The Importance of Access to Coal in the
Industrial Revolution
Kilian Heilmann
September 14, 2014
Abstract
I study the importance of access to coal by using historical population data as
a proxy for economic development. I calculate the cost-distance to coal fields by
incorporating geographical data on navigable rivers and turnpike roads. Regressing county-level percentage population growth on cost distance does not suggest
any significant relationship between coal access and economic growth. Replacing
population growth with life expectancy as the dependent variable confirms the zero
correlation with geographic features.
1
Introduction
There is little doubt that the Industrial Revolution ran on coal and data shows that
coal production increased by 1800% between 1700 and 1860 (Church, 1986), but the causal
effect of coal reserves on the vast increase of economic production during the 18th and 19th
century has been subject to discussion. Pomeranz (2000) argues that the availability of
coal near population centers was the key factor in explaining why the Industrial Revolution
happened in Britain and not for example in the coal-poor Netherlands which at that time
had similarly well-developed institutions as Britain. Church (1986) even claimed that
“[...] it is difficult to exaggerate the importance of coal to the British economy between
1830 and 1913.”1 Yet the importance of coal has been challenged by recent research.
Clark and Jacks (2007), using time series on coal prices at pithead, argue that Britain’s
access to coal reserves had a negligible effect and that demand for energy could have
easily been satisfied by traditional sources such as water, wind and firewood. This would
imply that coal was of minor importance for the onset of industrialization and that the
Industrial Revolution started in Britain for reasons other than geographical advantages
in coal resources.
Empirical work on this question has been hindered by the sheer unavailability of
economic data. Even if there exist estimates of economic production as in the case for
1
Church (1986), page 758, as cited in Clark and Jacks (2007).
1
Britain, the data is difficult to compare to the sparse records for other countries during
the period of interest. This paper tries to shed light on the question by using historical
population data for 19th century Britain and a consistent approach to measure proximity
to coal reserves. The main idea of the paper is the following notion: If the availability of
coal mattered on the big scale, it should have mattered on a small scale too. This means
that we should expect to see a positive relationship between distance to coal supplies and
economic output not only between countries, but also within countries, namely within
Britain. Comparing the economic performance of different cities and regions within a
single country further abstracts from other confounding country characteristics as all
British cities were subject to the same political and legal environment.
This project continues a research program that is concerned with geography’s effect
on economic outcomes. A variety of papers has argued for and against the importance
of geophysical features, with the effect being thought of either directly and indirectly
(through institutions) working on economic output. Crosby (1986) links the well-being
of many British colonies to their similarity with the British climate, allowing established
agricultural technologies to be easily implemented and thus spur growth. Mellinger et al
(2000) support this theory by showing that climate and coastal proximity are major determinants of economic development. Redding and Venables (2002) emphasize the important
negative effect of geographical isolation, such as distance to the nearest coast. They argue that remoteness drastically increases trading costs which in return negatively affects
economic prosperity. McCord and Sachs (2013) estimate the importance of geographical features and natural resources and show that alongside institutions and technology,
geography can explain countries’ timing of achieving middle income status. Acemoglu
et al (2002) in contrast argue that geography influences the economy at most indirectly
through institutions. Easterly and Levine (2003) support this view, finding that the climate in tropical regions indirectly determines cross-country income through institutions,
but they cannot document a direct effect.
This paper aims to contribute to the literature on the relationship between economic
development and geography by addressing coal resources, a geographical advantage which
has not been studied before. Proxying economic development with county-level population
data for the 19th century and regressing these on a cost distance measure to the nearest
coal fields, the results show that there is no significant effect of closer coal resources on
population growth. This suggests that the geography of coal fields did not determine
population patterns in Industrial Revolution Britain.
2
The paper proceeds as follows. Section 2 explains the data sources and data manipulation. Section 3 develops the empirical framework and presents the regression results.
Section 4 concludes and discusses the findings.
2
2.1
Data
Data Sources
The variable of interest in this study is local economic growth during the Industrial
Revolution. However, since there is no GDP data available for this period, I follow
the usual practice in the economic history literature to proxy economic development with
population growth (e.g. Hanlon and Miscio, 2014). British population data in a consistent
fashion is available for the 19th century and drawn from the British censuses conducted
every ten years starting from 1801. These decennial censuses enumerated the population
of the entire United Kingdom and even though the implementation of the census varied
over time, the data is believed to be of rather high quality (Lawton, 1978). In the absence
of direct measurement, population data is believed to capture economic development well
in the absence of migration barriers. This is true for 19th century Britain, a period in
which there was high internal migration between cities with little government influence
on industrial policy (Long and Ferrie, 2003).
The unit of investigation is the population number at the level of an Ancient County.
The Ancient Counties (also known as Historic Counties) are ceremonial divisions that
were created in the based on earlier kingdoms. Even though they ceased to serve as
administrative counties in 1889, they were still used in the 1901 Census and that census reports their population over the whole 19th century. The Ancient Counties have
the advantage that they experienced relatively little change in their boundaries. Even
though the Counties (Detached Parts) Act of 1844 eliminated many exclaves the county
boundaries featured, the area change was rather small.
To calculate a consistent measure of cost distance to the nearest coal fields, I use
historic geographical data drawn from several sources. Pope (1989) provides geographic
data for Industrial Revolution Britain. I am using the printed maps of navigable rivers
as of 1750, the turnpike road network as of 1750 and a map showing coal fields and coal
3
Variable
Table 1: Table of Means
Obs Mean Std. Dev.
Population Percentage Change
Cost Distance to Coal
Cost Distance to London
Initial Population in 1801
520
520
520
52
0.095
1
1
171,530
0.244
0.938
0.576
175,403
Min
Max
-2.321
2.406
0
4.584
0.064
2.433
16,300 859,133
output from 1700-1830. Geographical data on the exact borders of the Ancient Counties
are downloaded from the Historic County Borders Project.2
2.2
Data Manipulation
Processing the raw geographical data to obtain the dataset used in the regression is
performed in ArcMap 10.1. At first, I georeference the scanned maps using the inbuilt
georeferencing tools to obtain polyline shapefiles for the river and turnpike road network
as well as a polygone shapefile indicating the coal fields of Britain. The maps do not
come with a description of which projection is used, but projecting the national boundary
shapefile to an equi-area conic projection and creating a large set of control points yields
a very close congruence between the scanned map and the Ancient County shapefile.
After having obtained the shapefiles, I convert them to rasters and assign cost values
to each mode of transportation. As a first guess, I normalize river transport to a cost
of 1 and assign ocean transport the cost of 2, transportation via turnpikes the cost of 5,
and all other pixels the value of 7. I then combine the rasters to a single cost raster that
contains the cost of crossing each pixel in the raster. Using the Cost Distance calculator
embedded in ArcGIS results in a raster file containing the cost distance to the nearest
coal field for each point in the United Kingdom (Figure 3).
In a next step I create centroids for each ancient county and link them with the cost
to reach the closest coal field from the cost distance raster using the Extract Values tool.
In the same fashion, I calculate the cost distance to reach the capital in London.
As a robustness check I create a different measure of access to coal taking into account
not only the distance to the nearest coal fields, but also the size of it. I first convert
the polygon features for coal fields, rivers and turnpike roads into points and then use
2
Publicly available at http://www.county-borders.co.uk/.
4
the Kernel Density tool to measure the “density” of those features for each location in
the UK. For coal, this measure takes into account all surrounding coal fields where closer
fields have a higher weight than those farther away. Figure 4 shows that the strongest
concentration of British coal reserves is in the Midwest area which receives a much higher
weight than the smaller and more isolated fields of North East and South West England.
3
3.1
Econometric Setup and Results
Population Change
I calculate county-level percentage changes in population between each census and
regress them on the coal distance measure for the county’s centroid. Given London’s
importance as an administrative and economic center of the United Kingdom, I control
for the distance to London to capture any effects of the closeness to the capital. Since
the distance measures can only be interpreted relatively, I normalize the unit to be equal
to one standard deviation. The regression equation is given by
∆ log (popit ) = α + β1 distance coali + β2 distance londoni + it .
(1)
The results in Table 2 show that there is no significant relationship between either
distance variable and the percentage population change over the 19th century. The negative coefficients indicate that more remote regions experience less population growth, but
these two variables perform badly at explaining population patterns. Including decade
fixed effects does not change the results.
To account for changing importance of coal, I estimate the same coefficients for each
decade between the censuses and plot the results in Figure 2. It becomes evident that
the coefficient is imprecisely estimated for all periods except for the period preceding
the census of 1861 which seems to drive the results for the aggregate specification. In
conclusion, none of the estimates suggests a significant relationship between access to
coal fields and population growth on the county level. This is consistent with an earlier
finding that high skill workers left cities polluted by high industrial coal use (Hanlon,
2014).
5
Figure 1: Geographical and Infrastructure Features
Map combining rivers, turnpike roads, coal fields, and county features (Equidistant
Conic Projection)
6
Table 2: Regression Results
Dependent Variable: Percentage Change in Population
(1)
(2)
Cost Distance to Coal
-0.0112
-0.0112
(-1.05)
(-1.06)
Cost Distance to London
-0.0041
(-0.38)
-0.0041
(-0.39)
Constant
0.114∗∗∗
(4.70)
no
520
0.0024
0.137∗∗∗
(3.46)
yes
520
0.0452
Time Fixed Effects
N
R2
t statistics in parentheses
∗
p < 0.05, ∗∗ p < 0.01, ∗∗∗ p < 0.001
3.2
Alternative Specification: Life Expectancy
Instead of using population changes to proxy changes in economic development, I use
life expectancy. Previous research has shown that there is a significant positive relationship between life expectancy and GDP per capita over time (Preston, 1974). Szreter
and Mooney (1998) provide estimates for the expectation of life at birth for 16 large and
medium-size cities in five ten-year intervals starting from 1851-1860 and reaching until
1891-1900. There is considerable variation in life expectancy both between cities and
within cities over time (Table 4). Szreter and Mooney also estimate total population of
these cities for each decade. City sizes range in the initial from just 51,000 for Eccleshall
to a stunning 2.5 million people living in London.
I regress percentage changes of life expectancy on the cost distance to coal fields. The
simple correlation between those two variables is essentially zero and estimated with low
precision (See Table 3, Panel A). Adding control variables (total population of the city
at the beginning of each decade) and decade fixed effects does not change the results.
With the coefficients very close to zero and high standard errors, none of the regression
setups suggests any significant effect of access to coal on longevity. The large jump in the
R2 measure after adding decade fixed effects suggests that simple time trends that are
common for all cities drive the changes in life expectancy. The estimates suffer from a low
number of observations and low variation in the cost distance variable. As can be seen
from Figure 4, most of the cities are located very close or even within coal fields, thus
creating a cost distance of zero. Furthermore, the cities are very close to each other; in
7
Figure 2: Estimated Coefficients on Cost Distance to Coal
fact, what used to be separate entities back in the 18th century has grown to become one
urban area (e.g. Liverpool/West Derby, Manchester/Chorlton, Bristol/Clifton), further
increasing the problem of limited variation in the distance measure.
To mitigate at least the problem of zero cost distance, I rerun the regressions using the
alternative distance measure which provides more variation in the independent variables.
The distance measures created by the kernel density calculations result in different values
for the coal access variable depending on the density of coal fields close by. Controlling
for population size, access to roads, and navigable rivers, the weak and insignificant
relationship between coal access and longevity is confirmed. Again it seems that most of
the variation in life expectancy can be explained by decade-specific effects. This leads us
to conclude that the distance to coal did not have a significant effect on longevity for the
different English cities in the sample regardless of which distance measure is used.
4
Conclusion
Regressing the percentage change of population size for British counties during the
19th century on the constructed measures for access to coal did not suggest any significant
relationship between these two variables. Negative coefficients on distances show that
more remote counties experienced slightly lower population, but neither the least cost
8
Table 3: Regression Results
Dependent Variable: Percentage Change in Life Expectancy at Birth
Panel 1
Variable
Cost Distance
Population
Panel 2
-0.0000 -0.0000 -0.0000
(-0.11) (-0.96) (-0.23)
0.0000 0.0000
(1.00)
(0.18)
Decade FE
no
no
yes
Observations
64
64
64
R2
0.0002 0.0162 0.5298
Coefficient estimates of the regressions with
R2 refers to unadjusted R2 measure.
Variable
Coal Kernel
-0.0177 -0.0000
(-0.44) (-0.01)
Population
0.0000 -0.0099
(0.87) (-0.35)
River Kernel
0.0000 0.0000
(-0.89) (-0.06)
Turnpike Kernel 0.0000 0.0000
(0.76)
(0.92)
Decade FE
no
yes
Observations
64
64
R2
0.0220 0.5389
t-statistics given in parentheses.
distance to coal fields nor proximity to the capital in London can reasonably explain
different population patterns. The same result is true if the dependent variable is replaced
by life expectancy data, another proxy for economic development, where the variation in
longevity appears to be largely being explained by common time trends.
These results can be interpreted such that closeness to coal was not a necessary condition for economic growth during the Industrial Revolution. This is not to say that
geographical features were not important, but that inner-British variation in coal endowment played little to no role for county growth patterns. Areas further away from coal and
government centers grew at similar rates as those near coal fields and there are several
potential reasons that are consistent with this finding.
Firstly, these counties might have experienced economic growth in industries that were
not dependent on coal, confirming the notion that highly skilled workers fled the negative
health effects of the coal cities. Secondly, these counties might still have depended on coalintensive industries, but that the British transportation system at the time was efficient
enough to distribute coal from its sources to the production centers. This seems reasonable
given that proximity to the coast and the large amount of navigable rivers for most of
the country enabled coal to be shipped on a large scale by water as compared to costly
overland transport.
9
5
References
Acemoglu, Daron, Simon Johnson, and James A. Robinson (2002): Reversal of Fortune. Geography and Institutions in the Making of the Modern World Income Distribution, The Quarterly Journal of Economics 117 (4), p. 1231-1294.
Church, Roy (1986): The History of the British Coal Industry vol.3, 1830-1913, Oxford: Clarendon Press.
Clark, Gregory and David Jacks (2007): Coal and the Industrial Revolution 1700-1869,
European Review of Economic History 11 (1), p. 39-72.
Crosby, Alfred W. (1986): Ecological Imperialism. The Biological Expansion of Europe, 900-1900, Cambridge University Press.
Easterly, William and Ross Levine (2003): Tropics, Germs, and Crops. The Role of
Endowments in Economic Development, Journal of Monetary Economics, 50 (1), p. 3-39.
Hanlon, W. Walker and Antonio Miscio (2014): Agglomeration. A Dynamic Approach,
Working Paper.
Hanlon, W. Walker (2014): The Impact of Industrial Pollution on City Growth.
Lessons from the Dark Satanic Mills, Working Paper.
Lawton, Richard (1978): The Census and Social Structure. An interpretative Guide
to 19th Century Censuses for England and Wales, London: Routledge.
Long, Jason and Joseph Ferrie (2003): Labour Mobility, Oxford Encyclopedia of Economic History, New York: Oxford University Press.
McCord, Gordon G. and Jeffrey D. Sachs (2013): Development, Structure, and Transformation: Some Evidence on Comparative Economic Growth, NBER Working Paper
19512.
Mellinger, Andrew, Jeffrey D. Sachs, and John Luke Gallup (2000): Climate, Coastal
Proximity, and Development, in Clark, Gordon L., Maryann P. Feldman, and Meric S.
Gertler, eds.: The Oxford Handbook of Economic Geography, p. 169-194, 2000.
Pomeranz, Kenneth (2000): The Great Divergence: China, Europe and the Making
of the Modern World Economy, Princeton: Princeton University Press.
10
Pope, Rex (1989): Atlas of British Social and Economic History Since c.1700, London:
Routledge.
Preston, Samuel H. (1975): The Changing Relation between Mortality and Level of
Economic Development, Population Studies 29 (2), p. 231-248.
Redding, Stephen and Anthony J. Venables (2002): The Economics of Isolation and
Distance, Nordic Journal of Political Economy 28, p. 93-108.
Szreter, Simon and Graham Mooney (1998): Urbanization, Mortality, and the Standard of Living Debate. New Estimates of the Expectation of Life at Birth in NineteenthCentury British Cities, The Economic History Review 51 (1), p. 84-112.
11
Appendix
Table 4: Mortality and Population Data
City
Bristol
Clifton
Sheffield
Eccleshall
Newcastle
Gateshead
Leeds
Hunslet
Bradford
Birmingham
Aston
Manchester
Chorlton
Liverpool
West Derby
London
1851-60
35
42
34
40
34
37
34
38
37
35
42
30
37
27
38
38
Life Expectancy at Birth
61-70 71-80 81-90 91-1900
36
42
33
40
34
39
34
36
36
35
42
29
36
25
35
38
37
45
35
42
37
39
37
39
38
37
43
32
38
28
39
40
39
48
38
43
40
42
39
42
42
39
46
35
41
29
40
43
Data Source: Szreter and Mooney (1998)
12
43
50
39
46
42
44
41
40
44
38
45
36
42
30
41
44
Total Population (in thousands)
1851-60 61-70 71-80 81-90 91-1900
66
86
116
51
101
54
109
99
189
193
84
236
147
264
190
2,583
64
111
146
75
121
70
128
118
227
222
124
247
190
254
284
3,029
60
147
173
101
141
93
177
124
285
239
178
255
235
224
410
3,535
57
180
194
126
174
118
207
150
327
246
234
282
279
184
525
4,014
106
167
216
157
214
150
237
179
349
245
283
310
319
153
615
4,389
Figure 3: Cost Distance to Coal Reserves
13
Figure 4: Density Plots
Distribution of the density of turnpike road network, the navigable rivers network, and
coalfields as measured by the kernel density tool. More intense colors indicate a higher
concentration in the surrounding area.
14