Temporary Shocks and Persistent Effects in Urban Economies: Evidence
∗
from British Cities after the U.S. Civil War
W. Walker Hanlon
UCLA and NBER
February 22, 2015
Abstract
Can temporary shocks to city economies have persistent effects? If so, what mechanisms
drive these results? To shed new light on these questions, I study the impact of the
U.S. Civil War (1861-1865) on British cities. The Civil War caused a sharp temporary
reduction in the availability of raw cotton, the key input to Britain’s large cotton
textile industry. The result was a large temporary shock to cities with economies
specialized in that industry. Drawing on detailed new data, I show that this shock had
a persistent effect on the population of these cotton textile cities. I then document
the mechanisms behind this persistence. I show that the shock resulted in the loss of
firms, through bankruptcy in the cotton textile cities, with long-run implications for
industry employment. These effects were particularly pronounced in the machinery
& metals sector, which supplied capital inputs to cotton textile production. At the
same time, access to nearby cities with fast-growing economies pulled workers away
from the cotton textile districts. Understanding these mechanisms can help explain
the heterogeneous results found by previous studies in this literature.
∗
For their helpful comments and suggestions I thank Jeremy Atack, David Atkin, Leah Boustan, Ryan
Chahrour, Don Davis, Dave Donaldson, Pablo Fajgelbaum, Ronald Findlay, Richard Hornbeck, William Kerr,
Amit Khandelwal, Tim Leunig, Guy Michaels, Matthew Notowidigdo, Henry Overman, Humphrey Southall,
Eric Verhoogen, Jonathan Vogel, David Weinstein, Jeffrey Williamson and seminar participants at Caltech,
Columbia, Stanford, Vanderbilt, and UCLA. Reed Douglas provided excellent research assistance. Thanks
to Matthew Wollard and Ole Wiedenmann at the UK Data Archive for help with the data. Funding was
provided by the NSF (grant no. 0962545), the Economic History Association, the Ewing Marion Kauffman
Foundation, and the European Commission’s Marie Currie Initial Training Network (AMID Fellowship).
Some of the material in this paper has been previously circulated as a working paper under the title “Industry
Connections and the Geographic Location of Economic Activity.” Author contact information: 8283 Bunch
Hall, UCLA, 405 Hilgard Ave., Los Angeles, CA 90095, [email protected].
1
Introduction
How do cities respond to large temporary shocks to population or economic activity? Answering this question is central to understanding the forces that generate cities; if resource
endowments are the key determinant of city size then cities should rebound quickly in the
face of temporary shocks, while if increasing returns to scale is the key determinant then
sufficiently large temporary shocks should generate long-lasting effects. This has motivated
a substantial literature examining the long-term impacts of temporary shocks to cities. However, this literature has delivered mixed results, with some studies finding evidence of that
temporary factors can have long-lasting effects (e.g., Bosker et al. (2007), Bleakley & Lin
(2012), Redding et al. (2011), Kline & Moretti (2013)) while others show rapid recovery,
even from very large shocks (Davis & Weinstein (2002), Miguel & Roland (2011), Davis &
Weinstein (2008)).1 Further complicating matters, these studies consider a wide variety of
temporary events, ranging from wartime destruction to the loss of natural advantages. As a
result, it is hard to draw clear lessons from the existing literature.
Given these mixed and sometimes contradictory results, how can we move forward?
The answer offered by this paper is to document the long-run effects of a large temporary
shock in one setting, and then to dig into the mechanisms lying behind the observed effects.
Understanding the mechanisms can be useful in two ways. First, it can help us to understand
why previous studies have generated different results. Second, knowledge of the underlying
mechanisms can help us construct more accurate theories of city growth.
The event I consider is the large exogenous shock to British cities specialized in cotton
textile production in the 19th century resulting from the cotton shortage caused by the
U.S. Civil War (1861-1865). The lack of raw cotton caused a severe multi-year depression
in the cotton textile industry, putting hundreds of thousands out of work. This historical
setting offers a unique combination of features that allows me to assess both the persistence
of the temporary economic shock and the underlying mechanisms at work. First, while
the shock was severe, the underlying causes were also temporary; imports of raw cotton
dropped substantially during the civil war, and raw cotton prices increased dramatically,
but both cotton import quantities and prices returned to pre-shock levels within a decade
after the war. Second, there was substantial geographic variation in the incidence of the
shock, because the cotton textile industry was geographically concentrated in cities in the
Northwest region of England. This allows a difference-in-difference estimation strategy, and
in particular I am able to compare outcomes in the cotton textile cities to a set of similar
1
Another paper, Brakman et al. (2004), studies the impact of WWII bombing on German cities. They
find that the bombing had persistent effects for some cities but not for others.
1
cities with economies based on wool, linen, or silk textiles. These cities were economically
similar to the cotton textile cities, but their textile industries were not negatively affected by
the cotton shortage. Third, despite the magnitude of the event, there was little government
response due to the strong free-market ideology that dominated Britain during this period.
The limited government response eliminates an important potential concern for studies of
this type. Fourth, the direct impact of the Civil War was largely confined to the cotton
textile industry. This feature allows me to trace the secondary impacts as they cascade
through the local economies in order to identify some of the underlying mechanisms at work.
I begin by establishing that the cotton shortage had a long-term impact on the population
of cotton textile cities relative to other English cities. This is done by applying a differencein-difference approach to newly digitized data from the Census of Population. Specifically,
I find that population growth in the cotton textile cities dropped sharply during the 18611871 decade. The population growth rate in cotton textile cities rebounded after 1871, but
did not overshoot. As a result, the cotton shortage had a temporary growth impact and a
persistent impact on the level of city population. This result holds whether I compare the
cotton textile cities to all other large cities in England, or to the set of economically similar
cities with large textile industries based on inputs other than cotton (wool, linen, silk, etc.).
To strengthen these findings I show that similar results are obtained using synthetic control
methods.
Next, I examine the mechanisms driving the long-run geographic reallocation of population and economic activity that I observe using a combination of direct and indirect evidence.
I start by examining how events in the cotton textile towns led to reduced employment, focusing specifically on firm bankruptcies. The loss of local firms acted as a push factor,
driving population away from the cotton textile cities. Then, I consider the role played by
the availability of employment in nearby and economically similar towns. Economic growth
in nearby towns acted as a pull factor, providing an outlet for workers leaving the cotton
textile districts.
To analyze the impact of the cotton shortage on firms during the Civil War, new bankruptcy
data were collected covering all of England from 1862-1866, with over 40,000 entries in all.2
Because English bankruptcy law changed substantially at the end of 1861 it is not possible
to assess the overall impact of the Civil War on firm bankruptcies. Instead, I utilize an event
study approach based on two surprising news events related to the supply of raw cotton that
occurred late in the Civil War period. This allows me to provide clean causal evidence on
the link between expected raw cotton supplies and firm bankruptcies in the cotton textile
2
I cannot extend the data before 1862 due to an important change in bankruptcy law enacted in 1861.
2
cities. I show that these news shocks, which occurred in October, 1864 and March, 1865, led
to a spike in bankruptcies among cotton textile producers in those months. This was closely
followed by large numbers of bankruptcies in other sectors of the economy, particularly machinery & metals producers. The machinery & metals sector was a large and technologically
advanced part of the local economies which supplied the most important capital inputs to
cotton textile production, ranging from steam engines to textile machinery. That capital
suppliers were more affected by events in the cotton textile industry makes sense given that
we know that investment is volatile and procyclical.
The fact that the cotton shortage led to bankruptcies in industries other than cotton
textiles turns out to be particularly important. To see why, it is useful to contrast the
impact on the cotton textile industry with the impact on the machinery & metals sector.
The shortage of cotton affected cotton textile firms regardless of their location. As a result,
it did not differentially harm firms in the cotton textile towns. In the machinery & metals
sector, firms in the cotton textile towns supplied their local textile producers, but also
competed with firms in the non-cotton towns to supply the large national and international
markets. Because of their ties to local customers, this meant that machinery & metals firms
in the cotton towns were negatively affected by the cotton shortage while their competitors
in non-cotton towns were less affected. Thus, in this sector the cotton shortage differentially
harmed firms in cotton towns relative to their competitors in non-cotton towns.
To document the long-run impact of these differential effects, I use city-industry employment data from the Census of Population covering 1851-1891. Using panel data regressions
that compare industry employment in cotton textile cities to other cities, in pre-shock vs.
post-shock periods, I show that the machinery & metals sector suffered substantial negative
employment effects in cotton textile cities during the Civil War, and that these effects persisted through at least 1891. This result fits the patterns of bankruptcies observed during
the Civil War period, allowing me to link the contemporaneous bankruptcies in this industry
during the Civil War to long-term employment losses.
I also provide evidence that the availability of employment in other cities contributed to
the persistent impacts of the Civil War shock on cotton towns. To begin, I show that towns
that were both geographically close to the cotton textile cities and had similar economies
based on textile production experienced accelerated population growth during the Civil War.
This suggests that economically similar nearby cities benefited from the shock to the cotton
textile towns.
To connect the population gains in nearby cities to losses in the cotton textile cities, I
use data on the location of birth of city residents collected in each Census to infer internal
3
migration patterns. These data show that migration flows out of the cotton textile areas
increase during the Civil War decade. Net out-migration from cotton textile cities in 18611871 was equal to 0.7% of the total population of those cities, or 1.4% of the employed
population. At the same time, in-migration fell. There is no evidence of a reversal of these
patterns in later decades.
What factors determined the destination chosen by migrants leaving the cotton textile
cities during the Civil War? I consider four potential factors: (1) the geographically proximity
of destination cities to the cotton textile cities, (2) the economic similarity of destination
cities to the cotton textile cities, based on the size of the city’s non-cotton textile industry,
(3) previous migration flows from the cotton textile region, and (4) the growth in migrants
from other counties, which reflects overall economic growth in the city. I find that migration
patterns reflect all of these factors, but the only robust predictor of the destination cities
for migrants from the Northwest Counties was the growth in migrants into cities from other
districts. This indicates that pull factors related to the economic growth of cities was the
most important determinant of migration patterns, providing evidence that the availability
of other cities with growing economies contributed to the reallocation of population away
from cotton textile towns during the Civil War decade.
Within the literature on the impact of temporary events on city-size, my results are in
line with studies showing that temporary factors can have long-run effects, such as Bleakley
& Lin (2012) or Kline & Moretti (2013). My results contrast with the findings of Davis &
Weinstein (2002), who find no evidence that temporary shocks to city-size had long-term
effects. Understanding the mechanisms can help us reconcile these differences. My results
highlight that the existence of nearby cities that were not negatively affected by the cotton
shortage played an important role in generating persistent effects. Benefiting from weakened
competition, producers in these cities in sectors linked to cotton textiles, such as machinery
& metals producers, increased employment. This increased economic growth and drew in
new workers, many of whom came from the cotton textile cities. The relative gains made by
these towns during the Civil War period remained in the decades after the war.
These findings suggest that one explanation for why the findings of Davis & Weinstein
(2002) differ from my results, and those of some other studies, may be due to the lack of
alternatives to the affected cities in their context. This may be due to Japan’s mountainous topography (Head & Mayer (2004)), which contrasts with England’s much more subtle
geographic variation. Together, these studies suggest that locational fundamentals matter
when geography is sufficiently varied, but that settings where many locations share relatively
similar geographic features are characterized by multiple equilibrium city growth paths, as in
4
Krugman (1991), a story that is also consistent with the findings of Bleakley & Lin (2012).
This has implications for a growing literature focused on explaining city size and city growth
patterns (Gabaix (1999), Rossi-Hansberg & Wright (2007), Duranton (2007), Michaels et al.
(forthcoming), Desmet & Rappaport (2014)). As pointed out by Davis & Weinstein (2002), a
satisfactory explanation for city sizes or city growth should also be consistent with empirical
results documenting the impact of large temporary shocks on long-run city size.
My results can also be informative about the nature of agglomeration economies within
cities. Within this literature, there is ongoing debate about the importance of agglomeration
forces operating within an industry (sometimes called Marshall-Arrow-Romer spillovers) and
agglomeration forces across industries (Jacobs spillovers). Building on a large set of previous
work, a recent study by Ellison et al. (2010) provides evidence on the importance of crossindustry connections.3 This study contributes to this literature by providing clear causal
evidence of the importance of localized cross-industry connections. It is also the first study
within this literature to point out the particularly important role played by capital suppliers.
This study is closely related to previous empirical work examining the impact of temporary local labor demand shocks.4 Carrington (1996) studies the impact of the construction of
the Trans-Alaska Pipeline and focuses mainly on contemporaneous effects.5 Another closely
related paper is Black et al. (2005), which considers the impact of the coal boom and bust
on coal producing counties of the U.S. in the 1970s and 1980s. The key difference between
my study and these contributions is that I consider outcomes both during the shock period
and for several decades after.
There is also a growing literature examining the effect of trade on local labor markets
(Topalova (2007), Topalova (2010), Autor et al. (2012), Autor et al. (2013), Kovak (2013),
Dix-Carneiro & Kovak (2014)). Most existing studies in this vein consider permanent or longlasting changes in trade flows and show that they can have substantial economic effects. My
results contribute to this literature by showing that even temporary trade shocks can have
long-lasting effects.
The next section introduces the empirical setting, followed by the data, in Section 3, the
analysis, in Section 4, and conclusions in Section 5.
3
Other contributions to this literature include Greenstone et al. (2010), Faggio et al. (2013), and Hanlon & Miscio (2014). For a review of the broader urban literature on this topic, see Combes & Gobillon
(Forthcoming).
4
Another related strand of literature looks at long-run impacts to shocks to local resources. One example
is Hornbeck (2012), which shows that soil erosion during the Dust Bowl had a long run impact on local
population and economic activity. Michaels (2011) considers the impact of oil abundance on local economies
in the U.S. South. These studies differ from my work in that the changes they analyze were long-lasting.
5
Margo (1997) considers a similar temporary labor demand shock provided by the California Gold Rush.
5
2
Empirical setting
The cotton textile industry was a large and important sector of the British economy in the
second half of the 19th century. Cotton textile production was Britain’s largest manufacturing sector (by employment), cotton textile products were Britain’s most valuable export
good, and raw cotton was the country’s most important import. In 1861, just prior to the
U.S. Civil War, the industry employed 456,646 workers, equal to 2.3% of the total population
or England & Wales, or 9.5% of manufacturing employment.6
Links between the cotton textile industry and other sectors of the British economy play
an important role in this study. The most important of these were the linkages to machine
makers. Over time, many engine and machine making firms had grown up in the cotton
textile cities. These firms supplied their local textile producers with capital inputs, but also
competed with machine makers in other cities to supply the large national and international
markets. Cotton textile producers also purchased intermediate inputs such as coal, chemicals,
oils and soaps, and transportation and business services, while supplying downstream firms
in apparel, furniture, and papermaking.7
The shortage of cotton caused by the U.S. Civil War generated a shock that was both
large and temporary.8 Figures 1 and 2 illustrate the size of the shock as well as its temporary
nature. The left-hand panel of Figure 1 describes raw cotton imports into Britain. We can
see that the Civil War caused a sharp drop in the level of imports. At the same time, a
gap between imports from the U.S. and total imports opened up, as other suppliers such
as India and Egypt increased production. After the war, imports rapidly returned to their
original level. The right-hand panel of Figure 1 shows the import price. This price spikes
sharply during the Civil War. Prices remained high for a decade after the war, as U.S. output
recovered following the end of slavery, but by the mid-1870s prices had returned very close
to the pre-war levels. Figure 2 shows domestic raw cotton consumption, the best available
measure of output in the industry. This figure suggests that production dropped by as much
6
This figure includes only those employed in cotton textile manufacturing and excludes other closely
related industries such as cotton textile printing (12,556 workers) and cotton textile dying (4,772 workers).
Thus, it likely understates the industry’s importance.
7
Table 13 in the Appendix describes these upstream and downstream linkages in more detail.
8
This event was also largely unexpected. There is little evidence that the market anticipated the impact
of the Civil War. For example, raw cotton prices show very little increase in the first year of the war.
This is consistent with the historical record, which suggests that contemporary observers broadly failed to
anticipate the length and severity of the Civil War. For example, J.C. Ollerenshaw (1870, p.112), remarked
in his presentation to the Manchester Statistical Society that, “The American War commenced on April 5th,
1861, but for many months it had little effect on commerce - being generally regarded as merely temporary...”
A striking piece of evidence of this is underestimation of the magnitude of the impending conflict is the fact
that the initial Union Army enlistments were for only 90 days; it was assumed that the war would be over
before they expired.
6
as half during the war, but rebounded rapidly thereafter.
Figure 1: The impact of the U.S. Civil War on British cotton imports and prices
British raw cotton import quantities
Cotton prices on the Liverpool market
Import data from Mitchell (1988). Price data, from Mitchell & Deane (1962), are for the benchmark Upland
Middling variety.
Figure 2: Domestic raw cotton consumption in Britain
Domestic raw cotton consumption data, from Mitchell & Deane (1962), are the best available measure of
industry production.
The temporary nature of the changes in import quantities and prices may seem surprising
given the massive changes unleashed by the War, most importantly the end of slavery, which
led to a substantial reorganization in the U.S. cotton economy. The fact that this did not
have a major effect on import supplies or prices after the mid-1870s is due in large part to the
7
fact that new suppliers – particularly India – entered the market during the high-price years
of the Civil War and remained as suppliers thereafter. Contributing to this, new inventions
allowed British manufacturers take advantage of these new supplies (Hanlon (Forthcoming)),
while expanding rail networks opened up new productive regions. The result was that new
sources filled in for reduced U.S. supplies and import levels quickly rebounded. By 1876 the
benchmark cotton price had fallen back to the level observed in the year before the war.
An important feature of this setting is that the cotton textile industry was heavily geographically concentrated. Most cotton production took place in cities in the Northwest
region of England, comprised of the counties of Lancashire and Cheshire. According to the
1861 British Census, 82% of the cotton textile workers in England and Wales were located
in these two Northwest counties.9 In 1851, cotton textile production accounted for 29% of
employment in these counties. As a result, the impact of the cotton shortage was largely
confined to cities in the Northwest region. This fact is highlighted in Figure 3, which shows
the number of unemployed able-bodied workers seeking relief from local Poor Law Boards,
as a fraction of the total 1861 population, during the Civil War years. This is shown for the
cotton producing region (Northwest counties) the nearby wool textile region of Yorkshire
and all other English counties for which data are available.
Another important feature of this setting is that the shock was largely industry-specific.
This feature is important for my ability to identify the mechanisms through which the shock
generated persistent impacts. Figure 4 shows import and export data for Britain during the
Civil War period, where imports and exports related to textiles have been separated from
those related to all other industries. The left-hand panel shows that, once raw material
for textiles are removed, British imports suffered no noticeable effect from the U.S. Civil
War. Similarly, the right-hand panel suggests that once textile exports are excluded, British
exports also showed no negative effects.10
9
This pattern of concentration dates back to the late 18th century, and perhaps earlier. Using data from
the reports generated by the introduction of the Factory Acts in 1838, Crafts & Wolf (2013) show that this
pattern of geographic concentration was related to the availability of water power, the ruggedness of terrain
(which decreased the cost of land), proximity to a port (Liverpool), access to markets in other nearby cities,
and the area’s history of textile innovation in the 18th and early 19th century. While the Northwest region
also benefited from access to local coal deposits, many other regions did as well, and Crafts & Wolf find that
this had little impact on the location of the industry by 1838. By 1850 the importance of initial advantages
due to water power and cheap land had largely ceased to matter. Access to markets in other nearby cities
was also unlikely to have been an important concern in 1850 given that nearby markets provided only a
small fraction of industry demand. Thus, of the initial conditions that drove the location of the industry,
only access to the port of Liverpool still mattered by the period we study. Crafts & Wolf (2013) report that
Lancashire County contained 66% of Britain’s cotton textile spindles in 1850 and 79% in 1903. Their figures
appear to come from the British Parliamentary Papers for those years.
10
In the armaments and shipbuilding industries, where we would expect the war to have a more direct
effect, the impact was substantially muted by Britain’s policy of neutrality, which included a prohibition
8
Figure 3: Able-bodied relief-seekers as a share of 1861 population in Northwest counties,
Yorkshire, and all other English counties
Data from Southall et al. (1998).
A final useful feature of this empirical setting is that we can compare outcomes in the
cotton textile cities of Northwest England to outcomes in a set of similar textile-producing
cities with industries based on wool, linen, silk and lace, rather than cotton. Many of these
other textile cities were also geographically proximate to the Lancashire cotton cities, in
nearby Yorkshire County.11 Despite using different inputs, these other textile industries
shared many similarities with the cotton textile industry, including their technology, other
inputs such as coal and machinery, labor forces, employment practices, and organization.12
on providing weapons to either side (though we know that some producers managed to circumvent these
rules). One area where there was substantial changes was in ship transport, where many U.S. flagged vessels
were transferred to British ownership to avoid Confederate privateers. This may have had an impact on
the entrepot trade in port cities. Since none of the textile cities were major trading cities outside of their
textile-related imports and exports, this should not be a major concern when I confine the analysis to only
textile cities.
11
Wool textiles was the second most important textile industry in England during this period. The two
branches of this industry, Woolen and Worsted, employed 209,276 workers in 1861, equal to about 1% of
the total population of England and just over 4% of the industrial workforce. For historical reasons that
were likely similar to those that operated in Lancashire, though not as well studied, the industry was heavily
concentrated in Yorkshire. The 1861 census shows that 72% of the woolen textile workers and 90% of the
worsted textile workers in England and Wales were located in Yorkshire County. Wool textile production
accounted for 30% of employment in the industrialized West Riding region of Yorkshire. Other cities, such
as Derby, Norwich and Coventry specialized in silk, while Nottingham was a center for lace production.
12
Much of this similarity was driven by the adoption by other textile manufacturers of the innovations
generated in the cotton textile industry in the late 18th and early 19th century.
9
Unlike cotton, these other textile industries experienced little direct negative impact from
the U.S. Civil War, while they benefited from substitution away from more expensive cotton
textiles.13
Figure 4: British imports and exports 1851-1869
British imports
British manufacturing exports
Data from Mitchell (1988).
3
Data
The first data set used in this study describes the population of English cities every ten
years starting in 1841. This new database was collected and digitized from British Census
of Population abstracts. Because these data will be used to analyze patterns of overall
city population, it is important that we work with consistent geographic areas. To obtain
geographically consistent series, I take advantage of the fact that in each census report the
Census Office took the city boundaries for a set of major cities based in that year and then
went back to the previous census data and used the more geographically disaggregated data
to reconstruct the population within those boundaries in several previous decades. These
reconstructed city population data are available in two series, with the first spanning 18411891 and 46 cities and the second covering 1851-1901 and 55 cities.14 Thus, two or three
13
Graphs making this point are available in the Appendix. For example, imports of raw wool were
unaffected, since most of these imports came from Spain, Australia, South Africa, or South America. While
there was some effect on demand from the U.S., due to tariffs imposed to help fund the war effort, the U.S.
was a much smaller market at the time than it is today. Also, exports to European markets increased during
the period, particularly to France following a new trade agreement in 1860.
14
The data from the 1891 Census, which covers 1841-1891, reports population for 57 major cities based on
the 1891 city boundaries. Of these, I am able to identify the share of cotton textiles in total employment in
1851 for 46 cities. The data from the 1901 census reports population for 79 cities for the 1851-1901 period
10
observations are available prior to the U.S. Civil War, and it is possible to track impacts up
to 35 years after the War’s end.15
The Census of Population data also include information on the location of birth of the
residents of each location in each census year from 1851-1891. These data are used to estimate
the net flows of workers into a city from each English county over the course of each decade.
To analyze the role of specific industries, I use the occupation data, also from the Census
of Population reports, covering 1851-1891. The occupation data cover every individual,
including women and children, and in the vast majority of cases the reported occupations
closely correspond to industries. Examples include “cotton textile spinner”, “iron founder”,
or “boot and shoe maker”. Over time, the Census changed the set of occupational categories
reported, so to obtain consistent series I collapse from several hundred categories into 20
private sector industry categories that will be used in the analysis. These categories include
a substantial majority of the private-sector workers in the city economies.16 Table 14 in the
Appendix provides an overview of the industries included in the main city-industry dataset.
The set of cities for which occupational data are reported changes sharply in 1881, when the
census office begins reporting these data only for cities with a population over 50,000. Thus,
I will conduct analysis using two sets of city-industry data. One data set covers 1851-1871
and includes 71 cities, with 8 cities specializing in cotton textiles and 10 cities specializing
in non-cotton textile products. The second data set covers 1851-1891 and includes 31 cities,
with 6 cities specializing in cotton and 7 cities specializing in non-cotton textile products.
Unlike the city size data, the geographic extent of the cities borders used in collecting the
city-industry data expand over time as cities grow.
In addition to the Census data, I have also constructed a new set of data covering all
bankruptcies in England from 1862-1866. These data provide a uniquely rich and comprehensive way to track one important aspect of the shock in real-time. The roughly 40,000
bankruptcies in the database that I construct were hand-entered from announcements printed
in the London Gazette, England’s official government register. These data are available on
based on 1901 boundaries. Out of these, I can identify the 1851 cotton textile employment share for 55 of
these cities.
15
The 1861 observations were collected before the beginning of the U.S. Civil War and there is little chance
that these could have been substantially affected by expectations of the onset of the conflict. Thus, I treat
1861 as a pre-war observation.
16
The city-industry data used in all analysis exclude agricultural occupations, government workers, nonworkers, and a limited number of occupations such as ”Labourer” that do not clearly correspond to an
industry. The remaining categories correspond to those used in Hanlon & Miscio (2014) except that I have
collapsed the following similar categories to keep the set of results manageable: Road transport, Rail transport, Sea & canal transport, and Messengers, porters & storage into one Transportation Services category,
Chemicals & Drugs and Oils, Soaps, etc. into one Chemicals, Oils, etc. category, and Food, Drink and
Tobacco into one category.
11
a monthly basis and include the name, location, and occupation of the bankrupt individual.
The original data report around 15,000 unique occupations, which I have classified by hand
into industry categories that reflect those available in the Census data.
There are a few important facts to keep in mind when using the bankruptcy data. First,
there was a major change in bankruptcy law that came into effect in November of 1861.
This change expanded bankruptcy to cases where it had not applied before, leading to very
high levels of bankruptcy in late 1861 and 1862. As a result, in the analysis I use data
starting in June, 1863, when bankruptcies had returned to more normal levels.17 Second,
these were personal bankruptcy filings in a time in which the majority of firms were privately owned. Thus, they will capture most, but not all, firm bankruptcies. They may also
include bankruptcies by individuals who are not firm owners. For example, the category of
cotton textile related bankruptcies can include both the bankruptcy of a firm owner and the
bankruptcy of a foreman at the firm. In some cases, the occupation data allow us to identify
whether the person works for a business or is a business owner, but in general it is not
possible to systematically separate these two types of bankruptcies. This is not necessarily
a drawback of the data; even bankruptcies by individuals that were not business owners can
be revealing, since individuals are more likely to file for bankruptcy when they become unemployed, which in turn is often a reflection of the distress of their former employer. In one
respect this may offer a type of quality adjustment, since bankruptcies by larger firms are
more likely to result in bankruptcies by their workers. For further details on the bankruptcy
data, see Appendix A.2.
4
Analysis
The analysis is divided into two parts. The first part establishes that the temporary shock
had long-run effects on the size of the cotton textile cities relative to other English cities.
The second part seeks to understand the mechanisms driving these effects.
4.1
Persistent effects on city population growth
A good starting point for understanding the impact of the shock on city population growth
is to look at the average growth rates in the cotton cities and all of the other cities over each
decade in the 1841-1891 period. This is done in Table 1 using 46 cities. There are ten cotton
cities, defined as those with more than 10 percent of the working population employed in
17
See Figure 13 in the Appendix.
12
the cotton textile industry in 1851.18 There are eight other textile cities, defined as those
with more than 10 percent of the working population employed in any textile industry and
less than 10 percent employed in cotton textiles. In practice, these cities are dominated by
other textile industries and have fairly low levels of cotton textile employment (never more
than 2.8%). These definitions will be used throughout the paper.
The first pattern to take away from Table 1 is that, relative to all other cities or to just
other textile cities, the cotton cities in the Northwest of England suffered slower growth in
the 1861-1871 period. This was a reversal of the previous trend of faster growth in cotton
cities. After 1871, we see that growth in the cotton cities rebounds, but does not overshoot,
suggesting that population in these cities did not catch-up after the shock, at least through
1891.
Table 1: Average decadal population growth in cities
1841-1851
All cities (47)
25.1%
Cotton cities (10)
22.3%
Other textile cities (8)
19.0%
Non-textile cities(29)
25.8%
1851-1861
22.2%
20.4%
12.2%
22.7%
1861-1871
19.1%
10.8%
16.2%
21.3%
1871-1881
20.2%
18.3%
17.5%
20.8%
1881-1891
19.9%
13.0%
13.3%
21.8%
Figure 5 allows us to compare population trends in the cotton cities to the other textile
cities over the study period graphically. This figure presents the sum of log population for
the cotton and other textile cities across the entire 1841-1891 period.19 This graph shows
that there was little change in the population growth rate in the other textile cities over
this period, so that a trend-line based on the 1841-1861 period predicts population through
1891 reasonably well. For the cotton cities, we can see that the trend is fairly constant
in the 1841-1861 period, but slows substantially between 1861-1871. There is also weak
evidence that population growth remained lower in the cotton textile cities after 1871 than
would have been predicted based on the initial growth trend. Note that Figure 5 has been
constructed to allow comparison with a well-known graph from Davis & Weinstein (2002)
describing the population of Hiroshima and Nagasaki before, during, and after World War II
(their Figure 2). A quick visual comparison highlights the differences between their results
and the patterns that I observe.
18
This is a natural cutoff point, since there is a sharp drop-off in the share of cotton textiles in city
employment from Wigan, with 16.5%, to Warrington, with 7%.
19
Using the sum of log population here ensures that the patterns are not dominated by the large cities. It
also matches my empirical approach.
13
Figure 5: Population growth in cotton and other textile cities
Next, I explore these patterns using a regression approach. The baseline regression
specification is,
ln(P OPct ) − ln(P OPct−1 ) = β0 + β1 (CotT OW Nc ∗ Shockt ) + γc + λt + eit
(1)
where P OPct is the population of city c in period t, CotT OW Nc is an indicator variable for
the cotton cities, Shockt is an indicator variable for the shock period, γc is a full set of city
fixed effects, and λt is a full set of time effects. While I use a discrete indicator variable
for the cotton cities for the results presented in the main text, I have also calculated results
using a continuous measure of the incidence of the shock in each city based on the share of
employment in cotton textiles in the cities in 1851. Results based on a continuous measure
of city exposure to the shock, available in Appendix A.3, are very similar to those presented
in the main text.
The coefficient of interest in these regressions is β1 . In interpreting this coefficient, it is
important to keep in mind that, because all cities are operating within a connected economic
system, a negative shock to the cotton textile cities may generate positive effects for the noncotton cities. Thus, the β1 coefficient will reveal the change in relative growth rates between
14
cotton and non-cotton cities generated by the shock, which will be composed of both the
negative effects in the cotton cities and any positive impact of the shock on non-cotton cities.
Thus, this estimate will likely be larger than the change in the growth rate of the cotton
cities relative to their counterfactual growth in the absence of the shock. This would be a
serious concern if the goal was to estimate the exact magnitude of the impact of the shock
relative to a counterfactual of no shock. But it is not a serious concern if the goal is simply
to establish that there was a persistent relative change in city sizes. To be specific, the null
hypothesis I address is that the temporary shock had no impact on the population in the
cotton textile cities relative to other comparable English cities. This is different from the
null hypothesis that the temporary shock caused the population of the cotton textile cities to
differ from what it would have been in the absence of any shock. My empirical methodology
will address the first of these, but would not necessarily address the second.
Spatial correlation is a potential concern in this setting. To deal with this, I estimate
standard errors robust to spatial correlation up to 100km based on Conley (1999).20 These
spatial-correlation-robust standard errors are generally lower than the heteroskedasticityrobust standard errors, suggesting that errors are negatively spatially correlated. Thus, I
also present standard robust standard errors for all specifications. While serial correlation
can be an issue in panel data settings (Bertrand et al. (2004)), this is less likely to be a major
concern for the current study given that, in terms of observations (but not years covered),
the time-series dimension of the data is short relative to the number of cross-sectional units.
Table 2 describes results generated for different time periods using the specification in
Equation 1. Heteroskedasticity-robust standard errors are reported in single parentheses,
while spatial-correlation-robust standard errors are in double parentheses. Because I present
multiple standard errors for each regression, I will not follow the standard practice of using
* to indicate statistical significance levels.
The first column reports results from a placebo test using only data from 1841-1861,
prior to the war. We can see that the cotton cities do not exhibit statistically significant
differential growth patterns during this pre-period. The second column compares the decades
just before and just after the war. There is clear evidence that population growth in the
cotton cities fell in the 1861-1871 period relative to other cities. Column 3 expands the preperiod to include 1841-1851. The results in both columns 2 and 3 are statistically significant
at the standard 95% confidence level. Columns 4 and 5 use the full set of available data and
20
To implement this approach I follow Hsiang (2010). I have experimented with allowing correlation over
different distances and I did not find that this substantially affected the confidence intervals. I have also
experimented with allowing limited serial correlation based on the method from Newey & West (1987) and
this also does not seem to substantially impact the results.
15
estimate separate impacts for the 1871-1881 and 1881-1891 periods. Here, we are looking
for evidence that the cotton cities experienced faster growth after 1871 which might have
allowed them to catch-up to their previous growth path. The results provide no evidence
that any such catch-up took place, at least before 1891. This suggests that reduced growth
in 1861-1871 generated a persistent effect on the level of population in the cotton cities.
Table 2: Regressions of population growth in cotton vs. all English cities
(1)
Years
included:
Cotton cities
in 1851-1861
DV: City population growth rate in each decade
(2)
(3)
(4)
(5)
1851-1871
1841-1871
1841-1891
1841-1891
0.0161
(0.0351)
((0.0271))
1851-1901
-0.0864
(0.0271)
((0.0119))
-0.0783
(0.0293)
((0.0192))
-0.0783
(0.0294)
((0.0186))
-0.0702
(.0401)
((0.0246))
-0.0799
(0.0271)
((0.0217))
Cotton cities
in 1871-1881
-0.0056
(0.0311)
((0.0188))
0.0024
(0.0406)
((0.0247))
0.0015
(0.0312)
((0.0190))
Cotton cities
in 1881-1891
-0.0111
(0.0338)
((0.0219))
-0.0030
(0.0403)
((0.0272))
-0.0045
(0.0377)
((0.0191))
Cotton cities
in 1861-1871
1841-1861
0.0161
(0.0349)
((0.0223))
(6)
Cotton cities
in 1891-1901
-0.0462
(0.0417)
((0.0175))
City FEs
Yes
Yes
Yes
Yes
Yes
Yes
Time effects
Yes
Yes
Yes
Yes
Yes
Yes
Observations
92
92
138
230
230
275
Cities
46
46
46
46
46
55
Heteroskedasticity-robust standard errors in parentheses. HAC standard errors robust to
spatial correlation up to 100km in double parentheses. All specifications include a full set
of city fixed effects and year effects. The regressions in columns 1-5 use data from the 1891
census covering 1841-1891. The results in column 6 are based on a slightly different data
set from the 1901 census covering 1851-1901.
The results in Table 2 include the full set of cities for which sufficient data are available.
A potential concern here, as in all difference-in-difference analysis, is the parallel trends
assumption; for the procedure to deliver reliable results, the growth path of the untreated
cities, controlling for their initial growth rate, must provide a valid counterfactual for the
treated cities. We may worry that there is substantial variation in the underlying character16
istics of the cities that may lead to this assumption being violated. Seaports, for example,
may not be a good counterfactual for inland industrial cities. One way to address this concern is to confine the analysis to a subset of cities that are more similar to the treated cities.
This can strengthen our confidence in the parallel trends assumption, but it comes at the
cost of working with a somewhat reduced sample size. With this in mind, I now limit my
comparison to include only cotton textile towns and other towns with substantial textile
industries.
How economically similar were the other textile cities to the cotton textile cities? To shed
light on this question, I calculate the correlation across employment shares in 19 analysis
industries (treating textiles as one industry) between city pairs using city-industry employment data for 1851. I then look at the average, maximum, and minimum correlation across
all pairs of cities in a particular group or pair of groups. Table 3 describes these results.
The first row looks across all pairs of cotton textile cities. The second looks across pairs of
non-cotton textile cities. In the third row I look at all pairs comprised of one cotton textile
city and one non-cotton textile city. The fourth row does the same for pairs of cotton textile
cities matched with each non-textile city.
Table 3: Measuring the economic similarity across groups of cities
Summary statistics for industry employment share
correlations for city pairs of each type
Mean
Min
Max
Pairs of cotton cities
0.933 0.797 0.999
Pairs of non-cotton textile cities
0.833 0.452 0.997
Cotton and non-cotton textile cities 0.875 0.429 0.998
Cotton and non-textile cities
0.226 -0.081 0.780
The results in Table 3 show that the cotton textile and non-cotton textile cities were
economically very similar; the average correlation across pairs of cotton and non-cotton
textile cities is 0.875. This is nearly as high as the correlation between pairs of cotton textile
cities (0.933) and is higher than the correlation between pairs of other textile cities (0.833).
In other words, the non-cotton textile cities are, on average, more similar to the cotton
textile cities than they are to each other. The figures in the last row show that the cotton
textile cities are much less similar to the non-textile cities, with an average correlation of
only 0.226. These correlations suggest that the non-cotton textile cities provide a reasonable
control group for the cotton textile cities.
Table 4 presents results calculated by comparing the 10 cotton textile cities to the 8 other
17
cities where (non-cotton) textile production formed an important part of the economy. As
shown in Table 3, these non-cotton textile cities are economically very similar to the cotton
textile cities. This table shows that I obtain even stronger results when comparing the cotton
textile cities to this subset of economically similar cities.
Table 4: Regressions of population growth in cotton vs. other textile cities
(1)
Years
included:
Cotton cities
in 1851-1861
DV: City population growth rate in each decade
(2)
(3)
(4)
(5)
1851-1871
1841-1871
1841-1891
1841-1891
0.0486
(0.0576)
((0.028))
1851-1901
-0.1348
(0.0521)
((0.0394))
-0.1105
(0.0381)
((0.0348))
-0.1105
(0.0385)
((0.0403))
-0.0861
(0.0430)
((0.0365))
-0.1257
(.0523)
((0.0443))
Cotton cities
in 1871-1881
-0.0499
(0.0294)
((0.0237))
-0.0255
(0.0459)
((0.0164))
-0.0627
(0.0346)
((0.0155))
Cotton cities
in 1881-1891
-0.0611
(0.0333)
((0.0286))
-0.0368
(0.04998)
((0.0230))
-0.0765
(0.0373)
((0.0164))
Cotton cities
in 1861-1871
1841-1861
0.0486
(0.0566)
((0.0277))
(6)
Cotton cities
in 1891-1901
-0.0974
(0.043)
((0.0259))
City FEs
Yes
Yes
Yes
Yes
Yes
Yes
Time effects
Yes
Yes
Yes
Yes
Yes
Yes
Observations
36
36
54
90
90
90
Cities
18
18
18
18
18
18
Heteroskedasticity-robust standard errors in parentheses. HAC standard errors robust to
spatial correlation up to 100km in double parentheses. All specifications include a full set
of city fixed effects and year effects. The regressions in columns 1-5 use data from the 1891
census covering 1841-1891. The results in column 6 are based on a slightly different data
set from the 1901 census covering 1851-1901.
An alternative econometric approach to estimating the effect of this event is to use the
synthetic control method (Abadie & Gardeazabal (2003), Abadie et al. (2010), Abadie et al.
(2014)). Implementing this method involves combining all of the cotton cities into a single
composite cotton region.21 The composite cotton region is then matched to a synthetic
control constructed using a weighted combination of the available control units, where the
21
An alternative to aggregating the cotton cities to one composite cotton region is simply to run the
analysis on county-level data and combine Lancashire and Cheshire into one cotton textile county. However,
18
weights are constructed so that the synthetic control matches the composite cotton region
as closely as possible across a set of observable pre-treatment characteristics.
The synthetic control method offers two potential advantages in this setting. First, it
provides a transparent data-driven choice of control units and allows us to easily test how
well the synthetic control unit is matching the treatment cities across the set of available
features. Second, by treating the cotton cities as a single unit, it can help address concerns
that the cotton cities should be thought of as a single regional economy rather than a set
of independent city observations (despite substantial heterogeneity across the cotton cities).
The downside of the synthetic control method in this setting is that an analysis based on
region-level data involves relatively few observations, reducing the power of the exercise.
I implement the synthetic control method using city population data from 1841-1891,
aggregated to 9 regions, with one region, the Northwest, comprised of the cotton cities.22 I
consider two outcome variables, the log of population and the population growth rate. The
synthetic control is constructed by matching on industry employment shares in 1851 for 19
industries as well as the outcome variable in the pre-treatment period.
Figure 5 describes the results obtained using the synthetic control method. The top
panel of the figure describes results obtained with log population as the outcome variable.
The left-hand panel compares the actual log population in the cotton region to the synthetic
control. Starting in 1871 the actual population of the cotton region falls below the level that
we would expect given the synthetic control.23 The right-hand side of Panel A describes the
gap between the actual and synthetic control values for the cotton region (black line). To
provide a sense of the significance of these results, I also conduct a permutation exercise in
which synthetic controls are constructed for each of the treatment regions. The grey lines in
the right-hand graphs of Figure 5 describe the gaps obtained for the control regions.24 We
can see that the estimated gap for the cotton region lies below all of the placebo gaps.25
this approach generates misleading results because it ignores the substantial heterogeneity across cities within
these counties. In addition to the cotton textile towns, these counties also include important ports such as
Liverpool, rapidly growing industrial cities that had no cotton textile production such as Barrow-in-Furness,
as well as many smaller rural towns.
22
The regions are the Southwest, Southeast, London, East, West Midlands, East Midlands, Yorkshire, and
the North, plus the cotton cities in the Northwest.
23
A balance test, described in Appendix Table 19 shows that this combination comes close to reproducing
the actual cotton region values across most of the matching variables, with the exception of textiles, which
is impossible to match given that it is higher in the cotton region than elsewhere in England.
24
This permutation test follows Abadie et al. (2010). The gaps for two regions, London and the East,
are dropped from this figure following their advice, because the mean squared predicted error in the pretreatment period is more than 20 times larger than that of the cotton region, indicating that the synthetic
control for these regions does not perform well in this specification.
25
Given that there are 6 placebo gaps for which reasonable synthetic controls can be constructed, this is
consistent with a statistical significance level of at least 83.3 percent (1/6).
19
Table 5: Synthetic control results
Panel A: Results for log population
Cotton region actual population
Comparing cotton region population gap
and synthetic control
to alternative permutations
Panel B: Results for population growth
Cotton region actual growth
Comparing cotton region growth gap
and synthetic control
to alternative permutations
Top panel: Synthetic control based on matching industry employment shares in 1851 and region
population in the pre-shock period (1841-1861). Synthetic control weights are Yorkshire (0.524)
and London (0.476). Bottom panel: Synthetic control based on matching industry employment
shares in 1851 and population growth in the two pre-shock decades, 1841-51 and 1851-61. Synthetic
control weights are Yorkshire (0.647) and the Southeast (0.353).
The bottom panel of Figure 5 describes similar results obtained for city growth rates.
The left-hand panel suggests that the cotton region experienced substantially slower growth
in 1861-1871 than the synthetic control would lead us to expect. It is interesting to see that
growth in the cotton region falls in 1861-1871 while growth in the synthetic control jumps.
20
This reflects the displacement of workers from one region to the other, a feature that will be
revisited later. The right-hand panel shows that the (negative) gap between actual growth
and growth based on the synthetic control in 1861-1871 was lower in the cotton region than
we obtain when applying the synthetic control approach to any other region.26
This section has provided evidence the Civil War shock had a persistent impact on the
population of cotton textile cities in the following decades relative to other comparable cities.
There is no evidence that these effects diminished in the decades following the Civil War. In
the next section I look into the mechanisms that may be behind these persistent impacts.
4.2
Examination of the mechanisms
This section explores the mechanisms behind these effects. A natural starting point for this
analysis is to try to understand the impacts that the cotton shortage had in the cotton
textile towns during the Civil War period. This is done using bankruptcy data. Next, I link
these contemporaneous effects to long-run industry employment losses. Finally, I consider
the role played by non-cotton cities in generating the persistent city-size effects that I have
documented.
4.3
Bankruptcy analysis
This section examines the loss of firms during the Civil War period using bankruptcy data.
In particular, I focus on how events in the cotton textile industry generated bankruptcies in
other sectors of the local economies in the cotton textile cities. Drawing clear causal linkages
between bankruptcies in different sectors of the economy is difficult when firms operate in a
complex economic network. To overcome this challenge, I take advantage of two important
and unexpected moments of panic that occurred during the Civil War period and discretely
changed market conditions. These moments of panic can reveal the interlinked nature of the
local economies in the cotton textile cities and how the shock to cotton textile producers
spread across the broader local economies.
Specifically, I take advantage of events in October of 1864 and March of 1865 which led
to dramatic drops in the price of raw cotton, as shown in Figure 6. The first event was the
“peace panic” of October, 1864, which was generated by rumors that a peaceful end to the
Civil War was being negotiated. In a similar vein, the sharp fall in prices in the spring of
26
Given that there are 8 placebo gaps for which reasonable synthetic controls can be constructed, this is
consistent with a statistical significance level of at least 87.5 percent (1/8).
21
1865 was due to news that made it increasingly clear that the South could not win the war.27
Figure 6: Benchmark cotton prices, 1855-1869
Price data were collected from The Economist magazine. This price is for the benchmark variety, Upland
Middling cotton from the U.S., but the prices for other varieties show very similar patterns.
It may seem surprising that a fall in input prices led to panic and bankruptcy in the cotton
textile industry. However, manufacturers held a substantial amount of valuable cotton in
stock or in various stages or processing. The large drop in the value of these working capital
stocks led to bankruptcy, as described by Watts (1866) (p. 224) for the peace panic of 1864:
...a rumor crossed the Atlantic that men were meeting at Niagara Falls,
to try to arrange the terms of peace. Straightway men, instead of shaking
hands and throwing up their hats in thankfulness...looked into each other’s
faces with blank despair, as if peace, instead of war, was the greatest
curse upon earth. Nor was it without reason that this fear and terror was
expressed. Middling Orleans cotton fell from thirty-one pence to twentythree pence halfpenny...men who held largely of cotton, twist, or cloth
found their fortunes vanished in a night at the breath of this rumor. All
trade arrangements were in chaos. The workers on full-time were reduced
in two months by one hundred and forty-four thousand and fifty nine...
27
Watts (1866) writes, “In February, when it became evident that the Confederate government in America
must die, the fall of Richmond renewed the panic in this country, and again the prices of cotton and cloth
fell suddenly and considerably.”
22
We can get an idea of the loss in working capital value represented by the price declines in
October 1864 and March 1865 using data from Forwood (1870) on the quantity of inventories
held. Forwood reports that 266,866,000 lbs. of cotton and cotton goods were held in stock
at the end of 1864. If producers held a similar amount during the price fall in October 1864,
then their losses would have equaled 7.8 million pounds sterling. The price reduction in
March of 1865 was of a similar magnitude. For perspective, this implies that the lost value
of working capital during each of these price drops was equal to over 10% of the total value
of cotton goods produced in 1864 and 150% of the value of output less variable costs in that
year. Even compared to 1860, an exceptionally profitable year, the lost value was equal to
9.7% of total output value and 43% of the value of output less variable costs.
What this means for our purposes is that we can think of the timing of these two price
declines as largely exogenous, and driven by the changing fortunes of the cotton textile
industry. Thus, the responses observed in other sectors can be interpreted as the causal
effects of the shock working through linkages to the cotton textile industry.
Figure 7 describes the pattern of bankruptcies in cotton textiles (solid line) and in related
sectors (stacked bars) in major cotton textile cities during this period.28 The solid line shows
that there were high levels of bankruptcies in the cotton textile sector in both October of
1864 and March of 1865, as well as in June of 1866, corresponding to the three periods
in which the price of cotton rapidly declined. In contrast to the spikes in 1864 and 1865,
the increase in bankruptcies in mid-1866 was driven by forces outside of the cotton textile
sector, so this cannot be used to identify links between cotton textiles and other sectors of
the economy.29
Figure 7 shows a fairly clear pattern in which sharp increases in bankruptcies in the cotton
textile sector were accompanied, usually with a one-month lag, by spikes in bankruptcies
in other industrial sectors. Many of these additional bankruptcies are in the non-textile
manufacturing industries, though smaller increases also appear in construction, merchants,
and other textile industries. This suggests that the impact of events in the cotton textile
sector was rapidly transmitted to other sectors within the cotton textile cities.
28
This figure excludes a large number of bankruptcies by shopkeepers, beer sellers, farmers, government
employees and other miscellaneous sectors. Figure 14 in Appendix A.5 shows that there is at most a weak
relationship between bankruptcies in these other sectors and bankruptcies in the cotton textile industry.
29
Specifically, the panic in 1866 was precipitated by the failure of Overend, Gurney, & Co., a “bankers
bank” that played an important role in the British financial system at this time.
23
Figure 7: Bankruptcies in major cotton textile cities
The bar chart (left-hand axis) describes bankruptcies in a number of industrial sectors related to cotton
textiles. Solid line (right-hand axis) describes the number of bankruptcies in the cotton textile sector. The
cotton textile cities included in this chart are Accrington, Ashton-under-Lyne, Blackburn, Bolton, Burnley,
Bury, Manchester, Oldham, Preston, Rochdale, Warrington, Wigan, Stockport and Chester. All of these are
in Lancashire except for the last two, which are in Cheshire.
I now use a regression approach to estimate how these shocks to the cotton textile sector
impacted other local industries. Because I am dealing with count data that often take small
values, I use Poisson regressions, specifically the Poisson Conditional Fixed Effects Maximum
Likelihood Estimator (PCFE). The regression specification is,
BKcti = φc exp αi CotT OW Nc ∗ P anict + ηt
(2)
where BKcti is the count of bankruptcies in industry i, city c and month t, φc is a set of
city fixed effects, CotT OW Nc identifies cotton cities, P anict is an indicator variable for
November, 1864 and April, 1865 (the months after the cotton price drops generated a spike
in cotton textile bankruptcies), and ηt is a full set of time effects.30 Including the time
effects allows me to control for economy-wide factors, such as changes in the interest rate,
that may have affected the level of bankruptcies.31 This regression is run separately for each
30
31
In the Appendix I investigate some alternatives to the one-month lag used in the main specification.
A chart showing interest rate levels over Civil War period is available in the Appendix.
24
industry to obtain an industry-specific αi describing whether each industry was impacted in
the month after each of the panics.
The PCFE estimator has the advantage that it is robust to time dependent errors as long
as the conditional mean is correctly specified (Bertanha & Moser (2014)). Spatial dependence
may be a concern, but Bertanha & Moser (2014) show that as long as the pattern of spatial
dependence is time-invariant the standard sandwich PCFE estimator is consistent. My data
pass Bertanha & Moser’s test for time-varying spatial dependence, so the results below use
the standard sandwich estimator.
The industry definitions used in these bankruptcy data follow those used in the cityindustry employment data built from the Census returns. However, because of sparseness in
the data at the city-industry level, several of the smaller industries, those with few bankruptcies, have been aggregated.32
The results are shown in Table 6. Note that the number of observations used in each
regression changes, as some industries show no bankruptcies in some cities over the period
covered by these data, resulting in these cities being dropped. The first column of the top
panel shows that bankruptcies increase in the cotton textile cities in the month following the
panics. In particular, the incidence rate ratio for the arrival of a (non-cotton) bankruptcy
in a cotton textile city during a panic period is 1.59. The remaining columns look at the
response of particular industries in the month after the panics.
32
Wool, linen/flax, and silk textiles have been aggregated into a single “Other textiles” category, while
road, rail and sea & canal transport, and messengers, warehousing, etc. have been aggregated into a single
“Transport” category. Two similar manufacturing industries, “Oils, soaps, etc.” and “Chemicals & drugs”,
have been combined. Several industries with too few bankruptcies have not been included in the analysis:
“Earthenware & bricks,” “Mining related,” “Professionals,” “Shipbuilding,” “Water & gas service,” and
“Tobacco.”
25
Table 6: Bankruptcies by industry in the month following the cotton industry panics
Industry:
CotPanic
Obs.
Cities
Industry:
CotPanic
Obs.
Cities
All except
cotton tex.
0.468***
(0.169)
3,402
81
Agriculture
Construction
0.315
(0.886)
2,520
60
Chemicals
& oils
0.327
(0.571)
2,142
51
Food
processing
-0.550
(0.637)
3,276
78
0.157
(0.447)
3,318
79
Clothing
shoes, etc.
0.717
(0.665)
3,066
73
General
services
0.343
(0.565)
3,192
76
Instruments
& jewelery
0.665
(0.736)
1,890
45
Drinks
0.321
(0.624)
2,646
63
Leather
hair
0.659
(0.683)
2,016
48
Merchant
agent, etc.
0.608**
(0.294)
2,814
67
Metal &
machinery
1.305***
(0.323)
3,192
76
Industry:
Non-cotton
Paper &
Shopkeepers,
Textile
Transport
Wood &
textiles
publishing
etc,
finishing, etc.
furniture
CotPanic
0.741
0.894
0.425
1.076
0.875
-0.324
(0.650)
(0.711)
(0.419)
(0.859)
(0.722)
(0.502)
Obs.
1,260
2,226
3,360
1,680
2,310
2,478
Cities
30
53
80
40
55
59
*** p<0.01, ** p<0.05, * p<0.1 Poisson regressions run using the specification shown in Equation
2. Robust standard errors shown in parentheses. The data cover 59 cities from July, 1863 to
December 1866. Some observations may be dropped from a regression because some industries
show no bankruptcies in particular cities over the study period.
While most of the industries in Table 6 show an increase, only two – merchants, agents,
etc. and machinery & metals – exhibit statistically significant increases in bankruptcies. The
first of these provided business services to the cotton textile industry. The second provided
capital goods. The coefficient on bankruptcies by machinery & metals firms suggests an
incidence rate ratio of 3.7, implying a substantial increase in bankruptcies in that industry
in cotton textile cities in the month following each panic. A variety of robustness checks
confirming these results are presented in Appendix A.5.
The patterns documented here also reflect broader distress that extending even to firms
that were large and stable enough to avoid bankruptcy. For example, a history of the firm
Dobson & Barlow of Preston (a cotton textile city), the second largest textile machine maker
in England at the time, describes how “enforced ‘short time’ in the [cotton textile] mills reacted on the Textile Machinery Trade, and towards the end of the Civil War a large number of
the working classes were on the verge of destitution.”33 This caused numerous labor disputes
between the company and its workers as, “On the resumption of work the employees would
33
See dob (1927), pages 106-107.
26
naturally make some effort to remedy the loss of wages during the hard times.”
This section has provided evidence that events during the Civil War period led to
bankruptcies in the cotton textile cities, both among cotton textile producers, and, crucially, among their capital suppliers in the machinery & metals sector. Next, I tie these
bankruptcy patterns to industry employment during and after the decade spanning the Civil
War.
4.4
Industry employment effects
How did the bankruptcy patterns documented in the previous section translate into changes
in city employment during and after the Civil War decade? To address this question, this
section uses city-industry employment data from the Census of Population for 1851-1891.
As a first step, I quantify the contribution of employment growth in some broad industrial
sectors to overall city employment growth. I then turn to more detailed data and apply a
difference-in-difference approach to identify the industries most affected by the Civil War in
the cotton textile towns.
Table 7 decomposes city employment growth into the contribution of three broad sectors
of the city economies: cotton textile manufacturing, other textile manufacturing, and nontextile manufacturing.34 The tables is read as follows: if an industry represents 10% of
employment in a set of cities at the beginning of the decade, and grows by 10% over the
course of the decade in those cities, then the contribution of that industry to employment
growth in that set of cities is 1%. Thus, the figures in this table reflect both the initial size
and the growth rate of each sector of the economy, which together determine the sector’s
contribution to city employment growth. The first panel focuses on the six cotton textile
cities available in the 1851-1891 city-industry database. The second panel conducts a similar
exercise for the non-cotton textile cities, while the last panel includes all cities other than
the cotton textile cities.
In the cotton textile cities, employment growth prior to the U.S. Civil War was driven
by cotton textile manufacturing and other non-textile manufacturing. However, the growth
contribution made by the cotton textile industry dropped sharply in 1861-1871. Clearly the
direct shock to the cotton textile industry had an important effect on overall employment
growth in these cities. The contribution of cotton textiles to employment growth continued
to fall in 1871-1881, due in part to poor macroeconomic conditions from 1873-1879, a period
called the “Long Depression”. Cotton textiles employment growth then recovered after 1881.
34
A table showing a decomposition including all sectors of the economy is available in the Appendix.
27
Importantly, employment growth in non-textile manufacturing industries also slowed
down substantially during the 1861-1871 period. This fall cut an average of over 4% off
of employment growth in the cotton textile cities, which is roughly as large as the direct impact of the cotton textile industry itself. Non-textile manufacturing growth then increased
after 1871, but the growth contribution of this sector remained below the initial rate through
1891. This pattern suggests that the shock had an important indirect effect on city growth in
the cotton textile cities, through these other manufacturing industries. Comparing this pattern to that observed in the bottom panel shows that the reduced growth in the non-textile
manufacturing sector in the cotton textile cities was not driven by national-level changes in
the growth contribution of this sector.
The story is quite different in the other textile cities described in the middle panel. There,
the 1861-1871 period was characterized by a sharp increase in growth in non-cotton textile
manufacturing. An even more interesting pattern is visible for the non-textile manufacturing
industries, which experienced a substantial increase in their growth contribution in 18611871 and a sustained higher level through 1891. This persistent increase contrasts with the
persistent decrease in non-textile manufacturing experienced in the cotton textile cities.
Table 7: Decomposing the growth contribution by sector
Each cell of these tables represents the contribution of an industry to city growth over the period indicated.
So if an industry represents 10% of employment and grows 10% over the period, that industry’s contribution
to city growth is 1%.
The bottom line from Table 7 is that there was a substantial reduction in employment
28
growth during the 1861-1871 period in cotton textile cities, and that this slowdown was driven
in roughly equal shares by the cotton textile industry and other non-textile manufacturing.
While growth in both sectors rebounded in the post-war period, they often lagged behind
the growth rates achieved in other textile cities; there is no indication of the high growth
rates needed to catch up to the previous growth path.
Next, I conduct a more rigorous analysis using more detailed city-industry data. I start
by looking at short-run effects using data for 1851-1871 covering 71 cities, with 8 cotton
cities and 10 other textile cities.35 With only three time observations, I apply a differencein-difference approach:
ln(EM Pict+1 ) − ln(EM Pict ) = α +
X
ai (CotT OW Nc ∗ Y R1861−71 ) + ηit + ict
(3)
i
where EM Pict is employment in industry i, city c, and period t, CotT OW Nc is an indicator
variable for cities with more than 10% of employment in cotton textiles in 1851, Y R1861−71 is
an indicator variable for the 1861-1871 period, and ηit represents a full set of industry-time
effects.36 The ai coefficients in this regression allow me to identify the growth response of
each industry in the cotton textile cities in the post-shock period, controlling for national
industry growth rates.
To look at long-run employment effects, I use data covering 1851-1891 and run regressions
on levels, rather than growth rates. The specification is,
ln(EM Pict ) = a +
X
bi CotT OW Nc ∗ P OSTt + ηit + ict
(4)
i
where P OSTt is an indicator variable for the decades after 1861.
Both serial and spatial correlation in errors are a potential concern in these regressions.
To deal with these, I apply multi-dimensional clustered standard errors following Cameron
et al. (2011) and Thompson (2011). Errors are clustered by city-industry to deal with serial
correlation, by city-year, to allow correlated errors among industries within the same city,
and by industry-year, to allow cross-city spatially correlated errors at the industry level.
For both sets of regressions, I focus primarily on results comparing the cotton textile
35
Recall that cotton cities are those with more than 10% of their employment in cotton textiles. Other
textile cities are those with more than 10% of their employment in textiles but less than 10% in cotton
textiles. In general the share of cotton in these cities is low.
36
Note that I could potentially include city-year effects here. These are not included because I want to
allow for the possibility that many (perhaps all) industries grew more slowly in a particular city in a period,
a possibility that would be eliminated if we control for overall city growth in a year.
29
cities to the set of other textile cities.37 Table 8 describes the results. Columns 1-2 present
the coefficients and standard errors for the short-run difference-in-difference regressions (Eq.
3), while Columns 3-4 present the results for the long-run levels regressions (Eq. 4).
Table 8: Impact on industry employment growth in cotton cities
*** p<0.01, ** p<0.05, * p<0.1. Column 1 presents the ai coefficients for the specification in Eq. 3 with
corresponding standard errors in Column 2. Column 3 presents the bi coefficients for the specification in Eq.
4 with the corresponding standard errors in Column 4. The regressions include a full set of industry-year
effects. Standard errors, shown in parentheses, are clustered by city-industry, city-year, and industry-year.
The omitted industry in both regressions is “General services.”
Table 8 shows that the largest and most robust impact of the shock on employment
in cotton textile cities, relative to other cities, was the reduction in employment in the
machinery & metals sector. This finding fits nicely with the results obtained from the
bankruptcy analysis. It is also notable that I do not observe a reduction in employment in
the cotton textile sector in the cotton textile cities relative to other cities. This highlights a
key difference between the cotton textile industry and other sectors; cotton textile producers
37
Table 23 in the draft presents results obtained using all available cities in England. These also show
significant negative effects on employment in the machinery & metals industry in both the short and long-run.
30
were directly impacted by the cotton shortage regardless of their location. In contrast,
firms in other industries experienced a differential impact depending on the importance of
their ties to the cotton textile industry. Because machinery & metals producers located in
cotton textile towns were likely to have stronger ties to the cotton textile industry than
their competitors located in other textile towns, the impact of the shock on that industry
was larger in the cotton textile cities than in other textile cities. My findings suggest that
this resulted in a loss of competitiveness – for example through bankruptcy and the loss
of organizational capital – which had a long-term impact on the geographic location of
employment in this industry.
Given the consistent negative impacts I observe on producers in the machinery & metals
sector, it is worth considering this industry in more detail. This was an important sector
in the British economy during this period and one that was closely connected to the cotton
textile industry.38 It was also a major employer in the six cotton textile cities studied in
Table 8. Out of the 30 industries in the analysis database, metal & machinery firms made
up between 3.3 and 10.4 percent of employment in the cotton textile cities and were always
among the five largest employers.
Figure 9 describes the growth path of employment in the metal & machinery sector in
the cotton cities, the non-cotton textile cities, and all non-cotton cities, over the 1851-1891
period. Employment is given as the sum of log employment across the cities in each group,
so that the larger cities do not dominate the results, and for comparability each series has
been normalized by its 1851 value.39 Prior to the Civil War, we can see that this industry
was growing more rapidly in the cotton textile cities. After 1861, growth in the cotton textile
cities slowed down sharply, while growth in the non-cotton cities remained high. Overall,
it appears that the Civil War period marked a relative decline of the growth of metal &
machinery employment in the cotton textile cities.
38
Allen (2009) (p. 273) writes that, “the great achievement of the British Industrial Revolution was, in fact,
the creation of the first large engineering industry that could mass-produce productivity-raising machinery.”
Historical evidence suggests that the metal & machinery sector had important ties to the cotton textile
industry, ranging from textile machinery and tools to the steam engines that powered the textile factories.
Farnie (2004) writes that, “Textile engineering became the most important of all the ancillary trades [to
cotton textile production]. Its light engineering section supplied spinning machines and looms and a while
succession of related equipment, while its heavy engineering industry supplied steam engines, boilers, and
mechanical stokers.” There is also evidence that these producers gained from close proximity to their cotton
textile customers. Textile machinery makers in Bolton, for example, specialized in producing machines for
spinning fine thread counts, a product in which Bolton textile spinners were dominant, while in Oldham,
machine makers specialized in machinery for spinning the coarser thread count products that were largely
produced by textile factories in the Oldham area.
39
Note that focusing on log employment is also consistent with my econometric approach.
31
Table 9: Employment in the metal & machinery sector by city type, 1851-1891
Figure displays the sum across log employment in each set of cities, normalized by the 1851 level. The data
include the 31 cities for which consistent city-industry data are available from 1851-1891. There are 6 cotton
textile cities and 7 non-cotton textile cities.
How large were these effects quantitatively? One way to assess this question is to suppose
that employment in the metal & machinery sector grew at the same rate in the cotton textile
cities as in the other textile cities after 1861. In this case, it would have added over 13,000
metal and machinery workers to the economies of the 6 cotton textile cities in this analysis
in 1871, rising to 24,000 in 1881 and 22,000 in 1891. This is equal to 3-5.4% of total private
sector employment or 4.2-7.6% of manufacturing employment in these cities. These numbers
are not large enough to explain the full employment impact of the shock in these cities,
suggesting that either other mechanisms were also at work or there was a local multiplier
effect of losing employment in this industry. However, a long-term loss of 3% of total local
employment seems substantial, particularly in an innovative and relatively high-wage sector.
4.5
Pull factors, non-cotton cities and migration patterns
The previous section documents the short and long-run impacts of the cotton shortage on
cotton textile towns relative to other economically similar English cities. Now, I consider
how the availability of economically-similar nearby cities contributed to the persistent effects
I observe.
I start by looking at the characteristics of the non-cotton cities that experienced an
acceleration in growth during the Civil War decade, using the Census city population data.
32
After dropping the cotton textile cities, I consider two factors that predict accelerated growth
in the remaining cities: economic similarity to the cotton cities, as measured by the size of the
city’s non-cotton textile industry, and geographic proximity to the cotton textile district.40
The regression specification is,
∆ ln(P OPc1871 ) − ∆ln(P OPc1861 ) = b0 + b1 DISTc + b2 T EXc + b3 (DISTc ∗ T EXc ) + ec
where ∆ ln(P OPct ) is the change in log population in city c from decade t − 1 to t, DISTc
is the city’s geographic proximity to the Northwest Counties, and T EXc is the share of
non-cotton textile production in city employment.
The results are presented in Table 10.41 Columns 1-2 suggest that geographic proximity
and economic similarity to the cotton textile cities are correlated with accelerated city growth
during the U.S. Civil War, but these results are very weak. However, when these factors
are interacted, in Column 4, we see strong evidence that cities that were both geographically proximate and economically similar to the cotton textile towns experienced accelerated
population growth during the Civil War decade.
There are two potential explanations for the results in Table 10. On one hand, this
could be due entirely to push factors, i.e., population growth in these cities could be due
primarily to arrival of workers that were pushed out of the cotton textile cities due to lack
of employment during the cotton shortage. On the other hand, pull factors could have
also mattered. In particular, if producers in nearby non-cotton cities benefited from the
weakness of their competitors in the cotton cities, due, for example, to bankruptcies, then
this increased economic growth could have pulled workers out of the cotton textile cities.
Surely push factors played an important role here, but did pull factors matter as well? To
help answer this question, I now turn to data on migration patterns.
The migration patterns can help reveal the factors that contributed to worker movement
out of the cotton textile cities. The analysis relies on Census of Population data reporting the
county of birth of each resident in the analysis cities. Substantial changes in the population
of a city that report a particular county of birth between two census years can be used to
infer net population flows. Because these data are provided by city of residence but report
40
Liverpool is also dropped from this analysis, since the economy of that city was heavily reliant on the
cotton textile industry despite the fact that little actual textile production took place within the city.
41
These results are presented with robust standard errors. We may be concerned about spatial correlation
in these regressions. I have calculated additional results allowing for spatial correlation across cities within
100km of one another. These deliver similar but slightly stronger (more statistically significant) results than
those described in Table 12, though we may be concerned about how well spatially correlated standard errors
perform at the sample size available here.
33
Table 10: Factors related to increased growth during the Civil War decade among non-cotton
cities
DV: Growth in city population 1861-1871 relative to 1851-1861
(1)
(2)
(3)
(4)
Distance (std)
0.0266
0.0113
0.0392
(0.0201)
(0.0207) (0.0232)
Non-cotton textile employment share (std)
0.0382
0.0330
-0.0412
(0.0272) (0.0308) (0.0385)
Distance x Non-cotton textile employment share
0.0787**
(0.0335)
Constant
-0.00712 -0.00712 -0.00712 -0.0426*
(0.0196) (0.0190) (0.0193) (0.0244)
Observations
35
35
35
35
R-squared
0.052
0.106
0.113
0.233
*** p<0.01, ** p<0.05, * p<0.1 Robust standard errors in parenthesis. Regressions are run
across 35 non-cotton towns, with Liverpool excluded. The distance variable is exp(−distc )
where dist is the distance, as the crow flies, from city c to Manchester in thousands of
kilometers. The non-cotton textile employment share is based on data from 1851. Both the
distance and textile employment share variables are standardized.
only county of birth, they are informative about migration destinations but less informative
about the cities from which migrants originated.42
Before looking at the factors that determined migration destinations, I need to establish
that substantial migration from the cotton textile districts (the Northwest Counties) took
place. The left-hand panel of Figure 11 plots the number of residents of other English
counties who were born in the Northwestern Counties in each year. We can see that the
Civil War period was characterized by a sharp increase in residents of other counties who
were born in the Northwest, suggesting that out-migration took place during this period.
The number of people born in the Northwest but living in other counties increased by around
21,000 from 1861-1871, equal to a movement of about 0.7% of the 1861 population or 1.4% of
the 1861 working population of the Northwest counties. This compares to an increase of just
over 1,600 from 1851-1861. Note that, because these data span the 1861-1871 decade, they
will understate migration during the shock period if it was accompanied by return migration
before 1871.
Much of this out-migration flowed to the economically similar cities in nearby Yorkshire
County. The right-hand panel of Figure 11 describes the change in the number of Northwest
residents reporting Yorkshire as their location of birth and the change in the number of
42
The data do not provide information on the characteristics of migrants. It may be possible to obtain
information on migrant characteristics using the original household census data returns, but this would be
a very time consuming endeavor.
34
Yorkshire residents reporting the Northwestern counties as their location of birth, in each
decade. Both of these series reflect net migration from Lancashire to Yorkshire during the
Civil War period. Also, though the original patterns of migration resume after 1871, there
is no evidence of the overshooting needed to return the locations to the original positions,
suggesting that the migration flows that occurred in the 1861-1871 period were not reversed
in later decades.
Table 11: Evidence on migration patterns
Residents of other English counties
that were born in the Northwest region
Residents of the Northwest and
Yorkshire born in the other region
Data collected from Census of Population reports.
Next, I examine the factors that predict the destination of migration flows out of the
Northwest region. I use city-level data on the location of birth of city residents from 18511871. The specification is,
∆ ln(N W bornc1871 ) − ∆ ln(N W bornc1861 )
= α0 + α1 DISTc + α2 T EXc
(5)
+ α3 P AST migc + α4 ∆ ln(M IGotherc1871 ) + ect
where ∆ indicates a difference operator between decade t and t − 1, N W bornct is the count
of Northwest-born residents in city c and decade t, DISTc and T EXc are defined as before,
P AST migc is the count of current residents in city c who were born in the Northwest
Counties, and M IGotherct is the share of the city population made up of people born outside
the county in which city c, excluding those born in the Northwest Counties.43 The last term
43
The variable P AST migc is the share of city residents born in the Northwest Counties in 1851. This
35
is particularly important because it will help us differentiate between push and pull factors
in determining migration patterns. If push factors, together with features such as geographic
proximity or economic similarity, are the key driver of migration, then in cities that receive
more migrants from the Northwest Counties we should also observe fewer migrants from
other counties. In this case I should estimate α4 < 0. On the other hand, if pull factors
such as faster economic growth are an important determinant of migration patterns, then
cities that attract more migrants from counties outside of the Northwest region should also
attract more of the migrants leaving the Northwest Counties. Thus, an estimated α > 0 is
indicative of an important role for pull factors.
Results are shown in Table 12.44 Columns 1-4 describe regressions including each of
the migration determinants separately. These provide evidence that all four factors are
positively correlated with increased migration flows during the Civil War, though geographic
proximity and city growth, as indicated by in-migration from other counties, have the greatest
explanatory power. In Column 5 we combine all four factors. Only migration from other
counties into a city provide a statistically significant predictor of migration patterns from the
Northwest Counties during the Civil War. This suggests that pull factors – the availability
of fast-growing nearby cities – played an important role in providing an outlet for migrants
leaving the Northwest Counties.
This section has provided evidence that cities that were geographically close and economically similar to the cotton textile cities experienced increased growth as a result of the U.S.
Civil War, and that this increased growth acted as a pull factor that drew in migrants from
the cotton textile districts. The movement of workers during the 1861-1871 period was not
reversed in later decades, so the shock generated a permanent displacement of population.
These findings highlight the important contribution made by the availability of alternative
locations to generating the persistent changes caused by the temporary shock.
reflects past migration patterns, which have been shown to be an important determinant of future migration
flows (Bartel (1989), Altonji & Card (1991)).
44
These results are presented with robust standard errors. We may be concerned about spatial correlation
in these regressions. I have calculated additional results allowing for spatial correlation across cities within
100km of one another. These deliver similar but slightly stronger (more statistically significant) results than
those described in Table 12.
36
Table 12: Factors affecting migration destination of Northwest-born population during the
Civil War
DV: Decadal growth in share of Northwest-born residents in city population
in 1861-1871 compared to 1851-1861
(1)
(2)
(3)
(4)
Distance
0.289**
(0.111)
Non-cotton textile employment share
0.165**
(0.0810)
Initial NW pop.
24.20***
(8.573)
Growth of migrant population from other counties
0.306**
(0.120)
Constant
-0.0759
-0.0671
-0.307*
-0.113
(0.0959) (0.0993)
(0.166)
(0.0945)
Observations
48
48
48
48
R-squared
0.166
0.057
0.114
0.173
(5)
0.188
(0.146)
0.0150
(0.0658)
2.519
(10.72)
0.229*
(0.131)
-0.133
(0.147)
48
0.254
Robust standard errors in parentheses. Data cover 48 cities from 1851-1871. Liverpool is excluded because,
even though it did not produce cotton textiles, its economy was dependent on the cotton textile industry.
The distance variable is exp(−distc ) where dist is the distance, as the crow flies, from city c to Manchester
in thousands of kilometers. The non-cotton textile employment share is based on data from 1851. Both the
distance and textile employment share variables are standardized. Initial NW pop. is the share of NW-born
residents in total city population in 1851. Growth in share of population from other counties includes people
born in all counties other than the county in which a city is located and the Northwest Counties, as a share
of the city’s overall population.
5
Conclusions
This paper draws on a unique historical setting in order to provide well-identified evidence
of the persistent impacts of a temporary shock to city economies. My results show that the
economic shock caused by the Civil War resulted in a decrease in the population of cotton
textile cities relative to other English cities, and that these losses persisted through at least
1901 with no signs of diminishing. This pattern is consistent with models characterized by
multiple equilibrium city-size growth paths.
These findings contrast with some previous results, including the findings of Davis &
Weinstein (2002). Understanding the mechanisms at work in the setting I consider can help
us make sense of these differences. My results, together with previous findings, suggest that
locational fundamentals matter when geographic features are sufficiently varied, but that in
a settings where geographic variation across locations is modest, city growth is characterized
37
by multiple equilibrium growth paths.
This study provides a two useful ancillary findings. First, it shows that temporary trade
shocks can have long-term impacts on the geographic location of population and economic
activity. This result is interesting because economies are regularly exposed to temporary
shocks of this type. Further work is needed to determine the welfare consequences of such
events, though recent studies such as Notowidigdo (2013) and Yagan (2014) have made
progress in this direction.
Second, this study provides clear causal evidence that shocks to one sector of a local
economy can be rapidly transmitted to other sectors, and that links between producers
and their capital suppliers play a particularly important role in this process. Given that
investment is particularly volatile and pro-cyclical, it makes sense that capital suppliers may
be strongly impacted by shocks to their customers, but to my knowledge this is the first
study to make this point. This finding may be useful for informing the small but growing
theoretical literature investigating how shocks to one sector of the economy can generate
broader effects (Gabaix (2011), Acemoglu et al. (2012)).
38
References
1927. Samuel Cropton 1753-1827; A Brief Survey of his Life and Work with which is incorporated a short
History of Messrs. Dobson and Barlow Limited. Bolton, England: Dobson and Barlow Ltd.
Abadie, Alberto, & Gardeazabal, Javier. 2003. The Economic Costs of Conflict: A Case Study of the Basque
Country. The American Economic Review, 93(1), pp. 113–132.
Abadie, Alberto, Diamond, Alexis, & Hainmueller, Jens. 2010. Synthetic Control Methods for Comparative
Case Studies: Estimating the Effect of California’s Tobacco Control Program. Journal of the American
Statistical Association, 105(490), 493–505.
Abadie, Alberto, Diamond, Alexis, & Hainmueller, Jens. 2014. Comparative politics and the synthetic control
method. American Journal of Political Science.
Acemoglu, Daron, Carvalho, Vasco M., Ozdaglar, Asuman, & Tahbaz-Salehi, Alireza. 2012. The Network
Origins of Aggregate Fluctuations. Econometrica, 80(5), 1977–2016.
Allen, Robert C. 2009. The British Industrial Revolution in Global Perspective. Cambridge University Press.
Altonji, Joseph, & Card, David. 1991. The Effects of Immigration on the Labor Market Outcomes of LessSkilled Natives. In: Abowd, John M, & Freeman, Richard B (eds), Immigration, Trade and the Labor
Market. Chicago: University of Chicago Press.
Autor, David H., Dorn, David, Hanson, Gordon H., & Song, Jae. 2012 (July). Trade Adjustment: Worker
Level Evidence.
Autor, David H, Dorn, David, & Hanson, Gordon H. 2013. The China Syndrome: Local Labor Market
Effects of Import Competition in the United States. American Economic Review, 103(6), 2121–2168.
Bartel, Ann. 1989. Where Do the New US Immigrants Live? Journal of Labor Economics, 7, 371–391.
Bertanha, Marinho, & Moser, Petra. 2014 (August). Spatial Errors in Count Data Regressions. NBER
Working Paper No. 20374.
Bertrand, M, Duflo, E, & Mullainathan, S. 2004. How Much Should We Trust Differences-in-Differences
Estimates? The Quarterly Journal of Economics, 119(1), pp. 249–275.
Black, Dan, McKinnish, Terra, & Sanders, Seth. 2005. The Economic Impact Of The Coal Boom And Bust.
The Economic Journal, 115(503), 449–476.
Bleakley, Hoyt, & Lin, Jeffrey. 2012. Portage: Path Dependence and Increasing Returns in U.S. History.
Quarterly Journal of Economics, 127(May), 587–644.
Bosker, Maarten, Brakman, Steven, Garretsen, Harry, & Schramm, Marc. 2007. Looking for multiple equilibria when geography matters: German city growth and the WWII shock. Journal of Urban Economics,
61(1), 152 – 169.
Brakman, Steven, Garretsen, Harry, & Schramm, Marc. 2004. The Strategic Bombing of German Cities
During WWII and its Impact on City Growth. Journal of Economic Geography, 4, pp. 201–218.
Cameron, A. C., Gelbach, J.B., & Milller, D. L. 2011. Robust Inference with Multi-Way Clustering. Journal
of Business and Economic Statistics, 29(2), 238–249.
Carrington, William J. 1996. The Alaskan Labor Market during the Pipeline Era. Journal of Political
Economy, 104(1), pp. 186–218.
Combes, Pierre-Philippe, & Gobillon, Laure. Forthcoming. The Empirics of Agglomeration Economies.
In: Duranton, Gilles, Henderson, Vernon, & Strange, William (eds), Handbook of Urban and Regional
Economics, vol. 5. Elsevier.
Conley, T.G. 1999. GMM Estimation with Cross Sectional Dependence. Journal of Econometrics, 92(1), 1
– 45.
39
Crafts, Nicholas, & Wolf, Nikolaus. 2013 (August). The Location of the UK Cotton Textile Industry in 1838:
a Quantitative Analysis. University of Warwick Working Paper.
Davis, Donald R, & Weinstein, David E. 2002. Bones, bombs, and break points: The geography of economic
activity. American Economic Review, 92(5), 1269–1289.
Davis, Donald R, & Weinstein, David E. 2008. A search for multiple equilibria in urban industrial structure.
Journal of Regional Science, 48(1), 29–65.
Desmet, Klaus, & Rappaport, Jordan. 2014 (April). The Settlement of the United States, 1800-2000: The
Long Transition Towards Gibrat’s Law. Working Paper.
Dix-Carneiro, Rafael, & Kovak, Brian. 2014 (July). Trade Reform and Regional Dynamics: Evidence from
25 Years of Brazilian Matched Employer-Employee Data. Working Paper.
Duranton, Gilles. 2007. Urban Evolutions: The Fast, the Slow, and the Still. American Economic Review,
97(1), 197–221.
Ellison, G., Glaeser, E., & Kerr, W. 2010. What Causes Industry Agglomeration? Evidence from Coagglomeration Patterns. American Economic Review, 100(3), pp. 1195–1213.
Ellison, Glenn, & Glaeser, Edward L. 1997. Geographic concentration in US manufacturing industries: A
dartboard approach. Journal of Political Economy, 105(5), 889–927.
Faggio, Guilia, Silva, Olmo, & Strange, William. 2013 (October). Heterogeneous Agglomeration. Working
Paper.
Farnie, D.A. 2004. The Role of the Cotton Industry in Economic Development. Oxford University Press.
Chap. 18, pages pp. 557–584.
Forwood, WB. 1870. The Influence of Price upon the Cultivation and Consumption of Cotton During the
Ten Years 1860-70. Journal of the Statistical Society of London, 33(3), 366–383.
Gabaix, X. 1999. Zipf’s law for cities: An explanation. Quarterly Journal of Economics, 114(3), 739–767.
Gabaix, Xavier. 2011. The Granular Origins of Aggregate Fluctuations. Econometrica, 79(3), pp. 733–772.
Greenstone, Michael, Hornbeck, Richard, & Moretti, Enrico. 2010. Identifying Agglomeration Spillovers:
Evidence from Million Dollar Plants. Journal of Political Economy, 118(3), pp. 536–598.
Hanlon, W. Walker. Forthcoming. Necessity is the Mother of Invention: Input Supplies and Directed Technical Change. Econometrica.
Hanlon, W.W., & Miscio, A. 2014 (December). Agglomeration: A Dynamic Approach. NBER Working
Paper No. 20728.
Head, Keith, & Mayer, Thierry. 2004. The Empirics of Agglomeration and Trade. Handbook of Regional and
Urban Economics, 4, 2609–2669.
Hornbeck, R. 2012. The Enduring Impact of the American Dust Bowl: Short- and Long-run Adjustments
to Environmental Catastrophe. American Economic Review, 102(4), 1477–1507. Working paper.
Hsiang, Solomon M. 2010. Temperatures and cyclones strongly associated with economic production in the
Caribbean and Central America. Proceedings of the National Academy of Sciences, 107(35), 15367–15372.
Kline, Patrick, & Moretti, Enrico. 2013. Local Economic Development, Agglomeration Economies, and the
Big Push: 100 Years of Evidence from the Tennessee Valley Authority. Quarterly Journal of Economics.
Kovak, Brian K. 2013. Regional Effects of Trade Reform: What is the Correct Measure of Liberalization?
American Economic Review, 103(5), 1960–1976.
Krugman, Paul. 1991. Increasing Returns and Economic Geography. Journal of Political Economy, 99(3),
pp. 483–499.
Lester, V. Markham. 1995. Victorian Insolvency: Bankruptcy, Improsonment for Debt, and Company
40
Winding-up in Nineteenth-Century England. Oxford: Clare.
Margo, R. A. 1997 (June). Wages in California During the Gold Rush. NBER Historical Paper No. 101.
Michaels, Guy. 2011. The Long Term Consequences of Resource-Based Specialisation*. The Economic
Journal, 121(551), 31–57.
Michaels, Guy, Rauch, Ferdinand, & Redding, Stephen J. forthcoming. Urbanization and Structural Transformation. Quarterly Journal of Economics.
Miguel, Edward, & Roland, Gerard. 2011. The Long Run Impact of Bombing Vietnam. Journal of Development Economics, 96(11954), 1–15.
Mitchell, B.R. 1988. British Historical Statistics. Cambridge, UK: Cambridge University Press.
Mitchell, BR, & Deane, P. 1962. Abstract of British Historical Statistics. London: Cambridge University
Press.
Newey, Whitney K., & West, Kenneth D. 1987. A Simple, Positive Semi-Definite, Heteroskedasticity and
Autocorrelation Consistent Covariance Matrix. Econometrica, 55(3), pp. 703–708.
Notowidigdo, Matthew J. 2013 (March). The Incidence of Local Labor Demand Shocks. Working Paper.
Redding, Stephen J, Sturm, Daniel M, & Wolf, Nikolaus. 2011. History and Industry Location: Evidence
from German Airports. Review of Economics and Statistics, 93(3), 814–831.
Rossi-Hansberg, Esteban, & Wright, Mark L. J. 2007. Urban Structure and Growth. The Review of Economic
Studies, 74(2), pp. 597–624.
Southall, HR, Gilbert, DR, & Gregory, I. 1998 (Jan.). Great Britain Historical Database : Labour Markets
Database, Poor Law Statistics, 1859-1939. [computer file]. UK Data Archive [distributor] SN: 3713.
Thomas, Mark. 1987. An Input-Output Approach to the British Economy, 1890-1914. Ph.D. thesis, Oxford
University.
Thompson, Samuel B. 2011. Simple Formulas for Standard Errors that Cluster by both Firm and Time.
Journal of Financial Economics, 99(1), 1 – 10.
Topalova, Petia. 2007. Trade Liberalization, Poverty and Inequality: Evidence from Indian Districts. University of Chicago Press. Chap. Globalization and Poverty, pages 291–336.
Topalova, Petia. 2010. Factor Immobility and Regional Impacts of Trade Liberalization: Evidence on Poverty
from India. American Economic Journal: Applied Economics, 2(4), 1 – 41.
Watts, John. 1866. The Facts of the Cotton Famine. London: Simpkin, Marshall, & Co.
Yagan, Danny. 2014 (January). Moving to Opportunity? Migratory Insurance Over the Great Recession.
Working Paper.
41
A
A.1
Online Appendix (not for publication)
Empirical setting appendix
Table 13 describes the upstream and downstream industries linked to the cotton textile
industry.
Table 13: Industries with connections to textile production
Intermediate suppliers
- Coal mining
- Chemicals
- Soap
- Oils, paints, dyes
- Rubber
Customer industries
- Clothing manuf.
- Boot and shoe manuf.
- Furniture manuf.
- Paper manuf.
- Other misc. manuf.
Service suppliers
- Transportation services
- Business services
- Gas, electric, water service
Capital suppliers
- Textile machinery
- Steam engines
- Other metal industries
- Construction
This information is drawn from the input-output table constructed by Thomas (1987) and a variety of other
historical sources. The list excludes the main intermediate inputs – cotton, wool, flax, and silk – which were
not produced in British cities.
Figure 8 shows employment in the four main textile sectors over the 1851-1891 period
based on Census of Population data available every ten years starting in 1851. The cotton
textile industry grew rapidly from 1851-1861, but growth slowed from 1861-1871. In contrast,
wool textiles experienced a growth acceleration during the 1861-1871 period, due largely to
the lack of competition from cotton textiles. All of the textile industries experienced slow
growth from 1871-1881, a period that includes the Long Depression of 1873-1879. Over all
periods, cotton textiles did substantially better than the declining silk and flax/linen sectors.
42
Figure 8: Employment in cotton, wool, linen, and silk textile industries, 1851-1891
Chart includes data from the 31 cities for which city-industry data are available over the 1851-1891 period
(see Data section for details).
Figure 9 provides some additional data that allows us to construct a rough proxy for the
pattern of productivity growth in the cotton textile industry over the study period. The
solid line describes national cotton textile employment over the study period (in a log scale).
We can see that the pattern of national employment growth matches the pattern shown in
Figure 8 for the 31 cities contained in the main city-industry database. The top line in
Figure 9 gives cotton consumption in the year before each employment observation based
on data from Mitchell & Deane (1962). I.e., for 1851 I report cotton consumption in 1850.
This is done both because the Census employment data come from early in the year and
because it allows me to report consumption in 1860 rather than 1861, when consumption was
impacted by the beginning of the Civil War. This is the best available measure of industry
output. Finally, the dashed line in the middle of the graph reports cotton consumption, in
thousands of pounds, per cotton textile worker. This provides a rough proxy for productivity
growth in the industry. We can see that productivity grew strongly between 1851 and
1861, stagnated from 1861-1871, and then resumed growing at a somewhat slower pace.
This pattern may seem surprising given that Hanlon (Forthcoming) shows that there was
a burst in innovative activity in the cotton textile industry during the Civil War period.
However, the innovation undertaken during the Civil War period was primarily aimed at
43
mitigating the negative productivity effect of switching from higher-quality U.S. cotton to
lower quality cotton coming primarily from India. Thus, innovation during the Civil War
period served mainly to reduce the negative productivity effects of the war, rather than to
increase productivity relative to the pre-war period.
Figure 9: Cotton textile employment, consumption, and consumption per worker, 1851-1891
Total cotton textile employment for England and Wales is based on Census data. Cotton consumption data,
from Mitchell & Deane (1962), are for the year before the employment observation.
Figure 10 compares the agglomeration patterns of the textile industries over the study
period using the index from Ellison & Glaeser (1997) and firm size data from the 1907 Census
of Production. Somewhat surprisingly, the cotton textile industry was less agglomerated than
the other textile industries during the study period. This was due in part to the fact that
the industry was so large that it was forced to spread over numerous cities, while smaller
textile industries were easier to accommodate in just one or a few cities. Another pattern
visible in the data is that after 1861 the cotton textile industry became more agglomerated,
while the level of agglomeration tended to decline in all of the other textile industries over
the study period.
44
Figure 10: Geographic agglomeration in the textile industries over time
Agglomeration in the textile industries calculated using the index from Ellison & Glaeser (1997) and firm size
data from the 1907 Census of Production. Calculations use 30 of the 31 cities included in the city-industry
database, with London excluded.
Figure 11 describes the level of interest rates in England over the 1855-1870 period.
The two highlighted vertical bars mark the months in which panics led to high levels of
bankruptcies in the cotton textile industry.
Figure 11: Interest rates set by the Bank of England, 1855-1870
Interest rate data are provided by the Bank of England.
45
A.2
Data appendix
This section presents some additional details about the data used in this study. For the
Census of Population employment data, extensive details are available in the online appendix
to Hanlon & Miscio (2014).45
A.2.1
The city-industry database
Table 14 describes the industries included in the city-industry database together with their
1851 employment. These industries are very similar to those used in Hanlon & Miscio (2014)
except that in this project it is necessary to break out the textile sector into separate series
for cotton, wool, linen, and silk textiles. One consequence of this is that the useful data
for this project end in 1891; after that date it becomes more difficult to generate consistent
series for the separate textile industries.
Table 14: Industries in the primary analysis database with 1851 employment
Textile Manufacturing
Cotton textiles
Other textiles
Other Manufacturing
Chemicals & oils
Clothing, dress, shoes
Instruments & jewelry
Earthenware & bricks
Leather & hair goods
Metal & Machinery
Paper and publishing
Shipbuilding
Vehicles
Wood & furniture
136,197
179,449
30,702
328,669
31,048
19,580
26,737
167,052
42,578
14,498
9,021
69,648
Food, etc.
Food, drinks, tobacco
125,087
Services and Professional
Professionals
40,733
General services
460,855
Merchant, agent, etc.
58,172
Shopkeeper, salesmen, etc.
27,232
Transportation services
184,421
Others industries
Construction
Mining-related
137,056
24,505
Employment figures based on the 71 cities available in the 1851-1871 analysis dataset.
A.2.2
The bankruptcy data
The data on bankruptcies come from the London Gazette and were collected starting in Nov.
15, 1861, when the Gazette begins reporting bankruptcies filed under the new Bankruptcy
45
See www.econ.ucla.edu/whanlon/papers/hanlon miscio data appendix.pdf.
46
Act of 1861. To obtain the data, PDFs were downloaded from the London Gazette website,
and then the information was hand-entered off of each sheet. I have collected only information
from the “Notice of Adjudication of Bankruptcy and First Meeting of Creditors.” These
notices, which informed creditors that a person had filed for bankruptcy, represented the
first posted notice following a bankruptcy filing.
Figure 12 shows the first notice entered into the database. From each notice, we entered
the name of the bankrupt, their occupation, and their city and county. London counts as
both a city and a county. The occupation data was then hand cleaned and standardized.
In doing so, I tried to follow as closely as possible the categorizations used in the Census
occupation data.
Figure 12: Example bankruptcy notice from the London Gazette, Nov. 15, 1861
In many cases the notices list multiple occupations; I focus on the first occupation listed
(a similar approach was taken by the Census office when constructing the Census data
categories). Table 15 lists the number of bankruptcies in each of the industry categories
in England and Wales over the period from Nov. 15 1861 to the end of 1866. The largest
category of bankruptcies is among shopkeepers and salesman, followed by construction, food
processing (bakers, butchers, etc.), and merchants, agents and accountants (which includes
all commercial businessmen). Among manufacturing categories, we can see that the largest
number of bankruptcies occurred among clothing and shoe producers. This makes sense
given that this sector was characterized by many small local producers. There is also a large
number of bankruptcies in the metal and machinery category, which is characterized by a
mix between large and small firms. The number of bankruptcies in textile manufacturing
are relatively small, reflecting the fact that firms in that sector were larger.
47
Table 15: Bankruptcy counts by industry category, Nov. 1861 - Dec. 1866
The bankruptcy law in England underwent a major change with the passing of the
Bankruptcy Act of 1861. This Act merged bankruptcy and insolvency law (Lester (1995)).
Prior to the Act, only “traders” could file for bankruptcy, which allowed them to discharge
their debts without fear of imprisonment. The Act extended these bankruptcy protections
to all. This had the result of substantially increasing the number of bankruptcies, particularly during the first year after the law change. This pattern is clearly visible in Figure
13, which plots the number of bankruptcies by month in England and Wales from Nov. 15,
1861 through the end of 1866. These data begin in the middle of November of 1861 because
that was the first point at which the London Gazette begins reporting notices of preliminary
adjudication of bankruptcy under the new 1861 law.
To avoid the large spike in bankruptcies caused by the change in bankruptcy laws, I begin
my analysis with data starting in July of 1863. Thus, my analysis excludes the months highlighted in Figure 13. This Figure suggests that by mid-1863 the initial surge of bankruptcies
had passed and the overall level of bankruptcies had reached something resembling a normal
level.
48
Figure 13: Monthly bankruptcy counts, Nov. 1861 - Dec. 1866
Excluding data prior to July, 1863 leaves me with 25,123 bankruptcies. Because this
cutoff is somewhat arbitrary, I check the robustness of the results to the alternative of using
all of the data from 1863 (see Section A.5). Next, I limit the data geographically to ensure
that I am comparing the cotton textile cities to other urban areas. Specifically, I limit the set
of analysis cities to those in which a reasonable number (at least 30) bankruptcies occurred
of the period from mid-1863 through 1866. Table 16 shows the set of cities with more than
30 bankruptcies. I also exclude London from the main analysis. This is done both because
London is an outlier in terms of size, and because the data from London are more difficult
to deal with. This is because within London, people often listed their location as their
neighborhood, rather than simply as London. Also, in many areas that were part of the city
of London, people sometimes reported them as part of one of the home counties, particularly
Middlesex. As a result, it is necessary to include all of Middlesex as part of London. The
robustness exercises in Section A.5 show that including London does not substantially affect
my findings.
49
Table 16: cities included in the bankruptcy analysis
An important feature of the bankruptcy data set is that it covers all private bankruptcies.
Thus, it may include both bankruptcies by the owners of private business and bankruptcies
by workers who do not own a business. We may be concerned that including bankruptcies
by workers is misleading. To try to help address this issue, I have categorized bankruptcies
50
where the individual is obviously a worker rather than a business owner. These are identified
as occupations in which the bankrupt is listed as an “Assistant”, “Journeyman”, “Servant”,
“Labourer”, “Foreman”, “Manager”, “Machinist”, “Engineer” (which at this time often
means someone who operates an engine), and “Clerk.” In the robustness exercises in Section
A.5, I show that my main results are unchanged when bankruptcies containing any of these
worker identifiers are dropped.
For manufacturing industries, I have classified both makers and dealers into the same
category, a practice which follows the system used in the Census data. Thus, both a machine manufacturer and a machine parts dealer would be listed in the metal & machinery
category. However, much of the existing literature in this area has focused on manufacturers
and excluded dealers. To enable a similar divide in my data, I reviewed every occupation
in the manufacturing categories and divided them into makers and dealers. Many occupations include both producing and selling a good, and these I have classified as makers. In
the robustness exercises in Section A.5, I calculate separate results for the manufacturing
industries in which dealers are dropped from the data. I find that focusing exclusively on
makers does not affect the results in a substantial way.
Another issue arises for public companies. Bankruptcies of public companies will not be
reflected in the bankruptcy database. However, public companies were much less common
during the period I study than they are today; the vast majority of companies would have
been privately held and therefore would appear in the bankruptcy data.
A.3
Additional city population results
This section present some additional results documenting persistence in the impact of the
shock on city population. In the main text, I use a discrete cutoff of 10 percent of private
sector employment to identify the cotton cities. An alternative to this is to use a continuous
measure for the share of employment in the cotton textile industry. The two tables below
describe results corresponding to those shown in the main text (Tables 2 and 4) but using the
share of cotton textile employment in place an indicator for cotton textile cities. Specifically,
the key explanatory variables are constructed by interacting each city’s share of cotton textile
employment in total employment in 1851 with decade indicator variables.
51
Table 17: Regressions of population growth in cities with greater 1851 share of employment
in cotton textiles
(1)
Years
included:
Cotton emp. shr
× 1851-1861
DV: City population growth rate in each decade
(2)
(3)
(4)
(5)
1851-1871
1841-1871
1841-1891
1841-1891
0.0581
(0.0876)
((0.0658))
1851-1901
-0.181
(0.0670)
((0.0283))
-0.152
(0.0717)
((0.0509))
-0.152
(0.0721)
((0.0485))
-0.123
(0.0986)
((0.0589))
-0.167
(0.0677)
((0.0376))
Cotton emp. shr
× 1871-1881
0.0217
(0.0755)
((0.0465))
0.0507
(0.0999)
((0.0573))
0.0289
(0.0761)
((0.0407))
Cotton emp. shr
× 1881-1891
0.0232
(0.0912)
((0.0553))
0.0522
(0.101)
((0.0646))
0.0291
(0.104)
((0.0412))
Cotton emp. shr
× 1861-1871
1841-1861
0.0581
(0.0870)
((0.0557))
(6)
Cotton emp. shr
× 1891-1901
-0.0934
(0.123)
((0.0401))
City FEs
Yes
Yes
Yes
Yes
Yes
Yes
Time effects
Yes
Yes
Yes
Yes
Yes
Yes
Observations
92
92
138
230
230
275
Cities
46
46
46
46
46
55
Heteroskedasticity-robust standard errors in parentheses. HAC standard errors robust to
spatial correlation up to 100km in double parentheses. All specifications include a full set
of city fixed effects and year effects. The regressions in columns 1-5 use data from the 1891
census covering 1841-1891. The results in column 6 are based on a slightly different data
set from the 1901 census covering 1851-1901.
52
Table 18: Regressions of population growth in cities with greater 1851 share of employment
in cotton textiles (textile cities only)
(1)
Years
included:
Cotton emp. shr
× 1851-1861
DV: City population growth rate in each decade
(2)
(3)
(4)
(5)
1851-1871
1841-1871
1841-1891
1841-1891
0.125
(0.125)
((0.0459))
1851-1901
-0.269
(0.112)
((0.0631))
-0.206
(0.0881)
((0.0673))
-0.206
(0.0892)
((0.0766))
-0.144
(0.104)
((0.0617))
-0.250
(0.113)
((0.0804))
Cotton emp. shr
× 1871-1881
-0.0775
(0.0774)
((0.0314))
-0.0149
(0.110)
((0.0209))
-0.114
(0.0863)
((0.0336))
Cotton emp. shr
× 1881-1891
-0.0700
(0.103)
((0.0715))
-0.00736
(0.122)
((0.0533))
-0.112
(0.119)
((0.0442))
Cotton emp. shr
× 1861-1871
1841-1861
0.125
(0.122)
((0.0538))
(6)
Cotton emp. shr
× 1891-1901
-0.197
(0.141)
((0.053))
Yes
Yes
City FEs
Yes
Yes
Yes
Yes
Yes
Time effects
Yes
Yes
Yes
Yes
Yes
Observations
36
36
54
90
90
Cities
18
18
18
18
18
Heteroskedasticity-robust standard errors in parentheses. HAC standard errors robust to
spatial correlation up to 100km in double parentheses, except for the specification in columns
5-6, which uses 90km to avoid calculation issues. All specifications include a full set of city
fixed effects and year effects. The regressions in columns 1-5 use data from the 1891 census
covering 1841-1891. The results in column 6 are based on a slightly different data set from
the 1901 census covering 1851-1901.
A.4
Synthetic control result details
Table 19 describes the balance between the actual cotton region values for the matching
variables and the values for the synthetic control. We can see that overall the synthetic
control generally matches the actual values well, with the main exception being the share of
textile employment, which is impossible to match given that textile employment is higher in
the cotton region than anywhere else in the country.
53
Table 19: Synthetic control balance across matching variables
54
A.5
Additional bankruptcy results
This section provides some additional details related the the analysis of the bankruptcy data.
First, I provide additional charts, similar to Figure 7, describing the pattern of bankruptcies
in sectors unrelated to cotton textiles in the Northwest counties, in sectors related to cotton
textiles in the wool cities, and in sectors related to cotton textiles in all non-cotton cities.
Following that, I present a series of robustness checks related the results presented in Table
6.
Figure 14: Bankruptcies in other sectors in major cotton textile cities
Bankruptcy data were collected from the London Gazette. The bar chart (left-hand axis) describes bankruptcies in those sectors not included in Figure 7. These include many shopkeepers, beer sellers, government
employees, farming and mining firms, and some other miscellaneous occupations. Solid line (right-hand axis)
describes the number of bankruptcies in the cotton textile sector. The cotton textile cities included in this
chart are Accrington, Ashton-under-Lyne, Blackburn, Bolton, Burnley, Bury, Manchester, Oldham, Preston,
Rochdale, Warrington, Wigan, Stockport and Chester. All of these are in Lancashire except for the last two,
which are in Cheshire.
55
Figure 15: Bankruptcies in major wool textile cities
Bankruptcy data were collected from the London Gazette. The bar chart (left-hand axis) describes bankruptcies in a number of industrial sectors related to cotton textiles. Solid line (right-hand axis) describes the
number of bankruptcies in the cotton textile sector. The wool textile cities included in this chart are Bradford, Dewsbury, Halifax, Huddersfield, Leeds and Wakefield.
Figure 16: Bankruptcies outside of the cotton and wool districts
Bankruptcy data were collected from the London Gazette. The bar chart (left-hand axis) describes bankruptcies in a number of industrial sectors related to cotton textiles. Solid line (right-hand axis) describes the
number of bankruptcies in the cotton textile sector. The chart includes all of England outside of Lancashire,
Cheshire and Yorkshire counties.
56
Next, I conduct several robustness exercises that look at how my results are affected by
some of the decisions made when constructing the data and the empirical exercise. This is
done in Table 20. The first column presents results in which I construct the key explanatory
variable using a three month window rather than the one-month window used in the main
results. The new three month window is centered on the one-month window used in the
main text. Thus, for the panic in October, 1864, I look at effects in October, November, and
December of 1864. In the second column, I include London, which was excluded from the
results presented in the main text for reasons described in the Bankruptcy Data Appendix.
In the third column, I assess the impact of my choice to include only cities with more
than 30 bankruptcies during the period from mid-1863 through 1866. This column presents
results where instead I also include smaller cities with between 25 and 30 bankruptcies.
The fourth column addresses the fact that the bankruptcy data include both workers and
business owners. This is done by dropping bankruptcies with occupations that include
a term that clearly indicates that the bankruptcy is by a worker rather than a business
owner. These terms are “Assistant”, “Journeyman”, “Servant”, “Labourer”, “Foreman”,
“Manager”, “Machinist”, “Engineer” (which at this time often means someone who operates
an engine), and “Clerk.” The fifth column considers the impact of my decision to use only
data starting in July, 1863, to generate the results presented in the main text. To do this,
this column adds in the data for the first half of 1863. Finally, the sixth column addresses the
fact that, for manufacturing industries, the bankruptcy data pool together manufacturers
and dealers. I have gone back and reviewed all of the occupations in a set of manufacturing
industries in order to separate dealers from manufacturers. Column 6 presents results for
these manufacturing industries where all dealers have been dropped from the analysis data.
Overall, we can see that they key results presented in the main text – particularly the increase
in bankruptcies in metal & machine industries in the cotton textile cities following months
of panic in the cotton market – are robust to each of these alternative approaches.
57
Table 20: Robustness exercises for the bankruptcy results (continued on next page)
With
3 month
window
0.296**
(0.117)
0.478***
(0.160)
With
smaller
cities
0.473***
(0.149)
Dropping
worker
bankruptcies
0.442**
(0.182)
Using
all 1863
data
0.451***
(0.157)
Agriculture
-0.0124
(0.697)
0.649
(0.865)
0.588
(0.840)
0.413
(0.876)
0.367
(0.877)
Chemicals
& oils
0.454
(0.512)
0.473
(0.491)
0.494
(0.490)
-0.0815
(0.664)
0.174
(0.549)
Construction
0.390*
(0.208)
0.392
(0.434)
0.506
(0.412)
0.213
(0.444)
0.148
(0.452)
Clothing
& shoes
-0.0402
(0.412)
0.707
(0.640)
0.672
(0.621)
0.712
(0.656)
0.732
(0.659)
Drinks
0.204
(0.275)
-0.175
(0.585)
-0.232
(0.572)
0.301
(0.621)
0.154
(0.634)
Food
processing
-0.0448
(0.229)
-0.517
(0.611)
-0.540
(0.593)
-0.490
(0.635)
-0.531
(0.614)
General
services
0.454
(0.580)
0.376
(0.535)
0.315
(0.517)
0.345
(0.565)
0.292
(0.539)
Instruments
& jewelry
0.257
(0.598)
0.386
(0.662)
0.396
(0.662)
0.583
(0.732)
0.648
(0.753)
All
non-cotton
With
London
58
Makers
only
1.466
(1.534)
0.584
(0.610)
-12.37***
(0.566)
Table 21: Robustness exercises for the bankruptcy results (continued)
A.6
With
London
Leather,
hair goods
With
3 month
window
0.928*
(0.514)
0.894
(0.634)
With
smaller
cities
0.738
(0.634)
Dropping
worker
bankruptcies
0.597
(0.683)
Using
all 1863
data
0.701
(0.688)
Merchants
agents, etc.
Makers
only
0.695***
(0.158)
0.585***
(0.163)
0.575***
(0.168)
0.622**
(0.298)
0.645**
(0.303)
Metal &
machinery
0.408*
(0.217)
1.283***
(0.288)
1.202***
(0.307)
1.189***
(0.372)
1.358***
(0.338)
1.625***
(0.353)
Non-cotton
textiles
0.815**
(0.372)
0.283
(0.586)
0.290
(0.587)
0.773
(0.610)
0.676
(0.687)
1.253
(0.805)
Paper &
publishing
-0.191
(0.675)
1.083*
(0.623)
1.099*
(0.622)
0.910
(0.699)
0.981
(0.679)
Shopkeepers,
salesmen, etc
0.116
(0.198)
0.370
(0.378)
0.432
(0.373)
0.420
(0.421)
0.352
(0.409)
Textile
finish, etc.
0.789*
(0.432)
0.830
(0.736)
0.924
(0.710)
1.148
(0.907)
0.997
(0.843)
Transport
0.188
(0.380)
1.182*
(0.677)
1.156*
(0.655)
1.010
(0.714)
0.760
(0.638)
Wood &
furniture
0.0124
(0.369)
-0.371
(0.429)
-0.386
(0.451)
-0.341
(0.462)
-0.291
(0.465)
0.726
(0.732)
0.347
(1.166)
0.630
(0.588)
Additional results using the city-industry data
This section presents some additional results calculated using the city-industry database. I
begin with results corresponding to those shown in Table 7 in the main text but spanning
all sectors of the economy. I then present some robustness exercises related to the analysis
of city-industry growth presented in Tables 8 in the main text.
Table 22 describes the employment growth contribution of each broad economic sector
in the cotton textile cities, non-cotton textile cities, and all non-cotton cities. The top
three rows in each panel correspond to the results shown in Table 7 in the main text, while
the remaining rows describe the growth contribution of sectors of the economy outside of
manufacturing.
59
Table 22: Decomposing the growth contribution by sector – with all sectors
Each cell of these tables represents the contribution of an industry to city growth over the period indicated.
So if an industry represents 10% of employment and grows 10% over the period, that industry’s contribution
to city growth is 1%. Thus, the column sum of each table equals the total growth of the group of cities over
the period. Note that the column sums will differ from those in Table 1 both because we are working with
a different set of cities and because they represent employed population rather than total population.
Table 23 presents industry employment results corresponding to those in Table 8 in the
60
main text, but calculated using all control cities.
Table 23: Impact on industry employment growth in cotton cities using all control cities
*** p<0.01, ** p<0.05, * p<0.1. Column 1 presents the ai coefficients for the specification in Eq. 3 with
corresponding standard errors in Column 2. Column 3 presents the bi coefficients for the specification in Eq.
4 with the corresponding standard errors in Column 4. The regressions include a full set of industry-year
effects. Standard errors, shown in parentheses, are clustered by city-industry, city-year, and industry-year.
The omitted industry in both regressions is “General services.”
One feature of the city-industry database is that it is necessary to adjust the 1871 data
for the number of workers under 20. This is necessary because the data for 1851 and 1861
report employment divided into workers over 19 and workers under 20, and the data for
1881-1891 report only total employment, while in 1871 only employment for workers over 20
is reported.46 Because workers under 20 made up an substantial portion of the labor force,
it is important to adjust the 1871 data in order to obtain a consistent series. This is done
using the national data for 1871, which reports employment by industry divided into workers
over 19 and workers under 20. This is used to calculate a share of workers in each industry
that were under 20. This share is then applied to adjust employment in each city-industry,
46
An exception is London, where data are reported only for all workers in 1871.
61
under the assumption that the share of workers under 20 in any particular industry does not
vary too much across cities.
Because 1871 is an important year in the analysis, we may be worried that this adjustment
is influencing the results. One way to check this is to use only the data for workers over
19, which is available on a consistent bases for 1851-1871 without the need for adjustment.
Using these data, I obtain the results shown in Table 24.
Table 24: Impact on employment for workers over 20 in cotton cities, 1851-1871
*** p<0.01, ** p<0.05, * p<0.1. Table displays ai coefficients and standard errors based on the regression
specification in Equation 3. The data cover 1851-1871 and include only workers over 19 years of age. The first
set of results use data on 71 cities with 2096 city-industries and 4161 observations. The second set of results
use data on 18 textile cities with 534 city-industries and 1064 observations. In both sets, there are 8 cotton
textile cities. The regressions include a full set of city-industry and industry-year effects. Standard errors,
shown in parentheses, are clustered by city-industry, city-year, and industry-year. The omitted industry is
“General services.”
62