Anecdotal evidence, chocolate, Pret, and general equilibrium

I have been writing quite a lot about immigration recently - here and here.  But there’s an unfortunate tendency for dialogues on this topic (oral or in response to my columns) to go something like this:  I, or other economists who work in this area, say that the empirical evidence suggests that immigration has little or no impact on employment or unemployment overall. A respondent points to an example – they personally, or someone they know, has lost out on an employment opportunity to an immigrant. Or they note that a particular local business or sector seems to employ mostly immigrants – Pret a Manger in London is a frequent target.  We respond, saying that this is “anecdotal evidence”, sometimes (although I try not to use this phrase too much) referring to the lump of labour fallacy, and explaining again what the evidence says about overall or average impact.  The respondent then concludes, perhaps understandably, that economists live in a statistical world which has little or no connection with reality; and worse, that when confronted with reality they are not interested.


I’d like to try to explain why “anecdotal evidence”, in this specific context, is indeed irrelevant, not because it’s anecdotal, but because it is a partial rather than general equilibrium concept.  That is, by definition, it tells you only the local or partial impact of something, which may or not be offset by developments elsewhere in the economy, which – by definition – the anecdote cannot tell you anything about.

To illustrate this I think it’s helpful to look at a different example; not immigration, but inflation.  Suppose that the price of chocolate goes up, as it is doing now, because of bad weather in the Ivory Coast.  Now, what can we say about the impact on overall inflation:


  • in the short term, consumers  will adjust.  They may buy less chocolate and more of other things;
  • and, over time, so will other variables. The exchange rate may fall slightly (our terms of trade have worsened); 
  • and the Bank of England will adjust interest rates to ensure that we hit the inflation target.

Most economists will therefore agree on the following points:.

  • there may be some impact on inflation in the short-term. This will be less than one might expect just from the price rise, and the magnitude and persistence of the impact will depend on how the adjustment process works;
  • there will be no impact at all on inflation in the medium to long term.

Now suppose you want to know how quick the adjustment is – does the adjustment take a week or a year?  This is an empirical question and you need empirical evidence; and there are various ways you might think about finding that evidence. But the key point here is that the fact you're paying more for chocolate doesn't tell you anything about the answer to this question - the impact on inflation overall in the economy. It is not just that the price of chocolate it is “anecdotal evidence”; it is that the anecdotal evidence about chocolate tells you absolutely nothing about the question you are actually interested in, which is about the speed of  the adjustment process in the wider economy. 
The interesting, and difficult question, is not what happens in chocolate shops, but in all shops, for all prices.  The adjustment could take a week or a year, but looking at the price of chocolate alone - no matter how many shops you go into – can’t tell you.  This is because inflation is a general equilibrium concept; it is determined by the overall levels of demand and supply in the economy, and looking at one price – a partial equilibrium concept - can’t tell you anything.
The same is true for immigration and employment. Theoretically, it is quite possible that sandwich shops hiring immigrants increases unemployment for Britons in the short term, and it is equally possible that it doesn’t. But the point is that whether it does depends not on what you observe in the shops themselves, but what is happening in the wider economy.  The number of Britons employed in sandwich shops is a partial equilibrium concept, but the number employed overall – which depends not on sandwich shops, but overall levels of labour supply and labour demand – is a general equilibrium one.  So we need to know what happens to the Britons who are not employed in the shops, and what happens to overall levels of wages and of consumer demand.  Those are the factors that will determine the impact on unemployment.  No matter how many sandwich shops you go into, and how many “anecdotes” you accumulate about Poles working there, that tells you – literally - nothing about the question that you are actually interested in. 
So the point is not that anecdotal evidence is always invalid or irrelevant, or that it is not quantitative; it can often be very useful.  Going into sandwich shops and talking to management and employees about how the industry is organised can tell you a lot about the employment practices of that specific industry; and indeed NIESR publishes numerous high quality research studies that use systematic and structured qualitative methods to do exactly this type of research.  It is silly for economists to dismiss all evidence that doesn’t have numbers attached to it.  But, when looking at economy-wide variables – inflation, unemployment -  there is simply no alternative to looking at developments in the economy as a whole.


Gary's picture

If research shows no correlation between immigration and native employment/unemployment, then can the suggestion that immigration creates employment be ignored?

There is anecdotal evidence that some migrants create jobs.

There are claims by employers and universities that reducing immigration will cost UK jobs. Can these claims be ignored?

Or, as the aggregate effect is negligible, is there a distribution with some migrants causing unemployment and some creating employment? And better analysis, understanding and policies are required to promote employment creating immigration while discouraging displacement etc.

Jonathan Portes's picture

I'm not suggesting "ignoring" reasonable hypotheses/arguments. But yes, I don't think there is any empirical evidence that overall, on average, immigration creates jobs for native workers, and so I would indeed not attach much weight to claims to the contrary by employers or universities; however, in some sectors, including universities, I think there is very strong evidence that reducing immigration will reduce UK output and productivity, and hence our prosperity; those are much stronger arguments than "jobs" for a more sensible policy.

It is possible that, as you suggest, some immigrants create employment for natives and others reduce it; it is also possible that employment prospects for natives are simply more or less independent of immigration. I am inclined to believe the latter. But I do think some migrants enhance productivity and others don't, so I agree with your basic point here about selectivity.

J Smith's picture

Jonathan - interesting article but what we can't see is what would have happened to native employment rates if large scale migration from Eastern Europe didn't occur. Employers in sandwich shops and the like might have had to increase the wages and conditions on offer to draw more natives into the market.

I understand your point that the economy can expand with the influx of more workers but many of the jobs being filled are relatively low skilled and so should be accessible to the current 2.5 million unemployed in the UK. I'm not saying the migrants "took" the jobs but that without the migrant workers the economy would have had to adjust in other ways to draw in the UK unemployed.

Gary's picture

J Smith - 'what would have happened' (the counterfactual) is difficult to measure for obvious reasons. There may have been fewer Pret a Manger shops without numerate, motivated East European staff willing to work for £6.20/hour in London.

The research tends to show that overall migrants do not impact wages (with a very small drop for lower paid workers and very small rise for more well off workers - widening the gap between rich and poor?). So less migrants would not necessarily have led to rising wages and attracting more resident workers.

Of course this also means that cutting migration is unlikely to lead to wage inflation which has been used as a reason for migration.

The research also shows a very small increase in GDP per capita as result of immigration (the batting average effect) but probably no increase in GDP per capita for the resident population. So immigration probably does not help productivity greatly.

It does increase GDP by adding more people but that is probably irrelevant. A bit like saying the UK would be better off if Ireland rejoined because GDP of a larger UK would be higher (or alternatively if Scotland got independence).

Gary's picture

I wonder if there has ever been any research on internal migration of Brits between local authorities and unemployment. ONS estimate internal migration at 2.6 million a year (about 2 million working age) which is 3 to 4 times higher than immigration. I suppose if you ignored the lump of labour fallacy then net movements would increase unemployment in LAs receiving internal migrants and decrease unemployment in LAs losing internal migrants.

No one would really expect that to happen because people are much more likely to move to areas where jobs are being created. It does not really show what would have happened in an LA without internal migration. I suppose it does show the flaws in using spatial correlations to compare unemployment and migration.

Total views: 8,978