Migration and Mobility

"People leave the poorest places" does not fit the data

The tidiest story is a push and a pull: conditions here force you out, wages there draw you in, and the bigger the gap the more people move. Someone wrote a version of this more than a century ago, and it is still the default explanation in the news.

The trouble is that the poorest countries are not the ones sending the most migrants. In 2014 Michael Clemens reviewed forty-five years of research and set it out plainly: as a country's income per head rises from a low base, its emigration rate typically rises first, and only starts to fall once the country reaches the upper-middle-income range. That rise-then-fall curve is called the migration hump.

The reason is not mysterious: leaving costs money — documents, fares, agents, and the months of living expenses before the first wage arrives. People who are very poor are not uninterested in leaving; they cannot afford it. A little more income means a few more people who can. So across a long stretch, development is not the cure for emigration but its fuel. This rests on cross-country panel correlations, and individual countries are heavily shaped by policy and war, so do not take it as an iron law.

The demand side counts too. Michael Piore's Birds of Passage (1979) argued that wealthy economies grow a particular class of jobs — unstable, with no ladder, jobs locals feel demeaned by; but for someone planning to work a few years and go home, the meaning of that job is cashed out back there: a house built, a wedding held, younger siblings put through school. Labour does not simply flow to wherever there is a shortage.

What really explains the clustering is networks. The Mexican Migration Project began with four communities in 1982 and has run every year since 1987, going house to house collecting migration histories. From it Douglas Massey and colleagues drew out cumulative causation: each person who gets a foothold abroad lowers the cost for the next person from the same village — somewhere to sleep, someone to walk you to work, knowledge of which route works. Migration produces its own causes.

The boundary was found by the same researchers: using the same data on Mexican urban areas in 2004, Fussell and Massey found cumulative causation largely absent there, because urban networks are already dispersed. So push-pull is not wrong so much as wrongly scaled: it computes national average incomes, while what actually moves is specific people along specific relationships — and those relationships form a network in towns and villages, and do not in large cities.

They did not "leave" — they live in two places at once

Peggy Levitt did something unusual for her time: she lived at both ends. She did fieldwork in a Dominican village called Miraflores and also in Jamaica Plain in Boston, where people from that village had settled. The book, The Transnational Villagers, appeared in 2001.

One figure she recorded says a great deal: by 1994, more than sixty-five per cent of the village's 445 households had relatives in the Boston metropolitan area. Village decisions waited on what people in Boston thought; meanwhile how a child was raised in Boston, or who was seeing whom, remained subject to village opinion. This is not a village plus a migrant enclave — it is one village spread across two places.

What flows back is not only money. In 1998 Levitt proposed the idea of social remittances: practices and standards travel too — how a meeting should be run, who decides things at home, what counts as respectable. Money leaves a paper trail and these do not, but what they alter is the yardstick people judge by, and that is often harder to reverse than cash.

The money is not small either: the World Bank estimates that remittances to low- and middle-income countries reached roughly $685 billion in 2024, more than those countries received in foreign direct investment — and that counts only formal channels. As the piece on the anthropology of debt and money argued, where money goes is never only an economic question: remittances are at once income, obligation, and a position in the hierarchy back home.

The name for all this came from Nina Glick Schiller and colleagues in 1992: transnationalism, the argument that migrants do not move from one container into another but maintain a social field across both. A decade later she and Andreas Wimmer named a defect in the social sciences themselves — methodological nationalism, the default assumption that nation equals society equals the natural unit. Statistics are collected by border and surveys fielded by border, so anything spanning borders becomes invisible: not ignored, just absent from the table.

The counter-argument comes from inside the same camp. Surveying Colombian, Dominican and Salvadoran immigrants, Alejandro Portes and colleagues found that only a small share made cross-border business a regular livelihood; most put their weight on where they had landed. So: transnationalism is overstated as a general way of life, and sound as a category of phenomena the old framework missed entirely. Both sentences have to be said.

The children who "fit in" fastest do not necessarily end up doing best

For the first half of the twentieth century the expectation about immigrants' children was a straight line: each generation moves closer to the mainstream, with language, occupation and intermarriage following in order, and time solving the problem by itself. The idea came from the Chicago School's race relations cycle, systematised by Milton Gordon in 1964.

In 1993 Alejandro Portes and Min Zhou proposed something different, segmented assimilation: not one line but several forks. Which one you take depends on what kind of labour market you land in, whether there is still a functioning community around you, and which category other people sort you into.

They then ran a long-term study: from 1992 they surveyed more than five thousand eighth- and ninth-grade children of immigrants in schools in Miami and San Diego, following them into the mid-2000s. One result runs against intuition: children who kept their parents' language and stayed connected to an ethnic community did not on average do worse on academic and psychological measures, and on some did better. They call this selective acculturation.

Why would keeping the mother tongue help? Not because language has any magic in it, but because of parental authority. Children learn fast and parents slowly, and the household undergoes a role reversal: the child reads the contracts, makes the phone calls, talks to the teacher. The person who can speak becomes the person who decides, and the limits and backstop parents could provide stop working — not because they no longer want to provide them, but because they can no longer deliver them. What selective acculturation preserves is a channel through which things can still be discussed.

The counter-argument is equally strong. Richard Alba and Victor Nee's Remaking the American Mainstream (2003) points out that the mainstream itself expands — a great deal of what now counts as mainstream was carried in by an earlier round of immigrants and absorbed; emphasising the downward path too heavily understates how much of the long-run data still shows most groups moving up. The argument continues.

On evidence strength: that study covers two metropolitan areas, a particular period and a particular ethnic composition, so do not transplant it elsewhere as a law. What survives is the mechanism — acculturation is not a single slider from zero to one; both repertoires can be held at once, and both can fail to take hold at once.

There is a further layer, made in the piece on race and ethnicity: boundaries run in two directions. How you identify yourself is one thing; which category other people sort you into is another, and that one is not yours to change. The feeling the second generation most often reports, of falling slightly short on both sides, is frequently not an identity that failed to form but two systems of classification running at once, and failing to line up.