Most countries generally prefer higher quality immigrants. In most situations, this refers to educated, high-skilled, high-ability and motivated people who will be successful on the labor market. To increase the quality of the immigrant pool, many countries have turned to selective immigration procedures. This basically means that the higher the immigrant’s quality, the easier it will be for them to obtain a visa.
However, while skill and education levels are usually observable by immigration officers, other indicators of quality such as ability or motivation are not. This “duality” of quality characteristics poses some challenges: as we will see, increasing selectivity on observable quality characteristics may actually lead to a decline in overall immigrant quality.
This is the main theoretical prediction of a working paper by Bertoli, Dequiedt and Zenou (2014). Let us start with a quick description of their framework. Suppose there are two countries, an origin and a destination country. People from the origin country can be either educated or uneducated, and they can decide whether to work at home or in the destination country.
Wages vary across individuals but are generally higher for educated people in both countries. Migration to the destination country has costs, which are the same for all individuals with the same level of education. It is assumed that these costs are lower for educated people, which appears to be supported by other (empirical) research.
So level of education is the observable indicator of migrant quality in this model. The destination country can increase the number of educated immigrants relative to uneducated immigrants by decreasing migration costs for educated migrants relative to uneducated ones. This is positive selectivity on observable characteristics, as implemented in many countries.
We have an unobservable characteristic as well. This can be interpreted as the ability of the immigrants and varies from person to person. In the model, it is represented by the wage at destination (or more accurately by the difference between the stochastic components of an individual’s wages at destination and at origin).
Whether people with a higher ability will be more likely to emigrate or not is a parameter of the model. I.e. the authors can control it. The more likely scenario is that yes, people with a higher ability are more likely to emigrate. So self-selection into emigration by unobservables is positive.
So to summarize: we can have positive/negative selection on observables (education, skills) and also on unobservables (ability, motivation), and the nature of selection on these two sets of characteristics need not be the same. The former’s selectivity is chosen by the destination country (and is mostly positive in the real world), the latter’s is a parameter (input) of the model.
The main prediction of this model is that assuming positive selection on unobservables (i.e. higher ability individuals more likely to migrate), if the destination country increases selectivity on observables (e.g. education), it can happen that the overall quality of immigrants decreases (keeping the immigrant stock constant). This is a highly counterintuitive result.
Once again, it says that even though the destination country increases its intake of educated immigrants (and since the number of immigrants is assumed to be constant, it consequently decreases its intake of uneducated immigrants), the overall quality of immigrants decreases.
Quality here is defined as the average wage of immigrants (proxying for labor market success).
Why is this the case? Here is my attempt at an intuitive explanation. Selectivity on observables is increased by the destination country by decreasing the costs of immigration for educated migrants. This leads to a dilution of selection on unboservables. In other words, when migration costs for educated people were higher, unobservables (e.g. ability) were a more important determinant of whether a person will emigrate. Now that costs are lowered, unobservables will – so to speak – not matter so much for the migration decision. So educated people with worse unobservable characteristics (e.g. lower ability) will emigrate too.
Therefore, there will not be such a high self-selection into emigration on unobservables (e.g. by ability), i.e. selection on unobservables will be diluted. This will lead to more lower ability educated immigrants and can thus lower average wages (which proxies for quality).
At the same time of course, there will be fewer uneducated immigrants. So this has a positive effect on immigrant quality.
But recall that we’re talking about a scale-preserving change in selectivity, i.e. the overall number of immigrants does not change. For this reason, the drop in migration costs for educated people (and the consequent increase in educated immigrants) has to be offset by an increase in migration costs for uneducated people (and a consequent drop in uneducated immigrants’ numbers). But when that cost increases, we will have the opposite of the effect that we had for educated immigrants. Higher costs increase selection on unobservables (e.g. ability) and will push uneducated immigrants’ wages up.
The net effect of these two opposing forces on quality can vary. As the discussion above shows, the effect can be negative. But it can also be positive. The authors show that quality first increases as you increase selectivity on observables, but after some critical point it starts decreasing, as illustrated in the figure below.
This figure plots the difference between the migration costs of uneducated and educated workers on the horizontal axis as a measure of selectivity. I.e. as we go to the right, selectivity increases. On the vertical axis, we measure immigrant quality. The three curves represent three different levels of immigration. I.e. k_1 may refer to having 500,000 immigrants, k_2 to 1,000,000 immigrants and so on. The point is if we keep the number of immigrants constant (i.e. if we stay on the same k curve), we can see that selectivity has a hump-shaped relationship with quality.
As we can see that point where quality is maximized is increasing in returns to education both in the destination (difference between first set of mus) and in the origin (difference between second set of mus) countries. Furthermore, it is decreasing in the degree of selectivity on unobservables (gamma).
Let me reiterate once more that these results hold when there is positive selection by observables (e.g. education, skills) by the destination country, and there is also positive self-selection into emigration by unobservables (i.e. higher ability, more motivated people choose to emigrate more often than lower ability, less motivated people).
If for instance, there is negative selection on unobservables, then the result does not hold: increasing selectivity will always increase migrant quality. The intuition is similar to the one presented above for the positive self-selection case.
Finally, let me note that this is a pure theory paper, meaning that data may or may not back these claims up. The model does make sense and is likely to be true, so the more interesting question is whether any country in the real world is (or will ever be) in the region on the above curve where quality is decreasing in selectivity. This I believe is the main empirical question posed by the paper.
For instance, if selectivity on unobservables is positive but low in magnitude, then the peak of the curve may be so out to the right, that in the real world it may never be reached. So another interesting empirical question would be to identify how big the effects of unobservables are on migration decisions.
My hunch is though that it’s a plausible scenario that unobservables matter so much that some countries may (some day) find themselves in the perverse case of a downward sloping selectivity-quality curve.