Where slavery money came home

In 1833 Parliament passed the Act that ended slavery in most of Britain's colonies from 1 August 1834. It did not end it at once: most of the people it freed were made to keep working, unpaid, for their former owners as "apprentices" until 1838. And it paid compensation of £20 million, about 40% of government spending in 1833, not to the people who had been enslaved, but to the people who had owned them. The formerly enslaved received nothing. This page maps where in Britain those owners, and others who claimed the money, lived.

01 · The moneyWho was paid, and who was not

The Slave Compensation Commission's records list who owned enslaved people in the British Caribbean, Mauritius and the Cape at abolition and who claimed the compensation. Historians at University College London have spent years tracing those people and their families, and the Legacies of British Slavery database is the result. It puts particular weight on the "absentee" owners who lived in Britain rather than in the colonies, and it also records attorneys, mortgagees and legatees with a stake in the plantations. The enslaved people themselves appear in these records mostly as numbers and valuations. The database says so: "we do not have systematic information on the enslaved themselves."

The address file behind the LBS map holds 3,445 located addresses. 3,187 are in Great Britain, 80 in Ireland, 13 in the Channel Islands and 157 overseas. In Great Britain they are linked 5,331 times to 4,004 different people: 798 addresses are linked to two or more. Where a year is given, the median earliest year is 1834.

02 · WhereLondon first, then the Atlantic ports and the spas

1,158 addresses, 36% of those in Great Britain, are in London, 89% of them street addresses. Middlesex alone holds 33% of Britain's addresses and had 8% of its people in 1831, about 4 times its share. After London come Edinburgh (121), Bristol (106), Liverpool (100) and Bath (66), then Glasgow (32), Brighton (26) and Cheltenham (18). The five largest clusters hold 49% of all the addresses. Bristol, Liverpool and Glasgow were Atlantic trading ports; Bath, Brighton and Cheltenham were resort and spa towns.

Geocoded addresses by town, Great Britain.

03 · Per headMiddlesex, Edinburgh and the south

Set against the 1831 census, Great Britain had 2.0 located addresses for every 10,000 people. Among counties with at least 10,000 people, Middlesex had 7.8, Midlothian (Edinburgh) 5.7 and Moray 5.5 (from 19 addresses), then Gloucestershire, Sussex, Somerset and Kent. Tiny Nairnshire (9,354 people, 8 addresses) is higher still, at 8.6. Scotland as a whole (2.1) matches England (2.0); Wales is far lower (0.5). Half of all the addresses are in just five counties, which had 24% of the people. Across counties, a Gini coefficient of 0.51 (95% interval 0.50 to 0.53) measures how unevenly the addresses are spread compared with the population.

Shares of addresses and of 1831 population; Lorenz curve across counties.

Addresses per 10,000 people in 1831, by historic county.

04 · DistanceA pull towards London

I planned one test before looking: do counties near the great West India ports (London, Bristol, Liverpool and Glasgow) have more addresses per head? They do. In a count model with the 1831 population as the base, each doubling of the distance from a county's centre to the nearest of the four ports goes with a rate 37% lower (95% interval 25 to 46%, 84 counties). Being a port county adds nothing once distance is counted. The pattern holds when Middlesex is left out, with all LBS addresses including the unlocated ones, and with people instead of addresses. It weakens with a negative binomial model and loses its certainty when the whole South-East region (which in the LBS regions includes Middlesex) is left out. Distance to London alone fits better than distance to the nearest port, so most of this is a pull towards London.

Two checks say what the pattern is not. Street addresses follow it strongly; named country houses do not follow it at all. And Grade I and II* listed houses, a rough stand-in for old landed wealth, show no such distance pattern (0.98 per doubling, interval 0.82 to 1.16). The addresses are not simply where the rich lived.

Each dot is a historic county, sized by its 1831 population.

05 · Town and countryTown houses in London, country seats elsewhere

In London 89% of addresses are street addresses: a number in a square, a street or a terrace. Outside London only 32% are; 28% are named houses, such as a Hall, a Park or a Castle, and 39% give only a town or village. Scotland (26% named houses) looks like provincial England (29%).

What the address record names, by area.

06 · Still standingThe country houses on the lists

I took the 553 named houses outside London and outside the towns and looked for each on today's national lists of listed buildings, by name and within 2 to 5 km. 267 (48%) matched. A hand check of 100 cases found 86% of the matches right and a few houses missed, which puts the true share near 44% (interval 38 to 49%). 62 of the matched houses are in the top grade (Grade I or Category A). Many of the others have been demolished or were never listed. Across the counties of England and Wales, those with more addresses per head also have more of these houses per 100 listed great houses (Spearman's rho 0.53, 51 counties).

Country houses matched to listed buildings, by nation and grade.

07 · How I did thisMethod and limits

Addresses. The LBS map's address file, with each address's geocode (from OpenStreetMap), town, county and linked people. I used the counts of people and never their names. I mapped the county names to the historic counties of 1831 by hand and checked England and Wales against the nearest parish of 1851: 97% agree, the rest are near county borders. Population. County totals from the 1831 census. Address types come from the address text: a house number or a street word, a named house, or only a town or village. Listed buildings come from Historic England, Historic Environment Scotland and Cadw. The analysis plan was written before the results; anything added later is marked exploratory in the paper.

What this cannot say. The map shows where people lived, not how much money they received: the address file has no amounts. About 34% of the addresses are placed only at a town, village or parish. The located share varies by county, from 17% in Shropshire to 100% in Durham, though it is not related to the county rates (rho 0.08). Many people had more than one address. And none of this measures the harm done, which fell on people the records barely name.