVERYONE who believes in our birth right of FREEDOM.This is ” their” Brave New World. This is “their” Great Reset .This is all happening under the watch of undemocratic, WEF PUPPETS SUNAK & HUNT !

7 likes 4 comments

Analysis: The “great reset” was one of the most prominent conspiracy theories seen in this Guardian investigation. It was originally an initiative from the World Economic Forum calling for “fairer outcomes” and a sustainable future in response to the coronavirus pandemic. The concept was later hijacked by conspiracy theorists claiming it was an attempt by a small group to exert control.

These far-fetched ideas are able to take root due to the atmosphere of confirmation bias which is already taking place in these groups.

Van der Linden said the atmosphere of mutual agreement on issues such as immigration could be hijacked by more extreme “leaders” in the groups, who seeded conspiracies where people were already looking for answers to their more mainstream societal questions.

“When people feel that they’re connecting with others who have the same idea, it’s a community, but then when the community starts to have an agenda and leaders emerge it becomes a structure,” he said.

Deo agreed: “In some cases, the content about immigration is an entry point. From then onwards, people are exposed to a whole range of conspiracy beliefs and ideas.”

Sara Wilford, a lead researcher on the EU-funded research project Smidge (Social Media Narratives: Addressing Extremism in Middle Age) said that many conspiracies and misinformation posts contained some truth but were cynically distorted to exaggerate or exclude certain facts.

“I think our biggest problem as a society is that there are so many people who are taking that information and legitimising it,” she said, adding that conspiracy and misinformation were often spread by grassroots amplification originated by “bad actors” with large social media followings.

Van der Linden pointed to Pizzagate, the single conspiracy theory that turned into QAnon, as an example of how conspiracy theories could galvanise movements.

“When you have enough angry people and you get them together and you direct that anger towards a singular goal, for example immigrants, that can be and that’s what led to the Southport [and subsequent] riots.”

The combined membership of the groups across the network stood at 611,289 as recorded by the Guardian’s methodology on 29 July 2025: however, this figure almost certainly includes double counting of individuals can be members of more than one group.

Additional reporting by Olivia Lee and Carmen Aguilar García.

The full methodology including our use of OpenAI’s API can be found here. The Guardian’s generative AI principles can be found here.