Great fires

On 2 September 1666 a fire broke out in a baker's shop on Pudding Lane. By the time it was out, the City of London had lost some 13,200 houses, 86 or 87 parish churches and St Paul's Cathedral. The Rebuilding of London Act banned timber from the outside of new buildings, and London rose again in brick and stone.

London was not alone. Fire was the great urban disaster of early modern England. Timber houses with thatched roofs stood shoulder to shoulder along narrow streets, with bakers' ovens, maltings, smithies and tallow cauldrons among them. A dry summer and a strong wind could take a whole town in an afternoon. Northampton lost about 600 homes in 1675. At Wem in Shropshire in 1677 a girl named Jane Churm dropped a candle and 140 dwelling houses burned. At Blandford Forum in 1731 almost every building in the town was destroyed; an Act of Parliament followed, and the town was rebuilt in brick.

Visitors to Warwick, Blandford, Northampton or Wareham are told that their handsome Georgian streets are the fire's legacy. So I asked a blunter question: in the long run, did the towns that burned do better? Are they bigger, or richer, today than similar towns that never had a great fire?

01 · The recordA map of fires remembered

There is no complete gazetteer of English town fires, so I built a record from what Wikipedia remembers. I took the 1,681 articles linked from Wikipedia's lists of towns in England and Wales and of UK cities, pulled out every sentence that mentions fire, burning or a blaze (2,213 sentences in 748 articles), and read the 457 that carry a year before 1900. I added Wikipedia's List of urban fires and its articles on particular great fires.

A fire counts as a great fire if the text gives a year from 1000 to 1899 and describes damage to the town as a whole, not one building: much or most of the town, a third or more, the "Great Fire of" the town, or 20 or more houses. Burnings in war, raids and riots are kept apart and not counted. That leaves 102 great fires in 81 towns, 97 of them in England and Wales. Another 25 fires are "possible": dated town fires whose scale the text does not make clear.

The record peaks in the 17th century, with 33 great fires, and the 1700s are close behind with 31. Together the two centuries hold 63% of all the fires in the record. The busiest quarter century is 1675 to 1699, with 14. Then the fires fade: only 11 in the 1800s.

Part of that shape is real. Towns were growing in the 1600s while still built of timber and thatch, and the fading after 1800 matches what we know: brick, tile and slate replaced thatch, building acts forbade timber fronts after the big fires, and towns bought fire engines and insurance. Part of it is memory. Medieval fires are thinly recorded, and a Wikipedia editor is more likely to mention a fire that left a famous rebuilt High Street than one that left nothing to see. So read the shape as the fires that towns remember, not a count of every fire that happened.

The worst recorded losses

Away from London, the typical fire with a number lost 145 houses. Several towns burned more than once: 14 towns have two or more great fires in the record and 5 have three or more. Tiverton burned in 1598 1612 1731; Marlborough in 1653 1679 1690.

02 · WhereA southern and eastern disaster

The fires cluster in the old market towns of the South, the Midlands and East Anglia. In the Eastern region 23 of 141 towns (16%) have a great fire in the record. In the three northern regions it is 5 of 275 (1.8%), against 9.7% in the South and the Midlands (Fisher's exact test p below 0.001). Wales has only 2 of 154.

That fits the history as I understand it. Before the industrial age most of England's towns of any size were in the South, the Midlands and East Anglia, and many northern towns on today's map grew up with the mills, after the age of the great fires had passed. The panel hints at this too: 88% of the towns of the South and the Midlands lie within 2.5 km of a place in Domesday Book, against 37% of the northern towns (Domesday did not cover the far north, which explains part of that gap). Building stone may matter too: towns on the limestone belt and in the stone country of the North and West burned less easily than timber and thatch. I have not tested that here.

03 · The testDid the burned towns grow bigger?

Here is the main question. To compare like with like I need towns that are similar except for the fire. I used 1,079 towns in England and Wales outside Greater London with a Census 2021 built-up area: 71 with a great fire and 1,008 without.

The raw comparison flatters the burned towns. Their median 2021 population is 13,816, against 12,248. But they were already different before any fire: 92% of them have Domesday settlements of 1086 within 2.5 km, against 63% of the others, many of which are Victorian or modern towns. The burned towns were old, established places, the kind of towns that had enough houses to burn.

So I compared each burned town with towns of the same region and a similar Domesday footprint. In a regression that allows for region and Domesday households, burned towns are about 24.8% bigger in 2021, but the 95% interval runs from -6.4% to +66.4%, so no difference at all is well within the range (p = 0.13).

Matching gives the same answer. I paired each burned town with the unburned town in the same region whose Domesday households were closest. 40 of the 71 burned towns are bigger than their match, about what a coin would give (sign test p = 0.34). The median burned town is 1.25 times the size of its match, but the spread is wide.

And there is a catch that pushes the other way. The burned towns have longer Wikipedia articles: a median of 46,908 characters against 31,215, 1.50 times as long. Bigger and better known towns get longer articles, and longer articles are more likely to mention a fire. Once I also allow for article length the burned towns' edge disappears (-6.1%, 95% interval -24.0% to +16.0%). Among the 177 towns whose articles call them market towns, the most like-for-like group, the difference is -2.9% (95% interval -37.6% to +51.1%).

The fires made some of England's loveliest streets. They did not, on this evidence, make the towns that burned any bigger in the long run.

04 · TodayRicher?

Household income tells the same story. The ONS estimates for the financial year ending 2023 give the burned towns a median mean household income of £51,287, against £48,495 for the others: partly a southern premium, because the fires were southern. Allowing for region and size, the difference is -2.0% (95% interval -5.2% to +1.3%). In the matched pairs the burned town has the higher income in 28 of 70 cases, a median gap of -4.4%. If anything, the burned towns are a touch poorer than their neighbours of similar size, but the gap is small and not robust across methods.

05 · The old giantsThe biggest towns of 1377 and 1801

The Wikipedia article on England's historical town populations ranks the largest towns in the poll tax of 1377 and in the census of 1801. 5 of the 41 towns of 1377 had a great fire between 1377 and 1800 in my record: Bury St Edmunds, London, Northampton, Norwich and Shrewsbury. From 1377 to 2021 they grew a median 33.4 times, against 27.0 times for the rest (Mann Whitney p = 0.40). Of the 27 largest towns of 1801 still traceable today, 4 had burned before 1801; they grew a median 7.8 times to 2021, the others 8.8 times. Groups this small cannot settle anything, and both tables only list the largest towns, so the small market towns where most fires happened are not in them.

06 · The lookRebuilt in brick

Where the fires did leave a mark is on the streets. In 14 burned towns, the fire sentences themselves tie the town's buildings to the rebuilding, and 8 call the result Georgian. I also counted Grade I and Grade II* listed buildings with a date in Wikidata within 1.0 km of each town point. Of the 134 towns with at least three dated buildings, the 8 that burned between 1600 and 1799 have 29% of their dated top buildings from 1700 to 1837, against 20% in towns with no recorded fire (Mann Whitney p = 0.23). The direction fits the Georgian rebuilding, but only 17% of the top listed buildings near these towns have a date in Wikidata, and 8 towns are too few to be sure.

07 · What this cannot showMemory, matching and causes

A record of memory, not a gazetteer. This is a record of fires remembered in Wikipedia articles, not a complete list of the fires that happened. Many towns had great fires that their articles do not mention, and an article's length and care vary. Burned towns have longer articles, so the record leans towards better documented towns. I counted a fire only where the text gives a year and town-wide damage; the 25 possible fires and 30 war burnings are in the data file with the sentence that records them. Counting the possible fires as great fires changes the size estimate to +22.0% (95% interval -7.7% to +61.1%).

Reading, not keyword matching. I read every dated sentence and classified it by the rule above, with a review note on every fire. 4 great fires have no year in the quoted sentence: I took Chudleigh (1807) and Ilfracombe (1896) from their full articles, Leatherhead (1392) from "one year after", and Edinburgh (1824) from the article on its Great Fire.

Size before the fire. Only Domesday Book gives a size for almost every old town, and 1086 is a long way before 1675. Small towns are missing from the 1377 and 1801 tables. A town's 2021 built-up area is a different shape from the town that burned.

No causes. These are associations. A fire could help a town (new streets, new money) or hurt it (debt, people who never came back); the data can only say that, on average, burned towns ended up neither clearly bigger nor clearly richer than towns like them.