Is migration good or bad for indicator X (here, X could be wages, employment levels, self-reported happiness, crime, welfare state use, moral virtue, etc.)? The question, as posed, is ill-defined. The ambiguity could arise from different meanings or interpretations of indicator X. But there’s also considerable ambiguity in the “Is migration good or bad” part of the question. Good or bad for whom? Compared to what?

I was moved to write this post series after an aborted attempt at trying to synthesize what different people had said about the effects of migration. Often, the people were talking past each other, measuring slightly different things. There’s a question of what we should measure, i.e., what measurement is the most appropriate one. But a first step is knowing that it’s possible to be measuring many different things. This post series attempts to clarify the range of things one could be measuring and how they relate to one another.

The series is structured as follows:

1. Part 1: direct empirical measurement focuses on something that can be computed through direct empirical measurement: the performance of people who stay put in their countries, and the performance of people who migrate from one country to another.
2. Part 2: comparative statics, multiple matrices discusses how to compare different policy regimes or scenarios for migration. Such comparisons typically involve counterfactuals and cannot be settled completely by empirical data: we need a model, and there’s considerable model uncertainty even if the data is excellent.
3. Part 3: simplified model assuming no changes to non-migrants considers a simplified situation where we assume that migration at the margin primarily affects migrants and not the natives of either sending or receiving countries. The question is then about how migrants fare relative to the counterfactual where they are not allowed to, or were unable to, migrate. We’ll consider rank-ordering and quantitative comparison of natives of the source country, natives of the target country, potential migrants if they can migrate, and potential migrants if they cannot migrate.
4. Part 4: Models for migrant performance considers different models for how migrant performance might be predictable in terms of the performance of the source and target countries.
5. Part 5 discusses the descendants of migrants, and in particular the interaction with diaspora dynamics.
6. Part 6 wraps up by considering some subtleties that were omitted in the preceding discussion.

The series also includes a number of minor mathematical digressions. If you have a reasonable background in mathematics (up to basic calculus and linear algebra, to the level generally needed for social scientists) you should be able to follow these. But you can otherwise skip the mathematical digressions without loss of continuity.

Choice of analytical focus

Despite the bewildering array of possibilities we’ll consider, there’s a high chance that the model used in the series will remain wanting. We’ll defer a detailed (but still partial!) discussion of the shortcomings to part 6, but a few preliminary remarks might be helpful.

A broad remark worth making is that the analysis in the coming parts will focus heavily on the migrant’s country of origin/birth (the “source country”) as well as the migrant’s country of current residence (the “target country”). We’ll also consider the distinction between migrants and non-migrants. This suggests that there are only three important components to the person’s identity that carry importance in statistical aggregation: the source country, the target country, and whether the person is a migrant. While other attributes can vary, we’re not interested in using them as the basis for grouping, since we’re aggregating over them.

But the reality is more complicated. Religious and ethnic identities can be subnational or supranational. Lebanese Muslims and Lebanese Christians may be best viewed as separate groups (though there are some cultural similarities and they’re probably genetically close to identical). In the United States, the Native Americans (American Indians) may be better viewed as a separate subgroup. On the other hand, sometimes it may be better to consider ethnic or ideological groupings that cut across national lines, such as Scandinavian, Western European, Anglo-American, Arab Muslim, Sunni Mulim, Shia Muslim, sub-Saharan African, Hindu, or ethnically Chinese.

The reason for our singular focus on nationality is simply that immigration law as it currently stands gives extreme importance to national boundaries and national membership. It may be ironic that, on a website devoted to critiquing the existing global regime of borders and migration controls, and the rigidity of national identity enforced by laws, a series of blog posts so meekly follows the status quo. My only excuse is that one needs to start somewhere. But you should feel free to fill in your own variations of the ideas based on forms of identity that do not coincide with one’s place of birth and one’s current residence, rather than wait to get to part 6.

Where’s the data?

As I go over different aspects of the model, you might be tempted to ask: can one actually construct the data that’s needed to do the quantitative comparisons and answer the various questions I pose? Data does exist for some things but not for others. The data for the model discussed in part 1 is relatively good. For the model in part 2, there is considerable model uncertainty, so rather than standardized data, we generally have to rely on individual pieces of research that attack specific instances. Often, the absence of data will illustrate the underlying point, namely, that obtaining clear answers to some questions is hard. It’s best to view this conceptual framework more as a tool to encourage clear thinking than as something in which we can plug in numerical values and answer questions.

If you’re interested in learning about existing data sets on migration, take a look at the migration information web resources page on this website.