From 1663 to the 1830s, more than a thousand turnpike trusts took over England's main roads, put up gates, charged tolls and mended the road with the money. Their milestones still stand on the verges. Did being on a turnpike make a town grow?
In 1663 Parliament let the justices of Hertfordshire, Cambridgeshire and Huntingdonshire put up toll gates on the Great North Road, at Wadesmill, Caxton and Stilton, and spend the money on the road. It was the first turnpike. Over the next century and a half, local worthies petitioned Parliament for road after road, each run by its own trust: a body of trustees who could borrow, set up gates, collect tolls and pay surveyors. At their peak in the 1830s, Wikipedia's article says, over 1,000 trusts looked after some 30,000 miles of road in England and Wales.
The turnpikes carried the coaching age. The first mail coach left Bristol for London in 1784; Telford's rebuilt Holyhead Road cut the London mail's journey from 45 hours to 27. Inns, ostlers and horse dealers lined the routes. Then the railways came, and the coaching towns emptied. So I asked a simple question with the data I could get: were the towns on the turnpikes the ones that grew?
The honest short answer: turnpikes reached almost everywhere quickly, and once you allow for the fact that bigger places got them first, I can find little sign that being named in a turnpike trust made a place grow faster in the decades before the railway. The milestones tell a story of their own, though, and it is not quite the one you might expect.
Each coloured dot is a town or village named in the title of a turnpike trust, coloured by the year of the first such trust: dark green for early, ochre for late. Grey dots are the milestones and mileposts that survive and are recorded today; red ones are toll houses. Drag the slider or press play to watch the trusts spread. Search for any town in England or Wales to see when it was first named in a trust and how close the nearest surviving milestone is.
Wikipedia's eight regional lists name 1,245 English turnpike trusts, 1,119 of them with the year of their first act. The lists are marked incomplete, and the Sussex trusts have no dates, but the shape is clear. After a slow start (just six new trusts in the ten years from 1700), trust-making took off in the 1750s: 254 trusts in that decade and 189 in the 1760s. Between 1750 and 1772 alone, 495 trusts were set up, 44% of all those with a date. By 1800, 74% of the trusts existed. Only 34 were founded after 1830, when the railways began.
The trusts did not march out from London in a neat wave. The great roads out of London were turnpiked early, and towns named in trusts within 50 km of London had their first trust in a median year of 1748. But beyond 200 km the median was 1759, only about a decade later. Distance from London explains very little of the timing (Spearman rank correlation 0.14 across 457 towns). The North West was the latecomer, with a median founding year of 1789: 44% of its dated trusts were founded between 1790 and 1829, many of them for the roads between Lancashire's growing mill towns.
Some towns were the hub of a whole web of trusts. Leeds appears in the titles of 19 trusts, Manchester in 15 and Birmingham in 11. In all, 1,031 places are named in at least one trust title, 486 of them towns or cities on today's map. In England, 478 of the 1,060 places that Ordnance Survey calls a town or city (45%) are named in a trust. The rest were mostly reached by roads named after their neighbours.
The point of a turnpike was a better road, and the coaching trade made the most of it. These dates come from the Wikipedia articles on turnpike trusts, mail coaches and toll houses.
The turnpikes were a private fix for a public problem: local people borrowed against future tolls to mend their own road.
The trusts had to tell travellers how far they had come, and many of the stones and cast-iron posts they set up are still there. Wikidata and OpenStreetMap together record 8,214 milestones, 2,346 mileposts, 146 guide stones and 221 toll houses in the UK: 10,030 in England, 665 in Wales, 197 in Scotland and 35 in Northern Ireland. 5,406 are listed buildings or scheduled monuments.
Half the survivors stand within 45 m of an A-road. Most are dated only to a century, and most of those to the 19th: the coaching age was when trusts replaced old stones and added cast-iron posts. Of the 614 with an exact year, the median is 1820.
Where they survive is uneven. West Yorkshire has 20.7 milestones or mileposts per 100 km², Staffordshire 17.0 and Cheshire 16.1. Bedfordshire, which had some of the very first trusts, has just 1.1, and Northamptonshire 2.0. England as a whole has 7.5 per 100 km², Wales 3.0 and Scotland 0.1. Some of this is real loss to road widening and decay; much of it is recording. About half of the Wikidata milestones carry a listed-building designation, and how thoroughly the rest have been entered varies from county to county. Scotland's milestones are barely recorded here at all.
Still, the survivors do follow the old network. Across 401 squares of 20 km by 20 km in England, squares with more places named in trusts also have more surviving milestones (Spearman correlation 0.49). Milestones are a fair if patchy trace of where the turnpikes ran, but a count of survivors, not of the roads.
To test growth I need places with population counts from census to census. The best set I have is 345 places in England and Wales with census figures for 1801, 1831 and 1851 (326 from Wikidata, 19 inland towns from A Vision of Britain through Time). It is not a random sample: 259 of them are in Cambridgeshire, Huntingdonshire and the fens, and 54 in Dorset, Somerset and Devon, because that is where Wikidata's editors have entered parish series. Most are villages: the median place had 294 people in 1801.
I measured exposure to the turnpikes in two ways: whether the place is named in a trust founded by 1800 (28 places), and whether a surviving milestone, milepost or toll house stands within 1 km (85 places). Growth is the average yearly change in population, in percentage points.
At first sight the turnpike places did grow faster in the coaching heyday: from 1801 to 1831 the places named in a trust grew by a median 1.50% a year, against 1.07% for the rest (Mann-Whitney p = 0.003). But the named places were the bigger market towns, and bigger places grew faster everywhere. Once I allow for size in 1801 and for the region, the gap shrinks to 0.10 points a year, with a 95% interval from -0.13 to 0.34. A milestone within 1 km makes no difference (-0.06 points, -0.23 to 0.11).
If anything, the turnpike places lost ground when the railways arrived. From 1831 to 1851 the named places grew 0.31 points a year more slowly than similar places (interval -0.63 to 0.01, p = 0.057), and their growth fell back by a median 0.51 points between the two periods, against 0.20 for the rest (p = 0.018). That fits the coaching story, but it is a weak result from a small, lopsided sample, and I would not lean on it.
Stilton, on the first turnpike of all, shows the pattern people imagine. It had 509 people in 1801 and 793 in 1831, then stalled at 803 in 1851 and fell to 645 by 1881 and 535 by 1901, back where it started. But most rural places in this sample peaked around 1851 and shrank after, coaching or not, as people left the land for the towns: 69% of the places not named in a trust were at their largest in 1851 or earlier. Stilton's fall is steeper than most, which is suggestive, not proof.
The big towns tell the same story. 28 of the 29 largest towns of 1801 (London aside) are named in a trust title, so there is almost no contrast to test. Among the 22 with figures for both 1801 and 1861, neither an early first trust (Spearman 0.05) nor the number of trusts naming the town (0.10) is linked with faster growth. Manchester, Liverpool and Leeds grew for reasons of their own, and their turnpikes followed the traffic.
No map of the roads. The best evidence would be a dated map of every turnpike road, and one exists: the turnpike road GIS for England and Wales, 1700 to 1838, by Rosevear, Bogart, Shaw-Taylor and Satchell (openICPSR 195126). I could not obtain it for this study. My code is written to use it if it is supplied. Without it, a place counts as "named" only if its name is in a trust's title, which misses every town that a turnpike ran through without naming.
The lists are incomplete. Wikipedia flags its turnpike lists as incomplete; 59 Sussex trusts have no date, and Wales and Scotland have no dated lists at all. Seven of the eight lists were read with an automated web reader; for the South East list, held on disk as wikitext, the reader matched all 82 Berkshire and Kent rows I checked, name, year and act.
Milestones are survivors. They show where stones were set up, then kept, then recorded. Recording varies hugely between counties.
A small, lopsided growth sample. 345 places, mostly villages in two regions, and populations on census parish boundaries that changed. Associations, not causes: trusts were set up on roads where traffic already was.
Trusts. I took the tables of Wikipedia's eight lists of turnpike trusts by English region: trust name, the year of its first act and that act's short title. Rows without a trust name (later acts of the same trust) were dropped and a trust listed under two counties was kept once, leaving 1,245. I split each trust's name into place words ("Ashbourne and Yoxall Bridge" gives Ashbourne and Yoxall Bridge) and matched each to an OS Open Names town, village or hamlet of exactly that name, preferring one in the trust's region, then a town, then the one nearest the trust's other places. 1,084 trusts (87%) have at least one placed name.
Traces. Wikidata items in the UK that are milestones or toll houses, and OpenStreetMap features tagged as historic milestones, mileposts, toll houses or toll gates, deduplicated within 40 m. Wikidata's dates of exactly 1701, 1801 or 1901 are century dates. Distances to A-roads use OpenStreetMap trunk and primary roads with an A number.
Growth. Census figures for 1801, 1831, 1851 and 1881 from Wikidata (places with fewer than 50 people in 1801 left out) and A Vision of Britain through Time. Yearly growth is 100 × ln(P2/P1) divided by the years between. The regressions use ordinary least squares with robust standard errors, log population in 1801 and three areas (fens, south west, other).