Railways and Population Change
in
Industrializing England 

An Introduction to Historical GIS

Robert M. Schwartz
Department of History
Mount Holyoke College

 June 22, 1999

Draft 

Copyright Robert M. Schwartz

Chapter 1

Maps, Concepts, and Numbers

The Map Hunt

It is Tuesday morning.  I need to prepare for class.  I go to my bookshelf to find information on the economic history of Britain during the industrializing era. I take down the second edition of Peter Mathias’s classic text, The First Industrial Nation. An Economic History of Britain 1700-1914.  A glance at the table of contents draws me immediately to Chapter 10 in Part II on “The Railways.” It is short but informative with a useful table showing the miles of railways proposed and opened annually from 1832 to 1850. As I read on, I look for a map of the lines. There is none.

Puzzled, I return to the Table of Contents to find the maps in other chapters.  I find the list of tables, the list of figures, but . . .no list of maps. No maps?  Hmmm?  I look again and run my eyes slowly down the list of figures. Surely, I say to myself, the maps are included here. Not so. Disappointed, I repair to the appendices. They are impressive: richly documented with 40 tables! . . . And yet, there’s not a single map?  How can this be? 

            This surprising discovery prompts some reflections. I do not conclude that the book is second-rate. After all, it is reputed a classic, and the author is an economic historian of the first order.  The book attests to his admirable talent of combining quantitative data with economic analysis, historical detail, and lucid exposition.  Published in 1983, it also reveals a predilection for statistics at the aggregate level of the nation state.  These are statistics that are non-spatial in character, and it seems clearly to have been the author’s intention to focus on them, consciously excluding issues of geographic variation in social and economic development.

             Historical geographers, of course, have long attended to questions of this kind.  Social historians such as myself are at pains in their research to acquire “a sense of place” through the study of maps and contemporary descriptions.  That The First Industrial Nation and a number of other studies of the period proved to be “mapless” came as a surprise nonetheless because my interests and those of other historians and social scientists have changed.

 The personal computer has opened new kinds of research opportunities, ranging from access to on-line library resources to the use of images as primary historical sources in essays, books, and presentations, be they on paper or in digital form.  More specifically, the appearance on desktop computers of powerful GIS programs has greatly enhanced our ability to investigate geographic questions.  Previously the mapping of quantities in space and across time was so laborious and costly that the questions deemed feasible for study were much more constrained than they are today.  Using the technology available now, historical geographers and historians can map—and re-map—information with a speed and flexibility unavailable a decade ago in their search to identify and explain geographic patterns.  Consequently, new questions are arising and old ones are being re-examined as well. The geographic dimension in contemporary social inquiry and historical studies is being rediscovered and expanded in new directions.  Vive les cartes? Vive les maps!

The purpose of this booklet is to illustrate how this renewed interest and the technology of Geographic Information Systems go hand in hand in the revisiting of old questions and in the charting of new explorations.  The topic that we shall explore is a problem in environmental history:  the impact on the human and physical environment of new technology.  More specifically, we shall use GIS methods to examine the relationship between the development of the railway system and population change in industrializing England and Wales from 1851, the date of an accurate census, to 1914, when the rail system reached its peak in geographic coverage. 

 This study is possible only because of the pioneering work of Professor Humphrey Southall and Ian Gregory of the Geography Department at Queen Mary and Westfield College in London.  Working together, and often with a team of other specialists, they have created and continue to develop the Great Britain Historical Data Base.  To my knowledge, this is one of the largest and most advanced collections of historical GIS data in existence.  One of its wondrous features is a set of dynamic boundary files for the Census Registration Districts of England and Wales.  These files, or “coverages” include virtually all changes in the district boundaries from 1851 to 1911, the result of a vast effort of research and programming. Included also are attribute data from the decennial census enumerations of the same period[1]. With this expression of my appreciation and acknowledgement we can now consider the methods and technology of GIS.

What is GIS?

            Geographic Information Systems is the combination of computer-assisted cartography and data analysis. The abbreviated form, GIS, is typically used in writing and conversation. Each of the terms conveys a meaningful part of the whole:

 ·                    Geographic: geographic and spatial entities and realities, such as the location, size, and shape of ponds within a given area, or of rail lines and rail stations within districts of a state or country.

·                    Information: systematic collections of factual information or data as well as the use and the meaning of the data.

·                    Systems: the computer technology used to process, analyze, and display the data, including computers as well as the programs to make them function.

A GIS based research project is a process involving a number of phases or aspects. It begins with the collection of information and proceeds as follows. 

·                    Collection: the conversion of geographic information from maps to digital form for use in computer programs; the collection of data regarding the attributes of geographic entities, such as the water quality of ponds in an area.  

In addition to the data collected by Southall and Gregory, this part of the work included the digitizing of maps showing the rail lines as they existed in England and Wales in 1845, 1854, 1876, and 1914. Once in digital form, these data were combined with geographic coverages of the Census Registration Districts in England and Wales for each decade from 1851 to 1911.  The coverages were joined with attribute data on the districts.  Included for each decennial census year (1851, 1861, etc.) and for each district were the size of the population, the area of the district, and the estimated decennial change in the population size that is attributable to migration. 

·        Data Processing and Management: the storage and organization of the data in computer files and the retrieval of the data for analysis and display.[2]

·        Analysis: the investigation of research questions, those posed in designing the project  and those arising in the course of the study. Analysis refers to dividing larger questions into smaller components, searching for patterns in the data, interpreting their meaning, and identifying their likely explanations. In GIS research, the basic questions are geographic and spatial. What is where? Why is it there? Our key questions concern the geographic development and influence of railways.  In mid 19^th century England and Wales, where were the major rail lines located? What network of transport did the lines form?  And what combination of economic and geographic factors explains the geographic structure and growth of this network?·        Display and Presentation: the creation of maps, tables, and other displays of pertinent information that provide the support for the conclusions reached in the analysis; the presentation and reporting of the study results in a paper, essay, or book. In what follows, I shall concentrate exclusively on the last two components of GIS research, that is, on the analysis, display, and interpretation of data.  The investigation centers on the growth and impact of the railway in England and Wales from 1851 to 1914.  To proceed, we need to acquaint ourselves with the definition and application of some key geographic and statistical concepts.  It is fitting to begin by noting the significance of spatial as opposed to non-spatial analysis.  A good way to start, of course, is to look at a map! 

Map 1.1 England and Wales: Selected Census Registration Districts in 1861

            Map 1.1displays the name and location of eight Registration Districts, which were administrative units established for census enumerations, beginning in 1841. The map also shows the boundaries of the districts—632 in all.  This representation embodies the three components of GIS applications: points, lines, and polygons.·

Points: a point designates the coordinates of a geographic location; it has no spatial extension but marks a spatial position. In the map above the point features are hidden but are used to position the names of the districts.·

Lines:  a line is set of connected points that form a segment. When the geographic information on a paper map was converted to digital form, the district boundaries were built first by making points into line segments, and then by combining the segments to form a polygon. Lines possess the spatial properties of length and width as well as elevation. As we shall see, “the iron roads” of railways are represented by lines in maps that follow.

 ·        Polygons: a polygon is a closed, connected set of line segments. Each of the census registration districts is a polygon.

 There are other geographic properties associated with this map that we can consider next, noting how they differ from non-spatial attributes of geographic entities. The table below displays both spatial and non-spatial attributes. In a GIS program, a table like this serves to store and retrieve the information needed to create maps and tabular displays. Reading from left to right, the first three columns contain examples of spatial attributes—the Shape, Area, and Perimeter of the registration districts. The non-spatial attributes come next, beginning with the Name of the districts.

Table 1.1 Spatial and Non-Spatial Attributes

Spatial Attributes

Non-Spatial Attributes

Shape	Area km^2	Perimeter km	Name	County	Population 1861	Level of Pop. Size	Pop. Density 1861
Polygon	215	939	Alton	Hampshire	12063	Medium	56
Polygon	100	663	Barton Upon Irwell	Lancashire	39038	Medium	391
Polygon	393	1158	Bootle	Cumberland	5880	Small	15
Polygon	111	568	Haslingden	Lancashire	69781	High	630
Polygon	528	1344	Machynlleth	Montgomeryshire	12395	Medium	23
Polygon	5	123	Plymouth	Devon	62599	Large	12388
Polygon	1	51	St. Giles	London	54076	Medium	55164
Polygon	228	1160	Wolverhampton	Staffordshire	126902	Very Large	557
Scale of Measure-ment	Num.	Numerical	Nominal	Nominal	Numerical	Ordinal	Numerical

            As indicated in the last row of the table, the information that defines an attribute is of one kind or another, depending upon the properties of the information. In data analysis these properties are designated as an attribute’s scale of measurement.  In this study, we will distinguish three scales of measurement.

·        Nominal scale: the definition of a variable or attribute by a qualitative, as opposed to a quantitative, characteristic. The name of a district and the county to which it belongs are examples of nominal attributes.

·        Ordinal scale: the identification of an attribute in terms of its position within an ordered set of categories.

In Table 1 the Level of Population Size is measured on an ordinal scale having four categories that range from small to very large. Each of the districts is identified as belonging to one of the four categories. Classified this way, the values of attributes carry quantitative information about the relative size of district populations, but they do not carry information about the magnitude of the population differences between districts. The districts are ranked by size, but they cannot be compared further through the use of addition, subtraction, and so forth, as can be done with what we shall call numerical variables or attributes.

·        Numerical scale: the definition of an attribute in terms of a measurable, quantitative characteristic.  The area of a district, its population and populations density are good examples of numerical attributes.  Unlike their ordinal cousins, they can be compared mathematically, using addition, subtraction, multiplication, and so forth.

With these pre-luminaries in mind, let us consider next something further about the properties of geographic features.  Map 1.2 below will serve as basis for our continuing study, so it merits some introductory comments.

Geographic Properties and Attributes

 This map displays the names and locations of a sample of registration districts.  It is a random sample of districts that comprises 5 percent of the population of England and Wales in 1861. A sample of this kind has the advantage of being sufficiently representative so that the

Map 1.2.  Sample of Registration Districts in 1861

the statistical results of our work will be good estimates of the results that we would obtain if we analyzed all of the districts. Our sample consists of 32 districts, drawn at random from a total of 632 such units.  The entire collection of districts makes up, in statistical terms, the population of interest for this study.  In the next chapter, we shall work with maps based on the entire “population.”  But for the moment, we’ll work with the sample.  Working with a small collection of cases makes it easier to grasp the analytical methods and to make calculations by hand without a computer.

            An understanding of geographic properties will help us read and interpret our maps. To that end, a good way to proceed is to use the geographic concepts to pose and answer questions about our sample districts.  The table below will guide us through the process of applying the concepts.  Reading from left to right, we first familiarize ourselves with the concept. Then we move on to consider example questions. Finally, we use the map indicated to determine the answers.  In some cases the answers are given; for other questions the reader should search out the answers on his or her own.

Table 1.2  Geographic Properties

Geographic Properties		Example Questions		Answers (Based on Map 1.2.  Sample of Registration Districts in 1861)
Location ·        What is where? Where are the x features? ·        Which geographic entities are in a given region of the country?		Which Registration Districts (RDs) border the sea?  Which RDs are in the southeastern part of the country?		Cockermouth, Bootle, Pwllheli. Dolgelley, Machynlleth, Plymouth, Christchurch, Ringwood, Rochford
Proximity How close is x to y?  · How many  x  are within 30 kilometers of y?		Which two RDs are the closest to Greater London? (Refer to Map 1.3.)  According to the scale of the Map 1.3, what is the approximate distance in kilometers between Newton Abbot and Forehoe?  How many RDs are within 80 kilometers of Wolverhampton? (Use Map 1.3)		Hatfield (about 33 km) Rochford (about 60 km)  About 400 km
Size ·        How large is feature x? ·        Do small x features form a pattern?		Which RD appears to have been the largest in area? Did the smaller sized RDs tend to cluster in a particular region of the country?		Cockermouth
Adjacency ·        What is next to x? ·        Do x and y features share a common boundary in region z?		Which RDs shared a common boundary? Did RDs with shared borders tend to cluster near the sea as opposed to inland?		Cockermouth and Bootle Okehampton and Newton Abbot Dolgelley and Machynlleth Ringwood and Christchurch
Connection ·        Are x and y connected by z?  ·        Do those x and y features that are connected by z form a pattern?		Which RDs were more or less directly connected by rail line to London? (Use Map 1.5.) Were the RDs connected to London by rail other significant urban centers? (Use Map 1.4.)		Hatfield Ringwood Plymouth Wolverhampton Owestry
Containment ·        Does z fall within x? ·        How large is z within x?		Which RDs fall in the lowest level of population density, measured in persons per square kilometer? (Use Map 1.3.)  Which RDs fall within the 81 to 128 level of population density? (Use Map 1.3.)		Haltwhistle Bootle Dolgelley Machynlleth
Pattern ·        Does z fall within x usually when y is also present in x? ·        More often than not, is there a connection between x and y in one region but not in another?	What pattern or relationship seems to have existed between major population centers and the extent of the rail lines in 1854? (Use Map 1.4.) Which region was best served by rail service? London and the Southeast? The Midlands and lower Northwest? The Northeast? The Southwest? (Use Map 1.4 or Map 1.5.)		The Midlands and the lower Northwest.

Map 1.3 Population Density in Sample Registration Districts, 1861

Map 1.4 Urban Centers and Rail Lines, 1861

Map 1.5 Sample Districts and Rail Lines, 1861

Non-Spatial Attributes

Geographic features often contain characteristics that are non-spatial but important to study nonetheless, both on their own and as complements to geographic information.  As we saw in Table 1, non-spatial attributes are included in our GIS data base.  As we did when familiarizing ourselves with geographic properties, let us review some characteristics of non-spatial attributes, using the table below and the data in Table 1.4.

Table 1.3.  Population Size: Properties of a Non-Spatial, Numerical Attribute

Size or Magnitude ·        What is the maximum value of a set of numbers	Which of the RDs has the largest population, i.e., the maximum value of this attribute?	Wolverhampton in the County of Staffordshire
·        What is the minimum value?	Which has the smallest population, i.e., the minimum value of this attribute?	Ringwood in the County of Hampshire
Range: the difference between the minimum and maximum values.	What is the range of the population size for the sampled districts?
Center: the central tendency of a distribution of numbers, frequently summarized by either the mean or the median. ·        The mean is the average of a list of numbers ·        The median is the mid-point in a list of numbers arranged in order of magnitude	What was the mean population size of the 32 sample districts? What was the median population size of the districts? (Use Table 1.5)

Table 1.4 Sample Districts Attributes

Name	Area	County	Population 1861	Population. Density 1861 (persons per km^2)	Rail Density (km per km^2)
					1845	1854	1876
Alton	215	Hampshire	12063	56		4	61
Barton Upon Irwell	100	Lancashire	39038	391	82	174	328
Bootle	393	Cumberland	5880	15		62	90
Cheltenham	108	Gloucestershire	49792	462	14	13	13
Christchurch	129	Hampshire	10438	81		48	121
Cockermouth	690	Cumberland	41292	60	15	29	101
Dolgelley	639	Merionethshire	12482	20			86
Forehoe	156	Norfolk	12818	82		166	166
Haltwhistle	394	Northumberland	6693	17	45	90	90
Haslingden	111	Lancashire	69781	630		108	108
Hatfield	121	Hertfordshire	8400	69		80	214
Knaresborough	165	West Riding	17176	104		364	177
Lexden	284	Essex	22950	81	55	85	103
Lutterworth	240	Leicestershire	15515	65	65	74	76
Machynlleth	528	Montgomeryshire	12395	23			114
Newton Abbot	486	Devon	59063	122		59	102
Northwich	260	Cheshire	33338	128	93	94	189
Okehampton	538	Devon	18580	35			53
Oswestry	336	Shropshire	23817	71		71	149
Plymouth	5	Devon	62599	12388		469	1214
Pwllheli	379	Carnarvonshire	20908	55			43
Richmond (Surrey)	21	Surrey	18802	888	235	236	236
Ringwood	149	Hampshire	5357	36		109	162
Rochford	377	Essex	18282	49			39
Shaftesbury	149	Dorset	12986	87			56
South Molton	590	Devon	19209	33		9	49
St. Giles	1	London	54076	55164	641	738	738
Stamford	216	Lincolnshire	18213	84		108	243
Thrapston	225	Northamptonshire	14065	62		56	81
Warminster	240	Wiltshire	15942	67		9	77
Winchcomb	231	Gloucestershire	10082	44			11
Wolverhampton	228	Staffordshire	126902	557	24	143	168

Table 1.5 Rank Order of Population, 1861

Table 1.5 rearranges two columns from the preceding table by reordering them from the smallest to the largest in population size.  This makes it much easier to determine the median because we count down to the middle case.  Because there are an even number of observations, we have to calculate the median by taking the average of the observation on either side of the theoretical middle, in this example the 16^th and 17^th numbers on the list.  Hence, the median = (18213 + 18282) / 2 = 18248	Name	Population 1861
	Ringwood	5357
	Bootle	5880
	Haltwhistle	6693
	Hatfield	8400
	Winchcomb	10082
	Christchurch	10438
	Alton	12063
	Machynlleth	12395
	Dolgelley	12482
	Forehoe	12818
	Shaftesbury	12986
	Thrapston	14065
	Lutterworth	15515
	Warminster	15942
	Knaresborough	17176
	Stamford	18213
	Rochford	18282
	Okehampton	18580
	Richmond (Surrey)	18802
	South Molton	19209
	Pwllheli	20908
	Lexden	22950
	Oswestry	23817
	Northwich	33338
	Barton Upon Irwell	39038
	Cockermouth	41292
	Cheltenham	49792
	St. Giles	54076
	Newton Abbot	59063
	Plymouth	62599
	Haslingden	69781
	Wolverhampton	126902

Table 1.3 (Continued)

Frequency Distribution: an arrangement of the numerical values of an attribute showing how frequently they occur.

What is the distribution of population size in 1861?

Study Figure 1.1.

            Figure 1.1is a histogram displaying the distribution of the attribute, population size, in 1861. The values have been reorganized by classifying them into a set of defined categories. The first category is defined as values less than or equal to zero; the last or highest category, as values greater than 130,000. The others consist of intervals of 1000 each, beginning with the interval between 0 and 1000 (greater than zero, up to and including 1000). The scale of measurement here is ordinal. To facilitate the analysis, the original numerical attributes have been converted to ordinal ones. In the process, we lose information (the ability to be more precise, to calculate a mean, and so forth) and sacrifice complexity in order to construct a model, or a simplification, of the underlying distribution.  Such simplifications often prove revealing.

What does this histogram reveal? Clearly, the majority of districts have populations of between 10,000 to 20,000 people.  A glance at the  y axis and the number at the top of the largest bar shows that 16 of the districts—50 percent of the sample—are of this size.  Taking this category as “an anchor point,” we see that the distribution falls away to the right, as the number of districts within each category declines from the peak of 16 to 1 in the range between 120000 and 130000.  In statistical terms, the distribution is skewed to the right.

Figure 1.1.  Histogram of Population Size in 1861

This shape is common to many kinds of economic and demographic data. As in the past, so today, distributions of income and wealth usually resemble this structure. The incomes of most people fall in the lower range, while a small or tiny minority take home the “big bucks.” Many people work at McDonalds but few have the income of a Bill Gates. 

In the example before us, the great majority of districts contained populations in the lower range, while only one had more than 100,000 inhabitants.  This distribution reflects the evolving demographic landscape of industrializing and urbanizing England and Wales as a whole.

·        During the 19^th century the differences in size between the majority of communities and the few very large cities were enormous.  To cite one example, there were some 3,000 people living in the parishes of Blean near Canterbury (County of Kent),  compared to the nearly 3 million inhabitants of London around 1850.

·         By 1851, the majority of people in England and Wales lived in cities of 10,000 or more.

·        Nonetheless, cities over 100,000 in population continued to make up a minority of communities. 

This said, we need to bear in mind some characteristics of census registration districts.  They were not specific communities but administrative units that varied in physical size or area. Districts were centered on market towns or on wards within larger cities.  Hence, a rural district typically subsumed one or more villages while an urban district such as St. Giles in London contained only part of the entire city’s much larger population.

That districts were units of differing geographic size is another characteristic that deserves a comment.  Among the districts, this characteristic accounted for some of the variation in population size.  In some situations it is important to remove the effect that one attribute has on another in order to make standardized comparisons.  To neutralize the effect of unequal areas in our study, we can make comparisons using population density: the number of people per square kilometer. If we know the area and the population of a set of units, the calculation of density is straight forward. For each unit:

population density = population / area

We have already seen in Map 1.3 a standardized comparison showing different levels of population density among our sample districts. We shall return often to maps of this kind that display the geographic distribution of attributes that have been standardized by area.  In an exercise at the end of this chapter, the reader is asked to use the histogram above (Figure 1.1) as a model to create another one for the distribution of population density.  For the moment, however, let us return to our analysis of population size.

            In addition to a histogram, we can describe a distribution in terms of its center and spread.  Earlier we glanced at the mean and the median of a list of numbers, but a second look is needed now to understand the connection of these statistics with a frequency distribution. In this context, the concept of center carries the idea of typicality.  The mean and the median provide measures of the central or typical value of a distribution.

·        As a central value, the mean is the balance point in a distribution, the point at which the base of a histogram, like a teeter tooter, would remain balanced and level if it rested on the mean.

In Figure 1.1 the mean is the point that serves to balance the large cluster of districts in the lower range against the single district at the extreme right end of the distribution. Being a balance point, the mean is sensitive to extreme values. In this case the single extreme value in the 120000 to 130000 range serves to “pull” the mean away toward it.

·        Unlike the mean, the median is not sensitive to extreme values.

This “insensitivity” to extremes is reflected in the difference between median and the mean shown in Figure 1. At 18,248 the median population size of the sample is considerably smaller than the larger mean of 27, 154. Unmoved by the influence of extreme values, the median sits squarely in the middle position of an ordered set of numbers. It divides the set into two equal groups: Fifty percent of the numbers lie below the median; 50 percent lie above.

            Both measures of center are informative single-number summaries of a distribution. But the lesson here deserves underscoring: when the median and mean are quite different, this is a clear sign that the underlying distribution of values is skewed either to the right, as in this case, or to the left, as the distribution of ages at death would show. Few die young; most die in advanced age. Studying the mean and the median is a good habit to develop.

            After identifying the center of a distribution, we should note how the remaining values are arranged in relation to the center.  The term “spread” refers to this characteristic, and there are several measures of spread that are linked either to the mean or to the median.  The standard deviation is the measure of spread used with the mean; percentiles and quartiles are typical measures used with the median.

·        The standard deviation is the average of the distances between the values in a distribution and the mean. A small SD tells us that the values are more closely clustered about the mean than the values of another distribution with a large SD.

The spread of values around the median is indicated by several other measures that are analogous to percentages.

·        Percentiles divide a list of numbers arranged in order of size into 100 parts. Hence, the 10^th percentile is the value in a list below which are 10 percent of all the values.  On an exam, the 95^th percentile in a list of scores is the number that stands above 95 percent of the other scores.  It’s an “A.”  The 50^th percentile is none other than the median.

·        Quartiles divide an ordered list of numbers into 4 parts, each part consisting of 25 percent of the list.  The first, or lower quartile is the value below which are 25 percent of the numbers. The third, or upper quartile, is the value below which are 75 percent of the numbers.

·        Quintiles divide an ordered list into 5 parts, each of which comprises 20 percent of the values in question. We shall return to this concept later when considering the various ways of dividing a distribution when making and interpreting maps.

Now familiar with some geographic and statistical concepts we can apply them to continue our investigation into social and environmental history.  It is with the growth of the railways in England and Wales that the next chapter begins.

Exercises

1. On a copy of Map 1.3, draw two intersecting lines that divide England and Wales into four regions.  The horizontal line should run from Dolgelley to Forehoe such that Wolverhampton is below the line. The vertical line should proceed from Christchurch due north so that it runs between Knaresborough and Haltwhistle.  Using these regions of Northeast, Northwest, Southeast, and Southwest, do you find a pattern of population density that varies by region? Explain.

2. On a second copy of Map 1.3, shift the lines in any direction that seems to make sense.  To what extent does this change the results that you found in question 1?  Is there a regional pattern?  What lesson do you draw from this comparison?

3. The table below shows how the properties of attributes differ according to the scale of measurement.  A knowledge of attribute properties is important in choosing appropriate statistical procedures in an analysis.  It makes no sense, for example, to try to determine the mean of a nominal attribute, such as Name or Gender. Sometimes a degree of precision in information can usefully be sacrificed to get a clearer picture of an underlying pattern, as we saw in the case of the histogram displaying the distribution of population size.  To make the histogram, a numerical attribute (population) was transformed into an ordinal version (Level of Population Size) so the result would highlight a clearer, if less precise, picture of the distribution. To familiarize yourself further with the differing kinds of attributes, first study the table. Then define at least one additional attribute for each of the 6 types represented below in the cells of the table.  The new attributes can come from the data presented or suggested here, or from any other context.

Table 1.6.  Attributes and Scales of Measurement

	Attributes
Scales of Measurement	Geographic/Spatial	Statistical/Non-Spatial
Nominal	Shape: square, rectangle, circle, ellipse Feature: point, line, polygon Adjacent: yes, no	Name of District County Gender
Ordinal	Proximity: near (1-10 km), in region (10-40 km), distant (> 40 km) Area: small, medium, large	Level of Population Size Income Level: low, medium, high

Numerical

Position in coordinates of longitude and latitude Area in square kilometers Elevation in meters Proximity in kilometers Position of Rail Lines

Population Size Population Density Income in Dollars Number of Migrants Density of Rail Lines Number of Rail Stations Length of Rail Lines

4. Use Table 1.4 and Figure 1.1 to create a histogram showing the distribution of population density among the 32 sample districts.  Describe the pattern you see and explain what you think it reveals about rural and urban life in 19^th century England and Wales.

5. Map 1.4 displays major urban centers and the rail system as they existed around 1860.  It allows one to take up questions about geographic connections and geographic patterns based on spatial connections.  Use the map to “test” and discuss the soundness of the following generalization.

The building of railroads required so much capital that rail companies were at pains to ensure that their large investments would eventually return a profit.  Competition among them for the rights to build one route or another was also stiff.  Under these circumstances companies looked first to establishing lines providing transport of natural resources to industrial centers and finished products to markets. 

6. Study the figures on Rail Density in Table 1.4. 

a) Describe the advantage of using this attribute as opposed to one that records the length of rail lines in each district.

b) Create three histograms that show the extent to which railways developed in the sample districts.

c) Briefly state how you would display your results on a map?

Chapter 2

Railways and Population Change, 1851-1914

The social and economic transformation of nineteenth-century England and Wales is the classic example of western industrialization and urbanization.  Viewed from the perspective of social and environmental history, this transformation provides an interesting way to examine the impact of new technology on past human and physical environments.  One far-reaching example is the steam-powered railroad system which grew to reach nearly all corners of England and Wales from its beginning in the 1830s to its apogee on the eve of World War I. The landscape of the Victorian City was a monument to the railway age, with its huge train stations and rail yards, together with the great earthworks and tunnels that the rail network required.  To its stations, moreover, came more and more individuals and families who were moving to town in search of better opportunities, leaving the countryside behind and villages in decline.

Did the railways facilitate migration from countryside to town? What was the timing, extent, and geography of rural depopulation? Did rural men and women migrate in similar patterns?  Thanks to GIS methods, all of these questions can be taken up more effectively now than was previously possible.  To pursue them, it is advisable to break down the issues into smaller components and examine them first.  Looking into the growth of the rail system is a good place to start.

The Growth of the Rail System

            The Railway Age began in 1830 with the opening of the line between the port city of Liverpool and Manchester, the center of the revolution in mechanized textile production. Although locomotives pulling coal had earlier steamed along parts of the Stockton and Darlington line which opened in 1825, it was the Liverpool-Manchester railway that was the first to be truly successful.  Reflecting the enthusiasm that marked the opening of this new marvel of the age, there were lords, ladies, and gentlemen in attendance to witness the Duke of Wellington commemorate the occasion.  The exhilaration of the day, though, was marred by accident.  When William Huskinson, a government minister and rail enthusiast, caught sight of the Duke, he went to greet him, but in crossing the track he was struck and killed by a passing locomotive.  Thus did the dangers of the rail age make their mark on this historic day.  Huskinson’s untimely end, however, was soon overshadowed by the excitement surrounding the transportation revolution, capturing the Victorian public’s imagination in short order.

            Writing of his first train journey in 1830, the Reverend Edward Stanley recalled the elation he and his companions felt.

No words can convey an adequate notion of the magnificence ( I cannot use a smaller word ) of our progress. At first it was comparatively slow; but soon we felt that we were GOING, and then it was that every person to whom the conveyance was new, must have been sensible that the adaptation of locomotive power was establishing a fresh era in the state of society.…

The most intense curiosity and excitement prevailed... and ... enormous masses of densely packed people lined the road, shouting and waving hats and handkerchiefs as we flew by them. What with the sight and sound of these cheering multitudes and the tremendous velocity with which we were borne past them, my spirits rose to true champagne height.[3]

            As the years wore on and the rail system grew rapidly, some contemporaries saw in the railroad the reflected image of the technical and moral progress they so cherished.  Samuel Smiles was one notable example.  A self-educated man, his many books celebrated the feats of civil and mechanical engineers and the Victorian virtues their stories embodied.  Writing in 1859, he described the railway locomotive, one of their great feats, as nothing less than a moral force for good. 

The iron rail proved a magicians' road. The locomotive gave a new celerity to time. It virtually reduced England to a sixth of its size. It brought the country nearer to the town and the town to the country... It energized punctuality, discipline, and attention; and proved a moral teacher by the influence of example.

            Not everyone shared such sentiments.  Others saw something more menacing in the changes ushered in by the iron roads and smoky locomotives.  In Domby and Son, Charles Dickens took aim at the relentless noise and ugliness created by the digging of tunnels and the erecting of vast earthworks in London.

The first shock of a great earthquake had, just at that period, rent the whole neighborhood to its centre. Traces of its course were visible on every side. Houses were knocked down; streets broken through and stopped; deep pits and trenches dug into the ground; enormous heaps of earth and clay thrown up; buildings that were undermined and shaking, propped by great beams of wood. Here, a chaos of carts, overthrown and jumbled together, lay topsy turvy at the bottom of a steep unnatural hill; there confused treasures of iron soaked and rusted in something that had accidentally become a pond. Everywhere were bridges that led nowhere; thoroughfares that were wholly impassable; Babel towers of chimneys; wanting half their height...carcasses of ragged tenements, fragments of unfinished walls and arches, and piles of scaffolding, and wilderness of bricks, and giant forms of cranes, tripods straddling above nothing. There were a hundred thousand shapes and substances of incompleteness, wildly mingled out of their places, upside down, burrowing in the earth, ...moldering in the water and unintelligible as any dream...In short, the yet unfinished and unopened railroad is in progress. . . .

            In Dickens’s view, the city, the people, and nature itself were under attack.  But, if protest failed to stop the onslaught in London, it might succeed elsewhere.  Such were the thoughts of the great poet, William Wordsworth, when he marshaled verse to protect Nature. One year after being named Poet Laureate of England, he campaigned to stop the opening of a line from Kendall to Windermere, the heart and soul of his beloved Lakes District.  Dreading the inflow of careless tourists and the noise and smoke of locomotives, he denounced the scheme in a sonnet that appeared in The Morning Post, on October 16, 1844[4].

Is then no nook of English ground secure 

From rash assault? Schemes of retirement sown 

In youth, and 'mid the busy world kept pure 

As when their earliest flowers of hope were blown, 

Must perish;--how can they this blight endure? 

And must he too the ruthless change bemoan 

Who scorns a false utilitarian lure 

'Mid his paternal fields at random thrown? 

Baffle the threat, bright Scene, from Orresthead 

Given to the pausing traveler’s rapturous glance: 

Plead for thy peace, thou beautiful romance 

Of nature; and, if human hearts be dead, 

Speak, passing winds; ye torrents, with your strong 

And constant voice, protest against the wrong. 

            As with the coming of the industrial age generally, so with the railway, contemporaries reacted strongly to the transformation that was occurring around them.  But no matter how elegant the protest, the growth of the rail system from the 1840s quickened its pace. The locomotive proved virtually unstoppable. 

The pace of growth can be seen in the following graph which charts the total miles of rail open from 1832 to 1913 (data for 1852-1860 is unavailable).  A surge of construction marked the 1840s, which continued from the 1860s on, although the pace of growth gradually slowed, particularly after the mid-1880s.  Railway companies were surprised in the 1840s by the strong growth in of passenger traffic, which was unanticipated: they expected that traffic in freight would be the chief source of revenues and profits. As the receipts show, their expectations were eventually met in the late 1850s.  But to judge from the receipts, the upward trend in passenger traffic closely followed that for freight.

Figure 2.1 Growth of the Railway System, 1832-1913

            So much for the aggregate trends. What about the geography of rail service as railways expanded on the national scale?  Which cities and regions were served first and more extensively than others?  We can proceed with these questions in a number of steps.  Using some selected cases from our sample, Table 2.1 displays a selection of data, the density of rail lines in districts at four dates from the 1840s to 1914.  Following a method we used earlier, we can recode, or re-classify, these data by creating a set of simpler categories and then plot the frequency counts in a histogram.  The recoded measures of density appear in Table 2.2.  The histogram in Figure 2.1 displays the results, not only for the sample, but for the entire population of districts. 

Table 2.1 Data:  The Growing Density of Rail Lines in Registration Districts, 1845-1914

	Rail Density in km/km^2
Registration District	1845	1854	1876	1914
Alton		4.14	61.00	161.00
Barton Upon Irwell	82.25	173.81	328.00	547.36
Bootle		61.56	90.00	90.40
Haslingden		107.57	108.00	166.86
Machynlleth			114.00	126.66
Plymouth		469.15	1214.00	809.97
St. Giles	641.34	737.76	738.00	762.36
Wolverhampton	24.32	143.30	168.00	167.99
Plymouth		469.15	1214.00	809.97

Table 2.2 Data Recoded into 7 Categories

	Rail Density in km/km^2
Registration District	1845	1854	1876	1914
Alton	0	0-40	40-80	160-200
Barton Upon Irwell	80-120	160-200	>200	>200
Bootle	0	40-80	90.00	80-120
Haslingden	0	80-120	80-120	160-200
Machynlleth	0	0	80-120	120-160
Plymouth	0	>200	>200	>200
St. Giles	>200	>200	>200	>200
Wolverhampton	0-40	120-160	160-200	160-200

Figure 2.2

                This histogram is complex.  On the same chart, it plots four distributions of rail density, one distribution for each of the four dates for which the data is available—1845, 1854, 1876, and 1914.  Reading from left to right, it shows that the number of districts without any rail lines declined from 370 in 1845 to about 10 at the end of the period.  The most dramatic change took place between 1845 and 1854, as the number of districts without rail dropped by nearly two-thirds.  This conforms to the surge in construction evident in the graph we saw above.  What we learn from this display is that the surge had a significant geographic impact as new lines reached more and more districts.  And in the districts already served, the density of rail coverage grew.  The trend is visible in the range of 80 to over 200 kilometers of rail per square kilometer.  The density level of 80 to 120 was reached by 50 districts in 1845, 90 districts in 1854, about 150 in 1876, and some 140 in 1914.  The slight decline from 1876 suggests two opposing possibilities: 1) a reduction in rail service in some districts, and 2) an expansion of service that pushed some districts into the next higher category.  Overall, then, the general pattern is clear.  From the mid-century on the density of rail coverage grew in size and geographic extent.

            Since this was true, then what regions were well served? Which ones were not? And how did things change in time?  The histogram is of little help here because it is not designed to answer these kinds of questions?  But the following maps supply answers.  In addition, they illustrate the visual strength of GIS.  A glance at the maps conveys the immediate impression of continuing growth which reached farther and farther into the country as the years wore on.  Closer inspection reveals the outlines a relationship between railways and industrialization.  This is visible in the pattern where the industrial areas of Manchester and Leeds, in addition to London, were hubs of the rail network. Evident also are the growing concentrations in the northeast near Newcastle and, to the southwest, in the southern edge of Wales.  If these were hubs of the network’s main corridors, it is clear also that few districts were without rail lines by 1914.

Map 2.1

            One more question emerges.  Having examined the data in a table, a histogram, and a map, which of these displays seems most effective?  The best answer is the classic one: “it depends.”  It depends upon the question under study.  If the question calls for details, the table carries the day.  If the structure of the data is at issue, the histogram is a strong contender. By presenting a simplified version of the data in the form of a picture, it helps us find general patterns. When questions of a spatial nature arise, the map is best at conveying the information and its meaning.

The Railway Network:  Explaining the Pattern

            Having charted the course and geographic extent of growth, we can now search for likely explanations.  Some ideas have already arisen that point to the likely importance of industrial areas as magnets for railway development. As Map 2.2 shows, England and Wales were well endowed with natural resources needed for industrial development.  Coal was particularly

Map 2.2 Mineral Resources of England and Wales, 1851

Adapted from S.G. Checkland, The Rise of Industrial Society in England 1815-1885 (London, 1964), p. 151.

important in powering steam engines and in producing iron and steel—industries that railway development caused to expand tremendously. The great areas of coal production were centered around Newcastle in the northeast, Sheffield in the Midlands, and in south Wales near Cardiff and Swansea.

            That the transport of coal was a major force driving and shaping railway develop is clear when the geography of coal mining and the rail network are compared.  A good method for this is to overlay a map of the rail lines on a map showing the major industrial centers.  If an overlay is infeasible, an alternative methods is simply to place the maps side by side and note the correspondences that visual inspection brings into focus. 

Map 2.3  Overlay: Mineral Resources and Rail Lines in 1876

            The result is presented in Map 2.3.  Although the overlay is imperfect, it brings out clearly the geographic relationship between mining and rail transport.  By and large, the density of rail lines was greatest in the coal fields: those in the Northeast, in the area bounded by Manchester, Leeds, Sheffield, and in south Wales.  Areas of iron, lead, and tin were well served by rail also, as were the metals industries of the Midlands centered on Birmingham, Derby, and Manchester. 

The concentration of rail lines in the London area is equally clear.  But this concentration cannot be explained by the presence of natural resources or heavy industry.  The density of the London network stems from other factors.  London’s huge population made it the largest market for all matter of goods, which were shipped there from the producing areas, from manufacturing centers, and from the agrarian areas of the South and Southeast, as is indicated by the network of rail lines covering the Norfolk peninsula.  As a commercial and financial center London had no equal in the world during the period.  Its vast concentration of wealth and power drew vast quantities of goods and people from all parts of Britain.

One overlay, of course, can hardly be expected to offer a complete explanation of the course and extent of railway development.  To go further, additional overlays showing the geographic distribution of farming, manufacturing, and other factors would extend the complexity and completeness of the analysis.  But these matters must be left to the reader. As every author is quick to point out, a good study doesn’t answer all the questions. It answers a few and opens up many more for others to solve.  This said, we can move to another component of our problem, the question of migration.  How many people moved to new areas?  Was there a pattern to this movement?  How might this pattern be explained?

Modeling Population Change and Internal Migration

            In taking up these questions, it is advisable to consider first the limits of our information—a good first step in any research situation.  Our data come from the decennial census enumerations of 1851 to 1911.  These records do not offer direct information on migration and no information on the movements of people between the enumerations.  All is not lost, however.  What census records offer are the data needed to construct estimates of what is called “net migration” over the decade defined by one census and the next.

            Net migration is a residual component of population change. After establishing the total change in population between one census and the next, the change due to natural increase—the difference between births and deaths—is subtracted from the total.  The sum that remains is taken to be the estimated change due to migration.  For each registration district the calculations for the decade 1851 to 1861 run like this:

1.      Population in 1861 – Population in 1851 = Population change from 1851 to 1861

2.      Population change (1851-1861) – change due to natural increase (1851-1861) = the residual sum

3.      Estimated Net Migration = the residual sum

To get a feel for the nature of these estimates, let’s look at a hypothetical model made up of 5 districts of varying size.

Table 2.3 Hypothetical Model of Population Change

In district 1, the population in 1851 stood at 1000 and at 975 a decade later. This was a decline of 25, or of 2.5 percent from its initial size of 1000 inhabitants.  A glance at the other districts reveals a pattern: the small districts sustained population declines while the larger districts grew larger, roughly in proportion to their size.  Among the larger units 3, 4, and 5, the rule
	Reg District	Pop. 1851	Pop. 1861	Pop. Change	% Change
	RD1	1000	975	-25	-2.50
	RD2	1500	1400	-100	-6.67
	RD3	10000	10500	500	5.00
	RD4	30000	33000	3000	10.00
	RD5	100000	115000	15000	15.00

seems to be: the bigger the district, the bigger the increase. This is true both in absolute numbers and in terms of percentage increases from 1851 to 1861.  This model represents several aspects of urbanization in the industrial era.  As employment opportunities came to be more and more concentrated in urbanizing centers of mining, manufacturing, and commerce, these settlements grew, partly by natural increase and partly by the in-migration of people from smaller settlements where economic opportunities were in decline. 

            The nature and pace of migration were not uniform but varied in numerous ways.  It was bound to vary geographically because some regions of England and Wales, especially rural regions, were “sending areas,” while developing urban areas were “receiving zones.”  But, if there was variation in space, there was also variation over time, owing to the shifting fortunes of the economy.  A depression in the price of textiles, for example, gradually reduced the opportunities for workers in this industry.  Consequently, the city and region of Manchester gradually became less appealing to young men and women who set out from their rural parishes to find a better living.

            In the following table, hypothetical model 1 shows the effect of high levels of migration on our five districts. The small districts (RD1 and RD2) are in decline: inhabitants are moving away, leaving a smaller number of inhabitants behind, who are either barely reproducing themselves (RD1) or not reproducing themselves (RD2). This represents the situation in which the young are leaving and their parents are staying. The parents are no longer having children, and the few new families that form bring just enough children into the world to balance the deaths that occur, but no more than that.

            Meanwhile two other scenarios are playing themselves out.  In districts 3 and 4, population growth is robust and people are moving there in significant numbers.  Nevertheless, more of the growth is due to natural increase than to in-migration.  A third scenario is unfolding in district 5; it suggests a city in the midst of an economic and demographic boom.  In the decade since 1851, its population grew by 15 percent, an addition of 15,000 people to the 100,000 there at the previous census.  Here, in-migration outpaced natural increase.  To the newcomers at least, this district’s streets were paved with good prospects if not with gold.

Table 2.4 Models of Population Change and Migration

			Model 1: High Migration			Model 2: Low Migration
Reg District	Pop. Change 1851-61	% Change 1851-61	Total Migration	% Change Mig.	% Change Nat. Incr.	Total Migration	% Change Mig.	% Change Nat. Incr.
RD1	-25	-2.50	-25	-2.50	0.00	-12	-1.20	-1.30
RD2	-100	-6.67	-90	-6.00	-0.67	-30	-2.00	-4.67
RD3	500	5.00	200	2.00	3.00	100	1.00	4.00
RD4	3000	10.00	1000	3.33	6.67	500	1.67	8.33
RD5	15000	15.00	9000	9.00	6.00	1000	1.00	14.00

            The second hypothetical model tells a different story.  The smaller districts (RD1 and RD2) appear to be in full decline, as indicated by out migration and, even more, by the surplus of deaths over births.  Neither population is reproducing itself.  These districts appear to be made up of an aging population.  The young have mostly gone, and the parish bells toll more often for a funeral than for a birth.  If the district has rail service, perhaps the situation would be even worse, for its easier to leave.  Or perhaps it could be better: after all, rail service brought commercial opportunities and often lifted wages above those in more remote settlements that never saw the smoke of a nearby locomotive.  If there were a station nearby, perhaps more children would return for a visit, and residents could travel to see them in return.

            In the three larger districts, in-migration is perceptible but low.  The robust population growth is overwhelmingly the result of natural increase. A baby boom is afoot.

            One can imagine a third model that shows decline not only in the small settlements but in the larger ones as well.  A change of this kind occurred in the history of London and other cities.  After a long phase of growth, contraction set in, more or less gradually, as people began to move out of cities into their hinterlands or further a field.  During the 19^th century, the dominant demographic trend was the movement toward higher and higher degrees of population concentration.  But in the early decades of the 20^th century, this trend began to reverse itself, as a process of population dispersion took hold in London and other cities to varying degrees.

            To sum up, patterns of population change and migration varied in geographic space and over time.  The models that we’ve examined provide representations of spatial variation and of temporal variation too.  From the spatial perspective, models 1 and 2 illustrate different demographic patterns occurring in different geographic locations.  From the temporal perspective, the very same models illustrate also successive phases of population change during the 19^th and early 20^th centuries.  Now, if models are useful for grasping the essentials of complex social change, they are useful also as hypotheses that can be tested against data.  One of the exercises at the end of the chapter will set up such a test.

Geographic Patterns of Migration

            Several of the trends suggested by the models are born out by comparing the distributions of migration at the beginning and end of our period. In the course of six decades, migration into southern districts of England and Wales expanded, while in the north changes were less pronounced. The favored zones of in migration shifted from a concentration around the Manchester and Liverpool agglomeration toward the mining areas near steel-producing Sheffield and into the region of Hull, a port city and entrepot just to the northeast of Sheffield.  In sum, by 1911 the receiving areas in the North extended in a band from Liverpool to Hull.  In South Wales, the coal industry continued to attract workers.  By 1911, the south coast around Plymouth had changed from a sending to a receiving area.  And the continuing flows to the southeast were now touching a much broader circle of districts and communities.  This went hand in hand with the noticeable shift from population concentration to population dispersal.  A close look at the London shows how the process had taken hold there by 1911, possibly sooner.  Even though data for some districts in the 1851-61 decade is missing, it is clear that that movement out of greater London was well under way by the first decade of the new century.

Map 2.4 Changing Patterns of Migration

Young Men and Young Women on the Move

            Not surprisingly the most mobile segment of the population were young people in the late teens and early twenties.  Typically, they were the sons and daughters of agricultural laborers or village craftsmen who had little or no property to pass on to their children.  Moving away from home around 14 or so was a necessity among ordinary families in Victorian Britain, as it was elsewhere at the time.  This was not a new phenomenon but the continuation of a normal feature of daily life in the European past. What was new were the volume of movement and, to a lesser degree, the destinations to which the young went in search of a living.

            Bit if this much is clear, there remain interesting questions.  As the volume of migration increased and the northern industrial centers were expanding at full steam ahead, did the distances traveled tend to increase as well? Did Manchester or Birmingham draw their growing work forces from further a field than before? Unfortunately, the data we have at hand cannot supply an answer.  We can estimate the numbers of people who left a district and the number who moved in between one census and the next. But we cannot tell where those who left went, so the issue of a possible change from shorter to longer distance migration is beyond the limits of the evidence at hand.

            Nonetheless, we can pursue an intriguing question about the possible relationship between gender and migration.  Did the ebbs and flows of young women and men have a similar or different geographic distribution?

Map 2.5 will help us arrive at an answer. For the decade 1851 to 1861, it displays the change, due to migration, in the size of two segments of the same age group—females and males 15 to 24 years of age.  The degree of change is expressed as a percentage.  In the district of Forehoe in Norfolk, the figures look like this[RS1] :

Population Segment	Change due to migration, 1851-61	Percentage change due to migration, 1851-61
Females 15-24	-178	-15%
Males 15-24	-441	-36%

In the rural district of Forehoe in 1861, there were 178 fewer females 15 to 24 than ten years earlier, a decline of 15 percent that was due to out migration.  At the same date, the number of males in the age group showed a drop of 411 since 1851, a deeper decline of 36 percent due to out migration.  Young men were leaving in greater numbers than young women.

            The exodus of young men from Forehoe was indicative of the extensive out migration occurring in agrarian districts. The movement was especially pronounced in the band of communities running from the coast of Norfolk and Suffolk to the tip of Cornwall.  The same decree of out-migration was present in agrarian districts in south Wales and in a few scattered districts in the Midlands. The favored destinations for young men were the mining regions of south Wales, the Sheffield area, and the northeastern region centered on Newcastle and Hartlepool.  With its mines and foundries, the Birmingham region was a site of heavy in migration as well.     

            As for young women, the geographic scale of out migration was as extensive as it was for young men, but the intensity of their out migration was consistently lower than that of young men.  By and large, the main areas of female in-migration coincided with the areas favored by young men, yet there were notable differences.  In northeast around Hartlepool, the districts of female in-migration formed a tighter cluster than was the case for males; while in the Manchester-Liverpool region the reverse was true: the inflow of young women there was more extensive and more dispersed.  The most striking difference occurred in the London region.  The geographic extent and intensity of female in-migration was considerably greater than that of males.  These interesting differences cry out for explanation. And at the end of the chapter, an exercise with additional maps will help the reader find it.

Map 2.5 Population Change due to Migration, 1851-1861
 Females and Males 15-24 Years of Age

Railways and the Rural Exodus

          We come at last to a problem with which we began: the relationship between the extension of the rail system into the countryside and rural de-population.  Did the railways facilitate migration from countryside to town? 

Surprising as it may seem, there exists little systematic work on the question.  All studies agree that the Victorian and Edwardian eras saw the rail system reach its maximum extent and popularity as means of transporting people and freight.  But if this was the Golden Age of the railroads, how they affected the movements of people from village to town and from town to city is more often a matter of conjecture than demonstration.  Even less attention has been given to another question that we can address.  To what extent did the extension of the rail system into small towns and villages retard rural out migration and depopulation?  Rather than hastening a rural exodus, possibly the coming of railways brought new commercial opportunities and new jobs which served to stem the pace of rural out migration, at least in the short run[5]. It is the aim of this section to see if this hunch has any merit.

            That the rail system reached more and more rural districts during the Victorian era is beyond doubt. By 1876, the expansion into the countryside was well underway, a trend that reached its zenith in 1914. (See Map 2.6.)  In Forehoe, where we earlier studied the exodus of young men and women in the 1850s, the Railway Age reached the district somewhere between 1854 and 1876, probably in the 1860s. It was then that the Great Eastern Railway Company constructed several lines that brought rail service to the inhabitants.  Many residents produced food for the huge London market so the business potential was not easily overlooked by railway men and their investors.  The amount of service in the Forehoe area was rather good for rural communities: according to an 1876 map, four or five stations either fell within or were adjacent to the district.  One of the five was the junction station of Wymondham. (See Map 2.7 Rail Lines and Stations near Forehoe, 1876.)

Although the coming of the Great Eastern could not stop out-migration, it probably helped slow its pace.  Evidence of the slower pace of depopulation is visible in the figure below.  The rate of decline slowed little by little until the 1890s when a plateau was reached and the population remained stable. When the new century arrived, another sharp drop took place.

Map 2.6 Extension of the Rail System into Rural Districts, 1845-1914

Map 2.7 Rail Lines and Stations near Forehoe, 1876

Figure 2.3 Population Decline in Forehoe 1851-1911

            If all of rural England and Wales were like Forehoe, we could end here and trumpet the conclusion: very likely, the expansion of the rail system did retard out-migration.  And yet, was Forehoe representative?  Or just a familiar face in a crowd of unknowns? Clearly we cannot stop here.  We must look beyond one district and beyond the whole of Norfolk County to build an argument that will persuade. 

            The problem is to find a means of discovering whether a relationship existed between rail service and population change.  One way to proceed is look to the available data for variables or attributes that can serve as reasonable indicators of rail service and of population change caused by migration.  Rail stations would certainly be a good indicator of service, but, alas, information of that kind is unavailable. Lacking that, it is reasonable to use rail density as a stand-in for rail service.  The greater the density of lines, the more likely the existence of stations and of access to freight and passenger service.

            As for a measure of population change, the figures on population themselves will not do because some variation in their magnitude is the result of their geographic size, and their size changed over time.  All this makes it nearly impossible to know what is changing and when.  When faced with this perplexity, we are in the presence of what statisticians call “confounding.” We want to avoid this muddle.

Because population density neutralizes the effect of differing (and changing) district areas, it is a good choice. Regardless of whether boundary changes alter a district’s perimeter and area, standardizing our population figures by calculating the ratio of people per unit of area will provide us with a reliable measuring stick for gauging degrees of rural de-population.  Since we are interested in historical change, several new attributes are in order. For each district, the first measures the change in population density from 1851 to 1871; the second and third do likewise for the periods 1851-1881, and 1881 to 1914.  All three are so defined in order to distinguish the early and latter periods of railway expansion.  It was during the later period when the expansion was significant in the countryside.  In the earlier period, especially before 1870, rail service in rural areas was non-existent or limited.

One further task is at hand, which is to define a set of districts that can be classified as “rural.”  Here again, population density proves helpful.  After consulting a number of sources and the distribution of population density over the years, I have drawn a line in the sand, so to speak.  From here on, I shall consider that 74 persons per square is the dividing line between rural and non-rural. Those districts that fall on or below the line of 74 will be classified as “rural.”  One could argue with this definition, but it is not entirely arbitrary.  It is based on my inspection of the distributions of population density in the data and on the definitions in studies by other scholars.  Definitions of this kind are important matters because they will have an effect on the results.

With these matters out of the way, we can now see whether and how our indicators of rails service and de-population may be related.  We can begin with a map as a first, hopeful step in our exploration.

Up to now we have used maps to display the geographic distribution of one variable or attribute.  We now face the problem of constructing a map that will represent the possible correspondence between two attributes.  The following map is one possible display.  It is a choropleth map that presents in geographic space three different attributes: the location of the rail lines in 1876, the density of rail lines, and the change in population density from 1851 to 1881.  Five levels of population density are indicated by five shades of gray—a symbolization of an attribute defined by an ordinal scale, we may recall.  Rail density remains a numerically scaled attribute; a the height of a bar represent its various values. So much for the orientation. What does the map show?

Here is an interesting example of GIS run amok.  This map befits the epithet that Edward Tufte coined and thrust at the mania for pie-charts: “Chart Junk.”  There is too much going on in this map so our eyes are overwhelmed by too much information .  Our brains tire in attempting to abstract a recognizable pattern from an overload of detail.  To reduce the clutter, we could remove the rail lines, for they seem redundant and get in the way.  The result is a bit better but the new map is still a candidate for the dustbin.  The key problem is the inadequate representation of rail density, and this prevent us from drawing a meaningful conclusion.  Stymied, we may unwittingly start reading our favored ideas into the map—something all too normal but something to avoid as well.

The lesson is not lost, however. We learn something important about the limitations of maps as tools for testing relationships with complex data.  Moreover, trial and error is characteristic of fruitful exploratory analysis.  There is a cautionary tale here as well. Because the computer and GIS software can produce maps so much faster than we can begin to understand them, they excel at creating dazzling displays in marvelous color which prove meaningless or trivial.  Watch out for “Chart Junk!”

Map 2.8 Rail Density and Changes in Population Density in Rural Districts, 1851-81

            Perhaps an ingenious mapping solution to our problem is just around the corner; but since it’s not yet at hand, we need to take a different tack.  There is a statistical method called the “t test” that can be used to address the problem. This method tests the hypothesis that the means of an attribute belonging to two groups are so similar that they come from the same underlying population.  Recalling our question, we speculate that districts with high rail density will retain their populations to a greater degree than those with lower densities.  Hence we divide the rural districts into two groups.  In group 1 are the districts with low levels of rail density; in group 2, those with higher levels.  The line dividing “low” from “high” will be the lower quartile, an ordinal measure of position.  This means that 25 percent of the districts will be classified in the “low” group having little access to rail service; the remaining 75 percent fall in the “high” group with greater access.

            Next, we frame two versions of the hypothesis to be tested:

The Null hypothesis:

any observed difference between the means of population density change (1851-1871) in group 1 and group 2 is so small that the difference was very likely the result of chance or mere coincidence.  In other words, the two groups are similar and come from the same population of rural districts.  Differences among them changed densities of people are unrelated to the density of rail coverage.

The Alternative hypothesis:

the observed difference in the means of the two groups is sufficiently large that the difference was unlikely the result of chance or coincidence.  There was a relationship between the level of rail density and population density change: districts with high levels railway lines had larger numbers of inhabitants per square kilometer over the years studied. They were better able to slow the pace of rural out-migration.

One of the strange things about statistical hypothesis testing is that we have no viable means of “proving” that the so-called alternate hypothesis is true.  Typically this is the hypothesis we hope to confirm because our main interest is to discover relationships and connections that explain the way the world works. Discovering that no relationship seems to exist strikes us as a failure or at least not something on which to base a paper. (In fact, these discoveries are often quite important, even if they seem otherwise.)  In any event, what can be done is to determine whether the Null hypothesis is persuasive or not.  If the results of the test strongly suggest that it is untrue, our hopes rise if we are betting on the alternative one. Upon rejecting the Null hypothesis, we conclude that there are solid statistical grounds for accepting the Alternative as a sound version of reality—at least in the sample under test.

The result of the t-test shows that we can reject the null hypothesis with confidence, and conclude that rail service did help slow the pace of de-population, if only slightly.  From 1851 to 1871, districts with low rail density sustained an average decline in population density of 1.7 persons per square kilometer, compared to virtually no change, on average, in districts with greater rail coverage. 

The results for a second test are more telling because they concern the period from 1881 to 1911.  It was during this later period when the rail system’s effect on rural developments was more likely felt because by then train service was more widely available than before.  Although the average change in people per unit of area declined in both groups, the decline was significantly deeper (4.5) in the districts where access to railways was lower than the more modest decline (1.9) in districts having greater access. 

 Table 2.5  t-Test of  Changed Population Density by Level of Rail Density, 1851-1871

Group 1: Districts in lower quartile of rail density (1876) Group 2: Districts in upper three quartiles of rail density (1876)
	Group 1	Group 2
	Mean	Mean				Group 1	Group 2
Variable			t-value	df	p	Valid N	Valid N
Change in Pop. Density, 1851-1871	-1.667	.0903	-2.55	213	.011475	60	155

Group 1: Districts in lower quartile of rail density (1914)   Group 2: Districts in upper three quartiles of rail density (1914)
	Group 1	Group 2				Group 1	Group 2
	Mean	Mean				Valid N	Valid N
Variable			t-value	df	p
Change in Pop. Density, 1881-1914	-4.475	-1.871	-2.50	231	.0123	71	162

These findings bring our investigation to a successful conclusion.  They show that our hunch does indeed have merit.  Solid evidence and statistical reasoning confirm that the expansion of railways into the countryside did have a retarding effect on rural out migration and depopulation. 

Our results are promising and significant in a broader sense as well.  First, they illustrate how GIS methods combining mapping and data analysis make it possible to extend a study of an historical problem in new directions.  Second, though hardly the final word on the subject, they encourage further research and point toward refinements worth pursuing. 

To consider some such refinements will be a useful way of concluding, for that will offer a review of key aspects of the GIS approach.  The following table presents selected examples of this kind: aspects of spatial analysis that we have applied and others aspects that we could apply to extend the study. (See the appendix for the complete table.)  As noted in column three, one improvement would come through the addition to our data of information on rail stations: their location and the dates during which they provided service.[6]

Table 2.6  Review and Revision

Geographic Properties	Aspects Applied in the Analysis			Further Aspects That Could be Applied in Revising and Extending the Analysis
Proximity ·        How close is x  to y? ·        How many x are within 30 kilometers of y?	The distance from agrarian Forehoe to the food markets of London. The clustering of rail lines around industrial and commercial centers			The distance between the nearest major urban center and the nearest rail station by district to indicate the accessibility to urban markets for products and employment, a likely factor in explaining migration.  One could then examine whether out migration varied with degree of accessibility by rail. The number of rail stations within 30 km. of industrial and commercial centers as a better indicator of service offered to the public.  Comparing the results over time in a map and in a table would shed further light on the evolution of the rail network and the connection with population concentration as opposed to dispersion.
Connection ·        Are x and y connected by z? ·        Do those x and y features that are connected by z form a pattern?	In our sample, Hatfield, Plymouth, Wolverhampton, and Owestry were directly connected to London by rail.			When and to what extent did food producing areas such as Norfolk become connected by rail to London or to other urban markers for their products?  Mapping the results would show how rail transport and regional specialization of agriculture evolved together. Over time, as the national market in food emerged and grew, a fewer number of regions provided the bulk of the (non-imported) food supply.  One would expect these areas to be well served by rail transport.
Containment ·        Does z fall within x? ·        How large is z within x?	Districts with rail lines. The density of rail lines within a district as an indicator of rail service available to the inhabitants and the economic opportunities associated with railways			Districts with rail stations. The number of stations within a district as an improved indicator of rail service and its associated economic opportunities for the inhabitants.
Pattern ·        Does z fall within x usually when y is also present in x? ·        More often than not, is there a connection between x and y in one region but not in another?		In the evolution of the rail system, industrial areas and major population centers were connected first and more densely. Hence, the industrial areas of the Midlands and the lower Northwest surrounding Manchester were the best served by railways in the 1840s to 1860s. In the London area, female in migration was more pronounced and more extensive because the vast market for domestic servants (an employment in which women predominated) drew them there in greater numbers than it did males. Rural districts with greater access to rail service had, on average, lower levels of out migration than districts with less access, especially after 1880 up to 1900 when the expansion of rail into the countryside was well advanced.	Rural districts that were better served by rail stations for conveying passengers and freight were more apt to have lower levels of out migration and depopulation than districts less well served.

Location, proximity, connection, containment, and pattern—these five properties of the GIS approach and the maps that bring them to our eyes work together to enhance our understanding of significant relationships in geography and history.  Hopefully, this is clearer now than before, so we can pronounce this introduction to historical GIS a success.

Appendix

Geographic Properties		Aspects Applied in the Analysis		Further Aspects That Could be Applied in Revising and Extending the Analysis
Location ·        What is where? Where are the x features? ·        Which geographic entities are in a given region of the country?		The districts defined as rural, i.e., those with population density of less than or equal to 74 persons per square km., tended to lie on the eastern and western edges of the country, outside of the main corridors of development stretching from London to Birmingham, Liverpool and Manchester and to Sheffield and Newcastle, and including the mining areas of  South Wales.		Which districts were hilly or mountainous, and thus presumably more remote or inaccessible than those on the flat lands.
Proximity ·        How close is x  to y? ·        How many x are within 30 kilometers of y?		The distance from agrarian Forehoe to the food markets of London. The clustering of rail lines around industrial and commercial centers		The distance between the nearest major urban center and the nearest rail station by district to indicate the accessibility to urban markets for products and employment, a likely factor in explaining migration.  One could then examine whether out migration varied with degree of accessibility by rail. The number of rail stations within 30 km. of industrial and commercial centers as a better indicator of service offered to the public.  Comparing the results over time in a map and in a table would shed further light on the evolution of the rail network and the connection with population concentration as opposed to dispersion.
Size ·        How large is feature x? ·        Do small x features form a pattern?		The district of Cockermouth in the far northwest was largest in area. Smaller districts tended to cluster in urban areas because they designed to form parts of larger settlements
Adjacency ·        What is next to x? ·        Do x and y features share a common boundary in region z?		Cockermouth and Bootle in the northwest were among four pairs of districts in our sample that shared a common border. Cockermouth and Bootle, and the other pairs in our sample that shared borders tended to cluster near the sea as opposed to inland?		To what extent were districts without rail stations next to districts with stations?  Districts not adjacent to districts with stations would be more isolated and possibly sites of higher out migration.
Connection ·        Are x and y connected by z? ·        Do those x and y features that are connected by z form a pattern?		In our sample, Hatfield, Plymouth, Wolverhampton, and Owestry were directly connected to London by rail.		When and to what extent did food producing areas such as Norfolk become connected by rail to London or to other urban markers for their products?  Mapping the results would show how rail transport and regional specialization of agriculture evolved together. Over time, as the national market in food emerged and grew, a fewer number of regions provided the bulk of the (non-imported) food supply.  One would expect these areas to be well served by rail transport.
Containment ·        Does z fall within x? ·        How large is z within x?		Districts with rail lines. The density of rail lines within a district as an indicator of rail service available to the inhabitants and the economic opportunities associated with railways		Districts with rail stations. The number of stations within a district as an improved indicator of rail service and its associated economic opportunities for the inhabitants.
Pattern ·        Does z fall within x usually when y is also present in x? ·        More often than not, is there a connection between x and y in one region but not in another?	In the evolution of the rail system, industrial areas and major population centers were connected first and more densely. Hence, the industrial areas of the Midlands and the lower Northwest surrounding Manchester were the best served by railways in the 1840s to 1860s. In the London area, female in migration was more pronounced and more extensive because the vast market for domestic servants (an employment in which women predominated) drew them there in greater numbers than it did males. Rural districts with greater access to rail service had, on average, lower levels of out migration than districts with less access, especially after 1880 up to 1900 when the expansion of rail into the countryside was well advanced.		Rural districts that were better served by rail stations for conveying passengers and freight were more apt to have lower levels of out migration and depopulation than districts less well served.

Exercises

1.  Map 2.5 displayed information on migration among young women and men aged 15 to 24. Explain how Map 2.9 below can help explain the patterns for young women. In the 19^th century domestic service was the largest sectors of employment at the time for women.

Map 2.9 Domestic Service Employment, 1851 and 1911

                                1851                                                                      1911

Adapted from John Langton and R. J. Morris, Atlas of Industrializing Britain 1780-1914 (London, 1983) p. 143.

2. Map 2.10 shows how the same attribute can be displayed four different ways in a choropleth map, one that uses shadings to designate different levels within a distribution of numbers.  The purpose of the map is to display the geographic distribution of population density in England and Wales in 1851.  Each map employs a different method of dividing values of population density into categories which can then be represented by shadings of gray.  Each method of creating categories will produce a different display. The method of equal intervals divides the range (maximum-minimum) by the number of categories desired, each one being equal in size. “Langton’s intervals” on the map in the upper right quadrant refers to categories that were defined by historical geographers John Langton and R. J. Morris in their Atlas of Industrializing Britain 1780-1914 (London, 1986).  The method of “natural breaks,” used in the map in the lower right hand corner is the result of a mathematical calculation that attempt to create groups by dividing the distribution at maximum points of difference, i.e., at “natural breaks.”  One need not have a computer to use this method, for visual inspection of a distribution in a histogram is a fine method for locating “natural breaks” in data.  In the last map at the lower right, the quantile values are used to divide the distribution into 5 equal groups, each one comprising 20 percent of the distribution.  This requires arranging the data in order of magnitude, as was discussed in Chapter 1.

A good map should help bring to light a meaningful pattern in the data.  Which of the maps do you think is most effective in this regard.  Explain your reasoning.

3. Study the map from question 2 that you think is the most effective display of population density in 1851.  Choose three features of the geography of population density and offer an interpretation of their historical meaning or significance.

Map 2.10 Four Methods of Mapping Population Density in 1851

In the chapter there are two discussion that the reader should revisit and extend.  One is centered on two tables: Table 2.3 Hypothetical Model of Population Change and Table 2.4 Models of Population Change and Migration. A second discussion concerns the de-population in the district of Forehoe in Norfolk County.  The main point argued that pace of de-population slowed after the 1850s.  Draw on these discussions to address the following.

4.  To what extent does the data in Table 2.6 on Forehoe confirm either or both of  the hypothetical models in Table 2.4?

5.  In Forehoe did the pace of out migration slow after 1860? Explain your reasoning.

6.  Briefly describe in a set of bullet points the timing and magnitude of changes that occurred among young women and men 15 to 24 years of age, over the decades from 1851 to 1911.

Table 2.6 Population Change and Migration in Forehoe, 1851-1911

Decade End Date	Total Population at Beginning Date	Index 1861=100	Total Change due to Migration	Change in Number of Males 15-24	Change in Number of Females 15-24	Percentage Change Males 15-25	Percentage Change Females 15-25
1861	13565	100	-1953	-441	-178	-35.60178	-15.01267
1871	12818	94.5	-1787	-371	-264	-36.39706	-23.4862
1881	12308	90.7	-1540	-303	-140	-30.80448	-15.07551
1891	11971	88.2	-1406	-258	-104	-24.36321	-11.13197
1901	11988	88.4	-1598	-288	-142	-28.85385	-15.17666
1911	11329	83.5	-998	-129	-54	-13.26109	-6.17783

7.  Did demographic changes with this age group of the population accurately reflect the trend for the district as a whole?

8.  Use the Total Population at the beginning of the decade and the change due to migration to calculate for each decade an estimate of natural increase, the other component of population change that was discussed in the chapter. Explain what your figures on natural increase and those on the migration of young men and women given above suggest about the capacity of the district to reproduce itself.

9. How might the condition of the physical environment in Forehoe have been affected by the demographic changes you have discovered?

Further Reading

Baines, Dudley. Migration in a Mature Economy : Emigration and Internal Migration in England and Wales, 1861-1900. Cambridge: Cambridge University Press, 1985.

Carter, Ernest Frank. An Historical Geography of the Railways of the British Isles. London: Cassel, 1959.

Churton, Edward. The Rail Road Book of England: Historical, Topographical and Picturesque; Descriptive of the Cities, Towns, Country Seats, and Other Subjects of Local Interest, With a Brief Sketch of the Lines in Scotland and Wales. London: Sidgwick and Jackson, 1973-.

Davis, Bruce. GIS A Visual Approach. Santa Fe, New Mexica: Onword Press, 1996.

Friedlander, Dov. “Occupational Structure, Wages, and Migration in Late Nineteenth-Century England and Wales,” Economic Development and Cultural Change (1992): 295-317.

Gourvish, T. R. Railways and the British Economy, 1830-1914. London: Macmillan, 1980.

Langton, John and Morris, R. J.  Atlas of Industrializing Britain 1780-1914 (London: Methuen, 1986)

Lawton, Richard and Pooley, Colin G. Britain 1740-1950. An Historical Geography. London: Edward Arnold, 1992.

Lawton, Richard. "Rural Depopulation in Nineteenth-Century Britain." in Liverpool Essays in Geography. eds. Robert W. Steel, and Richard Lawton. London: Longmans, 1967.

Ottley, George. A Bibliography of British Railway History. London: Allen & Unwin, 1965.

Parris, Henry. Government and the Railways in Nineteenth-Century Britain. Routledge & K. Paul, 1965.

Redford, Arthur. Labour Migration in England, 1800-1850. Manchester: Manchester University Press, 1964.

Reed, M. C., comp. Railways in the Victorian Economy: Studies in Finance and Economic Growth. Newton Abbot: David & Charles, 1969.

Saville, John. Rural Depopulation in England and Wales, 1851-1951. London: Routledge & K. Paul, 1957.

Simmons, Jack. The Railway in England and Wales, 1830-1914. Leicester [Eng.]: Leicester University Press; 1978

———. The Railway in Town and Country, 1830-1914. London: David & Charles, 1986.

______. The Victorian Railway. London: Thames and Hudson, 1991.

Simmons, Jack, and Gordon. Biddle. The Oxford Companion to British Railway History From 1603 to the 1990s. Oxford ; New York: Oxford University Press, 1997.

Smiles, Samuel. The Life of George Stephenson and of His Son Robert Stephenson; Comprising Also a History of the Invention and Introduction of the Railway Locomotive. New York: Harper, 1868.

Southall, Humphrey R. “The Tramping Artisan Revisits: Labour Mobility and Economic Distress in Early Victorian England,” Economic History Review, 44 (1991): 272-296.

Thomas, Keith. Man and the Natural World : a History of the Modern Sensibility. New York : Pantheon Books, 1983.

Worster, ‘Donald. Nature's Economy : a History of Ecological Ideas.  New York : Cambridge University Press, 1985.

---

[1] Unless otherwise noted, all the maps of England and Wales and the bulk of the attribute data were created from the Great Britain Historical Data Base.  Thanks go to Jennifer Gieseking who digitized the rail lines from the map at the end of Jack Simmons’ book, The Railway in England and Wales 1830-1914 (Leicester: Leicester University Press, 1978). After she entered them into MapInfo, I converted them to ArcView shape files.

[2] For the project on railways and migration, a number of commercial software products were used: Microsoft’s Access, Excel, and Word, which provide respectively the functions of relational data base management, spreadsheet analysis, and word processing; and ESRI’s ArcView GIS program.  There are numerous counterparts to these programs available from other companies

[3] Blackwoods Magazine, November 1830.

[4] The copy I consulted is kept at the Public Record Office (National Archives of Britain) at Kew in Surrey.

[5] At first glance, the argument that railroads facilitated internal migration, the decline of the rural communities, and the growth of cities seems a truism, something hardly disputable.  A closer look at the literature and history suggests, however, that this is an argument that bears re-consideration. Although building of railway plants, junctions, and freight yards increased the populations where these works were located, the notion that railways were a primary cause of urban growth and rural depopulation has proved dubious and overly simple. As recent research has shown, new rail lines into major urban centers usually arrived after the population had grown significantly and not before. With the exception of Crewe in Lancashire and perhaps one or two other cases, railways in England did not create new towns as in the United States.

[6] The number of passengers each station served would be desirable as well for that would be perhaps the best indicator of the extent and utilization of rail service.  Unfortunately, such figures are scattered and incomplete.  The only fairly complete series of records are those for the Great Northeastern Railway, which I am in the process of incorporating in the project.

---

[RS1]

Population Segment	Number in 1851	Number in 1861	Change	Percentage Change
Females 15-24	1186	1008	-178	-15
Males 15-24	1239	798	-441	-36