Beyond the Tracks: re-connecting people, places and stations in the history of late-Victorian railways

Joshua Rhodes, Jon Lawrence, Kaspar Beelen, Katherine McDonough, Daniel C.S. Wilson

3.1 Introduction

The drawn-out process of industrialisation first witnessed in Victorian Britain was ineluctably accompanied by urbanisation: either thanks to the gravitational pull of existing cities, or in the creation of entirely new industrial ones. In the case of the former, cities such as Manchester, Newcastle, Birmingham and Leeds offered sites in which heavy machinery could efficiently operate in proximity to its fuel sources, as well as a disciplined, skilled labour force.^[1] In the case of the latter, dedicated new centres such as Middlesbrough, Barrow or Crewe announced themselves in different ways as exemplars of a new urban landscape. Transport, and the railway network in particular, was at the heart of these transformations. Railways transformed the economics of agriculture by opening up new urban markets, and the spatial basis of industry itself by making fuel, raw material and finished goods mobile in new ways. In doing so, they left their imprint on the shape of most towns and cities and influenced the whole country. Railways scythed their way through built-up areas creating new geometric forms, separating neighbourhoods, clearing slums as well as creating new ones. They brought noise, vibration, and smoke (once the preserve of the factory) into residential districts. But railways were also an immensely popular mode of passenger transport: by the end of the century over a billion journeys a year were taking people to the places they needed or wanted to visit.^[2] These incommensurable effects make the impact of rail difficult to evaluate. The challenge of producing any overall calculus does not prevent more local assessments of benefits and costs – which we might call ‘amenities’ and ‘disamenities’ – but it has not hitherto been possible to aggregate these impacts in search of broader patterns and trends. This chapter brings together different types of source material for the first time and suggests new ways of conceptualising the interaction of demographics with the spatial effects of the railway.

The physical extent of the railway, stretching as it does both within and between towns and cities, therefore makes the question of scale a key issue in any analysis. The regional, inter-city and national character of the railway – typically understood as a network of connected points – has drawn the attention of economic and transport historians and historical geographers, whose interests, for example in measuring traffic flows or economic development, have encouraged a particular set of methods for working in aggregate.^[3] These have predominated over the local or case-based approaches favoured by urban and social historians.^[4] Histories of railways with people at their core have proven elusive, not only thanks to this methodological divide, but also because important primary sources such as the census have, for technical reasons, been difficult to bring into conversation with other source materials relating either to transport, or to the spatial information found on maps. This chapter addresses some of these challenges by experimenting with interdisciplinary digital methods that allow new forms of convergence between source materials, combining detail at the level of the street, with analysis at a national scale.

The method we develop in this chapter should be seen as part of a long tradition of trying to understand cities. Despite their almost irreducible complexity, cities are synonymous with human culture and society at large, and in a time of dramatic material change such as nineteenth-century Britain, the city often came to stand for industrial modernity itself. For this reason, investigators and analysts of different stripes have been drawn to make sense of its new urban forms, sensing that it offered a key to unlocking more than just the particular case-study at hand.^[5] When Friedrich Engels wrote his classic study of mid-nineteenth-century Manchester, the urgency came from his belief that this ‘shock city’ of the Industrial Revolution was a portent for the future of cities, and human society, everywhere.^[6] Likewise, when attention turned to the slums of east London from the 1880s, it was because of a national outcry about the state of the nation, whose symptoms were found at their worst in its cities.^[7] Campaigning investigators such as Arnold Toynbee, Beatrice Webb and Charles Booth (among others) joined Engels in producing work that today might be called geography or sociology: empirical but also theoretical, and nearly always using maps to put forward spatial reasoning and argumentation in visual form.

This desire to grasp the city in its shifting totality recurred in the self-styled science of ‘civics’ advocated in the early twentieth century by Patrick Geddes, for whom the shape of urban development could not be separated from its history. In considering the morphology of the world’s then biggest example of the form, Geddes searched for an adequate metaphor: ‘This octopus of London, polypus rather, is something curious exceedingly, a vast irregular growth without previous parallel in the world of life – perhaps likest to the spreadings of a great coral reef”. Using a new visual tool for combining population and transport on one map (see Figure 3.1) of this ‘vast new unity’, Geddes noted the ‘new lines of union: the very word “lines” nowadays most readily suggesting the railways, which are the throbbing arteries, the roaring pulses of the intensely living whole’.^[8] Contemporaries like Edward Thomas had no doubt that the railway took the curse of ‘modernity’ out into the countryside, but whether it was the cause or effect of suburban growth, it was undoubtedly the key to understanding it.^[9] Geddes pleaded for contemporaries to take a unitary view of the whole, for which, as ever, he found existing frameworks lacking and coined what has become a keyword for the field:

To focus these developments [...] of the geographic tradition of town and country in which we were brought up [...] we need some little extension of our vocabulary [...] Some name, then, for these city-regions, these town aggregates, is wanted. Constellations we cannot call them [...] what of ‘Conurbations?’ That perhaps may serve as the necessary word, as an expression of this new form of population-grouping, which is already, as it were subconsciously, developing new forms of social grouping…^[10]

Figure 3.1. A Map displaying both population density and railway lines, used by Geddes, (original from Bartholomew’s Royal Atlas of England, 1899).

This suggestion of a causal relationship between spatial and social groupings became the focus of detailed attention, from the late 1950s for a loose collection of scholars including H. J. Dyos, John Kellett, Richard Dennis and Colin Pooley among others centred on the Urban History Group, who sought to explore the connections – as David Cannadine put it – ‘from shapes on the ground to shapes in society’.^[11] A series of influential studies by historians made use of the work, on the one hand, of geographers, and on the other, of sociologists, to explore how social categories such as class and occupation ‘had a clear spatial expression, and were reflected in distinctive social areas.’^[12] Despite voluminous studies exploring residential differentiation in paradigmatic cases from a variety of theoretical and empirical perspectives, the field appeared to get bogged down in methodological and definitional issues.^[13] To some extent this resulted from the divergent disciplinary commitments in play; but, equally, this is not so surprising, when we consider the highly heterogeneous nature of the data and evidence being used to make sense of supremely complex and over-determined processes of urban change.

John Kellett’s masterly work on the Victorian city – to which we are indebted in this chapter – did more than most to clarify the impact of the railway on the five key cities that comprise his case studies; however, Kellett is cautious not to generalise unwisely from his examples.^[14] The methods we outline in this chapter aspire, if not to generalise, then to propose a general framework for testing the effects of rail on residential differentiation nationwide, rather than just in large cities. Nonetheless, in so doing we remain mindful that the millions of individual actions that produced residential patterning were not always choices at all, and for most were choices constrained by economic and social factors; principally poverty, but also, for many, by dense networks of social obligation and reciprocity that underpinned their access to work, credit and, indeed, accommodation.^[15] When, towards the end of Ulysses, Leopold Bloom is depicted imagining his perfect house beyond Dublin’s city borders, James Joyce is reminding us that even for middle-class urbanites, unconstrained choice was the stuff of fantasy. It is, however, worth noting how Bloom, whose work required him to be a daily commuter, imagined his ideal location. His ‘thatched bungalow shaped 2 storey dwelling house’ would be set in extensive grounds and located ‘not less than 1 statute mile from the periphery of the metropolis’, and, perhaps most tellingly for our purposes, ‘within a time limit of not more than 5 minutes from tram or train line’.^[16]

This precision, even in fantasy, reminds us how, for workers like Bloom, an advertising agent, time and distance, amenity and disamenity, needed to be finely calibrated if money allowed.^[17] In this regard, Bloom’s outlook is strikingly different from that of the independently wealthy Margaret Schlegel in Howard’s End (1910). E. M. Forster portrays Margaret as viewing the rail network through the prism of leisure. For her, the great railway termini of London represent ‘our gates to the glorious and unknown. Through them we pass into adventure and sunshine, to them, alas! we return.’^[18] Saved from the tiresome need to work for their living, the Schlegels, like many independently wealthy Edwardians, could view the railway network purely as a means of opening up the whole country to their pleasure. For them it was pure amenity; their use was occasional and they could live where they pleased, within reason, because they could afford to be driven to the station when the need arose. The countervailing disamenity effects of rail have been acknowledged by urban and transport historians, but it has not previously been possible to explore these systematically alongside the rail network’s amenity effects and to examine their possible interaction. Although financial and legal factors inclined railway companies to develop new urban lines (and their associated infrastructure) in poor, predominantly working-class districts, Kellett is clear that, in turn, these incursions could exacerbate the decline of blighted urban areas. At the local scale, dense rail infrastructure could reduce connectivity with surrounding areas, as well as bringing new environmental disamenities associated with steam power. Writing about London, Charles Booth described pockets of poverty being ‘caught and held in successive railway loops’, and Kellett shows that such ‘barrier effects’ operated in cities across Victorian Britain to cause, or exacerbate, the deterioration of urban neighbourhoods.^[19] It is no accident that one of Gustav Doré’s most iconic images of Victorian poverty represents densely packed working-class streets framed at either end by towering railway viaducts, and enveloped in the effluvia of a passing train.

Figure 3.2: “Over-London by rail”, Gustav Doré print, from London: a Pilgrimage (Blanchard Jerrold, London, 1872) © The Board of Trustees of the Science Museum, London CC BY-NC-ND-SA 4.0

As well as precipitating slum clearances, the incursion of rail into the city offered passengers a new perspective looking down on the cramped quarters of the urban poor.^[20] This view became a familiar trope of Victorian literature that vividly encapsulated the class dynamics of slums and suburbs.^[21] Clearly, there were many other sources of environmental disamenity in Victorian cities, especially for the poor. The deleterious effects of living close by rail barely feature in late-Victorian poverty studies; industrial pollution and poor sanitation generally loom much larger. But railways also brought direct employment opportunities, especially for those in heavy industries and transport, which could make them a positive draw for some workers. Nonetheless, in the literature on urban history and town planning, rail is rarely conceptualised as an amenity, and its disamenity value remains highly subjective. In a seminal discussion, Dyos lists the types of amenity important in the establishment of Victorian suburbs: water, sewage, gas, retail, recreation, churches and schools, but not railway stations.^[22] With a few exceptions (such as in Glasgow or Nottingham which had interventionist municipalities) rail was conceived as something to which towns and cities were subjected by market forces. Conversely, an established legal framework existed for adjudicating harms caused by economic activities based on the concept of ‘nuisance’. As Martin Daunton puts it: ‘stopping a nuisance to one neighbour – say by an injunction against smells or noise from a workshop – could harm others by reducing employment. It became a matter of the balance of interests, and not absolute rights and wrongs. [... ] A nuisance might be justified on the grounds that the activity had a general utility to society’.^[23] but such arguments were rarely rehearsed in relation to railways, because the statutory basis on which companies had been granted access to land did not allow similar complaints to be made. Where such complaints could be raised, this depended on an understanding of relevant legal provisions. For instance, letters to newspapers by well-informed middle-class correspondents cited the infringements of rules made decades earlier in relation to the considerate operation of steam engines; by contrast, the voices of those, perhaps more directly at the receiving end of rail’s disamenity effects, living in the working-class districts, are rarely to be found in the archive. For this reason, among others, our notion of ‘disamenity’ is both subjective and qualitative. We infer that those in close proximity to railways could feel the sensory effects of the steam engine, including breathing its polluted air.

These effects impinged most heavily on the urban working class, because it was only the better-off who were in a position to make choices about how to balance amenities and disamenities, the impacts of which were not evenly distributed. Many Victorian writers held that sensibility itself varied by social class. Explaining the feelings of a middle-class wife who refuses to move with her husband to poor lodgings, Gissing writes that she ‘must have submitted to an extraordinary change before it would have been possible for her to live at ease in the circumstances which satisfy a decent working-class woman.’^[24] We do not claim similar insight into how people felt about the environmental disamenities associated with dense rail infrastructure, but our evidence does suggest that the push and pull effects of rail worked differently for different social groups.

It was in Britain’s great cities that the delicate balance between the amenity and disamenity effects of rail was hardest to strike, but our method is not confined to urban contexts. By developing a new method for mapping the physical imprint of railway infrastructure on the ground, we can explore patterns of residence on a broader scale.^[25] Our analysis goes beyond case studies that focus exclusively on cities and urban morphology by exploring the relationship between rail infrastructure and street-level demography in different residential contexts, both urban and rural. Realising this ambition has only been possible by converging datasets in wholly new ways.

3.2 Digital Convergence: The View from the Street

In this chapter, we combine three big, new and open datasets in a ‘convergence experiment’. The first dataset is digitized 1901 British census data that we have geo-coded at street level (StreetsGB). The second is a dataset derived from historic Ordnance Survey maps of Britain that captures the location of railway infrastructure (MapReader) around the turn of the twentieth century. The third contains the locations and details of railway passenger stations over time (StopsGB).^[26] We created and restructured these datasets with an eye towards convergence; establishing explicit spatial information for each source type laid the foundation for connecting them. The street operates as the central unit to which we link both socio-demographic data from the census and information about the railway network (proximity to passenger stations and rail infrastructure) we derive from the new StopsGB and MapReader datasets. Linking multiple datasets on a national scale at such a refined level of aggregation is a milestone for British historical research.

Although each dataset is derived from material, to different extents, available in the public domain, they were not previously in a form that was both machine-readable and historically meaningful. There is, therefore, great novelty and utility in making them public in new ways. The census and maps were the products of major governmental initiatives during the Victorian period, which have been notoriously difficult to analyse in their entirety due to their size, complexity and problems of access. Our contribution to transforming the immensely complex, and sensitive microcensus data is in shifting the unit analysis away from individuals (a level of data that cannot be shared publicly) and larger jurisdictions (like parishes or registration districts) to streets. Our third dataset, StopsGB, draws on the longstanding work of local historians and enthusiasts, previously available only in formats not amenable to computational methods. While each new dataset and the methods for their creation are described in more detail elsewhere, here we discuss the specific issues around data creation, aggregation, evaluation, and linking that inform our experimental design.

3.2.1 Population Census

A key component of our convergence dataset is geo-located street-level 1901 census data for Great Britain. We use Integrated Census Microdata (I-CeM), a digitised version of the individual-level returns from decennial censuses of England, Wales, and Scotland taken between 1851 to 1911.^[27] I-CeM data contains the individual returns of everyone enumerated in the census. Table 3.1 is a snippet of the dataset, showing the Heighes family, who lived on Harrow Road in Paddington, London. The structure of I-CeM mirrors how the census was administered. Information was collected on individuals, who were grouped within households (or institutions, if they lived in a workhouse, for example). There were then a series of successively larger administrative units grouping these further: enumeration districts, parishes, registration sub-districts, registration districts, and registration counties.

Each row of I-CeM stores information for a single person. Individuals are listed according to their position in a household, reflecting the patriarchal structure of society, with men, typically a husband or father, listed as the head of the household. Everyone else enumerated in the household was listed in relation to the household head, for example, as wives, sons, daughters, boarders, or servants. I-CeM also captures the rich information collected by census officials, including people’s names, addresses, ages, gender, occupations, and place of birth. In total, I-CeM contains over 100 variables for each person, combining information collected at the time and derived variables calculated after the digitisation and standardisation process.

Table 3.1: Sample I-CeM Data

Note: Data displayed ‘as is’ from I-CeM, including any transcription and spelling errors.

Despite the rich address information recorded in I-CeM, previous work has predominantly used aggregate administrative units (parishes, registration sub-districts, and counties) to map and spatially analyse the data. ^[28] Most recently, Bogart et al. have examined the macro-economic impacts of improvements in transport infrastructure over the nineteenth century by linking the parish/RSD-aggregated census data to GIS transport datasets (rail, road, and waterways).^[29]

Yet the individual-level address data in I-CeM offers the possibility of spatial analysis at much higher resolutions than afforded by aggregate units of parishes and RSDs.^[30] The challenge is linking the millions of textual descriptions of addresses in the census (e.g. 3 Harrow Road) to their ‘real-world’ coordinates. We have created an enhanced version of I-CeM, known as StreetsGB, which links (anonymised) individuals and their street addresses in I-CeM to one of two existing GIS datasets through an automated Python pipeline.^[31] The first dataset is Ordnance Survey (OS) Open Roads, a freely available and re-usable GIS of modern roads in Britain.^[32] We use this modern roads dataset because there is no comparable historical dataset of streets captured as line geometries for nineteenth- and early twentieth-century Britain. The second dataset is GB1900, a collection of geo-referenced, transcribed labels from the Ordnance Survey second edition County Series six-inch-to-one mile maps of Great Britain, published between 1888 and 1914.^[33] The advantage of GB1900 is that it contains the historical names of streets, contemporaneous to our census data. The drawback is that GB1900 contains only point geometries, which refer to the upper-left corner of the text label on the map, and so it does not represent the actual shape or length of the road as OS Open Roads does.

We link individuals at street-level and not to the specific house or property in which they live. Individuals living at ‘3 Harrow Road’ and ‘4 Harrow Road’ will therefore all be linked to ‘Harrow Road’ in OS Open Roads and GB1900. Linking at street level has several advantages over linking to individual properties or houses. Geo-coding individual properties may be geographically more precise, but this precision may be historically meaningless. House numbers in the census may bear no relationship to modern house numbering, either because streets have been renumbered or because the numbers refer to the order in which properties were visited by the census enumerator. Streets also provide both much higher spatial resolution than parishes or RSDs and capture the ‘character’ of an area through its occupational structure and demography. Creating our new geo-coded datasets using street-level rather than property-level data also enables us to publish the derived datasets open-source.

Table 3.2 shows the number and proportion of people linked to each dataset. In total, we link 72.3% of individuals to the street they lived on for the 1901 census. Overall, more people were linked to OS Open Roads than GB1900 but a substantial proportion of people (40%) were linked to both OS Open Roads and GB1900. Linking rates were highest for urban areas, where the OS Open Roads coverage is closest to historical street topography, and weakest for remote rural areas, especially in Wales where matching problems have been compounded by shifts in the preferred spelling of Welsh place names. Including GB1900 data significantly improved linking rates outside urban areas and also improved socio-economic representativeness nationwide, justifying their inclusion despite the limitations of their point data format.^[34]

Table 3.2. Number and percentage of individuals geo-coded in 1901 Great Britain Census.

Reflecting the slight urban bias to the geo-coded dataset, younger people (under 40 years old) and women are slightly over-represented compared to the full population. Similarly, predominantly urban occupational sectors (most notably manufacturing) are over-represented and trades located outside urban centres (e.g. farming and estate work and mining & quarrying) are slightly under-represented. We identified a particular over-representation of service, commercial and professional sectors, and an under-representation of general and factory labourers.^[35] Our hypothesis is that this reflects the weaker linking of minor streets and courts compared with main roads and substantial residential streets (due partly to subsequent demolition of these minor streets, leading to their omission from OS Open Roads, and partly to the absence of labels for these streets on the historical maps transcribed for GB1900). Overall, we are confident that the greater precision made possible by linking most people to the specific street they lived on outweighs the disadvantages associated with no longer being able to make statements about the whole enumerated population (as can be done when resolving people to their parish of residence).

3.2.2 Railway Infrastructure

Data from Maps

Like the census, individual map sheets have been widely used by historians; however, historical map series have never previously been used at the collection level. In their foundational study on urban history, James Johnson and Colin Pooley depict maps as objective, yet sporadic sources.^[36] Certainly, city maps were infrequently surveyed or printed. OS Town Plans, printed at an exceptionally large scale (1:500), cover only urban centres and few were resurveyed or revised at all. And yet for Britain, the early editions of large-scale six-inch and twenty-five inch to one mile OS maps contain rich details about pre-First World War landscapes. With continuous coverage of urban, suburban, and rural areas, these maps are key visual primary sources that contain evidence of industrial development. People have called for maps to be used to provide more than mere illustrations of claims made using evidence from textual sources, but it remains rare to see this in action beyond histories of maps themselves.^[37] With this in mind, we treat the content of OS maps as a ‘visual census’ to be examined computationally alongside the population census.^[38] Previously, urban history has been constrained by the practical challenge of visually comparing physical maps across multiple cases. Our computational methods make it possible to transcend the comparative case study approach, by instead assessing all towns and cities simultaneously alongside one another, and in the context of their suburban and rural hinterlands.

To make such research possible, we developed MapReader, a software library that creates reproducible, labelled data based on queries of large collections of homogeneous maps.^[39] MapReader takes digitised map images and their metadata as input, and produces labelled data for very large numbers of flexibly-sized ‘patches’, or grid-like regions of maps. This approach allows us to investigate one or more visual phenomena present on thousands of maps relatively quickly compared to less efficient, more complex computer vision tasks (such as object detection or image segmentation). While the patch is an artificial spatial construct, it is nonetheless sized by the researcher in relation to a concept of interest that has a visual manifestation on the map. And while patches are less precise than lines and points, using a classification task at the patch level is highly effective at identifying areas containing information of interest. Patches offer a coarse-grained approach whose benefits include flexibility and speed: as such, they are no replacement for reliable vector data, but they are a powerful tool when used to spot map features across large areas of the country. MapReader’s patch classification approach, therefore, lends itself well to making map content into flexible machine-readable data for research.

In this chapter, we use a MapReader dataset of patches identifying rail infrastructure across all of Britain, hereafter the ‘railspace’ dataset. After annotating roughly 62,000 patches manually, we fine-tuned computer vision models accessible in the MapReader pipeline, selected the best performing one (resnest101e)^[40], and were able to infer the presence of rail infrastructure on all 30.5 million patches for every second edition six-inch-to-one-mile OS map sheet (1.3% of the total patches were classified as railspace).^[41]

Figure 3.3. Detail of MapReader results for railspace patches (black) around Leeds [NB this figure to be replaced by enhanced image pending rights negotiations].

Our data is more comprehensive than other datasets representing historic railways because it contains tracks documented on maps that were not documented elsewhere.^[42] It also captures the full spatial footprint of rail infrastructure, including tracks, embankments, associated buildings, and any other visibly railway-related construction. This ‘railspace’ dataset offers a novel measure of how a key dimension of industrialisation impacted the physical, and therefore the social, economic, and cultural landscapes of communities. Exploring this ambient, or environmental, effect of rail is made possible by MapReader’s broader conception of railway infrastructure, i.e. how its patches capture the overall footprint of rail infrastructure beyond its stations and routes (typically abstracted into points and lines in standard vector data representations).

Data from a database of passenger rail stations

Combining this enriched ‘railspace’ output with other sources on British rail infrastructure allows new questions to be posed. We use a newly-created open dataset identifying over 12,000 railway passenger stations. Derived from Michael Quick’s monumental reference work Railway Passenger Stations in Great Britain: a Chronology, this information has been parsed and geo-coded to create a structured dataset suitable for spatial analysis.^[43] The dataset includes information about station opening and closing dates, which has been used to exclude stations not operating in 1901.^[44] In combination with MapReader’s predictions about ‘railspace’, the StopsGB dataset allows exploration of the relationship between the railway network’s overall footprint and its specifically passenger-facing aspects.^[45]

3.3 The Convergence Experiments

3.3.1 Overview

Triangulating these three datasets allows us to investigate the interplay of rail and residential patterns as twin facets of industrialisation. We use the following method to combine our datasets, in each case classifying streets (and hence the people who lived on them) by their relationship to the rail datasets. First, we calculate the distance from each street in StreetsGB to the nearest passenger station in StopsGB. This provides an indicator of proximity/access to railway stations as means of access to work or for general travel. Our analysis of proximity to stations uses this distance-to-station metric to group individuals, households, and streets at various distance bands from stations, e.g. Under 250m, 250-500m, 500-750m, 750m-1km and so on. Next, we categorise streets by their proximity to substantial railway infrastructure (tracks, sidings, depots, etc) as identified by MapReader. This provides an estimate of how strongly rail may have dominated the local environment. We define railspace proximity in terms of the percentage of MapReader patches classed as ‘railspace’ within 100 metres of each street.^[46] Since the majority of people lived on streets where there was no railway infrastructure within 100 metres, we include ‘No railspace’ as its own category, while streets near or within railspace are divided into bands from light (<20%) to very dense (>80%).

By combining our metrics for proximity to station and railspace we capture the different ways groups of individuals might use, interact with, and experience the rail network. For some, proximity to passenger stations offered amenity (the ability to traverse the rail network) but with varying degrees of disamenity depending on their proximity to heavy rail infrastructure and the associated nuisance of noise and pollution. For others, proximity to rail infrastructure, irrespective of distance to station, offered amenity through the employment opportunities that the railway brought. For this latter group, in contrast to rail network users, the amenity effect increased (at least in terms of employment opportunities) as railspace density increased.

3.3.2 People and Stations

We began our analysis of the datasets with a number of hypotheses about the likely demographic patterns associated with proximity to passenger railway stations and dense rail infrastructure. We assume that wealthier people possessed the resources to be selective in where they choose to live, and therefore chose to locate themselves in areas perceived as ‘desirable’ within their social milieu. This might include access to transport networks, and avoidance of locations where transport infrastructure degraded the local environment. By contrast, poor and disadvantaged people are assumed not to have been able to be as selective in where they chose to live, and therefore, by default, lived in areas deemed undesirable by those with greater economic power. These locations may have been characterised by limited access to transport networks and/or disruptive impacts on the local environment as a consequence, amongst other things, of transport infrastructure and its associated trades. Finally, we assume that some people (as a consequence of employment) may derive particular amenity from easy access to transport networks, and are therefore strongly motivated to live near transport nodes (including railway stations) despite some potential environmental disruption. Daily commuters to the central business district are the obvious example.

The census does not provide direct measures of either wealth or income, and so we have focused on demographic variables that provide surrogate markers, such as servant-keeping and occupation. According to Seebohm Rowntree, writing in 1901, employing live-in domestic servants represented the key social (and cultural) marker of the ‘middle class proper’.^[47] Exploring such wealth-sensitive census variables in relation to residential proximity to passenger rail stations revealed clear differences between urban and rural parishes.^[48] In urban areas, the percentage of households with live-in servants was highest close to stations (under 250 metres), falling sharply across distance bands between 250m and 750m, reaching its lowest point in the 1 to 1.25 km band, before rising again to a second, albeit lower, peak at 2 km from a station (beyond this the urban data become sparse and erratic, see Figure 3.4). By contrast, in rural parishes the mean number of servants is highest at greater distances from a station (over 2 km), although here too values are relatively high close to a station, and there is a notable fall between 500m and 1 km.^[49]

This allows us to draw important distinctions between the effects of rail infrastructure in urban and rural contexts, given that individuals living in urban areas were much more likely to live close to a station than their rural counterparts. In 1901, around 73% of the national population lived in parishes defined as ‘urban’, and this preponderance of urban residents was even greater, at 86%, among those living within 250m of a railway station (see Appendix, Tables 3.A.1 and 3.A.2). By contrast, rural dwellers predominated among those living more than 3 km from a station, accounting for 74% of households in this category (Table 3.A.2). Proximity to a station appears to have been much more important in urban and urban-fringe locations compared with rural areas, where distance to a station appears to exert a more muted influence on the residential patterns associated with high-status, servant-keeping households. This is consistent with the contrasting amenity uses associated with different factions of Rowntree’s ‘middle class proper’; with the different worldviews, and practical needs, of daily commuters like Leopold Bloom compared with independently wealthy individuals like Forster’s Schlegel family, for whom the railway network represented an occasional leisure amenity.

Figure 3.4: Percentage of households with live-in servants by distance to station and parish type, Great Britain (1901).

3.3.3 The Streetscape and Railspace Density

Alan Jackson’s Semi-Detached London nicely captures how finely balanced calculations about the contrasting amenity and disamenity effects of rail shaped social segregation in the late-Victorian suburb. In the typical suburb, he argues, ‘[M]ost of the houses would be within easy walking distance of the station, but the larger villas of the “carriage folk” were a little farther away, perhaps on higher ground. Nearer the station would be cottages of the serving classes and a row or two of shops with accommodation above for their owners’.^[50] As Table 3.3 shows, in 1901 most people did not live within 100m of rail infrastructure. Over 68% of people had no railway tracks within 100m of where they lived, and around 16% lived near light rail infrastructure (defined as between >0 and 20% of adjacent railspace patches). A further 16% lived within railspace densities of over 20%, but it was very rare for people to be surrounded by highly dense railspace (only 3% of streets registered adjacent railspace densities at over 60%). Individuals living in urban areas constituted the overwhelming majority of people living near each band of railspace above zero - in all cases above 84%.

Table 3.3: Number and percentage of individuals by railspace density, 1901 Great Britain (within 100m from street)

Finally, Figures 3.5 and 3.6 show the percentage of servant-keeping households by both railspace and proximity to a station. Perhaps the most striking feature of these graphs is that they demonstrate that, just as Jackson suggests, it was relatively common to live close to a station, but away from the direct environmental influences of railway infrastructure (the yellow ‘zero’ railspace line). Interestingly, urban households display little difference in levels of servant-keeping between streets with no railspace and those with some (green <20% line). Indeed, in all the urban bands less than one kilometre from a station, it is streets close to some railspace, rather than those with none, that register the highest proportion of servant-keeping households (perhaps because these streets tended to be associated with better network access for the well-to-do daily rail user). It is also striking that, as hypothesised, urban streets surrounded by dense railspace (over 40%) are associated with lower levels of servant-keeping, and that broadly speaking the denser the railspace, the fewer households with live-in domestic servants.

Figure 3.5: Percentage of urban households with live-in servants by distance to station and railspace density, Great Britain (1901).

Figure 3.6: Percentage of rural households with live-in servants by distance to station and railspace density, Great Britain (1901).

By contrast, the pattern is both different and less clear-cut among rural households. Here, except within 250m of a station, households with no nearby railspace consistently register the highest values for servant-keeping, while levels of servant-keeping oscillate wildly for all other household types, especially beyond one kilometre from a station (Figure 3.6). Here the influence of proximity to a station on the location of wealthy, servant-keeping households appears much more muted. The main exception is households on streets within 250m of a station, where, as in urban areas, servant-keeping households are slightly more common on streets with some nearby railspace (under 20%), than on those with none. Equally, between 250m and 750m from a station there is a general tendency for most rural household types to register falling levels of servant-keeping households with increasing distance from the station. This tends to suggest that in rural, as in urban areas, there were some people for whom proximity to a station carried a significant advantage. It is possible that this reflects customer-facing establishments such as inns, hotels and retail premises.

These patterns appear consistent with our hypothesis that, for well-to-do households, the amenity/disamenity calculus for living close to a station (and hence potentially also to rail infrastructure) would vary depending on how family members made use of the rail network (for daily work or occasional pleasure), and also how they travelled to the station (among Jackson’s ‘carriage folk’ the need to live very near a station would have been less pressing, even for the daily commuter). To explore these differences further, and also to explore the non-network amenities that rail might bring to some groups of workers (through employment opportunities associated directly with rail infrastructure), we shifted our focus to occupation. What type of worker was most likely to live close to a station but not close to dense rail? Did particular occupational groups tend to be concentrated in areas of dense rail, or were such areas simply left to those with least choice over where to live?

Given the complexity of occupational classification systems, we developed our own occupational clusters using the finest-grained occupational codes from the digitised census.^[51] We created occupational clusters for groups of workers deemed likely to have had distinctive relationships to either the amenity or disamenity aspects of rail. These consisted of four groups: workers whose employment was considered likely to be located in a Central Business District (CBD); workers directly employed on the railways, on railway building and maintenance or in making rolling stock (RW); workers in heavy manufacturing and extractive industries likely to require supporting rail infrastructure including the iron and steel and mining sectors (M); workers involved in the distribution and storage of goods (D).

Perhaps the most striking spatial pattern about the distribution of workers in these clusters was that the percentage of the workforce employed in heavy manufacturing increased as distance from the nearest passenger station increased (see Figure 3.7). If heavy industry was associated with rail transportation, this does not appear to have led to the creation of passenger rail stations, which seem unrelated to the location of workers in this sector. The other clear pattern to emerge was that, as predicted, the percentage of workers in occupations associated with the central business district declined with increasing distance from the nearest station (see blue line, Figure 3.7). It seems likely that these were the people most dependent on a daily commute for work; those for whom the amenity effects of rapid transit by rail were therefore greatest. Of course, there were many other ways to commute to work in late Victorian towns and cities, including omnibus, tram, bicycle and on foot.^[52] But even so, the pattern for this group, which includes workers in banking, insurance, commerce and the principal professions, is fairly conclusive. None of the other occupations appears to have a comparably strong association with proximity to railway stations, although distribution workers also show a broadly linear relationship of rising proportions closer to stations (Figure 3.8).

Figure 3.7: Percentage of all urban occupied workers in different customised employment clusters by distance from nearest station, Great Britain, 1901

Figure 3.8: Percentage of all urban occupied workers in three employment clusters by distance from nearest station, Great Britain, 1901

But what happens when we add the density of rail infrastructure—our ‘railspace’ measure—to this story? Does this shed light on the perceived disamenity effects of living close by rail, and, if so, who felt them? How did wealthier individuals, those with most choice in the matter, negotiate the balance between amenity and disamenity, and can we identify groups for whom the economic benefits associated with proximity to rail infrastructure may have led them to be overrepresented in areas of densest railspace? On a national scale, areas of densest rail infrastructure (over 60% railspace density) tended to be associated with more workers occupied in heavy manufacturing, railway employment and goods distribution, and with fewer workers associated with occupations concentrated in central business districts (Figure 3.9). For many occupations, greater railspace density was more strongly associated with the pull effects of employment in specific sectors dependent on rail, than with the push factors associated with railways as an environmental disamenity.

In saying this we do not wish to endorse Gissing’s class hierarchy of sensibilities. Economics alone is enough to explain why workers in some industrial sectors are more numerous in areas dominated by rail infrastructure. However, it is striking that only one of our occupational groups becomes markedly less numerous as rail density increased: workers associated with central business district occupations. We have already seen that these workers had a strong propensity to live close to passenger stations, and yet, their numbers fall at rail densities above 20%. Since many workers in these occupations would also have been wealthier than average it may be that those with sufficient means to do so (by no means all) were making finely calibrated decisions about the balance of amenity and disamenity from rail.

Figure 3.9: Percentage of workforce in custom occupational groups by railspace density, Great Britain (any distance from station), 1901

The pattern is similar when one looks only at urban parishes and focuses just on those people living within 2km of a railway passenger station (Figure 3.10). Again, the proportion of workers in CBD occupations rises between zero and 20% railspace density, and then declines. Indeed, if anything the fall is sharper suggesting that dense rail infrastructure may have exerted a stronger push effect for CBD workers in urban contexts, while the patterns for all other groups are broadly similar to the national picture. As with the analysis of the distribution of servant-keeping households, this suggests that modest (as opposed to no) railspace may have offered many daily commuters the optimum trade-off between the amenity and disamenity effects of railways in late Victorian Britain. It also reminds us that these are patterns of association, not iron social rules. We still find CBD workers living in areas of very dense railspace, just as we also find servant-keeping households on such streets (see Figure 3.5); but their numbers are lower. Overall, this national-scale analysis of patterns between the distance to station, railspace density, and the custom occupational groups highlights quite nuanced pictures for groups who had different desires, constraints, and opportunities. It is the combination of the national coverage of the data and its geolocation at the street level that makes it possible to track these residential trends at such a fine level, thereby highlighting the different experiences of those working in city centres compared with those working in rail-related industries.

Figure 3.10: Percentage of workforce in custom occupational groups by railspace density (urban only; living less than 2km from station), Great Britain, 1901

3.4 Bringing Place Back In

Having mapped these broad national patterns of association between street-level census data and rail infrastructure, we wanted to deepen our understanding of what was actually happening on the ground in specific places. To interrogate street-level patterns we therefore decided to identify four parishes where we already knew something about the quotidian experience of life alongside Britain’s rail infrastructure. We chose two well-known Victorian commuter suburbs: Camberwell and Upper Holloway; and two contrasting industrial districts: Salford and Middlesbrough. Harold Dyos’s classic study of Camberwell, Victorian Suburb (1961), includes a detailed account of the impact of suburban rail development from the 1860s, as well as a case study of the Sultan Street district, where the ‘barrier effects’ caused by constructing new urban rail lines contributed to the ‘creation of an urban slum’.^[53] Upper Holloway is where George and Weedon Grossmiths’ fictional City clerk Charles Pooter is said to live in his ‘nice six-roomed residence’. At the opening of his diary, Mr Pooter describes how his family initially feared that noise from the railway at the bottom of their garden would prove an inconvenience (and used this to secure a reduction on the rent). Pooter records that his landlord was right to argue that they wouldn’t notice the trains after a while, noting: ‘beyond the cracking of the garden wall at the bottom, we have suffered no inconvenience’.^[54] It reminds us that people with social pretensions (Pooter has plenty) could learn to adapt to the inconveniences associated with living close by rail, although we should not forget that Pooter is a figure of (gentle) fun, for whom life is generally an unhappy compromise. As Figure 3.11 illustrates, the streets of Upper Holloway tended to be particularly close to passenger rail stations, although Pooter himself (like so many Londoners), relied on the cheaper omnibus service to commute back and forth to his job in the City.^[55]

Middlesbrough and Salford, by contrast, were areas of heavy industry and docks, with vast areas devoted to the sidings and shunting yards required in an age when so much heavy freight needed to be moved by rail. In inner-city Salford, railways (and indeed waterways) had had to be punched through an already developed urban landscape,^[56] but in Middlesbrough rail infrastructure grew organically with the town during its great late-Victorian iron boom. Here, residential and industrial/transport sectors evolved in tandem and were therefore generally better segregated than across most of late Victorian Britain. Hence railspace density is only marginally higher for Middlesbrough than Holloway, whereas it is twice as high in Salford (Figure 3.11). We chose the ‘Regent Road’ parish of Salford because it included the cluster of terraced streets that Robert Roberts immortalised as ‘the classic slum’ in his semi-autobiographical account of the Edwardian working class.^[57] Roberts famously described the thirty or so streets and alleys that delimited his childhood social world as being ‘locked along the north and south by two railway systems a furlong apart’ (see Figure 3.18).^[58] Finally, the classic study of Middlesbrough in its industrial heyday is Lady Florence Bell’s study At the Works (1907), which includes vivid accounts of the dangers trains and trucks posed to working-class children living on streets close to the town’s great iron foundries; the segregation of railspace and residential space was far from absolute even when the two grew in tandem.^[59]

Figure 3.11 Mean railspace and distance to station by parish, Great Britain 1901.

Our first step was to understand how our four case-studies compared with other places. To do this, we generated a scatter-plot of 6,564 places plotting mean railspace density against mean distance to the nearest passenger railway station (Figure 3.11).^[60] Most places are clustered in the bottom left corner, reflecting the fact that by 1900 the bulk of the British population lived relatively close to a station and on streets with little or no adjacent railspace. However, even at the parish level there was evidently considerable variation. As noted, railspace was significantly higher in Roberts’s Salford than for the other three case-study locations.

But what can we learn by shifting our focus to visualising street-level patterns in the distribution of people and rail infrastructure? Are the broad patterns registered at national level amplified or obscured when visualised on the ground in our four case studies? In the maps that follow, MapReader’s railspace patches are represented by black squares. This is the physical footprint of railway infrastructure. White circles represent the passenger railway stations identified by StopsGB. Streets linked to the census returns in StreetsGB are shown in colour (dots for GB1900 point data, lines where the join is made using OS Open Roads data).

In the area of north London that includes Pooter’s Upper Holloway, railspace was densest in the south and much sparser in north (Figures 3.12 and 3.13). The type of rail infrastructure was different too, with railspace areas in the south including extensive sidings and terminuses with warehouses and locomotive sheds (see the Midland Works and the Brickworks near Gospel Oak station in the bottom left, and the goods and coal depots to the bottom right). Both these areas have since been developed with the former now light industry and warehouses, and the latter the site for a Premier League football stadium. By contrast, the railspace to the north east comprises an intersection of two lines. This consumes a smaller footprint and the nature of its use would have generated less noise, smoke, and disruption than the dense rail infrastructure in the south. The streets in the middle of the intersecting railway lines in the top right still exist. It makes sense that servant-keeping was high (red) in the north, across both non-railspace and railspace, but generally only high in the southern sections of the map in the middle area buffered from heavy railspace by streets dominated by households with lower levels of servant-keeping (Figure 3.12). Streets with the lowest (blue) and low (green) levels of servant-keeping intersect with rail in many places in the south. The density and type of railspace is therefore a key factor in this residential patterning. The distribution of CBD workers follows a similar pattern to servants in the northern suburbs, but the highest band (red) is widespread across the map, even close to the densest railspace in the south (Figure 3.13). Only in the bottom left (Gospel Oak) with its dense shunting yards was this less true. The differences in the distribution of servant-keeping households and CBD workers suggests that the servant measure is identifying subtle differences in social status among CBD workers. Though such workers are broadly present across most streets in north London, those with the means and space to employ live-in servants tended not to reside near the large rail sidings and goods sheds in the south.

Figure 3.12 Map of spatial distribution of servant keeping in relation to railspace and stations, Upper Holloway (London), 1901.

Figure 3.13 Map of spatial distribution of CBD workers in relation to railspace and stations, Upper Holloway (London), 1901.

The closest approximation to Pooter’s fictional house ‘The Laurels’ may be located on Pemberton Gardens/Road (now all Pemberton Gardens), which is just to the west of Upper Holloway station. The gardens on the south side backing on to the Tottenham and Hampstead rail line. Just under half the families on this street employed servants. CBD workers made up just over a fifth of the some 476 people living on the street. The houses got better towards Holloway Road (according to Booth) with the largest being up to 10-12 rooms with a bathroom and ‘one and often two servants kept’.^[61]

One of the two most deprived streets in the area (according to Booth’s reports) was located near the dense railspace running north-south from Finsbury Park to Holloway station. Queensland Road was right at the end of the sidings with a Great Northern Electric Light Station and a gas works. In the middle of Queensland Road, on the south side, was a Xylonite collar works, which produced plastic-coated shirt collars. Xylonite was a highly flammable substance and early in the 1890s there was a serious fire at a factory manufacturing it at Homerton, East London. Although they had ‘never had any fires here’, such close proximity to potentially hazardous sites indicates the disamenities brought both by rail infrastructure and the industries it attracted.^[62] The other heavily deprived street was Campbell Road, just west of Finsbury Park station. This was not a story of railway blight. Campbell Road was not hemmed in by rail, as Queensland Road was, and in fact was buffered from the rail lines and the Great Northern Goods and Coal Depot to the north of Clipton Terrace. Clipton Terrace itself was apparently showing ‘signs of going down’ by the 1890s, but a local police inspector could ‘give no certain reason for the street [Campbell Road] having become so bad’.^[63] According to Jerry White, ‘Campbell Bunk’, as it came to be known, became blighted because it was broken up into individual building plots, many of which languished undeveloped throughout the area’s boom years, in turn making the whole street unattractive to the ‘smaller servant-keeping class’ for whom it was designed.^[64]

In Camberwell, rail was generally less dense and mostly single or double rail lines except towards the north-east where there were sidings leading to a large number of goods sheds (just north of the image) and in the north-west, by the Thames, where there were four or more rail lines (Figure 3.14). Higher levels of servant-keeping (red) were not located near these dense rail areas. The exceptions were main roads which, due to their retailing businesses and pubs, had higher servant-keeping across areas of both low and high railspace density. For example, the only street near the goods area in the north-east was the Old Kent Road, notoriously the most impecunious address on the Monopoly board. Residential side streets to the south had high levels of servant-keeping not found north of Camberwell New Road and Peckham Road. These streets were near stations and some railspace (but not near warehouses and sidings). As at Holloway, CBD workers were more ubiquitous than high servant-keeping households, but at Camberwell they were mostly concentrated in southern and western streets and clustered in areas between main rail lines but still close to stations (Figure 3.15).

Figure 3.14 Map of spatial distribution of servant keeping in relation to railspace and stations, Camberwell (London), 1901.

Figure 3.15 Map of spatial distribution of CBD workers in relation to railspace and stations, Camberwell (London), 1901.

The location of Dyos’s slum is in the centre of the image indicated by the purple hashed area and includes Sultan Street (green line in Figures 3.14 and 3.15) and Hollington Street (blue dot in Figure 3.14, yellow in Figure 3.15), both immediately to the west of the rail line running north-south in the centre. This railway, built in the 1860s, cut access to Camberwell Road, impoverishing these two streets, but by 1901 few streets in this area, except Camberwell Road itself, had significant numbers of servant-keeping households.^[65] The unique situation of Sultan Street and Hollington Street may have made the coming of the rail line worse. One of Booth’s assistants, George H. Duckworth, commented that ‘these two streets are tucked away by themselves. There is a foot passage by the church out of the W[est] end of Sultan St. but none for carts’. The lack of cart access at that end was probably significant for limiting the disposal of waste (typically carried off in carts). Duckworth described Sultan Street as similar in ‘appearance and character’ to Hollington Street, which he noted had a ‘fearful mess in [the] street, bread meat, paper, vegetables, old tins [...] heavy rain but street full of children, bare head and bare feet [...] most windows broken’. The contemporary associations between deprivation and ethnicity are clear, with Duckworth noting that the inhabitants were ‘Irish cockney general labourers not the carters of Beckett Street [the street below but with no direct link to Hollington Street]’. The backs were narrow and the one or two roomed houses lacked gardens, and there was ‘much overcrowding’.^[66] The railway cutting off access to Camberwell Road may have led to these streets deprivation but by 1901 the main ‘disamentities’ were not the rail line to the east or distance from station (Wandsworth Road Station was less than 500 metres away). By the turn of the century, the significant environmental disamenity in this sub-district was the condition of the housing and overcrowding. The coming of rail may have tipped the balance, but at some point disamenities become mutually reinforcing and get baked in, as Jerry White shows for Islington’s ‘Campbell Bunk’.^[67] The Sultan Street area experienced similar cycles of deprivation. These streets now form the Wyndham and Comber Estate.

In Middlesbrough the dense rail follows the river on both banks of the Tees, but also cuts a swathe through the town centre heading towards Middlesbrough dock in the East. Streets with high rates of servant keeping were scattered across the town, although they tended to be on main thoroughfares and were largely absent in the East around the North Ormesby area just south of the main dock (Figures 3.16 and 3.17). There was some high levels of servant-keeping in the northern part of Middlesbrough, surrounded by river and rail, but in general servant employers here were customer-facing concerns including hotel managers on Albert Street (north-south immediately above Middlesbrough station) and Innkeepers and Publicans on West Street (to the west of Market Place). Depot Road, the most northerly road with high servant-keeping extending into the heavy railspace area, had only 2 households - a railway platelayer and a foreman steam engine fitter. The foreman employed a general domestic servant, accounting for the high (50 per cent) servant-keeping measure here. The distribution of CBD workers was much more clear cut, and shows them concentrated to the south of the town centre, especially far south near Albert Park, away from the dense railspace dominated by docks, sidings, and warehouses. One apparent exception is Charlotte Street (around 14 per cent CBD)—located to the west above the railway line near Middlesbrough station—but there were only nine individuals living on the street, one of whom was a CBD worker. By living further from the dense rail infrastructure, CBD workers also distanced themselves from Middlesbrough station, but it is important to note that the compact geography of Middlesbrough and lack of rail stations reflect a town where most journeys to work would be on foot or bus. For workers with the means to choose where they lived, the push-pull factors of disamenity and amenity did not therefore always act in tandem. In the case of Middlesbrough, it was the push (away from dense railspace) that appears to have dictated the decisions of most well-to-do residents. There were, of course, other environmental disamenities influencing the desirability of certain locations. For example, the area between Newport Road and the mass of sidings to the east of Newport Rolling Mill, regularly flooded in the 1890s and early 1900s. This included Marsh Road (railspace score of 0.49, indicating very high rail density nearby), where ‘the land slopes towards the river and thus acts as a drain to the higher parts of the town.’^[68]

Figure 3.16 Map of spatial distribution of servant keeping in relation to railspace and stations, Middlesbrough, 1901.

Figure 3.17 Map of spatial distribution of CBD workers in relation to railspace and stations, Middlesbrough, 1901.

At Salford, heavy railspace dominates around the docks in the south west and south of the bend in the Irwell river (Figures 3.18 and 3.19). Roberts’s ‘classic slum’ is located in the centre of the image, bounded by Gas Works to the west, sidings to the east and multiple mainline rail tracks to the north and south. There were few servant-keeping households except on Liverpool Street, a major thoroughfare with shops. Servant-keeping was generally low across the districts of Salford area between the two major clusters of rail infrastructure, with the exception of main roads such as Regent Road (which gives its name to the census registration sub-district covering this area). Higher rates of servant-keeping are generally only found in the suburban streets away from heavy rail infrastructure in the north-west and in the commercial heart of Manchester to the north-east. The distribution of CBD workers broadly follows servant-keeping with concentrations in the western suburbs and central Manchester. They are absent from Roberts’s classic slum area. Roberts was right when he recalled: ‘West of us, well beyond the tramlines, lay the middle classes, bay-windowed and begardened. We knew them not.’^[69] His ‘classic slum’ represents an extreme example of the disamenity effects that rail could bring. This was partly through its ‘barrier effects’, rail isolated the district from the surrounding cityscape, but this was doubtless reinforced when, as here, adjacent railspace was dominated by shunting yards and heavy industrial uses.

Figure 3.18 Map of spatial distribution of servant keeping in relation to railspace and stations, Salford (Manchester), 1901.

Figure 3.19 Map of spatial distribution of CBD workers in relation to railspace and stations, Salford (Manchester), 1901.

3.5 Conclusion

Structuring historically-rich digital sources so that they can be converged in novel and flexible ways opens new perspectives on the social history of industrialisation and urbanisation. Our method allows us to shift between big-picture macro analysis, and the fine-grained, human-scale exploration of social context at street level. Consequently, we have been able to assess the impact of rail on the British population at the turn of the twentieth century in more detail than previously possible.

In particular, we have illuminated the finely calibrated decisions of those able to make choices about where to live, including how proximity to rail infrastructure factored into these decisions. High status households with servants chose to live close to stations, particularly in urban areas, and showed a clear preference for being situated within one kilometre of a station. These same households also avoided particularly dense rail infrastructure. But in urban areas it was hard to avoid rail infrastructure entirely, and, for some, light rail infrastructure had to be tolerated in exchange for the convenience of being near a station. These same patterns held true for certain types of occupational groups, such as those in trades associated with commuting to a Central Business District, but not for others, such as workers in manufacturing, whose relationship to rail infrastructure was very different. Workers in heavy manufacturing were in fact more likely to live further from stations than near to them, but were disproportionately located in areas of high railspace, since although there were environmental disamenities, their livelihoods were tied to industries dependent on rail.

Our contrasting case studies of Camberwell, Holloway, Salford, and Middlesbrough highlight the differential impact of railways depending on the type of rail infrastructure involved. At street-level, the spatial distribution of CBD workers and high status servant-employing households mirrored national trends, with clear preferences for workers with the means to choose avoiding heavily industrialised areas with rail in preference for areas with no rail or little rail that were still near a station. The distinction between different types of railspace is important - extensive railway sidings and associated warehouses and factories represented a clear disamenity for those with means (albeit varying) to choose where they lived, but single or double rail tracks could be tolerated, even at close proximity. Holloway and Salford demonstrate most clearly the distinction between these two types of railspace and the character of the neighbours in the vicinity of each. Elsewhere, as in Camberwell, rail lines could exert a barrier effect that exacerbated perhaps already poorly connected streets thus leading to greater levels of deprivation than if rail had not been built nearby (notwithstanding the close proximity of stations). As Charles Booth’s notebooks underline, neighbourhoods near heavy railspace also display other markers of deprivation (e.g. overcrowding) that we have not considered in detail in this chapter, but that were clearly important environmental disamenities. There is more that could be explored about the relationship between pre-existing deprivation, how it was exacerbated by the advent of different types of rail infrastructure, and how the two intersected in rendering certain streets and neighbourhoods deprived.

This chapter is very much a first word on these themes using new data. There are many avenues for further exploration. Possibilities include considering the persistence of deprivation today in the same areas highlighted in our case studies. This raises the opportunity to use StreetsGB to look at historical antecedents of modern deprivation. It would also make it easier to explore how environmental disamenity interacted with ethnicity as well as class, since modern data collections register ethnicity in a way that historic census data does not (with historic data we only have information on place of birth). It would also be fruitful to identify new case studies using both the parish aggregated data to explore outliers, or by applying clustering techniques to identify streets across Great Britain that shared similar characteristics and relationships with rail infrastructure (patterns that would otherwise not be identifiable manually). There is also great scope to use these methods to explore change over time since all three sources (census, StopsGB and OS maps) have a temporal dimension.

Future MapReader experiments, with new training data and models might capture the tram, as well as the rail, network as the two systems evolved in urban and suburban areas. This would make it possible to compare both systems at the national scale for the first time. Bloom’s indifference to whether his rapid transit into central Dublin would be secured by train or tram probably was typical. It is widely recognised that the expansion of the municipal tram system from the late nineteenth century, alongside the development of suburban rail, played a decisive role in the social transformation of British cities, increasing levels of residential segregation by income/class and facilitating the rapid expansion that gave life to Geddes’s neologism ‘conurbation’ for contemporaries to describe their new regional mega-cities.^[70]

We openly release our code and data to ensure the reproducibility of the findings in this chapter and to support the development of future work.^[71] Throughout this research, we have made decisions that facilitate such openness. We hope that such open research practices for historical research will become standard, ensuring that historians can critically engage with the thought processes and interpretative decisions made throughout the processes of data creation, curation, and analysis.

Data Appendix

Table 3.A.1: Number and percentage of individuals by urban/rural classification, Great Britain, 1901.

Table 3.A.2: Number and percentage of individuals at different distances from stations by urban class (‘% Of distance band’, e.g. 86.33% of people less than 250m from a station lived in Urban parishes; ‘% of all Rural’, e.g. 9.93% of people in Rural parishes were less than 250m from a station).

Figure 3.A.1: Mean number of servants per household by distance to station and parish type (Great Britain, 1901).

1. Andreas Malm, Fossil Capital: The Rise of Steam Power and the Roots of Global Warming (London: Verso, 2016), esp. ch. 7; E. A. Wrigley, The Path to Sustained Growth England's Transition from an Organic Economy to an Industrial Revolution (Cambridge: Cambridge University Press, 2016). ↑

2. B. R. Mitchell, British Historical Statistics (Cambridge: Cambridge University Press, 1988), 547. ↑

3. See Dan Bogart, Xuesheng You, Eduard J. Alvarez-Palau, Max Satchell and Leigh Shaw-Taylor, ‘Railways, divergence, and structural change in 19th-century England and Wales’, Journal of Urban Economics 128 (2022); Stephan Heblich, Stephen J. Redding, and Daniel M. Sturm, ‘The Making of the Modern Metropolis: Evidence from London’, Quarterly Journal of Economics 135, 4 (2020): 2059–2133. For a spatial approach see, Robert Schwartz, Ian Gregory, and Thomas Thévenin, ‘Spatial History: Railways, Uneven Development, and Population Change in France and Great Britain, 1850-1914’, The Journal of Interdisciplinary History 42, no. 1 (2011): 53–88. ↑

4. With one exception, the journal Urban History has in nearly five decades published almost no research articles on the subject of Britain’s railways, see Simon T. Abernethy, ‘Opening up the Suburbs: Workmen’s Trains in London 1860–1914’, Urban History 42, no. 1 (2015): 70–88; an article which stands out by virtue of its attempt to connect transport and demography; an approach we attempt to follow. By contrast, The Journal of Economic History contains a range of work exploring the economic significance of railways, and is at the time of writing edited by a prominent railway historian. ↑

5. Monika Krause, Model Cases: On Canonical Research Objects and Sites (University of Chicago Press, 2021), see chapter 1 on changing predilections among American scholars in the twentieth century for focusing on Chicago or Los Angeles as emblematic of the ‘urban’, as well as the consequences. For greater detail on the former, see Thomas F. Gieryn, ‘City as Truth-Spot: Laboratories and Field-Sites in Urban Studies’, Social Studies of Science 36, no. 1 (2006): 5–38. ↑

6. Friedrich Engels, The Condition of the Working Class in England in 1844 ([1845] London: Sonnenschein & Co., 1892), esp. ch 2. ↑

7. E.P. Hennock, ‘Concepts of Poverty in the British Social Surveys from Charles Booth to Arthur Bowley’, in The Social Survey in Historical Perspective, 1880-1940, ed. Martin Bulmer et al. (Cambridge: Cambridge University Press, 1991): 189-216. ↑

8. Patrick Geddes, Cities in Evolution: An Introduction to the Town Planning Movement and to the Study of Civics (London: Williams & Norgate, 1915), 26-7.

   ↑

9. Edward Thomas, In Pursuit of Spring ([1914], Dorset: Little Toller, 2016), 46-8, 50-1. ↑

10. Geddes, Cities, 34. ↑

11. David Cannadine, ‘Residential Differentiation in Nineteenth-Century Towns’, in The Structure of Nineteenth Century Cities, ed. James Henry Johnson and Colin G. Pooley (London: Croom Helm, 1982). ↑

12. Cannadine, ‘Residential Differentiation’, 237 (citing Colin Pooley). ↑

13. Cannadine in 1982 lamented ‘the problem of the historian when he [sic] wheels his trolley around the supermarket of ideas in search of theoretical packages or bodies of data from other disciplines which may help him order his own inchoate mass of evidence: namely that in most other disciplines, from whose findings and theories he may wish to borrow, there is no agreement as to just what those theories and findings are.’ ibid, 238. ↑

14. John R. Kellett, The Impact of Railways on Victorian Cities (London: Routledge & Kegan Paul, 1969); the five cities are London, Birmingham, Manchester, Liverpool and Glasgow. ↑

15. Gareth Stedman Jones, Outcast London: A Study in the Relationship between Classes in Victorian Society (Harmondsworth: Penguin, 1971). ↑

16. James Joyce, Ulysses, (London: Penguin, 2000; 1922), 837-8. ↑

17. According to Alan Jackson, London estate agents and transport executives agreed that the optimum catchment area around suburban stations was the ‘magic half-mile’: Alan Jackson, Semi-Detached London: Suburban Development, Life and Transport, 1900-39 (London: Allen & Unwin, 1973), 220-21, 241-42. ↑

18. M. Forster, Howard’s End ([1910] London: Penguin, 1989), 27. ↑

19. Kellett, Impact, 293-5; Richard Dennis, English Industrial Cities of the Nineteenth Century: A Social Geography (Cambridge: Cambridge University Press, 1984), 127-32; David A. Reeder, Charles Booth’s Descriptive Map of London Poverty 1889 (London: London Topographical Society, 1984), citing Charles Booth, Life and Labour of the People in London, Third Ser. (London: Macmillan, 1902), 1:192.

    ↑

20. On the view from the railway carriage in general, see Schivelbusch, The Railway Journey; on slum clearances, see Kellett, The Impact of Railways on Victorian Cities, chs 10 and 11, and witness Friedrich Engels: ‘Of late, the Liverpool railway has been carried through the middle of them, over a high viaduct, and has abolished many of the filthiest nooks; but what does that avail? Whoever passes over this viaduct and looks down, sees filth and wretchedness,’ Condition of the Working Class, 92.

    ↑

21. Dyos and Reeder, ‘Slums and Suburbs’. E.g. Charles Booth, ‘Condition and Occupations of the People of East London and Hackney, 1887’, Journal of the Royal Statistical Society 51, no. 2 (1888): 282; George Gissing, The Nether World ([1889], London: J.M. Dent, 1973), 164.

    ↑

22. H. J. Dyos, Victorian Suburb: A Study of the Growth of Camberwell (Leicester: Leicester University Press, 1973), 138-168.

    ↑

23. Martin Daunton, State and Market in Victorian Britain: War, Welfare and Capitalism (Woodbridge: Boydell & Brewer, 2008), 92.

    ↑

24. George Gissing, New Grub Street ([1891], London: Penguin, 1985), 179. ↑

25. Kasra Hosseini, Katherine McDonough, Daniel van Strien, Olivia Vane, Daniel C.S. Wilson, ‘Maps of a Nation? The Digitized Ordnance Survey for New Historical Research’, Journal of Victorian Culture 26, no. 2 (2021): 284–99. ↑

26. Mariona Coll Ardanuy et al., ‘Station to Station: Linking and Enriching Historical British Railway Data’, Computational Humanities Research 2021 (Amsterdam: CEUR Workshop Proceedings, 2021): 249–65, http://ceur-ws.org/Vol-2989/long_paper29.pdf and the dataset itself: Mariona Coll Ardanuy et al., ‘StopsGB: Structured Timeline of Passenger Stations in Great Britain’ (2021), https://doi.org/10.23636/wvva-3d67. ↑

27. K. Schurer and E. Higgs, Integrated Census Microdata (I-CeM), 1851-1911. [data collection] (UK Data Service, 2020) SN: 7481, DOI: 10.5255/UKDA-SN-7481-2; K. Schurer and E. Higgs, Integrated Census Microdata (I-CeM) Names and Addresses, 1851-1911: Special Licence Access. [data collection]. 2nd edition (UK Data Service, 2022) SN: 7856, DOI: 10.5255/UKDA-SN-7856-2. ↑

28. L. Shaw-Taylor and E.A. Wrigley, ‘Occupational Structure and Population Change’, in R. Floud, J. Humphries and P. Johnson, eds., The Cambridge Economic History of Modern Britain: Volume I, 1700–1870 (Cambridge, 2014), 53–88; For a list of Campop’s working papers, see https://www.campop.geog.cam.ac.uk/research/occupations/outputs/preliminary/. ↑

29. Bogart et al, ‘Railways, divergence, and structural change’; Dan Bogart, ‘The transport revolution in industrializing Britain’, in R. Floud, J. Humphries, P. Johnson, eds. The Cambridge Economic History of Modern Britain: Volume I, 1700-1870. 4th ed. (Cambridge: Cambridge University, 2014); X. You, D. Bogart, E. Alvarez, A.E.M. Satchell and L. Shaw-Taylor, ‘Transport Development and Urban Population Change in the Age of Steam: A Market Access Approach’, Campop Transport Working Papers 9, https://www.campop.geog.cam.ac.uk/research/occupations/outputs/preliminary/marketaccessandsteam.pdf. ↑

30. T. Lan and P. Longley, ‘Geo-Referencing and Mapping 1901 Census Addresses for England and Wales’, ISPRS International Journal of Geo-Information 8, no. 8 (2019): 320; T. Lan and P. Longley, ‘Urban Morphology and Residential Differentiation across Great Britain, 1881–1901’, Annals of the American Association of Geographers 111, no. 6 (2021): 1796-1815. For alternate methods of working with historic census data at a sub-parish level, see Niall Cunningham & Ian Gregory, ‘Hard to miss, easy to blame? Peacelines, interfaces and political deaths in Belfast during the Troubles,’ Political Geography 40 (2014): 64-78. ↑

31. For a full discussion of the geo-coding method, see J. Rhodes, ‘Geo-coding historic census data: a new open methodology’, Historical Methods, forthcoming. CensusGeocoder pipeline code: https://github.com/Living-with-machines/CensusGeocoder ↑

32. https://www.ordnancesurvey.co.uk/business-government/products/open-map-roads ↑

33. “GB1900 Gazetteer” made available by the GB1900 Project. We acknowledge the Great Britain Historical GIS Project at the University of Portsmouth, the GB1900 partners and volunteers. https://www.pastplace.org/data/#tabgb1900 ↑

34. Rhodes, ‘Geo-coding historic census data’, forthcoming. ↑

35. See Rhodes ‘Geo-coding historic census data’, forthcoming. ↑

36. James Johnson and Colin G. Pooley, ‘The Internal Structure of the Nineteenth-Century British City: An Overview’, in The Structure of Nineteenth Century Cities, James Johnson and Colin G. Pooley, eds. (London: Croom Helm, 1982), 9. ↑

37. E.g. J.B. Harley, ‘Maps, Knowledge and Power’, in The New Nature of Maps: Essays in the History of Cartography, ed. Paul Laxton (Baltimore: Johns Hopkins University Press, 2002), 52. ↑

38. Hosseini et al. ‘Maps of a Nation?’, 286. ↑

39. Kasra Hosseini, Daniel C.S. Wilson, Kaspar Beelen, and Katherine McDonough, ‘MapReader: A Computer Vision Pipeline for the Semantic Exploration of Maps at Scale’, Proceedings of the 6th ACM SIGSPATIAL International Workshop on Geospatial Humanities (ACM: New York, 2021): 8-19. ↑

40. https://huggingface.co/Livingwithmachines/mr_resnest101e_timm_pretrain ↑

41. Kasra Hosseini, Daniel C.S. Wilson, Kaspar Beelen, and Katherine McDonough. ‘Mapreader_data_sigspatial_2022’, Zenodo, October 5, 2022, https://doi.org/10.5281/zenodo.7147906. For details on the dataset creation using MapReader see https://github.com/Living-with-machines/MapReader/wiki/GeoHumanities-workshop-in-SIGSPATIAL-2022. These are the same sheets from which the historical street data was transcribed in the GB1900 project. Some automatic post-processing of the predicted railspace results removed isolated patches (e.g. railspace patches with no neighbouring railspace patch within 205 metres were unclassified as railspace). 16,439 sheets produced about 30.5 million patches. This resulted in 487,360 railspace patches, or about 1.3% of all patches. It is important to state that this dataset is not error free: after fine-tuning a “weak model” (e.g. it was not trained using a large number of labelled patches), we used the results to visually identify systematic errors in the predicted patch labels. We then used annotation targeted on the areas of sheets containing such errors to add to the labelled dataset. ↑

42. M. H. Cobb, The railways of Great Britain: a historical atlas at a scale of 1 inch to 1 mile (Shepperton: Ian Allan Publishing, 2006). The Campop railway datasets are based largely on the Cobb atlas: J. Marti-Henneberg, M. Satchell, X. You, L. Shaw-Taylor, E. Wrigley, 1881 England, Wales and Scotland Rail Lines. [data collection]. UK Data Service, 2018. SN: 852993, DOI: 10.5255/UKDA-SN-852993; J. Marti-Henneberg, M. Satchell, X. You, L. Shaw-Taylor, E. Wrigley, 1861 England, Wales and Scotland Rail Lines. [data collection]. UK Data Service, 2018. SN: 852992, DOI: 10.5255/UKDA-SN-852992; and M. Satchell, E. Wrigley, L. Shaw-Taylor, X. You, J. Henneberg, 1851 England, Wales and Scotland Rail Lines. [data collection]. UK Data Service, 2018. SN: 852991, DOI: 10.5255/UKDA-SN-852991. ↑

43. Coll-Ardanuy et al, ‘Station to Station’; https://rchs.org.uk/railway-passenger-stations-in-great-britain-a-chronology/, version 5.02 released September 2020 by the Railway and Canal Historical Society [last accessed 14 September 2021]. ↑

44. Currently, we use 0.50% confidence score as the cut-off level for including a station in the dataset. ↑

45. StopsGB collapses the differences between passenger stations (e.g. urban, suburban, mainline), although we have information about which company operated trains from each station. ↑

46. The method for calculating the railspace score differs depending on whether the street geometry is an OS Open Roads line or a GB1900 point. For OS Open Roads, we split the road into a series of points at each vertex (whenever the road changes direction), calculating the proportion of MapReader patches identified as railspace within 100m radius of each point. The railspace score for the street is the mean of these scores. For GB1900, we simply calculate the proportion of MapReader patches identified as ‘railspace’ within 100m radius of the GB1900 point. ↑

47. B. Seebohm Rowntree, Poverty: A Study of Town Life (London: Macmillan, 1901), xi, 14; also Charles Booth, Labour and Life of the People of London: Volume 1, east London (London: Williams & Norgate, 1889), 60. ↑

48. The schema for Scotland lacks two transitional categories found in the England and Wales schema. To harmonise the schemas for Great Britain, we have classified ‘Urban transition’ parishes as ‘Urban’ and ‘Rural transition’ parishes as ‘Rural’. H. Smith, R. Bennett, and D. Radicic, ‘Towns in Victorian England and Wales: A new classification’, Urban History 45, no. 4 (2018): 568-594. doi:10.1017/S0963926818000020; Harry Smith and Robert J. Bennett, ‘Urban-Rural Classification using Census data, 1851-1911’, Drivers of Entrepreneurship and Small Business, Working Paper 6 (2017), visited 26 April 2022. ↑

49. The mean number of live-in servants per household follows the same trend, see Figure 3.A.1. ↑

50. Alan A. Jackson, Semi-Detached London: Suburban Development, Life and Transport (London: Allen & Unwin, 1973), 21. ↑

51. For details of the I-CeM occupational codings used to develop our clusters see Table 3.A.3.

    ↑

52. Walking remained a common mode of transport for all workers, and especially those clerks who could live in the inner suburbs, of whose number 22,000 lived in Camberwell by 1901, see Dyos & Reeder, ‘Slums and Suburbs,’ 368, 371. ↑

53. H.J. Dyos, Victorian Suburb: A Study of the Growth of Camberwell (Leicester: Leicester University Press, 1961), 69-74, 109-13. ↑

54. George and Weedon Grossmith, The Diary of a Nobody ([1892], Project Gutenberg eBook, 1997), chapter 1. ↑

55. Simon Abernethy, ‘Class, Gender and Commuting in Greater London, 1880-1940’, (Unpublished PhD thesis, University of Cambridge, 2015), Chapter 6, ‘On the buses’. ↑

56. The Manchester Ship Canal, which served Ordsall’s Salford Docks complex, only opened in 1894. ↑

57. Robert Roberts, The Classic Slum: Salford Life in the First Quarter of the Century (Manchester: Manchester University Press, 1971). ↑

58. Roberts, Classic Slum, 3. ↑

59. Lady Bell, At the Works: A Study of a Manufacturing Town (London: Edward Arnold, 1907), 14, 43, and 215, https://archive.org/details/atworksstudyofma00bellrich. ↑

60. Places (a mixture of parishes and sub-parish units) derived from I-CeM’s ParID variable. We excluded locations with fewer than 10 geocoded streets and 180 geocoded people. To aid graphic representation, we also filtered out remote rural locations where the average distance to a station was over 15km. ↑

61. LSE Library, BOOTH/B/349, 1897, George H. Duckworth's Notebook: Police and Publicans District 10 [Bethnal Green East], District 15 [South West Islington], District 17 [Upper Holloway], 179-81 ↑

62. LSE Library, BOOTH/B/348, 1897, George H. Duckworth's Notebook: Police and Publicans District 14 [West Hackney and South East Islington], District 15 [South West Islington], District 16 [Highbury, Stoke Newington, Stamford Hill], 235-9. ↑

63. LSE Library, BOOTH/B/349, 1897, George H. Duckworth's Notebook: Police and Publicans District 10 [Bethnal Green East], District 15 [South West Islington], District 17 [Upper Holloway]], 21-3. ↑

64. Jerry White, The Worst Street in North London: Campbell Bunk, Islington, between the wars (London: Routledge, 1986), 11-14. ↑

65. Dyos, Victorian Suburb, 112. ↑

66. LSE Library, BOOTH/B/365, 1899, George H. Duckworth's Notebook: Police District 32 [Trinity Newington and St Mary Bermondsey], District 33 [St James Bermondsey and Rotherhithe], District 34 [Lambeth and Kennington], District 35 [Kennington (2nd) and Brixton], District 41 [St Peter Walworth and St Mary Newington], District 42 [St George Camberwell], District 45 [Deptford], 103-5. ↑

67. White, Worst Street, 8-70. ↑

68. Northern Weekly Gazette, 3 Aug. 1895, https://www.britishnewspaperarchive.co.uk/viewer/bl/0003074/18950803/149/0007. ↑

69. Roberts, Classic Slum, 3. ↑

70. Dennis, Industrial Cities, 110-32; Geddes, Cities, 24, 34-45. ↑

71. A publicly available code repository is on GitHub [link pending]. Data is available here [link pending]. ↑