Lars Højsgaard Andersen on the Consequences of Lowering Welfare Benefits for Migrants and Their Families

Chancellor: Hello and thanks for joining us for the Poverty Research and Policy Podcast from the Institute for Research on Poverty at the University of Wisconsin–Madison. I’m Dave Chancellor. For this January 2020 episode, we hear from Lars Højsgaard Andersen of Denmark’s Rockwool Foundation about a policy change in Denmark that aimed to increase employment among refugees to the country by reducing public benefits. But there were some unintended consequences—and Andersen says his team’s findings suggest caution for other countries considering similar policies. When we started talking, I asked him to give us the background on this policy.

Andersen: During the 1990s, there was sort of an increased wave of refugee migration to the European countries and especially to the northern European Countries where you saw challenges related to the financial or fiscal burden to this issue for our welfare states, so as I’m sure you and people know we have very generous and universal welfare states in the Nordic regions and so when you receive sort of a large or a relatively large population group that has a high level of dependency on social welfare, well having a generous welfare state can become a fiscal burden. So that’s exactly what happened at that point in time. So Denmark received more refugees per capita than other countries that we think of as immigrant friendly, so Sweden, Germany, and the UK for example—at that point. And so we were talking about a real financial issue for the welfare state and sort of for the future sustainability of the welfare state. So, what the Danish government did in 2002 was sort of made an effort to incentivize refugees to take up employment, to sort of work their way around this issue, which is moving more people into employment, increase self-sufficiency, and also as a product of that, sort of improve social integration by having refugees move into the labor market and communicate with the natives in the workplace.

Chancellor: Andersen says that policymakers saw this as an all-around good thing to encourage refugees to move into employment. But critics were worried that this type of reform would produce more poverty and that it wouldn’t really make that big of a difference when it came to employment.

Andersen: That was back in 2002. Since then, the issue of the burden of refugee migration for the sustainability of welfare states in Europe has probably only gotten worse. Especially after the Syrian refugee crisis, right? But there are numerous policies aimed at increasing the labor market attachment of refugee populations in Europe. From 2002–2017, it’s 73 policies that are sort of directly targeted at welfare entitlements for migrant groups in the 27 EU countries. So it’s a tool that’s really being used a lot. And right now even, there are a number of countries who are in the middle of a process of implementing exactly the same type of reform.

Chancellor: Because Denmark has excellent population-level data, Andersen says that he and his team saw this as an opportunity to look at the effects of this policy change to inform similar policies in other countries around the world today.

Andersen: Because the similarity is just so big between that context in Denmark back then and the context you’re observing in other countries now today. So specifically what they did with the reform here was that they cut the level of social assistance that you were entitled to in case of unemployment, they just cut that by around 50 percent, so just lowering the level. So anyone who was granted the residency in the country before the reform date would be entitled to the pre-reform high level of social assistance if they were unemployed. And anyone granted residency after the reform would be entitled to the new and lower level of social assistance that was called Start Aid. And so this is a very neat discontinuity setup where you just have overnight, you just have a new policy regime for a very well defined group of people and this basically allows you to do an effect analysis of this reform on a number of margins that is fairly straightforward because the reform is very cleanly implemented. Technically we’re using what is known as a regression discontinuity approach. And so what we do here is we essentially compare refugees who are granted residency just prior to the reform to those who were granted residency just after, under the assumption that the only difference between these two groups of people that were granted residency very close in time, that the only difference was whether they were eligible for the older and more generous level of benefits or the newer and lower level. And in favor of that argument, the people who were granted residency around the time of the reform had already applied for asylum in the country on average 15 months before and the reform we’re looking at, the implementation period was like three months or so, so we think of that as a pretty strong argument for why there should be no other differences between the groups right around the reform date. This is backed up by a bunch of different types of statistical tests and visual analysis of the data. And so there are no signs that the setup shouldn’t work, so that’s very comforting.

Chancellor: I asked Andersen to give us a sense of the shift in the levels of generosity before and after the policy shift and how we might think of this in a U.S. context.

Andersen: That’s a good question because it’s so context dependent. I would say our social assistance levels are—the pre-reform ones were among, if not the highest in the world, then at least certainly among the very highest benefit levels you would get. It’s not a lot of money. I mean taking what we consider sort of a minimum wage job in Denmark would still give you a higher income than the social assistance level, but you are reasonably above what we would consider the poverty line. So that changed with the reform, the reform actually cut the levels down on par with what you typically think of as a poverty line measure, which you can kind of calculate as a subsistence minimum. It turns out that the subsistence minimum we find in Denmark, the amount in purchasing parity adjusted dollars actually corresponds very neatly to the U.S. that’s being—the official one from the U.S. Census Bureau. We have a group that now gets the social assistance level that corresponds to the U.S. subsistence minimum level before the reform they had twice that amount.

Chancellor: And, as Andersen mentioned before, part of what makes this kind of study possible is the quality and—expansiveness—of the data that Denmark has.

Andersen: Some people think about Denmark as this sort George Orwellian data nightmare because we basically record everything about people. So I usually describe the data as follows: So when you’re born in Denmark, the first thing that happens is this little machine prints out your social security number. It’s unique. It’s yours. No one else will ever be getting that particular number. All of these sort of birth related information is being recorded on that number. And this is a process that actually happens throughout your entire life. So enrollment in preschool, daycare, how you perform in school, who your mother and father are. As you progress through life, your tax records and your educational attainment, and so on, so on, where you live, and these sorts of things. All of that is recorded on this particular number. And so the beauty of the system is that these information are sent to Statistics Denmark, the national statistical agency, and they document the data and they sort of explain if there are discontinuities in variable descriptions over time and these sorts of things and then they make the data available for research. So this essentially gives you a full population individual level panel data of anything that happens to people’s lives. So I was born in 1980, which is when the major bulk of data is from. So you could basically track any of my institutional or administrative movements throughout my entire life. So that sort of in many ways is an unprecedented data source because you can select exactly the type of people that are targeted by a specific policy reform at a historical point in time. And you can follow them along various margins and across time if you want to. You can follow their family members and even into the neighborhood characteristics over time and these sorts of things, so it’s a very powerful data source for policy evaluation.

Chancellor: But it’s not just those people who are born in Denmark that receive an ID – or Civil Personal Registration number. Immigrants and refugees, for example, also receive a number once they enter the country and interact with public systems in some way.

Andersen: Anyone coming to the country—so if you came to the country and had contact with the doctor, you would actually be given a temporary number too. That would still be recorded, it’s just you wouldn’t really be able to link it to anything else because you wouldn’t have contact with any other institutions—so it’s basically all types of communication between people in the country and welfare institutions broadly defined that are being recorded. So we have all this information on the refugees too—most of it from the moment when they were granted residency. Some information about how they progress through the asylum system, but that’s a fairly new data sources.

Chancellor: Given all that, I asked Andersen what they actually found—what were the results of this policy change?

Andersen: One of the things about this paper is there are a lot of empirical results and that’s really where all the juice in paper is. The overall labor market response, the main labor market response that we find is that the policy actually worked in terms of increasing employment. So, within the first year after being granted residency, those who were granted residency before the reform had an employment rate of around 9 percent within the first year. Those who were granted residency right after the reform had 18 percent. So it’s a doubling of employment, of employment rates. So this is a really good thing, right. This is a sort of what the intention behind the reform.

Chancellor: Still, 9 percent and even 18 percent aren’t very high employment rates, so there were still issues with improving employment levels for this group. And Andersen says that, broadly, this lack of full employment can be linked to factors related to the policy and to factors related to the Danish labor market.

Andersen: If we look at sort of how the policy was implemented and administered and sort of in general how social assistance payouts work in Denmark, we sort of started with the point of observation that you have, almost parallel to the employment effect we were observing, you have a very similar effect on the amount of people who just drop out of the labor force altogether. Dropping out of the labor force is not synonymous with unemployment. You can be unemployed and looking for a job and receiving social assistance and here you would have to participate in integration courses, language courses, and so on. But you can also be in category that’s called “dropped out” and here you’re not employed, you don’t have a job. We don’t know if you’re looking for one but from sort of a welfare state point of view, you at least dropped out of the labor market system, especially concerning the point that you’re no longer required to participate in language courses and integration courses. You can participate in them, but there’s no financial incentive to do so. If you can choose between being at home on your couch or sitting in a democratic participation course, there’s no incentive for making sure that they actually go and participate in these types of courses.

Chancellor: And to be clear, those who are in this dropped out-and-not-participating-in-integration-courses category are ineligible to receive benefits except in the relatively small number of cases where they’re exempted because of disability. But another issue is that in Denmark, they have what are called mutual spousal support requirements which, in effect, can lead a fast drop off in benefit levels when a household’s combined wages go up.

Andersen: So let us say that my wife and I are both unemployed, we both receive social assistance. That’s fine. We both participate in labor market programs and so on. Then she gets a job. Then my social assistance will be lowered dollar for dollar as her earnings increase above the level of her own social assistance. What this means is that you have a very high marginal tax rate on the salary you’re bringing in relative to what you were getting in social assistance. And so what we’re observing when we’re starting to focus on couples instead of focusing strictly on individuals, because of the spousal support requirements. What you see is that all of the employment response is driven by males and all of the dropping out of the labor force response is driven by females. So, it appears as if there are within-household decision-making going on that sort of adapts the response of males and females to the way the policies are being administered. Having a very strong labor market response for males, I think it’s like a 17 or 18 percentage point increase—that’s a very good thing. But it’s paralleled by an equal exodus of women from the labor force. And they just basically stay at home and take care of the house then.

Chancellor: And it also may be that the structure of the Danish labor market may cap employment levels for this population.

Andersen: So what we see is that the full employment response is driven by the take-up of unskilled manual labor. So regardless of whether you have no educational credentials or whether you have more than 12 years of education when you arrive to the country, everyone who gets a job gets it in unskilled manual labor. And they also all only get it in jobs that pay more than what we call the minimum wage in the country. We have no formal minimum wage in Denmark, but we have very strongly unionized labor market. So that works de facto as a minimum wage—the unions set bargaining levels. You could probably see an even bigger employment response if the context that you’re incentivizing people into was more receptive to the type of skills that immigrants or refugees are bringing to the country. And this demand for the types of skills they bring, you can interpret that both as sort of—we’re not valuing the skills that they’re bringing, but also that we’re not really valuing the things that these people bring to the table, so more of a discriminatory kind of interpretation of that, of that issue. So essentially, when you have a highly unionized labor market, that doesn’t necessarily appreciate the types of skills people are bringing in, there’s a limit to how high you can expect employment effects to be. So essentially, put directly, you can incentivize individuals almost as much as you want, but if there are no sort of structural job—if the labor market they can enter into is strongly limited, incentivizing individuals is not going to do much because they’ll just be standing there looking around for jobs that are not there.

Chancellor: Andersen says that while the main part of their paper focuses on labor market outcomes—that was main intention of the reform—they looked at other outcomes as well.

Andersen: So what we wanted to do was to take sort of a broader social perspective on this type of reform because while we do see an increase in employment, we also see that the employment rate is still very low, right? That’s what we’ve been talking about until now. And so this means that 60 to 70 percent of the families that were granted residency after the reform have to live off of the now very low levels of assistance. And that’s something that you would expect to matter for people’s lives on a number of margins. I’m pretty you guys here at the Institute for Research on Poverty know that that can upset a whole bunch of responses that are not necessarily super good. So what we do is we focus on, as a behavioral outcome of adults we focus on crime. And for kids we focus on educational outcomes and on crime too. So, to take them step by step—so crime for adults, what we see is that with the reform there was a fairly steep increase in the average number of criminal convictions in the group. We see this as an immediate response, we see it in the first year, you see a higher conviction rate, and it persists at least through the first four years, you have a substantially higher level of criminal convictions. Now no matter whether we are looking at the entire adult sample, whether we are looking at people who are in couples, whether we’re looking only at men and whether we’re looking only at women, we just get the same result. There’s a pretty substantial increase in criminality. Now, interestingly, most of the response is driven by what we call property crime, so crimes against property broadly defined. And an interesting result is for the females, we see that a very large part of the crime response is driven by one specific type of property crimes, and that’s theft from supermarket as it is called. The interpretation that this could be related to food insecurity is fairly straightforward, but we don’t really know what is being stolen. we just know that it’s stuff that’s being stolen in supermarkets, so it could be clothes, it could be TVs or whatever. But either way this seems to be a response along the margin of what you could call subsistence criminality. Which makes sense when you think about the nature of the reform and what it did, cutting income to the level of the subsistence minimum.

Chancellor: And, as he mentioned, they looked at effects for children—but because that could be anyone between birth and age 18, there’s a lot of range in the sorts of relevant outcomes you might look at there.

Andersen: So what we do is we split the group into age groups sort of defined by what type of communication with the welfare state you have at specific age steps. And so for those who were granted residency in early childhood, 0 to 5 years old, we look at educational enrollment, preschool and daycare enrollment, and how they do in school when they start. One of the things there’s been a push for in Denmark is to make sure that immigrant children are enrolled into daycare and preschool. The argument is basically the same as I was talking about for the integration perspective of making people move into the labor market, into the workplace so they can communicate with folks in Danish, sort of learn the customs, learn the language—it’s the same idea with the children. You want them to be around native children and sort of communicate with them and learn the language, be exposed to a lot of Danish words and these sorts of things. And so what we see is there’s a strong decrease in the share of refugee children who are enrolled in preschool and in daycare. This is quite remarkable because, in many contexts you think, sure they have less money so why would you spend money having your kid in daycare when, especially if the wife is at home, which is one of the effects we saw on the labor market, especially for these low income groups, preschool and childcare is free. There’s no out of pocket payment for these sorts of things. Even for me, daycare and preschool and these sorts of things are highly subsidized, so I only would pay like 25–28 percent of the full price. So this is not a response that’s directly linked to out of pocket payments. This is more of a choice. We see these lower enrollment rates both during the first and the second year that the children are in the country. Then what we do is we follow them into the early school years when they are tested on their language skills in the early grades. And what we see here is exactly what you’d expect from what I’ve said before. The risk of these kids performing really badly on language tests just increases quite dramatically—it almost doubles. This is an effect that’s driven almost exclusively by what we call language comprehension, so the ability to hear sounds and translate them into Danish letters. So that response fits very well with what was the very point of having refugee children being enrolled in Danish daycare and preschools. This is sort of, again, not the type of effect of the reform that you were probably looking for as a policy maker.

Chancellor: For those children aged 10 to 13 when they were granted residency, they were able to look at what their educational outcomes were as of 2016.

Andersen: So this is at a point where you would sort of have expected them to progress more or less through the education system. So you would get sort of a picture of educational attainment. And what we see here is that for those who were granted residency right after the reform, they achieve on average half a year less schooling by 2016 than comparable kids who were granted residency right before the reform. Interestingly, this effect was driven exclusively by boys. So you have boys who end up having almost a full year less education by 2016 than their pre-reform counterparts. So what’s happening is they move out and take low level jobs. Or they receive social benefits themselves. And we interpret this as sort of a way to chip in on the household financial situation basically, so again, it’s sort of a response—it makes sense that this type of response could be linked to the reform in itself. The educational margin is pretty important. So we’re looking at a difference in years of schooling that sort of corresponds to whether you have the possibility of completing vocational training or not. This is something that can matter quite dramatically for lifetime income and for wellbeing in general, studies have shown.

Chancellor: Finally, the last group they looked at were the children who were granted residency during their teenage years.

Andersen: And here we look at criminal convictions by age 20. And here we see a very strong response again, so much higher levels of involvement with the criminal justice system for those are granted residency after the reform compared to those before the reform. Both groups have much higher rates than natives, but you still see this very substantial effect within the group. The effect is driven by those—looking at teenagers broadly sort of provides you with a challenge that people who are younger than the age of criminal responsibility, 15 in Denmark, they’re not formally processed in the system so you have a bunch of people there who by definition have zeroes in the criminal justice system outcomes. So if we look only at those who are older than the age of criminal responsibility during their first year in the country, what we see is that the response is indeed driven by them, and it’s an even stronger response naturally. In crime, there’s this very strong gender effect. Almost all criminality is driven by men, young men in particular, so if we do this analysis only for boys who were older than the age of criminal responsibility when they were granted residency we see a doubling of the average number of criminal convictions for those who were granted residency right after the reform compared to the reform just before. Once again, this is a very strong behavioral response to the reform and one that was not very likely to be intended by the policy makers.

Chancellor: At the beginning, Andersen told us that one of the reasons this study is relevant is because there are so many other countries that are exploring similar policies right now, and so I asked him what advice they would offer based on what they’ve learned.

Andersen: Our urge to them is to make sure to pay attention to detail, pay attention to details in the way these reforms are implemented, pay attention to the incentives that the reform provides relative to what disincentives that might be built into the context that you’re reforming it into. Pay attention to context. Context is super important, so maybe in addition to thinking about incentivizing individuals, you might want to think about incentivizing the context to provide jobs that would make sense for these people to take. Or, in one way or another, at least pay attention to the structural constraints of the society you’re implementing it in. And then there’s the question of all of these dynamic or derived effects, all of these other responses that are offset by this type of reform is that, if we implement policies that are meant to incentivize people to take up employment but you’re not getting, in principle, full employment out of those reforms, you should probably pay attention to the dark side of the reform. So those who will have to live off of lower income levels just because that’s the main tool we have for incentivizing people, and so our results show some pretty dramatic effects along a number of very important margins, margins that are particularly important for the future of inequality within the country. And so this is something you might want to think carefully about as you implement these types of policies. So how do we balance individual level incentives and labor market responses with the types of consequences that income reductions can have for families?

Chancellor: Thanks to Lars Højsgaard Andersen for sharing this work with us. The paper, with co-authors Christian Dustmann and Rasmus Landersø, is available as Study Paper 138 from the Rockwool Foundation. This podcast was supported as part of a grant from the U.S. Department of Health and Human Services, Office of the Assistant Secretary for Planning and Evaluation but its contents don’t necessarily represent the opinions or policies of that Office, any other agency of the Federal government, or the Institute for Research on Poverty. To catch new episodes of the Poverty Research and Policy Podcast, you can subscribe on Apple, Stitcher, or Google Play Podcasts. You can also find all of our past episodes on the Institute for Research on Poverty website. Our theme music for this episode is “Staring Straight” by Maarten De Boer. Thanks for listening.