by Ian Preston
Centre for Research and Analysis of Migration, Dept of Economics, University College London
Two reports published on the same day last week contain evidence on the impact of recent immigration on employment of the UK-born. One of these (NIESR 2012) claims to find no evidence of an effect; the other (MAC 2012) suggests that certain sorts of immigrants in certain periods may have depressed employment. The difference in conclusions makes it difficult to identify exactly what economic evidence really shows. This note offers some comments on what can be learnt from these two reports.
The reports usefully complement each other. The methods are broadly similar. Both compare changes in employment rates and levels of immigration across spatially defined local labour markets, correcting as well as possible for other effects. However they use different data sources over different periods and at different levels of spatial aggregation so it is not surprising that results differ and it is informative to compare the ways in which they do.
One report (MAC 2012) uses a large labour survey that has been running for many years and the authors are therefore able to assess things over a longer period. However because the survey covers fewer workers they need to aggregate up to larger spatial units. The measure of immigration, based on change in numbers of foreign-born workers, keeps track of immigrants leaving as well as entering regions.
The other report (NIESR 2012) uses recent administrative data of much wider coverage and these authors are therefore able to focus on smaller spatial units but only over a shorter period. The measure of immigration, based on NI registrations, is focused on employment-related immigration but has the weakness that it tracks only immigrant arrivals. Since flows of immigrants between regions and out of the country are not small relative to inflows this seems a significant countervailing drawback.
Some media coverage has taken the line that it is only common sense to think immigration must lower employment prospects for those already in the country and all that has happened is that one study has finally uncovered what should have been obvious all the time. Of course, if it were so obvious then it would hardly have been necessary for so much research effort to have been expended on the question. But it is not obvious.
When, say, an immigrant secures a vacancy for which a UK-born applicant was considered, it is often visible and understandably motivates grievance, even though it may not be obvious that the UK-born applicant would have been offered the job otherwise (and even though leaving the vacancy unfilled might have had consequences for employment of others). When spending by the employed immigrant adds to local demand, that, by contrast, is far less visible but no less important. Mere movement of workers from one country to another does not in itself change the amount of demand for goods in the world. The demand for internationally tradable goods produced in the UK can come from anywhere in the world and is not obviously tied to how many people are in any particular country and as migrants move into the country they bring demand for nontraded goods and services with them. It is perfectly reasonable – indeed arguably common-sense – to start from an expectation that immigration should bring just enough demand into the country to enable immigrants to be employed without any adverse effect on the UK-born.
This expectation is only a starting point - a more sophisticated view will recognise the potential for immigration to affect prices and wages in ways that may feed back onto employment in one way or another. Processes of equilibration take time and there may be short term employment effects even if these disappear in the long run. In the absence of any clear presumption about which way the final effect should go, the matter calls out for empirical research. We should not be surprised if that research fails ultimately to find evidence of any substantial impact.
There have been studies in several countries and the preponderance of evidence is strongly suggestive that employment effects are small if they exist at all. Some of that research, cited in both reports, was done by researchers including myself at CReAM (CReAM 2005); that work failed to find convincing evidence for employment effects in the UK but that, as with much work in the area, was arguably as much because of the difficulty of making estimates precise as it was because they were clearly small. Other people have found similarly scant evidence. In this context, new research such as these two reports is a welcome addition.
The MAC paper looks for effects from immigration of several different sorts in several different periods. Many of these estimated effects are unexciting; typically at most one type of immigration in one subperiod is found to be individually statistically significant. Focussing on individual statistical significance can be misleading when there is no reason to think that the particular type of immigration concerned has drawn attention for any reason other than its having the strongest attached statistic. A better question to ask would arguably have been whether or not one can statistically reject the hypothesis that all effects are jointly zero. It is not at all clear that such a test would support the conclusion that immigration harms employment.
The main results in the NIESR paper are based upon much smaller spatial units. This gives more data points, enhancing statistical precision, but also makes it less easy to be confident that the boundaries of these spatial units reliably contain the labour market effects studied. At this level of spatial aggregation, no evidence of employment effects is found.
The paper also, however, reports results estimated at regional level, the same as in the MAC report. The NIESR authors do not concentrate on these, because they prefer the results at the lower level of spatial aggregation on methodological grounds, but the result do make a striking contrast to those in the MAC report since they both appear statistically not badly determined and actually point to immigration being associated with lower unemployment and being so more strongly in periods of downturn. This contrast diametrically with the MAC conclusions and does not encourage confidence in the robustness of conclusions that can be reached with regional level data.
The MAC report itself is extremely open about the fragility of results to several robustness checks such as removal of outliers. This lack of solidity does not necessarily mean that they are not picking up real effects but does call for caution before using conclusions as a guide to policy-making. Furthermore, insofar as there may be employment effects of immigration, some of these robustness checks hint they may be short term and therefore compatible with being transitional rather than part of an effect on long run equilibrium in the economy.
Recent work on the closely related question of how immigration affects wages in the UK (CReAM 2012) shows that weak effects on the average can be compatible with different effects along the distribution of skills. The NIESR report points to investigation of the possibility of differential effects on employment prospects of high and low skilled UK born workers as a priority for future work. This seems a sensible judgement.
The body of the MAC report is typically careful not to claim causality, preferring to refer to "associations" (even though talk of "displacement" is suggestive of a causal impact). What has been found is that a particular subclass of immigration has tended in part of the data to be higher in regions and periods with rising unemployment than in other regions. This doesn’t mean that it caused the change in employment. Predictably, much media coverage has been less reticent.
It is impossible to interpret these associations without thinking about the processes driving immigration. Immigrant location is not random. There is a lot of evidence that immigrants tend to go to areas where immigrants of similar types already live. This is easily understandable as there may be family connections and associated employment networks in established communities of similar origin. This effect is likely to be weaker among recent EU immigrants who come from populations with less history of immigration to the country.
Immigration is at the same time responsive to economics and this is what raises problems with causal interpretations. Employment booms may draw immigration into an area and this will tend to mask any negative employment effect. Temporarily high wages may also draw in immigration and if that inhibits hiring then this will tend to inflate the appearance of negative employment effects. The problems which this raises are conceptually very difficult to overcome. One response is to see whether similar association is also there between employment changes and that part of immigration flows which can be predicted from plausibly noneconomic drivers, like prior concentrations of immigrants in an area. Both NIESR and MAC check this. While the NIESR conclusions are reportedly robust, the precision of the negative employment effects estimated by MAC tends to deteriorate sharply.
Viewing the totality of new evidence, it is difficult to see a persuasively robust empirical case for long run harmful effects of immigration on employment of the UK-born.
NIESR 2012: “Examining the relationship between immigration and unemployment using National Insurance Number registration data”, Paolo Lucchino, Chiara Rosazza-Bondibene and Jonathan Portes, National Institute of Economic and Social Research, 2012
MAC 2012: "Analysis of the Impacts of Migration", Migration Advisory Committee, 2012
CReAM 2005: “The impact of immigration on the UK labour market”, Christian Dustmann, Francesca Fabbri and Ian Preston, Economic Journal, 115, pp F342-F358, 2005
CReAM 2012: "The effect of immigration along the distribution of wages", Christian Dustmann, Tommaso Frattini and Ian Preston, Review of Economic Studies, forthcoming, 2012