Day 37 · Migration and Diaspora

The Shape of a Human Tide Is Cut by a Document

Monday, 3 August 2026 · BigCat's Time Machine
Diaspora tends to get written as tide — a force of nature, nobody's responsibility. But in the four episodes below, whether people went or stayed, where they landed, and who was kept out can each be traced to a specific document, a person who signed it, and a trade-off that could have gone the other way.
EVENT · 01

The Expulsion Order Was Really Aimed at Those Who Had Already ConvertedThe Alhambra Decree · Castile & Aragon, 1492

1492Granada · Salonica · AmsterdamKamen · Yerushalmi

Granada fell on 2 January 1492 and the Reconquista ended. On 31 March, Ferdinand and Isabella signed the Alhambra Decree: Jews had until 31 July to convert or leave. The negotiator was Isaac Abravanel, royal finance official and leader of the Jewish community; across the table stood the Inquisition, established in 1478.

Abravanel offered a large sum to buy the decree back, and the crown — fresh out of a ten-year war and financially strained — had reason to listen. But the real mechanism sits in the text of the decree itself: what it keeps returning to is not Jews but conversos, the New Christians forcibly baptized in waves since 1391. The Inquisition had concluded they still kept contact with Jewish kin and retained Jewish practice, and that the "source of temptation" therefore had to be removed. This was a quarantine policy aimed at people who had already converted, with the cost paid by those who had not. The numbers remain contested: the traditional figure is 300,000, while Henry Kamen, working from tax rolls, estimates 40,000–80,000 actually left, most choosing baptism instead — the decree manufactured more conversos than it expelled Jews. Those who left went to Salonica, Livorno and later Amsterdam; Jonathan Israel shows that the cross-border credit network of seventeenth-century Sephardic merchants was precisely this dispersal rewired.

Had the crown taken the money and withdrawn the decree, Spain would have kept its credit and tax-collection networks — but the converso problem would have been untouched. What was truly irreversible was not the decree; it was the rule that lineage must be proved, which once established cannot be switched off. The debate runs along the same seam: Benzion Netanyahu holds that most conversos had converted sincerely and that the persecution was an early form of "purity of blood" (limpieza de sangre) racism; Yerushalmi and Kamen disagree, arguing the Inquisition did find extensive Judaizing practice. The core question is the same either way: did the persecution create the identity it claimed to be eliminating?

The deadlock inside loyalty vetting: once a group is defined as needing to prove its own innocence, any evidence can be read as camouflage, and the process can never close.

A policy that asks people to prove what is inside them will never reach a verdict.
Who in your organization sits in the position of having to keep proving themselves? Who is counting that cost?
EVENT · 02

After the Massacre, the VOC Did Not Expel the Chinese — It Gave Them a JobBatavia 1740 and the Kapitan System

1740Batavia (present-day Jakarta)Blussé · Wang Gungwu

From the late seventeenth century the Dutch East India Company (VOC) expanded sugar production around Batavia, recruiting large numbers of migrants from Fujian. When Caribbean output rose in the 1720s and sugar prices collapsed, Javanese mills failed in batches and unemployed Chinese workers accumulated outside the city. Governor-General Adriaan Valckenier (in office 1737–1741) was already in a factional fight with his deputy, van Imhoff.

On 25 July 1740 the VOC council resolved that Chinese without residence permits would be deported to plantations in Ceylon. Once the round-ups began, a story spread outside the walls: those put on ships never reached Ceylon but were thrown overboard. Leonard Blussé's Strange Company (1986) notes the rumour cannot be verified — but it determined what followed. On the night of 8 October armed Chinese gathered and attacked the city; the next day the VOC ordered house-to-house searches for weapons inside the walls, and within hours the search became a massacre that ran for three days, leaving almost none of the roughly ten thousand Chinese inside the city alive. Valckenier was dismissed and jailed the following year and died in prison in 1751; no verdict was ever delivered. Yet the VOC did not expel the Chinese afterwards — it could not do without them. Chinese residents were confined to Glodok outside the walls, and the Kapitan Cina system was reinforced at the same time: the VOC appointed Chinese headmen to collect taxes and farm monopolies, institutionalizing a migrant community as the tax-collecting middle layer — protected from above, extracting from the society below. The structure lasted two centuries.

Had the VOC chosen to resettle the unemployed workers locally — costlier, but proposed at the time — those three days probably would not have happened. The real fork, though, is not the three days but how tax collection was farmed out: as long as the middle-layer structure exists, the conditions for periodic violence stay loaded — the riots of 1965 and 1998 run along the same fault line. Blussé stresses the contingency of panic and distorted information; scholars in the line of Wang Gungwu stress structure. The difference is whether 1740 reads as an accident or a symptom.

The middle layer of the platform economy — resellers, outsourced moderators, channel partners: they absorb the friction and the anger, and when something breaks they are the first to be dropped by both sides.

Institutionalizing a group as the middleman is signing it to a long-term scapegoat contract.
Who absorbs the friction between you and your users? When something goes wrong, who protects them?
EVENT · 03

A Quarter of an Acre: Converting Hunger into Permanent LandlessnessThe Gregory Clause · Ireland, 1847

1845–1852Ireland · Grosse ÎleÓ Gráda · Kinealy · Trevelyan

Potato blight reached Ireland in September 1845, where roughly three million people lived almost entirely on potatoes. Prime Minister Robert Peel quietly bought Indian corn to hold prices down and left office in June 1846; the incoming Russell government handed relief to Charles Trevelyan, Assistant Secretary to the Treasury.

The Poor Law Extension Act passed in June 1847 carrying a clause added by the landlord MP William Gregory: anyone holding more than a quarter of an acre was, along with their family, ineligible for any poor relief. The logic reads well — relief should go to the poorest. The effect was to force more than three hundred thousand smallholders into one choice: give up the land in order to eat. Landlords acquired a legal instrument for clearing their estates, and roughly 250,000 people were evicted between 1847 and 1854. Trevelyan meanwhile shifted the cost of relief onto Irish local rates, and the western counties that needed relief most had the thinnest rate base. Between 1845 and 1852 about a million people died and 1.5 million left; mortality on the 1847 crossings was high enough that the quarantine station at Grosse Île in Canada buried five thousand in a single year.

Without the Gregory Clause, a tenant could have drawn relief and held onto the land to wait for the next season. That is not fantasy: the blight was milder in 1847 and the harvest partially recovered, yet that is precisely the year of the largest clearances and the largest emigration. It was policy, not the pathogen, that translated "famine" into "permanent departure." The debate is about culpability: Cormac Ó Gráda's Black '47 and Beyond (1999) argues it was not deliberate extermination but that laissez-faire plus providentialism produced a culpable dereliction; Christine Kinealy puts more weight on the deliberate element of policy.

The cliff effect in asset-tested welfare: eligibility requires applicants to first sell the car, the tools and the thin savings — repairing today by dismantling next year.

Eligibility rules often decide where people end up more than the disaster does.
Which threshold you have designed requires the applicant to destroy their own capacity to recover?
EVENT · 04

In 1951, the Word "Refugee" Was Written NarrowDrafting the Refugee Convention · Geneva, July 1951

1951GenevaHathaway · Betts · Collier

At the Évian Conference of July 1938, thirty-two states discussed taking in Jewish refugees from Germany and Austria; almost none committed apart from the Dominican Republic, and Nazi propaganda promptly announced that nobody wanted them. After the war every government understood that an institution was needed — and equally did not want open borders. The conference of plenipotentiaries convened in Geneva in July 1951.

The fight was not over whether to protect but over how wide to write the definition. The American delegation pushed narrow and insisted on two gates: the convention would cover only events occurring before 1 January 1951, and a signatory could declare it applicable only to events occurring in Europe. Article 1 A(2) as adopted defines a refugee as someone with a well-founded fear of persecution for reasons of race, religion, nationality, membership of a particular social group, or political opinionpeople fleeing war, famine or environmental disaster are outside it, and so is anyone who has not crossed a border. Only the 1967 Protocol removed the two gates; the United States never signed the convention itself. Today more than 120 million people are displaced worldwide, the largest share of them internally — for whom the convention has not a word.

Had the definition read "forced to leave home for any reason," tens of millions of internally displaced and climate-displaced people would have legal standing today; but written that way in 1951, the main receiving states would likely have refused to sign and the convention would have been a dead letter — a genuine dilemma of design: enforceable and narrow, or wide enough to cover, but not both. James Hathaway points out that the five grounds map almost exactly onto the people the West was glad to receive during the Cold War (East European dissidents) — an instrumental design. Alexander Betts and Paul Collier, in Refuge (2017), judge the system to have failed while arguing that widening the definition would destroy what political viability it has left.

The definitional fight in AI governance: thresholds for "frontier model" or "high-risk system" written narrow are enforceable but miss most of the field; written wide, nobody signs. The same structure, replayed seventy years on.

The real boundary of a regime is written in its definitions, not in its declarations.
In rules you helped write, who does the definition exclude — and where did those people end up?

Four Dispersals: The Usual Explanation, and the Clause That Actually Worked

Push and pull are both present, but where people went was decided by the few lines written down at the time.
Site / Date
Usual explanation
What actually decided the outcome
Spain · 1492
religious zeal expels the Jews
the real target was conversos; expulsion cut off the "temptation"
Batavia · 1740
ethnic hatred suddenly erupts
tax-farming locked migrants into a middle layer; risk carried forward
Ireland · 1847
famine drives people out of Ireland
the quarter-acre clause converted hunger into permanent landlessness
Geneva · 1951
humanitarianism establishes refugee protection
a Cold War definition excluded most of today's displaced in advance

Going Deeper

1. Why does the push–pull model fail to explain these four episodes?
Push–pull treats migration as flow under a pressure differential. It predicts direction but not timing. The blight was milder in 1847 than in 1846, yet emigration was higher; the unemployment at Batavia had existed for a decade, yet the massacre came three months after the deportation order. The switch is the moment a document strikes an option off the menu.
2. Can the structural risk of a middleman minority be dismantled?
The danger is not the group but the functional design in which taxes are collected by outsiders — it lets a ruler's costs be carried on someone else's face. Two remedies have worked: bring the function into the regular bureaucracy, or give members of the middle layer political representation of their own. The test is whether the role has protection from above but no channel of accountability below. If so, both sides will eventually drop it.
3. Beyond narrow-or-wide, is there a third path for definitions?
Yes, and each trades strength of obligation for breadth of coverage. One is tiering: keep the core definition narrow with strong duties, and build lighter-duty complementary protection around it (the EU's subsidiary protection took this route). The other is triggered extension: write the definition narrow but pre-commit to temporary mechanisms that activate automatically in defined circumstances. The general move: never optimize enforceability and coverage on the same dimension.
4. Written as one model, what are the variables across all four?
Who gets drawn outside the boundary × who bears the cost of drawing it × whether an exit option exists × whether the receiving end opens. The easiest to overlook is the second — in 1492 the unconverted paid, in 1847 the tenants paid, in 1951 the future displaced paid, and in no case were the decision-makers on the list. The fastest indicator of how far a policy will go is whether the people deciding sit downstream of the cost.