When it comes to migration there’s no such thing as a pull-factor - at least that’s what Médecins sans frontières has to say. The Refugee Council goes further saying that some three-quarters of migrants do not even dain (or as we may say later, dane) to claim awareness of the existence of a welfare state. The European Council on Refugees and Exiles again draws the same conclusion - with articles concluding that both full factors and any sense of a homo economicus does not exist.

They have all been presented, in one form or another in the Guardian, the Guardian again, or the BBC as evidence that welfare migration not only does not exist but cannot exist. Seemingly the large majority of press communication arguing that not only is there no such thing as a pull factor, but that migrants were not even aware of the existence of the things supposedly pulling them, stem from papers developed by one of two people - Migration Professor Heaven Crawley and Sociology Lecturer Lucy Mayblin.

What is regrettable in the research though is that it is in essence pure expressed preference. Crawley and the Refugee Council’s findings are dependent on interviews with 43 refugees and asylum seekers coupled with a focus group. Whilst qualitative research no doubt has a role it seems, shall we say, methodologically interesting for journalists to then present findings from a sample with a 14.94pp error as the Guardian’s gospel truth. Mayblin’s research, much of it cross-referencing the work of Crawley, reaches the same pre-ordained conclusion that pull-factors simply do not exist.

Raw simple correlational quantitative research comparing welfare levels to migration rates (and underpinned on the presumption that only levels, not access, play a role), has supported these ‘findings’.

The conclusion is of course a helpful one to draw ideologically. After all, if there is no such thing as a pull factor, there is no way to manage border crossings - indeed, it becomes essentially foolish to try. The rational position is my position, whereas you are merely seeking to constrain the inevitable. Border policy, regardless of what the public may want, becomes like attempting to micro-manage the rising of the sun. I am rational, you are an Aztec sacrificing people on the altar of Huitzilopochtli.

Thankfully however, and given the wonders of disciplinary imperialism, we have papers that do what our previously reviewed papers do not - they actually look at revealed preference and they actually have a quantitative grounding.

One such paper is the Welfare Migrant Hypothesis: Evidence From An Immigrant Welfare Scheme in Denmark, from three Danes working on migration economics in the US.

The paper’s essential method is quite simple - it does not use current interviews to justify the previous decisions - rather, it studies empirically what actually happened in Denmark when welfare levels were altered; the natural experiment arising from a policy being introduced, repealed, and then re-introduced.

The policy studied and its context is this - following Denmark’s 2001 election in which the right (the blue block) triumphed winning 52.3% of the vote and 94 of the Folketing’s 179 seats a new centre-right government was formed under Anders Fogh Rasmussen (famed both for being NATO Secretary-General and the 2nd of 3 back-to-back Danish Prime Ministers named Ramussen). This combined the conservative liberal Venstre and the liberal conservative Conservative People’s Party, and was propped up by the right-wing populist Danish People’s Party. This government, elected at the start of what has become known as the consensual closing, was eager to shore up Denmark’s public finances and address popular concerns over immigration and so tried a novel approach - reducing migrant welfare.

More specifically, they enacted Start Help, a specific welfare payment for new non-EEA citizens to replace access to the mainline existing cash assistance (kontanthjælp) program that native Danes were able to benefit from. The process of adopting the new benefit regime allowed substantial cuts to take place - up to 50% for some - with, for example, the benefits of a childless couple falling from the equivalent of nearly $2,500 a month to a little shy of $1,300 or a single person with a child from $1,500 to $900.

Source: Agersnap, Ole, Amalie Jensen, and Henrik Kleven. "The welfare magnet hypothesis: Evidence from an immigrant welfare scheme in Denmark." American Economic Review: Insights 2, no. 4 (2020): 527-542.

When the right-wing government lost its re-election bid in 2011, and the blue block’s left-wing counterpart the red-block under the leadership of Borgen-analogy, and Corbyn-era TV interview fame, Helle Thorning-Schmidt, formed a new government, these cuts became one of the first things to be rolled back. The government of social democrats, social liberals, democratic socialists, propped up by ecosocialists and now-topical Greenlandic left-wing parties, took issue with the Start Help scheme arguing that the policy had failed in its objectives of integration and immigration management.

After four years of a troubled left-wing government, the right was back, with the Danish People’s Party 2nd on 21.1% of the vote, and with it the Start Help scheme. That natural experiment underpins the paper - introducing the scheme, repealing it, and reintroducing it. That policy chaos helps remove confounding effects too - the effects of similar legislation introduced at the same time can be discounted here given the policy uniquely happened thrice. The same strategy counters the 2015 migration crisis shock that coincides with one of the reintroductions - as does running the sample both without Syrians and in comparison to the other Nordics who had not introduced such legislation. Most importantly, its application to only non-EEA immigration, given EU law on migration access, allows for both a treatment and control which, though not compositionally the same, can still have a difference-in-differences approach applied to them.

Difference-in-differences is a relatively simple calculation where the impact is the treated group endpoint value minus the treated groups beginning value, with the same formula for the control group then minussed from it. Migrants from the European Economic Area then (or more accurately the EU pre-Brexit plus EFTA) form the control. What is the before and what is the after is created by applying the linearity observed in the historic trend to future migration with the trend forming the before and the reality the after. Knowing the trend then allows for both trends to be removed and the study to focus on a direct comparison between the two groups (ie 0 in the detrended data being exactly what the linear trend would predict).

What they find from that approach is clear - where for the two decades prior to the reform both EEA and Non-EEA immigration detrended evolved in parallel, it diverged sharply in 2002, returned to the same trend level post 2012, then diverged sharply again at the 2015 reintroduction. Also important here is that the scheme took 4 years to have an effect and repeal took the same amount of time, meaning 2015 was not a mere spike. Its symmetrical pattern is proof against confounding.

This detrended data then suggests that the cuts involved in the welfare scheme reduced relevant net migration by some 5,000 a year - a near 50% annual cut versus trend. A pattern that persisted for years.

To check their validity they trialled a second approach too - constructing a synthetic control from Denmark’s neighbours. Here Norway, Finland, and Sweden’s data is weighted and averaged to form a synthetic Denmark. These countries' data points, normalised as a percentage of the existing migrant population, are matched with Denmark’s. They find in the period pre-reform that these trends evolve in parallel (the essential point for robust findings) until the 2002 Danish reform where only Denmark diverges from the Nordics.

The parallel evolution allows for that synthetic Denmark to be estimated - removing concerns that the spike after the 2012 abolition is confounded by the 2015 migration crisis given all Nordic countries were affected by it but all other Nordics see their migration spike in 2016 not 2015 alongside Denmark (the last year pre welfare scheme reintroduction). As the paper says this then is “consistent with a model in which refugees choose where to seek asylum based in part on benefit levels, and where the reduction in Danish welfare benefits causes some Syrian migrants to forgo Denmark in favor of other destination countries”.

This method of estimation implies an either larger cut in migration than the prior approach - though issues with substitutability of the Nordic countries (ie that if deterred from Denmark one is likely to head to Norway whose society is structured the same and whose primary language is derived from it - æ, ø et al) do create a challenge in using it further chiefly that if deterred from Denmark and shifted to Norway the same person would count twice. Regardless, running both approaches allows an accurate estimate of impact from the former, and a check for confounding from the latter.

Returning to estimation method one, the difference in differences between Denmark and trend for EEA and non-EEA migration, allows for a specific elasticity of migration demand to welfare supply to be estimated.

Here, they create twelve groups based on the conditions associated with their welfare payments (single or a couple; childless, one child or two and more), and whether they are EEA or non EEA. Each group can then have their own welfare responsiveness measured through applying the same-detrended approach used above on the specific welfare cuts the non-EEA group faced versus the EEA-group. These findings can then be brought together and weighted based on the share of the overall applicable population to find a central estimate.

The findings then are this - for immigrants over 30 an additional $100 in benefits increases the number of new immigrants by 0.44% of the pre-existing number. That is to say that if benefits rise by $100 and there are 100,000 existing immigrants prior to the benefit rise, then next year you would expect an additional 440 immigrants on top of whatever the existing baseline was. The actual scheme reduced benefits by up to $800 or a 3.5% cut to the existing number, i.e. if the stock is 100,000 then next year you would expect 3,500 fewer new immigrants versus the baseline.

In converting the data to an elasticity, what it yields is quite large. With Syrians included it is 1.29 or without Syria, and with the age parameter being reduced to 18, it is 1.51. A 1% reduction in benefits then can reduce relevant migration by up to 1.53%. This is a high elasticity, in business for example a good is considered luxury where its elasticity is higher than 1, 1.53 then is nearly 5x higher than private school demand elasticity and ranks it alongside oysters and in the short-term new cars.

It should also be noted that 91% of the net migration reduction is achieved through reducing future arrivals rather than encouraging emigration (unsurprising given pre-2002 arrivals were exempted). It is also driven overwhelmingly by asylum claims and chain migration - not through deterring workers or students meaning the most welfare-intense are deterred.

Reducing welfare then clearly deters migration, with welfare cuts providing a strong means to reduce both the scale of the individual net fiscal loss and its growth in future. How transferable these findings are then to other contexts, where benefits are already notably lower (ie differing marginal elasticity) or where the relative importance of other pull factors such as language or ease of chain-migration may be higher, can be debated but the general principle that welfare migration is very much real cannot.

If the general elasticity finding from Denmark were to be linearly applied directly to the UK, something that is not particularly statistically sound as outlined, then one would expect that if Rachel Reeves were to cut the welfare accessible to new asylum seekers by £50 per month (23.5%) then one would expect 35.48% fewer new asylum seekers versus trend. Impacts then are clear and substantial.

Of course, other states have learnt this Danish lesson. The UK has already introduced a No Recourse to Public Funds, though of course it does not mean that there is in fact no recourse to public funds. France’s much talked about Migration Law attempted to implement a similar rule and Germany looks set to implement large cuts to Bürgergeld (a 6000€ a year payment to some 5.5 million people of which half are non German). Once again, Denmark has become a laboratory for immigration experiments, setting out a framework for reducing welfare migration and proving that policy levers are not futile attempts to control the sun. Not bad for a country that the likes of our aforementioned friends at the Guardian has so routinely praised for its progressiveness.

Welfare Migration; Yes It Exists

The (Un)Employment Rights Bill

Britain’s ILR Emergency