Urbanization, Internal Migration, and Occupational Mobility in Victorian
Britain
Jason Long
Current version: November, 2001
Victorian Britain experienced the most rapid and thorough urbanization the world had yet seen. People flocked from the
countryside to the cities to such an extent that, by mid-century, Great Britain had become the first predominantly urban
society in history. This study addresses three related questions: Who were the rural-to-urban migrants, what forces drove
them, and did moving to the city pay off in terms of attaining higher socioeconomic status? I answer these questions with
the aid of a new dataset of 28,000 individuals matched between the 1851 and 1881 Censuses of the Population of England
and Wales. I build a structural econometric model to assess the effect of migration and the nature of migrant selectivity,
controlling for the endogeneity of the migration decision. I find that those who left the countryside for the cities did not
come from the bottom of the economic and social ladder; rather, they were better off initially than those who remained in
rural areas. Urban migrants were positively selected: their prospects in the urban labor market were superior to those who
did not move. Moreover, their prospects in the rural labor market were also superior. Finally, the decision to move was,
by and large, a fruitful one. The treatment effect of urban migration was large and positive across all socioeconomic
strata. On average, people who moved to the city were substantially more successful in improving their socioeconomic
status than they would have been had they remained in rural areas, and they were more likely to experience upward
intergenerational occupational mobility.
When a man is tired of London, he is tired of life; for there is in London all that
life can afford.
Samuel Johnson1
The scope and acceleration of urbanization over the last two centuries are striking. In 1800, only 2 percent of world
population was urbanized (living in places of 20,000 or more inhabitants). Since then the rate of urbanization has
dramatically risen, so that today approximately half of the world’s populace lives in cities. In recent decades this
urban growth has been driven largely by the developing world, and nearly all of the projected expansion of the next
30 years is expected to accrue to less developed countries. In many cases, the migrants who fill these cities come
from and enter into poverty. Though this has recently begun to change, for most of the 20th century rural-to-urban
migration has been dominated by poor, unskilled individuals with essentially non-existent opportunities in their
Jason Long, Department of Economics, Northwestern University, 2003 Sheridan Rd., Evanston, IL 60208-2600.
E-mail: [email protected].
I would like to thank my advisors, Joel Mokyr, Joe Ferrie, and Joe Altonji, for their helpful comments. I also benefited from
conversations with Henry Siu, Jim Sullivan and Chris Taber, and from input from workshop participants at Northwestern University. Justin
Hayes and Humphrey Southall provided me with data. This research was greatly aided by a Northwestern University Graduate Research
Grant and a Pew Younger Scholars Program Fellowship.
1
Boswell, Life.
places of origin. According to Robert McNamara, former President of the World Bank: 2
The cities are filling up and urban unemployment steadily grows…the “marginal men,” the wretched
strugglers for survival on the fringes of farm and city, may already number more than half a billion, by
1990 two billion. Can we imagine any human order surviving with so gross a mass of misery piling up at its
base?
Empirical evidence supports McNamara’s assessment. Recent work has revealed that internal migration within a
developing country works like a lottery: most individuals who move experience negative returns while a few reap a
very large payoff.3 Considered as a group, nearly all rural-urban migrants fare very poorly, and experience negative
returns to migration. In short, rural-urban migration in the modern developing world is at best a rational but highly
risky proposition, and at worst an irrational, detrimental decision.
Just as urbanization over the last fifty years has been largely a phenomenon of the developing world, so it
was in the 19th century, when Great Britain was at the forefront of the then developing world. Through the 1800s,
Great Britain experienced the most rapid and thorough urbanization the world had ever seen.4 The population
underwent a tremendous redistribution in the wake of the Industrial Revolution. The 1851 Census of the Population
of England and Wales revealed a watershed event: for the first time in the history of any significant nation, more
people lived in towns than in the countryside.5 Furthermore, 38 percent of Britain’s populace lived in cities of
20,000 or more, and by 1881, this figure topped 50 percent. From 1801 to 1911, the urban population increased
nearly by a factor of ten, from 3.5 to 32 million. “Galloping urbanization” is Bédarida’s description.6 The primary
city in Great Britain, indeed in the world, was London. In 1825 it passed Beijing to become the world’s largest city,
and by 1900 it counted 6,480,000 occupants, 42 percent more than second place New York.
It is difficult to say with any precision how important were the dual forces of natural increase and ruralurban migration to the growth of the cities. What is clear is that both were strong. It is the latter that is the focus of
the current study. Rural exodus had been going on for some time (fueling the growth of London as early as the 17th
century), but it picked up considerable speed in the 1830s and 1840s and did not slow until the 1880s. It has been
estimated that from 1841 to 1901, the rural areas of England and Wales lost more than 4 million people from
internal migration, 3 million of whom headed for the towns, at a rate of more than half a million per decade.7 They
2
Qtd. in Todaro, Economic Development, p. 264.
Tunali, “Rationality of Migration.”
4
The Netherlands experienced one of the first sustained periods of substantial urbanization, first in the south from 1475 to 1550 and then in
the north, to an even greater degree, between 1580 and the 1670s. Only the province of Holland, however, was ever more than half urban.
The country as a whole reached its maximum level of urbanization, 42 percent, in 1675, then fell to a level of 35 percent by 1815. Neither the
rate nor the level of Dutch urbanization matched that of England, Wales, and Scotland in the 19 th century. See de Vries and van der Woude,
First Modern Economy, pp. 60-63.
5
Towns being places of more than 2,500 inhabitants. See the “Data” section for a discussion of the definition of “towns” and “cities.”
6
Bédarida, Social History.
7
Crouzet, Victorian Economy, p. 93.
3
2
left to escape disaster. They left to seek an education, a spouse, adventure, anonymity, or just the change in locality
itself. Most importantly, they left in search of employment—either a better job or simply any job at all.8
As Britain’s cities grew during the 19th century, so did its economy. From 1830 to 1900, Britain’s GNP
grew at an annual rate of 2.5 percent, while GNP per capita grew 1.44 percent per year. 9 Certainly the technological
advancements of the Industrial Revolution were a primary force behind this growth, but other factors were at work
as well. Among these were urbanization, and rural-urban migration in particular. Modern economic growth theory,
with its emphasis on increasing returns and external scale economies, has recognized that the localized information
and knowledge spillover of urbanization make cities the “engines of economic growth in an economy.”10 The
movement of people from Britain’s rural areas to the cities represented a movement of labor from areas of lower
marginal product to higher, resulting in efficiency gains for the national economy. Rural-urban migration was an
important ingredient in the rise of Great Britain to wealth and economic modernity; as such, it has received
extensive attention in the literature since the very days during Victoria’s reign when it was occurring. There is still
much we do not know, however. We know very little about the migration decision from the level of the individual
migrant: what individual-specific factors drove them, whether they were the best or the worst of the rural labor
pool, and most significantly, what were the returns to moving from a rural to an urban area. We do not know these
things because we have always lacked adequate longitudinal data with which to observe the actual migration
decision.
In this study I fill this gap by using nationally representative, individual level, longitudinal data to examine
rural-urban migration in 19th century Great Britain for the first time. To do this, I have constructed a dataset of
approximately 28,000 individuals matched between the 1851 and 1881 Censuses of the Population of England and
Wales. Three questions provide central motivation: (1) Who were the rural-urban migrants, (2) what forces drove
them, both from an individual and a locational perspective, and (3) what was the effect of moving to a city on the
attainment of socioeconomic status? In order to answer these questions, I have developed a structural econometric
model with which to analyze the data on matched individuals. I find that the urban migrants of Victorian Britain
were not the “wretched strugglers,” those with poor to non-existent rural opportunities; rather, they came from the
middle of the social and economic ladder. They in fact had better rural prospects than did those who remained
behind, and they were positively selected for migration: they performed better in the urban labor markets than the
rural persisters would have had they instead chosen to move to an urban place. Economic forces, specifically the
lure of higher quality jobs, were chief among the determinants of migration. Finally, the effect of moving to a city,
the returns to urban migration, was positive and large across all socioeconomic strata.
8
For evidence supporting this assertion, see Pooley and Turnbull, Migration, p. 152.
Crouzet, Victorian Economy, pp. 33, 37.
10
Black and Henderson, “Urban Growth,” pp. 252-53. Also, Lucas, “Mechanics.”
9
3
Previous Research
Internal migration in Great Britain after the Industrial Revolution has been the subject of a literature dating back to
the 19th century itself. 11 E. G. Ravenstein wrote the earliest important work on internal migration in 19th century
Britain in 1885. He used birthplace data from the 1881 census to draw up seven “laws of migration,” which have
served to frame much of the subsequent work on internal migration within Britain and elsewhere. The first and
most important of Ravenstein’s laws is that most moves covered only a short distance, and that exceptions to this
rule generally involved Britain’s “great centres of commerce or industry.” He also posited that urban natives were
“less migratory” than their rural counterparts, and that “females are more migratory than males.” Ravenstein’s laws
have largely stood the tests of time and modern scholarship. Another major early work is Redford’s Labour
Migration in England, 1800-1850, which confirms the short-distance nature of 19th century British internal
migration and refutes any notion of large-scale movement from the overpopulated South to the recently
industrialized North. Redford also highlights the attractive power of higher wages and better opportunities for
employment in the towns and cities as the primary force behind the rural-urban moves. Perhaps the most
comprehensive study of rates of migration within and emigration from Victorian Britain is Baines’ Migration in a
Mature Economy. He uses census birthplace information along with mortality statistics from the Annual Reports
and Decennial Supplements of the Registrar General to estimate for every county in England and Wales, for every
decade from 1860 to 1900, the rate of outward-bound overseas emigration and the rate of outward-bound migrants
who remained within England and Wales. Measuring internal migration this way, Baines found that about 8.4
percent of the population of England and Wales undertook an internal, intercounty move between 1861 and 1880.12
His numbers also indicate that outward-bound inter-county migration rates were higher for rural counties than for
urban: roughly 11 versus 7 percent.13 Finally, Baines finds that most intercounty migrants were young: 81 percent
were aged 15-34, and the rest were children.14
Though still significant, the literature that aims specifically to analyze the causes of migration in 19th
century Britain is smaller than that which simply describes it. Hunt is an early example.15 He finds a moderately
significant positive relationship between county-level earnings and net gain from migration for 1871-1891 and
1891-1911. Many other studies similarly find a relationship between internal migration levels and economic
indicators. Boyer and Hatton verify the positive relationship between earnings (and expected earnings, taking
unemployment into account) and migration, and add a consideration of the so-called “friends and relatives” effect.
11
See Boyer and Hatton, “Migration” for a recent, thorough survey of the field.
No doubt Baines found this out; he did not, however, report it. His book contains many detailed decompositions of migration numbers, but
it nowhere gives an overall rate of internal migration. This figure was calculated using Baines’ raw migration data, reported in Appendix 1 of
the book.
13
Baines defines rural counties as those in which fewer than 35% of the population lived in towns of more than 20,000 occupants. So
defined, there were 34 rural counties and 18 urban.
14
Others have verified this result. See, for example, Friedlander and Roshier, “Internal Migration,” and Williamson Coping, pp. 40-42.
15
Hunt, Regional Wage Variations.
12
4
They find internal migration flows to be positively correlated with the stock of previous migrants, which indicates
that the migration decision depends upon the availability of information as well as anticipated earnings gains.
All of the studies above—16 books and papers on the subject are surveyed by Boyer and Hatton—are
concerned with internal migration within Great Britain. In all of them, migrants from the rural areas of the land to
the cities play a central role. In none of them are any such migrants actually observed. All of these studies analyze
migration at some aggregate level, most often that of the county.16 It is sensible to study internal migration at the
county level; counties gain and lose population, and county-level indicators certainly affect migration flows. On the
other hand, it is people who do the moving. If we cannot see these people, who they are and what they do, then we
are missing much of the story of why migration occurs. We also miss the entire story of the extent to which moving
to a city in Victorian Britain “paid off.” Supposedly, people moved in response to average wage gaps and
unemployment rates. But did they receive those higher wages? Did they get better jobs? If we don’t see the people
who moved, as well as those who did not, then we cannot know.
Two studies do consider individuals, but not in such a way as to answer these questions. Williamson uses
1851 census data to compare the (imputed) earnings of those born and living in cities with those born outside the
cities but living in them in 1851.17 He finds that those born elsewhere earn about as much as natives, and that both
groups earn more than did the average fully employed farm worker. Though these data contain information on
individuals, it is only at a point in time; nothing is known of the people before 1851, so the migration decision itself
cannot be analyzed, nor can the consequences of the migration decision (i.e. the treatment effect) be assessed.
Pooley and Turnbull use an interesting dataset to look at individuals over time: survey responses from genealogists
and family historians from all over England and Wales. Their data are chronologically diffuse, containing 16,000
people born between 1750 and 1930, and they use them to examine the history of migration broadly from the 1700s
to the present. One finding of their study which is particularly relevant for the work at hand is that, according to the
documentary evidence provided by their survey respondents, work-related reasons were the single most important
cause of all internal migration among their sample, accounting for approximately 38 percent of all moves.18 The
book is unique and important, but it does not ask the questions that economists have asked: there is no formal
analysis of the determinants or consequences of the migration decisions of individuals, nor is rural-urban migration
specifically addressed. Also, the nature of their dataset calls into question its accuracy and representativeness. It is
only accurate insofar as the family historians’ information is accurate, and it is in no way a random sample of
information from the population of England and Wales.
What is needed, and what this study provides, is a large, nationally representative dataset of individuals
who are observed both before and after they decide to move to the city (or to remain in the countryside). Only with
16
Friedlander, “Occupational Structure,” analyses migration at the somewhat finer level of the registration district.
Williamson, Coping, ch. 5.
18
Pooley and Turnbull, Migration, p. 152.
17
5
such information is it possible to analyze properly the role of both individual and locational factors in the ruralurban migration decision and to gauge the effect this decision had on the ability of people to obtain high quality
occupations.19
The Data
The data used here come from a new sample of nearly 30,000 males linked from the 1851 Census of the Population
to the 1881 census. The population censuses have long been recognized as the most important source of individuallevel data for Great Britain from the 19th century. They are large, and they are nationally representative; in fact,
they are (essentially) the whole population. One critical limitation of the censuses, however, is that each covers
only a point in time. Successive censuses can be used to examine changes in the nation over time but not changes in
the lives of individuals over time. No continuity exists for individuals between the censuses. But continuity can be
created. It has recently become possible to search for specific individuals among the vast records of the census
enumerators’ books which include information on every person living in England and Wales at the time of
enumeration.
The data source that makes such a fine search feasible is rather remarkable. The Genealogical Society of
Utah in conjunction with the Federation of Family History Societies has computerized the entire 1881 census of the
population of England, Wales, and Scotland. The compilers of the data have made the raw data files for the
enumerators’ books from England and Wales available through the Data Archive at the University of Essex.20 With
these data any individual or group of individuals can quickly and easily be searched for throughout the entire 1881
census. In this case the search was for individuals from another census: that of 1851. The dataset used was a
computerized 2 percent sample of the 1851 census, compiled principally by Anderson, Collins, and Stott.21
The information available for each individual relates, of course, to the questions asked in the two censuses:
name, address, relationship to head of household, marital status, age, sex, occupation, place of birth, and whether
blind or deaf and dumb. Those born and currently residing in England or Wales were asked to give town and county
of birth. The same was true of Scotland. Those born in England or Wales currently residing in Scotland and those
19
It is important here to note that such data have been assembled and analyzed for the U.S. in the 19th century by Ferrie. See “New Sample,”
“Migration,” and “Down on the Farm.” Ferrie has employed data from the 1850-70 U.S. censuses to construct a linked sample of individuals
similar to the one constructed for the present study. Eventually it will be possible to compare the causes and consequences of rural-urban
migration in Britain directly to those in the U.S. A presentation of this joint research will be given at the at the 2002 International Economic
History Congress in Buenos Aires, for the session titled “Making a Career: Individual Work-Life Histories and Labour Market Structures”.
20
Study number 3643
21
It is a stratified two percent systematic cluster sample from the enumerators’ books. For England and Wales, settlements with fewer than
2,000 inhabitants are sampled in their entirety, on the basis of one settlement in fifty. For the remainder of these countries, and for all of
Scotland, the sampling unit is the enumeration district, every fiftieth successive enumerator’s book being selected. In its entirety the sample
contains 945 clusters and 415,000 individuals. Regarding the construction of the sample, it is noteworthy that the clustering procedure
ensures that family units remain intact; thus, for every individual contained in the sample, we have the complete census information for each
member of that individual’s household, including immediate family members and anyone else residing in the same dwelling place (servants,
lodgers, visitors, etc.). For a full description see Anderson,National Sample. This dataset is also available through the Data Archive, as study
number 1316.
6
born in Scotland currently residing in England or Wales were asked to give country of birth only. Name, age, and
birthplace information was used to link individuals between the two censuses. In order to be considered a true
match for an individual from 1851, an individual from 1881 had to have either the same name or a close phonetic
variation thereof (for example, Aitken and Aitkin were considered to be equivalent), a year of birth different by no
more than five years, and the same county and town of birth. Also, none of this information could be missing from
an individual’s record. The variation in birth year was allowed in order to account for age misreporting, and the
town of birth, like the name, was allowed to vary slightly (Bowden was considered equivalent to Gt. Bowden, for
example).22 If by these criteria an individual from the 1851 sample had more than one match from the 1881 census,
then that individual was discarded.
Applying this matching process to an initial pool of 187,117 English and Welsh males from the 1851 2
percent sample yielded a set of 28,474 men observed both in 1851 and 1881, a success rate of approximately 15
percent.23 The data come from two nationally representative sources, so as long as the matching process does not
skew the sample, the set of matched individuals should also be representative of the population of England and
Wales. Table 1 shows a comparison between the sample of matched individuals and the entire group of males from
the 1851 2 percent sample.24 There are differences between the groups throughout the categories, but they all stem
from one important factor: the matched men are younger on average. The average age for the matched men is 18,
while for the entire group it is 24. This should be expected. Life expectancy at birth was only 39.5 years in England
in 1851; younger men were more likely still to be alive in 1881. From this one difference follow the others:
matched men were more likely to be sons rather than household heads; they were less likely to be married; more
likely to be students and weavers and less likely to be anything else; and more likely to still be living within their
county of birth (having had less time to move away).
The only noticeable difference that probably is not due to age is an under representation of people living in
London in 1851. The most likely explanation for this anomaly is that the concept of a “town of birth” is a bit more
slippery for those born in London. There is such a multitude of place names within the city that those born in
London were probably more likely to report something different as their town of birth between the two censuses,
particularly if they lived in London in 1851 but had moved by 1881.25 For the most part, the sample is a good
representation of the population of Victorian England and Wales, skewed somewhat toward the young.
22
Age misreporting was fairly common in 19 th century Britain. A discussion of this issue is included below, in the “Estimation” section.
38,504 Scottish men were also included in the search, but only a handful was matched. Non-migrant Scots could not be matched, as the
1881 data cover only England and Wales, and even migrants could not be matched with much success in the 1881 census since they were not
specifically required to give more than simply their country of birth. A small minority did give their full birth information, were able to be
matched, and are included in the sample of 28,000.
24
It also includes the smaller subsample of individuals to be used in the econometric analysis. This group is discussed below.
25
Much as a modern suburbanite would respond with her specific town name if asked about her residence within the metropolitan area, but in
another part of the country might well simply give the city name as her home.
23
7
TABLE 1
COMPARISON OF ALL MALES, MATCHED SAMPLE, AND ESTIMATION SUBSAMPLE
All males (%)
24.36
Matched sample (%)
18.09
Estimation subsample (%)
16.02
15.84
11.22
3.01
2.91
2.58
2.11
1.68
1.61
1.29
1.12
22.30
11.96
2.20
2.30
2.11
2.66
1.59
1.68
1.68
1.27
27.61
18.76
0.45
2.81
3.05
3.26
1.51
1.80
0
1.75
44.56
33.80
5.69
1.89
1.86
59.86
25.24
3.46
3.69
1.32
100
0
0
0
0
Married
Unmarried
Widowed
47.72
47.66
4.18
42.31
55.61
1.45
4.12
95.21
0.22
Region
East
Lancashire-Cheshire
London
London Environs
Midlands
North
South
Wales
York
6.58
11.31
8.55
12.70
19.89
5.06
18.49
4.93
12.50
7.88
12.43
5.14
13.71
20.94
5.02
20.57
4.61
9.70
9.86
8.77
0
13.99
25.12
5.25
26.50
3.13
7.39
Lived in county of birth in 1851
71.56
88.41
95.12
187,117
28,474
3,774
Age (mean)
Occupation
Student
Ag. Laborer
Farmer
Laborer
Miner
Weaver
Tailor
Carpenter
None
Shoemaker
Relation
Son
Head
Lodger
Servant
Visitor
Marital status
N
Notes: All values are from 1851. The occupations listed are the 10 most prevalent among the sample of all males. Students were children
either attending school or receiving formal instruction at home.
Sources: 2 percent sample of 1851 census and new sample of matched individuals.
In order to examine the rural-urban migration decision, a smaller subsample must be extracted from the
28,000 matched individuals. If we want to see what caused some to move to the city and others to remain behind,
and what happens to those who go compared to those who stay, then we obviously want to limit ourselves to
considering individuals who began the period outside of any city. There is no hard definition for what constitutes a
“town” or a “city.” The United Nations has recommended that all places with more than 20,000 inhabitants living
close together be considered as “urban.” Not all countries use this definition. In fact, the U.S. Census defines an
“urban place” as any locality with more than 2,500 people. Within the literature these two numbers are often
considered to be meaningful cut-off points. In 19th century Britain, both places of 2,500-20,000 and those of more
than 20,000 were on the rise. The population of the former quadrupled from 1801-1901, and that of the later
8
increased by a factor of eleven; for the years 1841-1901, the multiples were two and four, respectively. So while
both were growing, it was the cities of more than 20,000 that were experiencing the most rapid expansion, and it
will be these that constitute “urban” for the purposes of this study.26,27
In order to judge the effects of moving, or not moving, to the city, the sample must also be limited to those
individuals for whom there is solid economic information from both censuses. Unlike their U.S. counterparts, the
Victorian censuses do not provide any quantitative economic variables such as personal or real estate wealth. What
they do include is fairly detailed occupational information. A distinction was made between employer and
employee and master and apprentice, and the number of persons employed was recorded. The instructions given to
the enumerators in 1851 are as follows:
In trades the master is to be distinguished from the Journeyman and Apprentice, thus- ‘(Carpenter, master employing
[6] men)’; inserting always the number of persons of the trade in his employment on March 31st.
In trades where women or boys and girls are employed, the number or each class should be separately given. Where the
master is one of a manufacturing or mercantile firm, the entry should be after this form: – ‘Cotton manufacturer – firm of
3, employ 512 men, 273 women, 35 boys, and 272 girls.’ 28
The occupation variable contains similarly detailed information for farmers and agricultural workers. The
enumerators were told that
The term FARMER is to be applied only to the occupier of land, who is to be returned – ‘Farmer of [317] acres
employing [12] labourers,’ the number of acres, and of in and out-door labourers, on March 31st, being in all cases
inserted. Sons or daughters employed at home or on the farm, may be returned – ‘Farmer’s Son,’ ‘Farmer’s Daughter.’ 29
There were literally thousands of occupations listed in the enumerators’ books; there are in the neighborhood of
8,000 unique occupations given in the set of 28,000 matched individuals.30 Making sense of all of this occupational
information requires some work. In the absence of wage information, some system of ranking jobs according to
their desirability must be employed. Clearly nothing like the 65-tier, thoroughly modern Hope-Goldthorpe scale
will be applicable to eighteenth century occupational data.31 Fortunately, fairly strong consensus has arisen in
support of a ranking scheme proposed by Armstrong.32 He argues that only the Registrar General’s social
classification schemes of 1921 and 1951 satisfy the dual requirements of being not too refined for the rough (by
modern standards) Victorian census data and at the same time including exhaustive published lists whereby the
26
“Urban” being so defined, the terms “rural” and “countryside” will be used throughout the paper to refer to all other places; hence, the
terms are used rather loosely, as towns of, say, 15,000 hardly could be considered truly “rural” in mid-19th century Britain.
27
Town and city populations from the 1881 census were used both for places in 1851 and in 1881, to ensure that the nationwide effects of
increasing population between the two dates did not cause people to appear to urbanize simply by living in a growing town.
28
Qtd. in Higgs, Clearer Sense, p. 108. Instructions for 1881 were not substantially different.
29
Ibid, p. 104.
30
This after substantial cleaning of the data. A first pass reveals 19,139 non-identical occupation responses from the 28,000 men.
31
For a description of the Hope-Goldthorpe scale and its application to modern British occupational data, see Mayhew and Rosewell,
“Occupational Mobility.”
32
Armstrong, “Information.”
9
multitude of occupations listed in the census can be classified into established categories. Armstrong recommends
the 1951 scheme on the grounds that it is simpler to implement, but for the present study the 1921 scheme was
chosen for its earlier date. It is, after all, thirty years closer to the Victorians. This scheme consists of five ranked
classes of occupation:
Class I.
Class II.
Class III.
Class IV.
Class V.
Professional etc., occupations
Intermediate occupations
Skilled occupations
Partly skilled occupations
Unskilled occupations
This system of classification is clearly aligned to capture job desirability. It makes a normative statement:
higher-class jobs were better than lower class. Two sources, both published by the General Register Office of Great
Britain, are needed to classify each occupation fully. The Classification of Occupations, 1921, lists approximately
16,000 different occupations and gives a three-digit code number for each. The Decennial Supplement for 1921
then provides the appropriate class rank for every code number.33 Armstrong provides several simple,
straightforward, and most importantly, objective modifications to this system to bring it into somewhat better
harmony with the nature of nineteenth century occupational and class structure. Probably most important is his use
of the employee information given under the occupational field in the census. Regardless of job title, all employers
of 25 or more are to be placed in Class I, and all people with Class III or IV occupations employing at least one
person other than a family member is to be placed in Class II. Since Armstrong first proposed this scheme in 1966
and further developed it in 1972, it has become the standard for ranking occupations in Victorian Britain.34
Armstrong’s basic scheme has been modified in one way, in order to take full advantage of the information offered
by the census. Since entire households are observed both in 1851 and 1881, including not only family members but
servants as well, the ratio of servants to household members can be calculated for each household. The job class
ranking of some individuals was upgraded (never downgraded) according to the following scheme, proposed by
Royle: all heads whose households contained at least one servant per household member were placed in Class I, all
others with one servant per three household members in Class II, and any others that employed at least one servant
in Class III.35 Formulated in this way, the job class variable may be thought of as a representation of socioeconomic
status, and the terms occupational class and socioeconomic status are used interchangeably throughout the study.
33
Warm thanks to Dr. Humphrey Southall for providing me with a computerized dictionary with which I was able to code approximately one
third of the occupations. This dictionary, along with many new entries which I have provided, is available through the Great Britain
Historical Database. See http://www.geog.qmw.ac.uk/gbhgis/database/db_index.html. Armstrong recommends using the 1951 scheme for the
sole reason that both sets of information needed to code the occupations are contained within one book, the Classification of Occupations,
1951. Furthermore, this book is much more readily available in libraries throughout the country than are the two 1921 volumes.
34
Armstrong demonstrates that job class, defined according to this system, is positively correlated with the employment of servants and
negatively correlated with the incidence of shared accommodation. “Information,” p. 212.
35
Royle, “Social Stratification.”
10
The final cut taken from the data also has to do with the amount of information available for individuals. In
1851 the men in the sample were either heads of a household (25%), sons of the head (60%), grandsons, nephews,
or brothers of the head (4%), or non-family members (servants, lodgers, and visitors being the most common, in
that order). Sons are by far the largest group, which is good, because they offer the richest source of information.
For sons (if they are old enough) we can observe not only their class and industry of occupation, but also those of
their fathers. Then as now, the concept of intergenerational occupational mobility was an important one. The
occupation and class of the father were powerful influences on the occupation and class of the son. We want to
know the extent to which moving to the city allowed people to get better jobs—better jobs than they themselves
used to have, and better jobs than their fathers before them had. So the analysis will focus on the largest group
within the sample: the sons. As for considerations of age, the minimum legal working age was nine years, so only
those sons who were nine years old or older in 1851 will be included in the sample. Only a small minority of sons
in the sample were in their thirties or older. These were excluded, as the forces governing their migration decision
might be significantly different from those of young men. Out of the 28,474 matched individuals, 3,774 are sons
living with their father in a rural area in 1851, with both son and father reporting solid occupational information.
They are shown in the third column of Table 1 for comparative purposes, though of course this subsample is
unrepresentative of the male population of England and Wales by construction. It is this group which will be used
to analyze rural-urban migration.
A Model of Rural-urban Migration
Migration to the city is not, of course, a randomly assigned treatment; an individual’s expectation of his labor
market prospects will influence his decision of whether or not to move. The model used to analyze the migration
decision, then, is one of “regime switching with endogenous switching,” to paraphrase Maddala’s terminology.36
The basic model is
Regime 0:
y 0i = β 0′ X 0i + ε 0 i
if Di = 0
(1)
Regime 1:
y1i = β 1′ X 1i + ε 1i
if Di = 1
(2)
Di = 1
if γ ′ Z i + u i ≥ 0
(3)
Di = 0
otherwise
(4)
where y (generally a continuous variable) represents the labor market outcome and D the decision of the regime in
which to participate. X and Z are the factors that influence an individual’s labor market outcome and participation
36
Maddala, Limited-Dependent, pp. 223-28. Also referred to as a Type 5 Tobit Model; see Amemiya, Advanced Econometrics, pp. 399-402.
For a good overview of the identification and parametric and semiparametric estimation issues associated with this class of model, see
Heckman, “Varieties.”
11
decision, respectively, β and γ are vectors of coefficients, and ε and u are unobservable factors. A wide range of
empirical questions has been examined with this model, including union-nonunion wage analysis, housing-demand,
and migration.37 A modification of this model will be employed here. The model cannot be adopted wholesale
because in the present case the labor market outcome variable yi is not a continuous variable like wage or wealth,
but rather it is the discrete variable job class (or socioeconomic status). The standard model involves the estimation
of two linear outcome equations, (1) and (2), and one nonlinear decision equation represented by (3) and (4). The
model developed here allows for equations (1) and (2) to be nonlinear as well.
The model is defined as follows. There exists a continuum of job quality, Y * ∈ (− ∞, +∞ ) . Individuals have
(
)
a utility function U y * , Θ , where y* is a realization of Y* and Θ is a vector of other inputs. All individuals prefer
higher quality jobs: ∂ U ∂y * > 0 . The maximum quality of job that an individual can attain, y i* , is defined to be a
linear function of that person’s observable traits and skills, Xi , and unobservable characteristics, εi . The function
may be different in the urban regime than in the rural, though the relevant elements of Xi are assumed to be the
same in each:
Rural:
y 0*i = β 0′ X i + ε 0 i
(5)
Urban:
y1*i = β 1′ X i + ε 1i
(6)
The decision facing each individual in this model is whether to migrate to a city or remain in a rural setting. The net
benefit of moving, Di* , is defined to be a linear function of (1) Zi , a vector of observable individual and location
specific characteristics, (2) the difference between the individual’s maximum attainable job quality in the urban
(
)
regime versus that in the rural, y1*i − y 0*i , and (3) ui , unobservable characteristics:
(
)
Di* = γ 1′ Z i + γ 2 y1*i − y 0*i + u i
(7)
Only the outcome of the choice, Di , is observed: Di = 1 if Di* ≥ 0 , Di = 0 otherwise.38 To this point, the model is
identical to the base model specified in (1) – (4). The difference is that in the base model, either y0i or y1i is observed
for each individual, depending on the value of Di , whereas in the current model neither y 0i* nor y1i* is observed for
any individual. What is observed is the job class, yi , of every individual:
37
See Trost, “Demand”; Lee, “Unionism”; Robinson & Tomes, “Self-Selection”; and Ferrie, Yankeys.
Note that the decision is not whether to move or stay, as is typical in most migration studies. The decision is whether to move to an urban
area or not. So an individual who moves from one rural area to another would have Di = 0.
38
12
y 0i ∈ [1,2,3,4,5]
if Di = 0
y1i ∈ [1,2,3,4,5]
if Di = 1,
(8)
The following relationship is assumed to exist between job class, yi , and job quality, y i* :
yri =
5
if
-∞ < y ri* ≤ k1
4
if
k1 < y ri* ≤ k2
3
if
k2 < y ri* ≤ k3
2
if
k3 < y ri* ≤ k4
1
if
k4 < y ri* ≤ +∞
(9)
r = 0,1
where k1 – k4 are constants. Class I jobs are the most desirable, and Class V the least. Under this formulation, all
jobs within a class are not equivalent; indeed, the best job in any class is only marginally inferior to the worst job in
the next highest class. Also, the quality/class structure is identical in the cities and the countryside, since the four
threshold levels k1 – k4 do not vary between the two regimes. This is as it should be; otherwise, an individual could
change job class without changing jobs simply by moving from one regime to the other.
The three-equation system defined by (7) and (9) is estimated by Full Information Maximum Likelihood
(FIML), which proceeds in two stages. First, by substituting (5) and (6) into equation (7), (7) can be written in
reduced form as
Di* = γ ′Wi + v i
(10)
where W contains all the elements of X and Z. A standard assumption is made regarding the disturbance terms: ε1 ,
ε2 , and v are assumed to be i.i.d. draws from a trivariate normal distribution with mean vector zero and covariance
matrix
σ 12

Σ=


σ 12 σ 1v 

σ 22 σ 2 v 
1 
(11)
13
With this distributional assumption, (9) becomes a standard ordered probit model with five outcomes and four
threshold levels, with the outcome depending on the latent index variable y*.39 The model, therefore, is one of
“switching ordered probits,” rather than the common switching regression model. The three equations in (9) and
(10) are jointly estimated by maximum likelihood. The likelihood function is
L( β 0 ,β1 , γ, k1 , k 2 ,k 3 ,k 4 ,ρ0 , ρ1 ) =


 F(k − β ′ X ,γ ′W, − ρ )

′
′
′
′
′
′
1
0 0i
0 ∏ F( k j − β 0 X 0 i ,γ W, − ρ0 ) − F( k j −1 − β 0 X 0 i ,γ W, − ρ0 ) ∏ F( β 0 X 0 i − k 4 ,γ W,ρ 0 ) 
∏
yi =5
yi = j
yi =1
j = 2 ,3 ,4


[
1− Di
]


 G( k − β ′ X , − γ′W,ρ )

′
′
′
′
′
′
1
1 1i
1 ∏ G( k j − β1 X 1i , − γ W,ρ1 ) − G( k j −1 − β1 X 1i , − γ W,ρ1 ) ∏ G( β1 X 1i − k 4 , − γ W, − ρ1 ) 
∏
yi = j
yi =1
 yi =5

j = 2 ,3 ,4
[
Di
]
where F and G are, respectively, the bivariate normal distribution functions of (ε0i , vi) and (ε1i , vi) and ρ0 and ρ1 are
the correlation coefficients of the two distributions. β0 and β1 are assumed to include a constant term, so the four
threshold levels, k, are not all identified, and k1 is normalized to 0. This estimation procedure produces consistent
and asymptotically efficient estimates of the parameters of interest, β0 and β1 .40
The second stage is the structural estimation of equation (7). With estimates of β0 and β1 in hand, predicted
values of both y 0i* and y1i* are obtained for each individual as yˆ ri* = βˆ r′ X i , r = 0,1. These predicted values are
substituted into (7) in place of the unobservable y 0i* and y1i* . (7) is then estimated by probit maximum likelihood to
obtain estimates of the structural parameters γ1 and γ2.
The estimates of β0 , β1, γ1, and γ2 reveal the determinants of job class attainment and urban migration. To
get at the central issues of the selection of urban migrants and rural persisters and especially the treatment effect of
migration, it is necessary to define several more parameters to be estimated. The selection of urban migrants (ϕ1)
and the selection of rural persisters (ϕ0) are given by
39
The distributional assumption is standard, but it is not innocuous, as the growing literature on semi- and non-parametric estimation of
selectivity and treatment effects demonstrates. See Heckman, “Varieties,” Manski, “Nonparametric,” and Newey et al. “Semiparametric.”
40
Standard “two-step” estimates of the parameters in a self-selection model are consistent but not efficient. Their advantage lies in ease of
computation; indeed, in econometrics texts one often encounters phrases like “maximization of the likelihood function…can be cumbersome”
(Maddala, Limited-Dependent, p. 224). No such problems were encountered with the likelihood function specified above, however. Given
appropriate starting values and the technical trick of maximizing the inverse hyperbolic tangent of ρ instead of ρ itself (in order to constrain it
between its logical bounds of -1 and 1 and for numerical stability during maximization), convergence was not problematic. At any rate, the
nonlinearity of the second stage job class equations precludes standard two-step methods for this model.
14
(
= E (y
) (
| D = 0 ) − E (y
)
| D = 1)
ϕ 1 = E y1* | D = 1 − E y1* | D = 0
ϕ0
*
0
*
0
(12)
They are calculated conditional on the observed variables using the estimated β coefficients and conditional means
of the X variables:
~
~
ϕ 1 = X 1 βˆ1 − X 0 βˆ1
~
~
~
~
ϕ 0 = X 0 βˆ 0 − X 1 βˆ 0 , where X 1 = E (X | D = 1) and X 0 = E (X | D = 0)
(13)
If ϕ1 > 0 then, conditional on the observables, urban migrants were positively selected; i.e., they achieved higher job
quality in the urban labor market than the rural persisters would have had they chosen to move to the city. The
analogous characterization holds for ϕ0 and the selection of rural persisters.
The treatment effect of rural-urban migration, τ, is defined to be
(
)
τ = E y1* − y 0* | D = 1
(14)
~
~
τ = X 1 βˆ1 − X 1 βˆ 0
(15)
and is calculated as
Like the selection parameters, τ is calculated conditional on the observables of the model. The treatment effect is
~
considered in terms of the average urban migrant: X 1 represents the endowments of the skills and attributes
available to the average migrant. If the treatment effect is positive, τ > 0, then migration to the city yielded a
positive return in terms of job quality for the average rural individual in England and Wales of the type who
typically chose to move to the city from the countryside.41
Estimation of the Model
The determinants of maximum attainable job quality in 1881—the elements of X in (5) and (6)—include class rank
in 1851, father’s class rank, age, age squared, “age discrepancy,” whether the individual lived in a town in 1851,
whether the individual was an eldest son, whether the father was a farmer or employer, an interaction between the
41
Conditioning the selection and treatment effect parameters on the observable variables of the model entails the following assumptions: (1)
E(ε1 | D = 1) - E(ε1 | D = 0) = 0, (2) E(ε0 | D = 1) - E(ε0 | D = 0) = 0, and (3) E(ε1 - ε0 | D = 1) = 0. See Heckman, “Varieties,” for a theoretical
and empirical discussion of these assumptions.
15
previous two terms, the industrial classification of the individual’s occupation in 1851, the degree of age-heaping
for the county of residence in 1851, and region in 1851. Summary statistics for these variables, and for those in the
migration equation, are presented in Tables 2 and 3. The first table summarizes the attributes of all of the
individuals in the estimation sample in 1851 and 1881. The second compares the attributes of those who decided to
move to a city by 1881 with those who remained in a rural place, also giving values for both 1851 and 1881. This
table will be discussed in more detail in a subsequent section. Several of the variables merit some discussion. The
inclusion of the individual’s class rank in 1851 allows us to measure the effect of moving to the city on the change
in an individual’s job class rank; it controls for individual-specific effects on the level of job quality. This is
analogous to running a regression on change in wage or wealth without constraining the coefficient on the first time
period value to be equal to one. Formulating the model in this way offers several advantages. First, if change in job
class were the independent variable in the model, the set of possible realizations of that variable would not be the
same for all individuals. For example, only people who started in Class V in 1851 would not have the possibility of
a downward move. Second, it is interesting in its own right to measure the effect of class in 1851 on class in 1881
as an indicator of the rigidity of occupational class structure in Victorian Britain. Finally, we might expect that the
influence of class in 1851 on class attainment in 1881 would be different for someone who switched regimes than
for someone who made no switch. A priori we might expect that the relationship would be weaker for those
individuals who moved to the city, given that they would be entering a somewhat different job market and would be
sure to experience some sort of change in occupation (even if the change were as small as being a spinner in a
different factory or a carpenter in a different town). The job class of the father is also included. Just as estimating
the effect of the individual’s class in 1851 on his class in 1881 allows us to measure the degree of occupational
mobility, so including father’s class in 1851 reveals the degree of intergenerational mobility, an important question
itself. So the model is really one of change in job class, where part of the change is from the individual’s own class
in 1851 to his class in 1881, and part is an intergenerational class change from father’s class in 1851 to son’s class
in 1881.
One “class” included in 1851 is not a proper occupational classification at all: that of “student.” It is the
second most common class in 1851, and the single most commonly reported occupation, in this sample as well as in
the census as a whole. It does not fit into the Registrar General’s occupational classification system, but it may be
regarded as a class of its own. For the 1851 census, parents were instructed to report their children as students
(“scholars” in the language of the Victorian censuses) if they were older than five and were “daily attending school,
or receiving regular tuition under a master or governess at home.” The instructions for what constituted a student
were more stringent for the 1851 census than for any other of the nineteenth century: the requirement of being
under a master or governess if at home was dropped from the 1861—1881 censuses, and no instructions at all were
given on the subject in 1891. Furthermore, compulsory education did not become law until the Education Act of
1870, so there was no legal incentive for parents to lie about their children’s schooling. Though no substantive
16
TABLE 2
SUMMARY STATISTICS
Means
INDIVIDUAL CHARACTERISTICS:
Job class
1 - Professional
2 - Intermediate
3 - Skilled
4 - Partly skilled
5 - Unskilled
Student
Father's class
1 - Professional
2 - Intermediate
3 - Skilled
4 - Partly skilled
5 - Unskilled
Age discrepancy (years)
0
1
2
3
4
5
Age
Eldest
Inheritance
Married
Not in town of birth
Industry
Agriculture
Building
Distributive
Mining
Textiles
Iron & Steel
Other Manuf.
Other
LOCATION CHARACTERISTICS:
Living in city
Living in town
Region
Lancashire
London
London Environs
Wales
Yorkshire
Others
Age-heaping
Distance to city
Wage gap
Previous migrants
Unemployment
Nearby cities
Pct. in agriculture
Pct. in manuf.
Notes: N = 3,774. Variables defined in text.
Sources : See text.
1851
Frequency
Percent
14
63
1144
1046
436
1071
0.37
1.67
30.31
27.72
11.55
28.38
61
621
1445
1255
392
1.62
16.45
38.29
33.25
10.39
1881
Frequency
Percent
89
581
1713
944
447
---
2.36
15.39
45.39
25.01
11.84
---
1853
1095
386
205
116
119
49.10
29.01
10.23
5.43
3.07
3.15
16.02 years
2166
453
93
980
57.39
12.00
2.46
25.97
907
123
71
145
292
84
560
1592
24.03
3.26
1.88
3.84
7.74
2.23
14.84
42.18
906
304
330
182
168
146
764
974
24.01
8.06
8.74
4.82
4.45
3.87
20.24
25.81
0
1918
0
50.82
896
1524
23.74
40.38
210
0
528
118
279
2639
5.56
0
13.99
3.13
7.39
69.93
307
156
511
121
286
2393
8.13
4.13
13.54
3.21
7.58
63.41
61.0 (index)
23.35 km
4.79 sh/wk
13,716
7.97%
32.56
29.29%
23.58%
17
TABLE 3
SUMMARY STATISTICS, URBAN vs. RURAL
Mean, 1851
Moved to city
Remained rural
INDIVIDUAL CHARACTERISTICS:
Job class
1 - Professional
0.007
0.003
2 - Intermediate
0.017
0.017
3 - Skilled
0.324
0.297
4 - Partly skilled
0.195
0.303
5 - Unskilled
0.102
0.120
Student
0.356
0.261
Father's class
1 - Professional
0.026
0.013
2 - Intermediate
0.170
0.163
3 - Skilled
0.453
0.361
4 - Partly skilled
0.267
0.353
5 - Unskilled
0.085
0.110
Age discrepancy (years)
0
0.486
0.493
1
0.284
0.292
2
0.085
0.108
3
0.067
0.050
4
0.044
0.027
5
0.036
0.030
Age
15.3
16.3
Eldest
0.566
0.576
Inheritance
0.107
0.124
Married
0.020
0.026
Not in town of birth
0.325
0.239
Industry
Agriculture
0.146
0.270
Building
0.031
0.033
Distributive
0.019
0.019
Mining
0.021
0.044
Textiles
0.123
0.063
Iron & Steel
0.027
0.021
Other Manuf.
0.151
0.148
Other
0.482
0.403
LOCATION CHARACTERISTICS:
Living in town
0.626
0.472
Region
Lancashire
0.111
0.039
London
0
0
London Environs
0.150
0.137
Wales
0.018
0.035
Yorkshire
0.083
0.071
Others
0.640
0.718
Age-heaping
61.6
60.8
Distance to city
19.6
24.5
Wage gap
4.98
4.73
Previous migrants
15,339
13,211
Unemployment
8.84
7.70
Nearby cities
36.8
31.2
Pct. in agriculture
0.264
0.302
Pct. in manuf.
0.259
0.229
Notes: N = 3,774. 896 moved to an urban area; 2878 did not. Variables defined in text.
Sources : See text.
Mean, 1881
Moved to city
Remained rural
0.031
0.141
0.564
0.116
0.148
---
0.021
0.158
0.420
0.292
0.109
---
0.036
0.099
0.127
0.027
0.061
0.059
0.229
0.362
0.304
0.075
0.075
0.055
0.039
0.032
0.194
0.226
0
0.530
0.199
0.174
0.135
0.017
0.088
0.387
0.045
0
0.136
0.037
0.072
0.711
18
details of their instruction can be known, children older than nine who were listed as students were receiving an
education.42 Many of their peers were not; they were working.43 By including the students as a separate class, we
can gauge in a very rough fashion its efficacy for future job quality attainment as compared to being employed at a
young age.
Literacy and specific educational information are not observed, but an interesting proxy is available. “Age
discrepancy” is defined to be
Reported Age 1881 − (Reported Age 1851 + 30)
As discussed above, the age matching criterion is that an individual’s reported age in 1881 cannot be more than five
years different from what it “should” be (the reported age in 1851 plus 30).44 Roughly half of all individuals in the
sample reported consistent values for age in the two censuses. Another quarter was off only by a year, and the
remaining quarter was off by two to five years. This was a time before systematic record keeping, and many people
had only an approximate idea of their age.45 The observed age discrepancy for each individual gives an indication
of that individual’s familiarity with arithmetic (“numeracy”) and the precision of his concept of time. In general it
may be regarded as a proxy for literacy, education, intelligence, and the like, and therefore be expected to have a
positive impact on job quality attainment. 46,47
For the population as a whole, this tendency of people not to know their age precisely manifests itself as
age heaping—the tendency of any population to over-report rounded ages. Hence, the census will record many
more 30-year olds than 29- or 31-year olds, for example. The precise degree of age heaping can be measured for
each county in England and Wales by comparing the observed age distribution (using the entire population of each
county from the 1881 census) to an approximation of the true distribution obtained by applying any of a number of
42
See note on following page regarding the correlation between student status and inconsistency in age reporting in the census.
See Tuttle, “Role of Children,” for a discussion of child labor in 19th century Britain and the high incidence of gainful employment for
young children.
44
In most cases for this sample, the age in 1851 would have been reported by the parents. It could be possible, therefore, for an individual to
report his correct age and still have a non-zero value of age discrepancy, if his parents had incorrectly reported his age in 1851. This seems
unlikely, though, as most people would learn their age from their parents.
45
For a discussion of age-enumeration in the Victorian censuses, see Higgs, Clearer Sense, ch. 7.
46
There is, in fact, a correlation between age discrepancy and being a student in 1851. Students were 8% more likely to know their exact age
than non-students; among those whose age reporting was not consistent between the two censuses, students were 4% more likely to be off
only by a year, rather than by 2–5 years.
47
Under reporting was more common at every level than over-reporting, by about 23% overall. There may be reason to think that those who
under-report their age are less likely truly to be ignorant of their exact date of birth and/or of arithmetic than those who over-report; it may be
simple vanity at work. To check this, the model was re-estimated with age discrepancy defined as the actual discrepancy, rather than the
absolute value, with the following formulation:
43
… + β-2(discrep ≤ -2) + β-1(discrep = -1) + β1(discrep = 1) + β2(discrep ≥ 2) + …
Neither of the hypotheses that β-2 = β2 and that β-1 = β1 could be rejected with any confidence.
19
smoothing techniques to the observed distribution. The technique used here is Graybill’s weighted moving average
of the Sprague coefficients, which produces drastic smoothing of the heaped distribution.48 An index of age heaping
is then calculated for each county relative to the county with the highest degree of heaping (the Isle of Man,
followed by Jersey and Radnorshire; lowest were Sussex, Westmoreland, and Yorkshire). Age heaping has been
shown to be correlated with important measures of economic development like per capita income, so in this context
it is designed to capture otherwise unmeasured community effects that might influence the labor market
performance of individuals.49
The remaining variables are more straightforward. People living in a town in 1851, as opposed to those in
the truly rural countryside, may have possessed a form of “locational” human capital that would ease their
transition into the big-city labor market. This effect is captured by the inclusion of a dummy variable which equals
one if the individual lived in a town of between 2,500 and 19,999 inhabitants. The dummy variables indicating
whether a person in the sample was an eldest son and whether his father was a farmer or an employer are designed
to capture an inheritance effect. Fathers who owned land were likely to leave it to their eldest son, as the system of
primogeniture was the norm with regard to property inheritance in most of England. Likewise, sons of fathers who
owned any sort of business were likely heirs to an occupational inheritance, though in this case the inheritor well
could have been any son rather than just the eldest. Finally, industrial and regional dummy variables are included to
test whether certain industries and locations offered particularly advantageous or disadvantageous starting points
for job quality attainment. Industrial classification is according to the well-known scheme developed by Charles
Booth for his survey Life and Labour of the People in London, written in nine volumes from 1892 to 1897. Booth’s
system consists of 10 divisions and 49 subdivisions, the subdivisions generally being aggregated in some fashion to
facilitate analysis. The aggregation used in this case, along with the relative frequency of each group in the data, is
given in Table 2.
In addition to his expectation of his future job prospects in both regimes, the factors that influence an
individual’s decision whether to move to the city—the components of Z in (7)—include both individual- and
location-specific elements. The individual-specific characteristics include age and age squared, marital status, age
discrepancy, whether the person lived in a town in 1851, whether the person was living in the town of his birth in
1851, and industrial classification. The human capital interpretation of migration (that people migrate in order to
maximize lifetime net benefit from moving) would suggest that the likelihood of migrating will decrease with age,
as individuals have a shorter time span over which to reap the gains from moving and as they make more locationspecific investments in their place of residence. More educated, knowledgeable people would have a lower
tendency to report their age inconsistently; these same people should be better able to gather information on
48
For a complete description of this and other smoothing methods employed by demographers, see Shryock and Siegel, Methods and
Materials, pp. 696–702, 878.
49
For a more thorough discussion of the economic usefulness of age heaping information, particularly in a migration analysis setting, see
Mokyr, Ireland.
20
potential urban moves. Moving to a big city might have been a less drastic change, and so have carried a lower
psychic cost, for those already living in towns in 1851. There are only two new variables in this group. The first is
marital status, the expected effect of which is not clear a priori. While studies of overseas migration generally find
the typical migrant to be a young, single male,50 this may not be the case for shorter-distance (indeed often local)
moves from countryside to city. Indeed, during the second half of the nineteenth century, the decline in rural
employment affected the job prospects of women more than those of men, so women became more likely than men
to leave the countryside in order to find work in the city, most often as domestic servants.51 The pull of the urban
labor market being strong for women, a married young man whose wife was looking for work outside the home
might have been more likely to move to the city than an unmarried man. The second new variable is an indicator of
whether an individual was living in his town of birth in 1851. Approximately 26% of the individuals in the sample
were not; the household had therefore already made a move prior to 1851. These individuals (or at least the heads
of their households) had already demonstrated some willingness to move, and their ties to their community in 1851
might not have been as strong as those who were still living in their town of birth. For both reasons, these
individuals are expected to be more likely to decide to move to the city by 1881.
The elements of Z that depend on the individual’s location in 1851 include the distance to the nearest city,
the number of large cities within 100 km, a proxy for the total number of previous migrants in each nearby city, the
average rural-urban wage gap, the urban unemployment rate, an interaction between the unemployment rate and
distance to nearest city, the percent of the male workforce engaged in agriculture and in manufacturing in the home
county in 1851, the degree of age heaping of the home county in 1851, and the regional dummies. Of these
variables only age heaping and the regional dummies are in X. Distance to the nearest city and number of large
cities nearby were calculated using modern grid point references for place names from the 1991 British Census.52
Distance should be negatively correlated with tendency to migrate, as it would be both more costly to move to a
distant city and more difficult to gather pertinent information; conversely, number of nearby cities should exhibit a
positive correlation. Knowing people who had already moved to a particular city was another important source of
information for the potential migrant. A proxy was constructed for this “friends and family” effect. First, the county
of birth of all urban residents in England and Wales in 1881 was tabulated by city using the 1881 census data.
Second, the set of most likely destination cities for each individual in the sample was defined to be the nine closest
cities to his town of residence in 1851 plus London.53 Finally, for each individual i with likely destination cities
50
For example, see Hatton and Williamson, “Mass Migrations.”
Baines, “Population.”
52
These data are readily available only to academics working within the U.K. I am very grateful to Justin Hayes with MIMAS at the
University of Manchester for personally providing me with the geographic data from the 1991 census (after receiving all proper permissions,
of course). Karen Dennison with the U.K. Data Archive at Essex University and Ann Janda with the Social Science Data Services at the
Northwestern University Library were also very helpful.
53
Hence, all individuals from the same town will have the same set of likely destinations, which is a desirabl e property for the set. A direct
result of this property is that it is possible that an individual’s actual choice of destination city will not be among his likely choices. London
51
21
j = 1,…,10, the proxy for the stock of previous migrants is defined to be
previous migrants i =



j =1 

10
∑

m
−1  ij
d 
j ij 
d ij−1
∑
(16)
where mij is the number of people living in city j in 1881 born in the county of residence of individual i, and dij is
the distance between city j and the place of residence of individual i. This is a proximity-weighted average of the m
values of all likely destination cities. The average urban-rural wage gap and the urban unemployment rate were
calculated along similar lines. The wage gap is defined as the average wage of laborers in the building trade in
nearby cities minus the average wage of agricultural laborers in the origin county.54 The urban unemployment rate
was calculated using data on joblessness among members of the Amalgamated Society of Engineers, available for
56 cities in England and Wales from 1858 through 1909.55 It is calculated as the proximity-weighted average—as in
(16)—of the rates from 1862, 1868, and 1879 for the two nearest cities and London. This is the urban
unemployment rate, so its expected effect on tendency to move to the city should be negative; people would be less
likely to move to the city if there were high unemployment there. However, we do not observe the relevant local
unemployment rate and so cannot control for it. Therefore, since the urban unemployment rate is taken from the
cities nearest to each individual’s place of origin, it will proxy, to an extent, for the local unemployment rate facing
each individual. Insofar as this is true, the effect of unemployment would be positive.56 To test for this effect, an
interaction term between the unemployment rate and distance to nearest city is included. We expect this term to
have a negative coefficient: for any given unemployment rate, an increase in the value of the interaction term
indicates that the unemployment is “farther away,” thus more truly the unemployment of the destination city and
less that of the place of origin. Finally, the fraction of the male workforce in the county of origin engaged in
agriculture and in manufacturing is taken from the occupational tables from the 1871 census of the population. 57
The three parameter vectors β1, β2, and γ are identified under the assumptions of the model; however, if the
variables in X and Z are identical, this holds only if the structure and normality assumptions of the model are
attracted migrants from a great distance much more than did any other city in Great Britain, so it is considered to be a feasible option for all
potential migrants.
54
Agricultural laborers’ wages are from Hunt, “Industrialization,” Table 6. They cover the years 1867-1870. Building laborers’ wages are
from Hunt, Regional Wage Variation, Table 1.5, for the year 1886. As this series includes only 36 cities in England and Wales, the set of
likely destination cities was limited to London plus the closest two, rather than the closest nine. City wages were proximity-weighted as in
(16).
55
These data are from Southall, “Regional Unemployment,” Table 3. For a discussion of its appropriateness as a proxy for the unemployment
experiences of other groups, see Southall and Gilbert, “Good Time to Wed.”
56
Unfortunately, the occupational information from the 1851 census cannot be used to determine individual unemployment within the
sample. Specific instructions regarding the unemployed were not given to enumerators until 1861; many people out of work in 1851 reported
the fact, but many simply reported the job they most recently held.
57
Lee, British Regional Employment Statistics.
22
exactly correct, “almost certainly too thin a reed upon which to base inference.”58 Fortunately, in this case there are
some reasonable exclusion restrictions. Z includes quite a few variables not present in X: marital status, whether the
person was living in the town of his birth in 1851, and all of the location-specific variables save age heaping and
the regional dummies. X does not include quite so many unique variables: the eldest and inheritance dummies,
father’s job class, and own job class in 1851. These, of course, do influence the individual’s migration decision, but
only through their effect on
(y
*
1i
)
− y 0*i . With these exclusion restrictions, the parameters of the model are
identified, even if the assumptions of the model do not hold exactly.
Empirical Results
Occupational Class Attainment
Results from estimating the two job class equations, (9), as well as from what may be considered the “baseline”
model, an ordered probit estimation of job class in 1881 on X plus a dummy variable for moving to a city by 1881,
are presented in Table 4. The columns labeled “Urban” show the estimates of β1 , the coefficients for those who
chose to move to an urban area by 1881, and the columns labeled “Rural” show the estimates of β0 , the coefficients
for those who remained in a rural place throughout the time period.59 The ancillary parameters ρ and k do not vary
between the two groups. Interpretation of ordered probit results is not exactly straightforward. The coefficient
estimates represent the effect of a change in the explanatory variable on the unobserved latent variable y*, job
quality in this case. Job quality is a theoretical construct of the model; it has no units or ready interpretation. The
coefficient estimates reveal the sign of the overall effect of the explanatory variable, the precision with which the
effect is estimated, and the statistical significance level of the effect. For an understanding of the magnitude of the
effect on the “real world variable,” here job class, the marginal effect of each dependent variable must be
calculated. These are shown in Table 5, with the “Urban” and “Rural” columns as in Table 4. Also shown are the
baseline probabilities of attaining each job class conditional on moving to a city or remaining in a rural place. These
are calculated at the mean values of all explanatory variables. Each explanatory variable has a different marginal
effect for each possible outcome of the ordered probit, so there will be a different effect for each of the five possible
job classes that an individual could attain in 1881. Note that the sign of the overall effect of each explanatory
variable is unambiguous (though perhaps not statistically significant from zero), but the sign will vary across 1881
job classes. A variable which positively influences job quality will positively influence the probability of being in a
high job class but will negatively influence the probability of being in a low job class. As expected, the 1851 class
58
59
Johnston and Dinardo, Econometric Methods, p. 450.
Again, these individuals did not necessarily remain in the same place; they could have moved from one rural area to another.
23
TABLE 4
MAXIMUM LIKELIHOOD RESULTS: DETERMINANTS OF JOB CLASS ATTAINMENT
Job class, 1881
Variable
Constant
Job class
Job class, 1881
Urban
Rural
(1)
(2)
(3)
0.449
-0.131
0.699
(.319)
(0.758)
(0.359)
(†††)
(†††)
(†††)
1,2 - Prof., Intermed.
1.371***
1.958***
1.156***
(0.147)
(0.289)
(0.172)
3 - Skilled
0.354***
0.403**
0.317***
(0.089)
(0.193)
(0.101)
4 - Partly Skilled
0.136
0.311
0.054
(0.109)
(0.227)
(0.126)
0.515***
0.785***
0.384***
(0.074)
(0.162)
(0.086)
(†††)
(†††)
(†††)
1 - Professional
1.427***
0.930***
1.630***
(0.159)
(0.283)
(0.203)
2 - Intermediate
0.463***
0.444**
0.366***
(0.101)
(0.198)
(0.121)
0.342***
0.213
0.348***
(0.064)
(0.142)
(0.074)
-0.134**
-0.238
-0.117*
(0.065)
(0.148)
(0.072)
0.201***
0.133
0.224***
(0.046)
(0.095)
(0.053)
-0.008
-0.028
0.004
(0.05)
(0.103)
(0.057)
0.052**
0.167***
0.009
Student
Father's class
3 - Skilled
4 - Partly Skilled
Age discrep. (years)
(†††)
0
1
Age
Age 2 / 100
Eldest
Inheritance
Eldest*Inheritance
Age heaping
Living in town
Job class, 1881
Variable
(1)
Industry
(0.027)
(0.059)
(0.031)
-0.505***
-0.023
(0.072)
(0.162)
(0.081)
0.03
0.111
0.002
(0.039)
(0.081)
(0.045)
0.335***
-0.102
0.552***
(0.11)
(0.213)
(0.13)
-0.039
-0.252
-0.012
(0.108)
(0.237)
(0.121)
Urban
Rural
(2)
(3)
-0.128
(†††)
(†††)
Agriculture
-0.181*
-0.199
(0.106)
(0.234)
(0.122)
Building
0.280**
0.305
0.301**
(0.128)
(0.273)
(0.144)
Distributive
0.319**
0.126
0.361**
(0.148)
(0.308)
(0.168)
Mining
Textiles
Iron & Steel
Other manuf.
Region
Lancashire
London Environs
Wales
(†††)
-0.152**
Job class, 1881
Yorkshire
Moved to city
0.065
0.167
0.083
(0.122)
(0.309)
(0.136)
-0.114
0.033
-0.200*
(0.102)
(0.211)
(0.119)
-0.022
-0.049
-0.016
(0.144)
(0.289)
(0.166)
0.204**
0.271
0.192
(0.089)
(0.189)
(0.101)
(†)
(†††)
0.041
-0.015
-0.037
(0.089)
(0.155)
(0.12)
-0.075
-0.353***
0.001
(0.053)
(0.109)
(0.061)
0.166
0.113
0.219*
(0.116)
(0.313)
(0.126)
0.126*
0.282*
0.045
(0.075)
(0.154)
(0.086)
-0.005
(0.043)
ρ1
-0.053
(0.100)
ρ2
-0.324***
(0.088)
k2
0.959***
(0.028)
(0.031)
k
3
2.429***
2.408***
(0.038)
(0.050)
k
4
3.665***
3.642***
(0.060)
(0.076)
0.948***
-0.003
-0.005
-0.003
(0.003)
(0.007)
(0.004)
0.051
-0.084
0.050
Pseudo R2
0.092
(0.038)
(0.083)
(0.045)
Log likelihood
LR χ2
-4572.715
-6457.798
925.530
1284.230
Prob > χ2
0.000
0.090
0.000
* = Significant at the 10 percent level ** = Significant at the 5 percent level *** = Significant at the 1 percent level
† = Jointly significant at the 10 percent level †† = Jointly significant at the 5 percent level ††† = Jointly significant at the 1 percent level
Notes: N = 3,774. The two columns of equation (1) represent all 3,774, (2) represent the 896 who moved to a city, and (3) the 2,878 who remained in a rural place.
2
Standard errors are in parentheses. All explanatory variables use 1851 values. Pseudo R ≡ 1 - L1/L0, where L1 is the log likelihood of the model and L0 is the log
likelihood of the “constant only" model. The LR χ statistic represents a test of all parameters of the model being equal to 0. Omitted dummies are Class, Father's
class: 5, Age discrepancy: 2-5, Industry: Others, Region: Others. k1 is normalized to 0.
2
Source : Sample of matched individuals, 1851-1881.
24
TABLE 5
MAXIMUM LIKELIHOOD RESULTS: DETERMINANTS OF JOB CLASS ATTAINMENT
Marginal effects of explanatory variables
d[Pr(y=1)] / dx
d[Pr(y=2)] / dx
d[Pr(y=3)] / dx
d[Pr(y=4)] / dx
d[Pr(y=5)] / dx
Urban
Rural
Urban
Rural
Urban
Rural
Urban
Rural
Urban
Rural
(1)
(2)
(1)
(2)
(1)
(2)
(1)
(2)
(1)
(2)
Job class
1,2 - Prof., Intermed.
0.3639
0.0921
0.3075
0.2739
-0.3423
-0.0452
-0.2437
-0.2151
-0.0855
-0.1057
3 - Skilled
0.0161
0.0081
0.0888
0.0596
0.0335
0.0515
-0.0847
-0.0636
-0.0538
-0.0555
4 - Partly Skilled
0.0120
0.0012
0.0682
0.0097
0.0273
0.0098
-0.0656
-0.0106
-0.0419
-0.0100
Student
0.0402
0.0103
0.1797
0.0734
0.0334
0.0591
-0.1594
-0.0774
-0.0939
-0.0654
Father's class
1 - Professional
0.0811
0.2007
0.2297
0.3490
-0.0668
-0.1784
-0.1720
-0.2581
-0.0720
-0.1131
2 - Intermediate
0.0206
0.0108
0.1022
0.0726
0.0237
0.0510
-0.0930
-0.0751
-0.0535
-0.0593
3 - Skilled
0.0074
0.0084
0.0453
0.0640
0.0228
0.0591
-0.0450
-0.0690
-0.0306
-0.0626
4 - Partly Skilled
-0.0072
-0.0024
-0.0480
-0.0202
-0.0316
-0.0224
0.0494
0.0225
0.0374
0.0225
0
0.0044
0.0049
0.0277
0.0396
0.0155
0.0413
-0.0279
-0.0436
-0.0197
-0.0421
1
-0.0009
0.0001
-0.0059
0.0007
-0.0034
0.0007
0.0060
-0.0008
0.0042
-0.0007
0.0055
0.0002
0.0349
0.0016
0.0197
0.0017
-0.0352
-0.0018
-0.0248
-0.0017
-0.0166
-0.0005
-0.1055
-0.0040
-0.0595
-0.0042
0.1065
0.0044
0.0750
0.0043
0.0036
0.0000
0.0229
0.0004
0.0134
0.0004
-0.0233
-0.0004
-0.0166
-0.0004
Inheritance
-0.0031
0.0202
-0.0206
0.1167
-0.0136
0.0568
0.0213
-0.1139
0.0160
-0.0798
Eldest*Inheritance
-0.0065
-0.0002
-0.0477
-0.0021
-0.0402
-0.0022
0.0509
0.0023
0.0435
0.0022
Age heaping
-0.0002
-0.0001
-0.0010
-0.0005
-0.0007
-0.0006
0.0011
0.0006
0.0008
0.0006
Living in town
-0.0028
0.0011
-0.0176
0.0088
-0.0099
0.0093
0.0177
-0.0098
0.0125
-0.0094
Agriculture
-0.0059
-0.0025
-0.0398
-0.0218
-0.0275
-0.0251
0.0413
0.0244
0.0318
0.0251
Building
0.0138
0.0092
0.0704
0.0607
0.0173
0.0400
-0.0646
-0.0624
-0.0369
-0.0475
Distributive
0.0047
0.0119
0.0276
0.0745
0.0117
0.0431
-0.0268
-0.0750
-0.0172
-0.0544
Mining
0.0065
0.0020
0.0368
0.0153
0.0142
0.0143
-0.0354
-0.0166
-0.0222
-0.0149
Age discrep. (years)
Age
Age 2 / 100
Eldest
Industry
Textiles
0.0011
-0.0035
0.0069
-0.0322
0.0037
-0.0425
-0.0069
0.0365
-0.0048
0.0417
Iron & Steel
-0.0015
-0.0003
-0.0100
-0.0028
-0.0062
-0.0030
0.0102
0.0031
0.0075
0.0031
Other manuf.
0.0110
0.0049
0.0605
0.0362
0.0209
0.0310
-0.0573
-0.0388
-0.0351
-0.0333
Region
Lancashire
-0.0005
-0.0008
-0.0032
-0.0064
-0.0018
-0.0070
0.0032
0.0071
0.0023
0.0071
London Environs
-0.0089
0.0000
-0.0658
0.0001
-0.0582
0.0001
0.0704
-0.0001
0.0625
-0.0001
Wales
0.0042
0.0060
0.0246
0.0427
0.0108
0.0324
-0.0240
-0.0449
-0.0156
-0.0363
Yorkshire
0.0122
0.0010
0.0642
0.0082
0.0188
0.0081
-0.0598
-0.0090
-0.0353
-0.0083
0.0127
0.0078
0.1456
0.1105
0.5183
0.4905
0.2436
0.2807
0.0799
0.1104
Pr[y =class | X =E(X )]
Notes: Dependent variable y is job class in 1881. Effects are calculated at mean of x for continuous variables; for discrete variables, effect is Pr(y=c|x=1) - Pr(y=c|x=0), c=1,…,5.
variables, both own and father’s, are strong influences on subsequent job quality attainment. In each case save one,
the effects are as anticipated: being in a higher class in 1851 and being the son of a father in a higher class strongly
predict ending up with a higher quality job in 1881. For example, Table 5 indicates that people who began in Class
I or II and moved to the city were 31 percentage points more likely to end up in Class II than all others who moved
to the city, and they were 36 percentage points more likely to end up in Class I. Given that the two baseline
attainment probabilities are 15 and 1 percent, respectively, it was very difficult indeed to move up into a Class I or
II job from a lower initial class. The only anomalous class is Class IV, for which the own-class effect is not
significant and the father’s class effect is negative. This has a simple explanation: by far the most common Class IV
job outside of cities was agricultural laborer. For agricultural laborers (or their sons) who moved to a city, the most
commonly attainable job was that of “general laborer,” which could be all sorts of things (very often simply
unskilled factory work) and which carried with it a class ranking of five. As a result many of the Class IV workers
who migrated to the city ended up with a lower class job. It should be noted that they did not end up with lower
paying jobs, on average. The average farm laborer in England and Wales earned ₤29.04 per year, while the average
general, non-agricultural laborer earned nearly two thirds more, ₤44.83.60 It could be argued on this ground that
general laborers should be ranked at least as high as agricultural laborers. Perhaps they should. There are other
factors to consider, however, such as difficult to measure payment-in-kind to farm laborers, arguably better job
quality of agricultural work, and perhaps occupying a higher, somewhat more dignified position on the social
ladder (though still quite low). For the present purposes, the rankings will not be modified from Armstrong’s
standard scheme.
Perhaps most interesting with regard to the class variables is to compare their effect on those who moved to
urban areas versus the effect on those who remained in rural places. A chi-square test reveals no significant
difference between the combined effect of both own and father’s class on those who moved versus those who
remained. However as Table 4 shows, for those who moved to a city, the effect of their own class in 1851 was
stronger than the effect of their father’s class. The reverse was true for those who remained behind, the job class of
their father being a stronger influence on their 1881 class attainment than was their own class in 1851. Also, the
effect of own class was stronger for those who moved than those who did not, and the effect of father’s class was
weaker, though only the former effect is statistically significant (the two χ2 statistics are different from zero with 97
and 80 percent confidence, respectively). It appears that leaving the countryside and moving to the city offered
migrants a better chance to escape the intergenerational career trajectory begun with their father’s job. For those
who moved, their own position mattered more than that of their father. It is also interesting to note that the benefit
of being a student in 1851 was nearly twice as great for those who moved to a city as it was for those who remained
60
Tuttle, “Children,” p. 173.
26
rural. Students who moved to the city were 18 percentage points more likely to end up with a Class II job than nonstudent urban migrants; for those who remained in a rural place, the advantage was only 7 percentage points.
As for other variables of interest, people who reported their age with consistency between the two censuses
performed better in the labor market. The effect was particularly strong, and statistically significant, for nonmigrants, who were 8 percentage points more likely to end up with a Class II or III job and 4 percentage points less
likely to find themselves in a Class V job than non-migrants who inconsistently reported their age by more than one
year. The benefit did not apply to those who were off by only a year; they did not do appreciably better than those
off by more than a year. Age positively influenced job quality attainment for urban migrants, though at a declining
rate—17 years of age in 1851 was optimal. Having a potential inheritance was a strongly positive force, but only
for non-migrants, as was expected. The effect was particularly positive for the attainment of Class II jobs, where it
increased the probability by 12 percentage points. The industrial classification of the individual’s job in 1851 was
also significantly influential only for persisters; starting out in the building and distributive trades (the latter being
“dealers” of one sort or another) was beneficial to eventually getting a high quality job, while beginning in the
textile industry was detrimental. For those who moved to an urban place, having an initial job in the agricultural
sector was poor preparation for eventual success and starting out in any branch of manufacturing other than steel,
iron, and textiles was helpful, though neither effect was statistically significant. Conversely, region of origin was an
important factor only for those who moved. York seemed to be a good place from which to hail, while for whatever
reason the immediate environs of London were disadvantageous.
Self-Selection and the Determinants of Migration
Who were the urban migrants? Table 3 gives an overview of their characteristics. 896 of the 3,774 individuals in
the estimation sample moved from a rural to an urban area between 1851 and 1881, for an overall urban migration
rate of 24 percent. Out of the entire sample of 28,474 matched individuals (which is younger than the population as
a whole because of the survival effect in the matching process), 18,740 (66 percent) were living in a non-urban area
in 1851. Of these, 4,387 moved to a city by 1881, for a rate of 23 percent over the thirty-year period. As for the
characteristics of these migrants, most strikingly, they were not those clinging to the bottom of the economic and
social ladder, desperate for any sort of a change. We see relatively more people with Class III jobs in 1851 and sons
of fathers with Class III jobs among the pool of urban migrants than among the non-migrants. Just the reverse is
true for Classes IV and V, which are represented more heavily among the rural persisters. Sons who were described
as students in 1851 were also more prevalent within the group of urban migrants than the group who remained in
rural places. It was decidedly not the case that the truly destitute and poverty stricken, those occupying the society’s
lowest class, were pouring from the countryside into the cities. Rather, it was those who were middle-of-the-road—
children who had received schooling, sons of fathers with skilled occupations, young men who themselves were
employed in a skilled occupation—who left the rural areas of England and Wales and headed for the cities.
27
Table 6 presents results from the probit estimation of the urban migration decision, which yields a more
(
accurate and detailed picture of the characteristics of urban migrants. 61 The large, positive coefficient on yˆ 1*i − yˆ 0*i
)
indicates that in fact people were moving to the cities in order to improve their socioeconomic status and that the
pull of this factor was strong relative to other factors. Interpreting the magnitude of the effect is not straightforward,
as y* , job quality, is not a “real” variable. Every unit increase in ant icipated job quality difference between the rural
and urban regimes increased the odds of an individual moving to a city by 7 percentage points. The estimated cut
points from Table 6 shed a little light on this figure. The range of y* between different job classes varies from 0.95
for Class IV to 1.46 for Class III. So the prospect of being one class higher in the city than in the countryside
(roughly an increase in y* of 1 to 1.5) increases the odds of migration by about 7-10 percentage points—a large
increase considering that the baseline predicted probability of moving is 22 percent. Understanding the other
determinants of migration is simpler. As expected, people who were not living in their place of birth and those
living in towns were more likely to move to a city by 1881, by 5 and 8 percentage points, respectively. Taken
together, this suggests at least the possibility of a sort of intergenerational step -wise migration, with fathers moving
from the truly rural areas to small towns, and their sons subsequently moving to the bigger cities. This is purely
speculative but not inconsistent with the data. 62 Age did not yield the usual negative effect, almost certainly because
the sample is already restricted to the young. Nor did consistent age reporting yield the expected positive
informational effect. Those working in agriculture and mining in 1851 were considerably less likely to move to a
city by 1881.
The location-level variables were also important factors in the urban migration decision. The economic
incentive to move was not entirely captured by the effect of job quality; the average wage gap was also a significant
influence. For every shilling per week of expected wage difference between the city and the countryside, an
individual was about 1 percentage point more likely to move. As anticipated, the effect of the unemployment rate
was not clear-cut. The effect of local urban unemployment by itself was significantly positive, indicating that it may
have been proxying for local unemployment. The test for this is to interact unemployment with distance to nearest
city. The coefficient is negative, indicating that the farther away was the nearest city, the more the effect of high
urban unemployment was indeed negative. Surprisingly, the number of nearby cities exerted a significant, though
small, negative influence. Those living in more agricultural counties in 1851 were slightly less likely to move to a
city, and those living in the vicinity of London in 1851 were about 6 percentage points more likely to be living in a
61
The standard errors in this table are calculated by bootstrapping. Standard analytical estimates are not appropriate due to the fact that
yˆ − yˆ 0*i is estimated rather than observed.
*
1i
62
In future work I will look into this more closely, since the father’s birthplace is observed as well.
28
TABLE 6
PROBIT RESULTS: DETERMINANTS OF URBAN MIGRATION
Move to city
Variable
coefficient
(individual-specific)
Move to city
dF/dz
-1.296***
coefficient
(location-specific)
(1)
Constant
Variable
---
Distance to city
(0.501)
y 1i * – y 0i *
0.239*
Wage gap
0.035**
Previous migrants
-0.016
-0.044
-0.013
(0.073)
Age
0.020
-0.093
0.092
0.157***
0.281***
-0.270***
0.025
0.010
Mining
-0.493***
0.005
-0.002
0.002
-0.685*
-0.204
(0.370)
0.508
0.151
(0.553)
Region
(†)
Lancashire
-0.076
0.108
0.033
(0.189)
London Environs
0.007
0.194**
0.061
(0.095)
Wales
0.003
(0.209)
-0.0004
(0.005)
0.083
(0.155)
Distributive
0.013
(0.002)
Pct. in manuf.
(0.087)
Building
-1.32E-06
(0.000)
Pct. in agriculture
(†††)
Agriculture
-0.006***
0.048
(0.056)
Industry
Nearby cities
Age heaping
(0.055)
Living in town
-0.001***
0.028
(0.163)
Not in town of birth
0.043***
Unemp*Distance
-0.028
(0.109)
Married
-4.43E-06*
(0.013)
0.006
(0.038)
Age 2 / 100
0.011
(-3.12E-06)
Unemployment
(0.065)
1
0.0003
(0.016)
Age discrep. (years)
-0.055
0.001
(0.004)
0.071
(0.127)
0
dF/dz
(1)
-0.230
-0.063
(0.184)
Yorkshire
-0.12
0.156
0.049
(0.139)
(0.170)
Textiles
0.056
0.017
Pseudo R2
0.065
-1933.311
0.025
Log likelihood
LR χ2(94)
Prob > χ2
0.000
Pr[D =1 | Z =E(Z )]
0.222
(0.107)
Iron & Steel
0.083
(0.170)
Other Manuf.
-0.040
(0.082)
270.290
-0.012
* = Significant at the 10 percent level ** = Significant at the 5 percent level *** = Significant at the 1 percent level
† = Jointly significant at the 10 percent level †† = Jointly significant at the 5 percent level ††† = Jointly significant at the 1 percent level
Notes: N = 3,774, of which 896 moved to a city by 1881 . Standard errors are in parentheses, calculated by bootstrapping via data resampling with 250 repetitions. All
dependent variables are 1851 values. Omitted dummies are Age discrepancy: 2-5, Industry: Others, Region: Others. dF/dz are marginal effects, F is Pr(D=1)
evaluated at the mean of z, except for discrete variables, where the marginal effect is Pr( D=1|z=1) - Pr(D=1|z=0).
Sources : see text.
29
city (not necessarily London) in 1881. Distance to nearest city had a positive effect, but it was very small and not
statistically significant. The “family and friends” effect of previous migrants also was not significantly different
from zero.
We have seen that the urban migrants did not begin the time period as the dregs of the labor market. We
turn now to the question of selection: whether the urban migrants and rural persisters were positively or negatively
selected. Table 7 presents estimates for the three migration effect parameters defined in (12) and (14) .63 The first
two summarize the selection process.
TABLE 7
MEASURES OF THE MIGRATION EFFECT
(Dependent variable is latent job quality in 1881, y*)
estimate
standard error
90% Confidence Interval
80% Confidence Interval
(ϕ 1 )
0.1337
0.0400
0.0676
0.1998
Selection of Urban Migrants
0.0824
0.1850
Selection of Rural Persisters
(ϕ 0 )
-0.0982
0.0308
-0.1490
-0.1377
-0.0474
-0.0587
Treatment Effect
(τ )
0.2087
0.1563
-0.0494
0.0082
0.4668
0.4092
Notes: Standard errors and confidence intervals are calculated by bootstrapping via data resampling with 250 repetitions.
Urban migrants were positively selected while rural persisters were negatively selected. Both results are statistically
significant at the 1 percent significance level. Not only did the migrants perform better in the urban labor market
than the persisters would have, they also would have outperformed the persisters in the rural labor market had they
chosen not to migrate. In this sense, then, urban migrants were the “cream of the crop.” They were those whose
labor market prospects were brightest.64
Treatment Effect
The central question of the present study concerns the effect of moving to a city on the ability of migrants to attain
high quality jobs; i.e. the treatment effect of moving to a city. The estimate of the treatment effect, along with
standard error and confidence intervals, are shown in Table 7. The treatment effect is positive. Though the estimate
of τ is not statistically significant according to the standard two-tailed test, the null hypothesis that τ1 ≤ 0 can be
63
As before, the standard errors are calculated by bootstrapping. Typical analytical standard errors are inaccurate since the X values used to
calculate the migration effect parameters are conditional means rather than observed values. This distortion is small, however, and the
reported bootstrapped standard errors are very close to the analytical ones.
64
In this aspect they appear to be different from rural-urban migrants in the 19 th century U.S. Ferrie reports that urban migrants in the U.S.
between 1850 and 1860 were negatively, rather than positively, selected. They fared worse in the cities than rural persisters would have had
they chosen to move. See Ferrie, “Down on the Farm,” p. 11.
30
rejected at the 10 percent significance level. Moving to the city allowed the average individual to obtain a better job
than he would have been able to get had he remained in a rural place.
Magnitude is best interpreted by considering job class transitions rather than effects on the latent variable
*
y . Also, it is informative to examine exactly for whom the treatment effects were largest. This information is
presented in Table 8. Here the treatment effect of moving to a city is considered in terms of the effect of a move on
occupational mobility, both own mobility and intergenerational. The table gives the probabilities that an individual
will attain either a higher or lower occupational class (respectively, “up” and “down” in the column headings),
conditional either on moving to an urban place or remaining in a rural area.65 In the first five rows, “higher” and
“lower” are interpreted in terms of the individual’s own class in 1851, and in the second five in terms of father’s
class.
Class in 1851 = 1 or 2
Class in 1851 = 3
Class in 1851 = 4
Class in 1851 = 5
Class in 1851 = student
Father's class =
Father's class =
Father's class =
Father's class =
Father's class =
TABLE 8
TREATMENT EFFECT BY CLASS
(Dependent variable is job class in 1881)
Pr(up|urban)
Pr(up|rural) Pr(down|urban) Pr(down|rural)
0
0
0.1976
0.3874
0.2131
0.1641
0.2533
0.3148
0.5188
0.4282
0.1598
0.2215
0.8374
0.8260
0
0
0.2510
0.1793
0.2153
0.2941
1
2
3
4
5
0
0.0377
0.2225
0.5326
0.8940
0
0.0341
0.1602
0.4219
0.8140
0.8376
0.7067
0.2432
0.1516
0
0.7432
0.7224
0.3206
0.2264
0
Net Gain
0.1898
0.1105
0.1523
0.0115
0.1505
-0.0944
0.0193
0.1398
0.1855
0.0800
Net gain (of moving to a city) is defined as
[Pr(up | urban) - Pr(up | rural)] + [Pr(down | rural) - Pr(down | urban)]
(17)
For example, the average individual who held a Class IV job in 1851 had a 49.12 percent chance of moving up to a
Class I, II, or III job if they moved to a city and a 41.18 percent chance if they chose to remain in a rural place. That
same person would have a 17.73 percent chance of falling to a Class V job in the city and a 23.42 percent chance of
falling in the countryside. So moving to the city confers a 7.94 percentage point boost in the probability of making
65
For the purpose of this exercise, individuals’ classes as students in 1851 were considered to have moved up if they obtained a Class I or II
job by 1881, and they were considered to have moved down if they held Class IV or V jobs in 1881.
31
an upward move and offers a 5.69 percentage point lower likelihood of falling to a lower class, for a net gain of
13.63 percentage points.
The first important thing to note from Table 8 is that the gains from urban migration were realized across
all socioeconomic strata. The only measure of net gain that is not positive is that of intergenerational mobility for
sons of fathers with Class I jobs (less than 2 percent of the sample); they were less likely to experience downward
mobility if they remained in a rural area. All individuals were more likely to improve their own socioeconomic
status if they migrated to a city, and sons of fathers belonging to any class other than I were more likely to end up in
a higher strata by 1881 than their fathers occupied in 1851 if they moved. The second feature to note from the table
is those groups for whom the treatment effect of migration was the largest. The highest gains in occupational
mobility are realized by those who hold Class I or II jobs in 1851; they are much less likely to fall to a lower class if
they move to a city than if they do not. As Table 2 shows, however, this group represents a small minority (about 2
percent) of the sample. By far the largest groups are those who begin with a Class III or IV job or as a student. The
net gain of moving to a city was large for all these individuals. For the average student, Class IV worker, and Class
III worker, the net gain of moving to a city was 15.05, 13.23, and 11.05 percentage points percent, respectively.66
About 70% of the sample had fathers with Class III or IV jobs. These men realized net gains to urban migration, in
terms of intergenerational occupational mobility, of 13.98 and 18.55 percentage points, respectively.
Consider, then, the average person who held a Class IV, partly skilled job in 1851, lived in the countryside,
and subsequently decided to leave and move to a city. Such an individual was 21 percent more likely to improve the
quality of his occupation than if he had remained in a rural area (52 versus 43 percent), and he was 39 percent less
likely to fall into a Class V, unskilled occupation (16 versus 22 percent). The average rural son of a Class IV
worker was 26 percent more likely to attain a higher status than his father if he moved to an urban place than if he
remained in the countryside, and he was 49 percent less likely to find himself with a Class V occupation. These are
substantial gains. For people like this, moving to the city was an important way to improve their standing in the
labor market.
Conclusions
Large-scale rural-urban migration was one of the most prevalent demographic features of 19th century Britain.
Cities grew more than twice as fast as the nation as a whole during this time, and in the process Britain became the
world’s first predominantly urban society. Of the young men moving from the countryside to the cities, the skilled
and educated were the most prevalent. They were neither the poorest nor the richest rural residents. In one
66
As a group, individuals with Class 4 jobs in 1851 realized strong net gains from moving to urban areas even though many of them made
the downward move from agricultural laborer to general laborer. But as has been noted, these people actually received 65 percent greater
annual earnings, on average. If agricultural and general laborers were put into the same class, or if general laborers were ranked above
agricultural, then the treatment effects would be even stronger.
32
important sense, they were the best that the pool of rural labor had to offer. Urban migrants were positively
selected; they were those whose labor market prospects were brightest, whether in the cities or in rural areas. In this
aspect, the nature of rural-urban migration in Victorian Britain is very different from what it was during much of
the 20th century, when the urban migrants typically were those with the poorest prospects. In 19th century England
and Wales, these migrants left for the cities for a variety of reasons, but chief among them was a desire to improve
their economic condition. Expected increases in socioeconomic status over what could be obtained in rural England
and Wales and higher average wages in urban than in rural areas drove people from the countryside and small
towns into the cities. People left to find better jobs, and they found them. Urban migrants were more likely to
improve their socioeconomic status (occupational class) than they would have been had they remained in a rural
area, and they were more likely to experience upward intergenerational occupational mobility. The effect was
strong for the average rural resident of Great Britain, and it was even stronger for the sort of person who was in fact
inclined to move. Individuals from all socioeconomic strata realized these positive returns to migration. In this
aspect as well then, the experience of the urban migrant in Victorian Britain was very different from that of the
migrant in the modern developing world, who is quite likely to experience negative returns to urban migration. It is
beyond the scope of the current study to attempt to explain these differences, though it is possible that British
industry and the British economy as a whole was in many ways more developed in the mid to late 19th century than
were those of many Third World countries in the late 20th century, and that this difference is germane in some way
to the effect of urban migration. Perhaps even more importantly, the institutional structure of 19th century Britain
surely was very different from that of the modern Third World. Victorian Britain represented the height of laissez
faire political economy. The same certainly cannot be said of most of today’s Third World, particularly in Africa
and Latin America. The free market economy of Britain from 1851 to 1881 may have provided more stable urban
employment opportunities and more potential for advancement across socioeconomic strata than is true in the
modern developing world, where a small minority of migrants experience tremendous gains and the large majority
suffer loss.
References
Amemiya, T. Advanced Econometrics. Oxford: Basil Blackwell, 1986.
Anderson, Michael. National Sample from the 1851 Census of Great Britain: Introductory User Guide. University
of Edinburgh, Dept. Of Economic and Social History, September 1987.
Armstrong, W. A. “The Use of Information about Occupation,” in E. A. Wrigley, ed., Nineteenth-century Society.
Cambridge: Cambridge University Press, 1972.
Baines, D. Migration In A Mature Economy: Emigration and Internal Migration in England and Wales, 1861-1900.
Cambridge: Cambridge University Press, 1985.
_______ “Population, Migration and Regional Development, 1870-1939” in R. Floud and D. McCloskey, The
Economic History of Britain Since 1700, vol 2. Cambridge: Cambridge University Press, 1994.
Bédarida, F. A Social History of England, 1851-1975. London, 1979.
33
Black, D., and V. Henderson. “A Theory of Urban Growth.” Journal of Political Economy 107, no. 2 (1999): 25284.
Booth, C. (ed.) Life and Labour of the People in London, 9 vols. London, 1892-97.
Boswell, J. Life Of Samuel Johnson. New Haven: Yale University Press, 1994.
Boyer, G. R., and T. J. Hatton. “Migration and Labour Market Integration in Late Nineteenth-Century England and
Wales.” Economic History Review 50, no. 4 (1997): 697-734.
Crouzet, F. The Victorian Economy. New York: Columbia University Press, 1982.
de Vries, J. and A. van der Woude. The First Modern Economy : Success, Failure, and Perseverance of the Dutch
Economy, 1500-1815. New York : Cambridge University Press, 1997.
Encyclopædia Britannica Online, “Population.” [http://www.eb.com:180/bol/topic?eu=115183&sctn=23]
[Accessed 3 Sep 2001].
Ferrie, J. P. “A New Sample of Americans Linked from the 1850 Public Use Sample of the Federal Census of
Population to the 1860 Federal Census Manuscripts.” Historical Methods 29 (Summer 1996).
_______. “Migration To The Frontier In Mid-Nineteenth Century America: A Re-Examination Of Turner’s ‘Safety
Valve.’”
prepared
for
the
World
Congress
of
Cliometrics,
July
1997.
[http://www.faculty.econ.nwu.edu/faculty/ferrie/papers/munich.pdf]
_______. “How Ya Gonna Keep ‘Em Down on the Farm [When They've Seen Schenectady]? Rural to Urban
Migration in the U.S., 1850-70,” prepared for the Cliometrics Conference, May 1999.
[http://www.faculty.econ.nwu.edu/faculty/ferrie/papers/urban.pdf]
_______. Yankeys Now: Immigrants in the Antebellum United States, 1840-1860. New York: Oxford University
Press, 1999.
Floud, R. and D. McCloskey. The Economic History of Britain Since 1700, 2nd ed., 3 vols. Cambridge: Cambridge
University Press, 1994.
Friedlander, D. “Occupational Structure, Wages, and Migration in Late Nineteenth-Century England and Wales.”
Economic Development and Cultural Change 40 (1992): 295-318.
Friedlander, D., and R. J. Roshier. “A Study of Internal Migration in England and Wales: Part I.” Population
Studies 19 (1966): 239-79.
Great Britain, General Register Office. The Registrar-General's Decennial Supplement. England and Wales.
London: H.M.S.O., 1921.
_______. Census of England and Wales, 1921. Classification of Occupations. London: H.M.S.O., 1924.
_______. Census of England and Wales, 1951. Classification of Occupations. London: H.M.S.O., 1956.
Harris, J. R., and M. P. Todaro. “Migration, Unemployment, and Development: A Two-Sector Analysis.” The
American Economic Review 60, no. 1 (1970): 126-142.
Hatton, T. J., and J. G. Williamson. “What Drove the Mass Migrations from Europe in the Late Nineteenth
Century?” Population and Development Review 20 (September 1994): 533-59.
Heckman, J. “Varieties of Selection Bias.” The American Economic Review 80, no. 2 (1990): 313-18.
Higgs, E. A Clearer Sense of the Census. London: HMSO, 1986.
Hunt, E. H. Regional Wage Variations in Britain, 1850-1914. Oxford: Oxford University Press, 1973.
_______. “Industrialization and Regional Inequality: Wages in Britain, 1760-1914.” Journal of Economic History
46, no. 4 (1986): 935-966.
Johnston, J., and J. DiNardo. Econometric Methods, 4th ed. New York: McGraw-Hill.
Krugman, P. “The Current Case for Industrial Policy,” in Salvatore, D., ed., Protectionism and World Welfare.
Cambridge: Cambridge University Press, 1993.
Lee, C.H. British Regional Employment Statistics, 1841-1971. Cambridge: Cambridge University Press, 1971.
Lee, L. F. “Unionism and Wage Rates: A Simultaneous Equation Model with Qualitative and Limited Dependent
Variables.” International Economic Review 19: 415-33.
Lucas, R. E., Jr. ‘‘On the Mechanics of Economic Development.’’ Journal of Monetary Economics 22 (July
1988): 3–42.
Maddala, G. S. Limited-Dependent and Qualitative Variables in Econometrics. Cambridge: Cambridge University
Press, 1983.
34
Manski, C. F. “Nonparametric Bounds on Treatment Effects.” The American Economic Review 80, no. 2 (1990):
319-23.
Mayhew, K., and B. Rosewell. “Occupational Mobility in Britain.” Oxford Bulletin of Economics and Statistics 43
(1980): 225-255.
Mokyr, J. Why Ireland Starved. London: George Allen & Unwin, 1983.
Newey, W. K., J. L. Powell, and J. R. Walker. “Semiparametric Estimation of Selection Models”Some Empirical
Results.” The American Economic Review 80, no. 2 (1990): 324-28.
Pooley, C. and J. Turnbull. Migration and Mobility in Britain Since the 18th Century. London: UCL Press, 1998.
Ravenstein, E. G. “The Laws of Migration.” Journal of the Royal Statistical Society 48 (1885): 167-227.
Redford, A. Labour Migration in England, 1800-1850, 2nd ed. Manchester: Manchester University Press, 1964
(1926).
Robinson, C., and N. Tomes. “Self-Selection and Interprovincial Migration in Canada.” Canadian Journal of
Economics 15, no. 3: 474-502.
Royle, S. A. “Social Stratification from Early Census Returns: A New Approach.” AREA 9 (1977): 215-219.
Shryock, H., and J. Siegel. The Methods and Materials of Demography, 2 vols. Washington: U.S. Dept. of
Commerce, Bureau of the Census, 1973.
Southall, H. “Regional Unemployment Patterns Among Skilled Engineers in Britain, 1851-1914.” Journal of
Historical Geography 12, no. 3 (1986): 268-286.
Southall, H., and D. M. Gilbert. “A Good Time to Wed?: Marriage and Economic Distress in England and Wales,
1839 to 1914.” Economic History Review 49 (1996): 35-57.
Todaro, M. P. Economic Development in the Third World. New York : Longman, 1989.
Trost, R. P. “ Demand for Housing: A Model Based on Inter-Related Choices Between Owning and Renting.”
Unpublished Ph.D. dissertation, University of Florida.
Tunali, I. “Rationality of Migration.” International Economic Review 41, no. 4 (2000): 893-920.
Tuttle, Carolyn. “The Role of Children in the Industrial Revolution.” Unpublished Ph.D. diss., Northwestern
University, 1985.
United Nations Centre for Human Settlements (Habitat), “General Assembly, Special Session for an Overall
Review and Appraisal of the Implementation of the Habitat Agenda.” New York, June 2001.
[http://www.unchs.org/Istanbul+5/bg10.htm]
Williamson, J. G. Coping with City Growth During the British Industrial Revolution. Cambridge: Cambridge
University Press, 1990.
35