The macroeconomic effects of immigration were a topic of heated discussion during the Brexit referendum and the US presidential election. A common argument was that immigrants might limit job opportunities for natives. In the Brexit debate, discontent about the wave of immigration that followed the EU enlargement to include Eastern European countries was one of the key factors in determining the outcome of the referendum.
Many have studied the effects of immigration flows on employment and wages. They have used mostly disaggregate data (see Kerr and Kerr 2011 for a survey). But the impact of immigration on the macroeconomy has not been investigated systematically. Our recent research aims at filling this gap using Norwegian data (Furlanetto and Robstad 2016), as Norway is one of the few countries for which a quarterly net immigration series is available from the early 1990s. The case of Norway is interesting because, while immigration was a marginal phenomenon in the 1990s (see Figure 1), it became the dominant driver of population growth after EU enlargement to include Eastern European countries. In 1990 3.5% of the population were immigrants. By 2014, this was more than 12%, driven by large migration flows from Poland, Sweden and Lithuania. The immigration share for Norway in 2014 is comparable to that of Denmark and Sweden, and higher than that of the UK and Germany, both slightly lower than 10%.
Figure 1 Annual change in Norway’s population (%)
The availability of a long time series allows us to include a net immigration in a vector autoregression (VAR) model, which is the most widely used empirical model for macroeconomic analysis. Immigration is a fully endogenous variable in our model, responding to exogenous immigration shocks but also to a series of macroeconomic disturbances driving the business cycle. The sample period is the first quarter of 1990 to the second quarter of 2014, and we achieved identification by imposing a limited number of impact sign restrictions on the response of macroeconomic variables to shocks.
We imposed two key sign restrictions to disentangle immigration shocks from other disturbances. First, that an exogenous increase in immigration would have a negative impact effect on wages, in keeping with the effects of labour supply shifts in standard macroeconomic models, and with recent microeconomic studies on the impact of immigration in Norway (Bratsberg and Raaum 2012, Bratsberg et al. 2014). Second, that an increase in immigration has a positive effect on the labour force participation rate. Ideally, we would have liked to concentrate only on immigrants that rapidly enter the labour force, irrespective of their geographical origin, but the quarterly immigration series is only available by country-groups, and so we cannot isolate job-related immigration. Therefore, we restricted our focus to immigration from western countries (EU/EFTA countries, North America, Australia, New Zealand and Eastern Europe) that nevertheless constituted the bulk of the recent immigration boom, which is mostly job-related.
Additional annual data from Statistics Norway confirms that work was the major motivation for immigration from the countries included in our analysis, whereas this is not the case for the countries excluded from our analysis, and data from 2014 show that the immigrants included in our series had on average higher employment and unemployment rates than natives, thus suggesting a higher participation rate. In contrast, immigrants from non-western countries exhibited an employment rate substantially lower than natives, and migrated to Norway mainly because of family reunification or as asylum seekers. As these immigrants probably entered the labour force with a delay, including them would have invalidated our identification assumption for immigration shocks.
In Figure 2 we present the responses (measured in percent) of GDP for the mainland economy, real wages, the immigration rate (defined as the ratio between the stock of immigrants and the active population) and the unemployment rate (measured in percentage points) to an exogenous increase in immigration over a horizon of 36 quarters (the horizontal axis). An expansionary immigration shock has persistent effects on GDP, real wages and the immigration rate, despite only restricting the impact response for all these variables.
The unemployment rate declined on impact, and more so after few quarters. The size of the response is remarkable and may highlight some complementarities between immigrants and natives. This effect on unemployment is not necessarily surprising if we consider that many immigrants (in particular from Eastern Europe) come to Norway with a job offer, but when we re-estimated the model using a measure of unemployment for native Norwegians, we found tit also decreases. Therefore, according to our analysis, an increase in job-related immigration seemed to create new jobs for natives, not crowd them out. A decline in unemployment in response to an increase in immigration is consistent with previous estimates for the US (Peri 2012).
Figure 2 Responses to an exogenous increase in immigration
Note: Impulse responses to a one-standard-deviation immigration shock. The dashed-dotted line represents the posterior median at each horizon, and the shaded area indicates the 68th posterior probability region of the estimated impulse responses.
The hump-shaped response in the immigration rate may be explained by a combination of three factors. First, it may reflect registration delays. Registration is mandatory only for employment contracts longer than six months, and it is possible that many workers started working in Norway on short-term contracts (contributing to GDP) and registered once they obtained a longer contract. Second, it may be related to family reunifications. While work is the main reason why immigrants included in our sample came to Norway, the importance of family reunifications was not negligible. We can imagine that it may have taken some time before the whole family moved following labour-motivated immigration. Finally, the peak after 10 quarters may also reflect network effects – immigrants from the same country tended to follow each other.
GDP response peaked after four quarters, well before the peak in the immigration response. This may reflect composition effects leading to a decline in productivity. When we introduced output per hour as a measure of labour productivity, we estimated a negative response of productivity to a positive immigration shock. This decline in productivity was mainly driven by a strong drop in capital intensity reflecting the adoption of less capital-intensive and more unskilled efficient technologies (Lewis 2011).
We considered several other variables in our study. The burden that immigrants may place on public ﬁnances is another popular argument that is used to oppose immigration, as social security programs in host countries (and in Norway in particular) are more generous than in immigrants’ countries of origin. We found that a positive immigration shock lowered public expenditure in the short run, but increased it in the long run, perhaps reflecting family reunifications. But fiscal revenues also increased, and so the net effect on public finances was slightly positive in the short run, and neutral in the long run.
We also found that an immigration shock had no effect on house prices or household credit growth. Therefore, immigration does not seem to have played a big role in driving the recent housing boom in Norway. One possible explanation is that many immigrants were Eastern European workers who were active in the construction sector. While many of these workers were unlikely to buy a house in the short run, they still contributed to the supply of new houses.
Finally, we included different measures of prices: an expansionary immigration shock had no effect on domestic prices, but increased in the Consumer Price Index (CPI) in the medium run through exchange rate depreciation. Positive immigration shocks seemed to have expansionary effects on key target variables for monetary policy, such as the unemployment rate and CPI inflation.
Our research did not find any support for the macroeconomic arguments that have recently been used against immigration. In our model, immigrants do not limit job opportunities for native workers, and an increase in immigration has no negative effects on the fiscal balance (if anything, we find a small positive effect). It is important to stress, however, that our results refer only to immigration from western countries, and so largely capture job-related immigration. They cannot be extended to evaluate the impact of other kinds of immigration, such as an increase in refugees. While fears for employment and the balance of public finances seem misplaced, we also identified a negative effect on productivity. This may be a worry for long-term growth.
Authors’ note: The views expressed in this column are those of the authors and should not be attributed to Norges Bank.
Bratsberg, B and O Raaum (2012), “Immigration and wages: evidence from construction”, Economic Journal 122, 1177-1205.
Bratsberg, B, O Raaum, M Røed and P Schøne (2014), “Immigration wage eﬀects by origin”, Scandinavian Journal of Economics 116, 356-393.
Furlanetto, F and Ø Robstad (2016), “Immigration and the macroeconomy: some new empirical evidence”, Norges Bank Working Paper 18/2016.
Kerr, S P and W Kerr (2011), “Economic impacts of immigration: a survey”, NBER working paper 16736.
Lewis, E (2011), “Immigration, skill mix, and capital skill complementarity”, Quarterly Journal of Economics 126, 1029-1069.
Peri, G (2012), “The eﬀect of immigration on productivity: evidence from U.S. states”, Review of Economics and Statistics 94, 348-358.