<<

. 8
( 14)



>>

Percent arable “0.4610* 0.2442 “0.4596 0.4565
(0.2615) (0.3152) (0.2991) (0.3655)
Milk cows/rent 0.0510** “0.0267** 0.0554** “0.0381**
(0.0100) (0.0121) (0.0115) (0.0146)
Other cattle/rent 0.0017 0.0046 0.0038 0.0097
(0.0055) (0.0066) (0.0062) (0.0077)
Horses/rent 0.0935** 0.1412** 0.0983** 0.1500**
(0.0201) (0.0242) (0.0229) (0.0282)
Sheep/rent “0.0147 “0.0368 0.0246 “0.0419
(in hundreds) (0.0405) (0.0488) (0.0456) (0.0565)
Women™s wage “0.2173** 0.1793 “0.2884** 0.1984
(0.0905) (0.1091) (0.1046) (0.1262)
Men™s wage 0.0934** “0.0199 0.1061** “0.0132
(0.0411) (0.0496) (0.0468) (0.0571)
N 220 220 220 220
R2 0.367 0.249
Log-likelihood “285.46 “317.96

Standard errors in parentheses.
* ¼ signi¬cantly different from zero at the 10% level
** ¼ signi¬cantly different from zero at the 5% level
Normalized coef¬cients are presented for the Tobit estimations.
Source: Young, Northern Tour and Eastern Tour.


similar results, and the Tobit estimations are similar to the OLS esti-
mations, indicating that the results are robust.
Employment patterns followed the predictions of comparative
advantage. The number of cows on the farm increases the number of
women hired, but not the number of men. The percent arable land
increases the number of men hired, but not the number of women.
Horses increase both types of labor, perhaps because, other things being
equal, more horses meant more work to be done. However, the presence
of horses does have a greater effect on male employment than female
employment, which would be expected if horses were associated with
arable agriculture. Other cattle and sheep, for which we had no a priori
expectations, have little or no effect on the employment of either men or
women. These results increase our con¬dence in the estimation results.
196 Gender, Work and Wages in Industrial Revolution Britain

The effect of a higher wage for either men or women should be to
reduce the employment of that type of labor because it is more expen-
sive. This prediction holds up well for women; the women™s wage always
has a signi¬cantly negative effect on women™s employment. Men™s
employment, however, does not seem to respond to the men™s wage.
While I will be able to test whether the wage is exogenous in the
women™s equations, I do not have an instrument for the men™s wage.
We are left with the unsatisfactory own-wage effect in the men™s
equations.
The men™s wage coef¬cients in the women™s equations indicate that
an increase in the male wage was associated with an increase in female
employment. The cross-price effect is always signi¬cantly positive.
When I control for the type of farming activity by including variables
such as the number of farm animals and “percent arable,” the number of
women hired still responded to the price of hiring men. Even after
deciding how many cows to keep and how much arable land to work, the
farmer™s decision about how many women to hire still responded to the
price of men, suggesting he or she was willing to substitute men and
women within tasks rather than, for example, always assigning dairy
work to women. To assess the size of the estimated effects, Table 4.5
presents elasticities at the mean. For women™s employment, the cross-
wage effect has a larger elasticity than the other explanatory variables,
suggesting that the substitution effect was relatively strong. The employ-
ment of women was more responsive to changes in male wages than it was
to changes in the number of cows on the farm. Farm employment was not
simply determined by gender roles, but responded to wages.
While the cross-price effect in the women™s employment equations is
signi¬cantly positive, the cross-price effect in the men™s equations is
not signi¬cant in the ¬rst speci¬cation and only marginally signi¬cant
in the second speci¬cation. In Table 4.5 we see that the size of the
response of male employment to the women™s wage is smaller than the
response of male employment to the number of horses on the farm.
However, these relatively low elasticities are consistent with the higher
elasticities for women™s employment. As noted in Appendix 4.1 (see
p. 344), the input with the larger factor share will have a lower cross-
price elasticity. The average farm spent £16 per year on men servants,
but only £5 per year on women servants. This is enough of a difference
to explain the difference in elasticities.20 Thus the smaller cross-price

20
These factor shares imply that the ratio of the cross-elasticities should be 0.31. The
ratio of the estimated elasticities in Table 4.5 is actually larger than this, ranging from
0.43 to 0.76.
Testing for occupational barriers in agriculture 197
Table 4.5. Elasticities

Elasticity at the mean

Dependent variable Independent variable OLS Tobit

Speci¬cation one
Women Percent arable “0.022 “0.029
Cows 0.214* 0.230*
Horses 0.329* 0.340*
Women™s wage “0.619* “0.737*
Men™s wage 0.623* 0.710*
Men Percent arable 0.128 0.266
Cows 0.063 0.105
Horses 0.825* 0.863*
Women™s wage 0.268 0.324
Men™s wage 0.167 0.371
Speci¬cation two
Women/rent Percent arable “0.179 “0.178
Cows/rent 0.389* 0.423*
Horses/rent 0.446* 0.469*
Women™s wage “0.587* “0.779*
Men™s wage 0.617* 0.701*
Men/rent Percent arable “0.019 0.171
Cows/rent “0.197* “0.281*
Horses/rent 0.652* 0.693*
Women™s wage 0.469 0.519
Men™s wage “0.127 “0.084

Elasticities for Tobit estimations are elasticities of the index.
* ¼ coef¬cient is signi¬cant at the 5% level
Source: Young, Northern Tour and Eastern Tour.

elasticities in the men™s equations do not necessarily contradict the
results from the women™s equations. Since men had a larger factor
share, their cross-price elasticity should be lower. In this case, the
cross-price effect is low enough that it does not show up as statistically
signi¬cant.
My estimates of cross-wage elasticities are valid only if the wages are
exogenous, and not in¬‚uenced by the farmer™s employment decisions.
The employment measures in this data set are for individual farms which
are likely to be price-takers. The wages that Young reports are the wages
prevailing in the local labor market, which were in¬‚uenced by demand
conditions, but if each farm was small relative to the market, the
employment decisions of an individual farmer would not be enough to
determine the wage rate. Fortunately, I do not have to rely on this
explanation alone, but can test whether the wages in my data set are
198 Gender, Work and Wages in Industrial Revolution Britain

exogenous. Using wages in alternative employments such as spinning as
an instrument for female agricultural wages, I ¬nd that a Hausman
speci¬cation test fails to reject the null hypothesis that female wages
were exogenous to the farmer™s employment decisions.
If wages were not exogenous, the supply and demand equations would
be a simultaneous system:

Nf ¼ a1 þ a2wm þ a3wf þ X 0 b þ e
Demand:
Nf ¼ b1 þ b2wf þ b3wA þ u
Supply:

where Nf is female employment, wm is the male wage, wf is the female
wage, X is the matrix of controls, and wA is the wage in a woman™s
alternative employment. If I were measuring the employment of married
women, then the labor supply equation would include the male wage,
since a married woman™s labor supply would be affected by her hus-
band™s wage. In this case, however, the dependent variable is female
servants, who were overwhelmingly single, and thus the male wage does
not affect women™s labor supply.
If this simultaneous system is the correct speci¬cation, then I need to
use instrumental variables to estimate the demand equation. Young
provides an instrument for women™s wages by including information on
alternative wages in such statements as, “The employment of the poor
women is spinning of ¬‚ax: a woman can earn from 3d. to 6d. a day.”21
These statements mainly described the work of married women, but
spinning was also the alternative to farm service for the single woman
and thus affected her labor supply decision. In almost all cases the
alternative wage that Young supplied was for spinning. In one case it
was for mining, and in one case it was for lace-making. I also include
towns where Young speci¬cally noted there was no non-agricultural
employment for women, assigning a wage of zero to the alternative.22
Since the wage is only given in a few cases, I also use a dummy variable
for the presence of alternative work. “Spinning” is a dummy indicating
the presence of spinning work, the wage for which averaged 5d. a day.
The “high wage” dummy indicates the presence of either mining or lace-
making, which each paid about 1s. a day,23 or more than twice as much

21
Young, Northern Tour, vol. I, p. 326. Where a range of wages is given I take the mid-
point, so this wage would be recorded as 4.5 pence a day.
22
The absence of employment was indicated by such statements as, “The poor women
and children in total idleness.” Ibid., vol. II, p. 174.
23
Young gives one wage quote for each of these occupations. Mining paid 1s. a day in
Reeth, North Riding, and lace-making paid 11d. a day in Maidenhead, Berkshire. Ibid.,
vol. I, p. 357, and vol. III, p. 9.
Testing for occupational barriers in agriculture 199

as a woman could earn spinning. The omitted category is towns spe-
ci¬cally noted to have no non-agricultural work for women. Using these
dummies increases the sample size, and still provides an instrument
correlated with the agricultural wage.
The wage in alternative work is a valid instrument for women™s wages;
it enters the supply equation but not the demand equation. Work
opportunities in other industries were an important determinant of the
labor supply curve facing the farmer, but should not affect his or her
demand for labor. Using these instruments, I can calculate the Two-
Stage Least Squares (2SLS) estimates and use a Hausman speci¬cation
test to determine whether wages were exogenous.24 Under the null
hypotheses of exogenous wages, the 2SLS estimates are consistent but
inef¬cient, while under the alternative that wages are endogenous, OLS
estimates are inconsistent. Thus, I will use the Hausman test to compare
OLS and 2SLS estimates. Table 4.6 presents two sets of 2SLS esti-
mates, one using the alternative wage, and one using the dummy vari-
ables. The own-price coef¬cients are no longer signi¬cantly negative,
and in the dummy variable speci¬cation the cross-price effect is not
statistically signi¬cant, but otherwise the results are similar to the OLS
and Tobit results. The Hausman tests do not reject the null, so we
cannot reject the hypothesis that the wage is exogenous.25 To test the
quality of the instruments I am using, I also ran 2SLS estimations
including the residuals from the ¬rst stage. The residuals are not stat-
istically signi¬cant in either case, and including the residuals did not
change the conclusion that female employment increased when male
wages increased.
The results are robust; men and women farm servants were substi-
tutes. The employment of women clearly responded to the wages of
men. Men™s employment responded less strongly to women™s wages, but
this is consistent with the relative factor shares of men and women
workers. Employers were willing to hire more women if men™s wages
increased, indicating that they made hiring decisions based on pro¬t
maximization rather than strictly according to gender roles. In 1770 the
market for farm servants was not segregated by gender; men and women
were not hired in completely isolated labor markets. Bergmann™s occu-
pational segregation model is not a good description of employment in
the agricultural labor market of 1770.

24
J. A. Hausman, “Speci¬cation Test in Econometrics,” Econometrica 46 (1978),
pp. 1251“71.
25
The possibility remains, however, that large standard errors may cause us to accept the
null when we should not.
200 Gender, Work and Wages in Industrial Revolution Britain
Table 4.6. Two-stage least squares estimates and speci¬cation test

1st stage 2nd stage 1st stage 2nd stage
women™s wage women women™s wage women

Constant 1.6258** “0.0419 1.7770** 1.1054
(0.3161) (0.8945) (0.3163) (0.7276)
Acres “0.0320* 0.0221 “0.0432* 0.0089
(in hundreds) (0.0172) (0.0363) (0.0220) (0.0315)
Acres squared 0.0013** “0.0026** 0.00077* “0.0017**
(in 10,000s) (0.0003) (0.0008) (0.00040) (0.0006)
Percent arable 0.1121 “0.0932 “0.1223 “0.0541
(0.2037) (0.3756) (0.1949) (0.2699)
Milk cows 0.0156** 0.0482** 0.0068 0.0147**
(0.0063) (0.0119) (0.0055) (0.0074)
Other cattle 0.0021 0.0002 0.0012 0.0074**
(0.0027) (0.0051) (0.0026) (0.0034)
Horses “0.0194* 0.0890** “0.0103 0.0539**
(0.0109) (0.0204) (0.0108) (0.0146)
Sheep “0.0306** 0.0358 “0.0013 0.0295
(in hundreds) (0.0128) (0.0253) (0.0173) (0.0236)
Men™s wage 0.1874** 0.1309* 0.1931** 0.1683**
(0.0264) (0.0695) (0.0296) (0.0536)
Predicted wage “0.2345 “0.5288**
(0.2887) (0.2439)
Alternative wage 0.0892**
(0.0136)
Spinning 0.3021**
Dummy (0.1196)
High wage 1.1213**
Dummy (0.2022)
N 81 81 184 184
R2 0.755 0.718 0.362 0.471
Speci¬cation test 0.004 1.491
Critical value for 5% test 16.92 16.92

Standard errors in parentheses.
* ¼ signi¬cantly different from zero at the 10% level
** ¼ signi¬cantly different from zero at the 5% level
Source: Young, Northern Tour and Eastern Tour.



II. Wage correlations
While the previous section used data from Arthur Young™s tours to test
whether men and women were substitutable in agricultural work in
1768“70, this section will use wage correlations to test for substitut-
ability over a broader range of years. If women did not face occupational
barriers, then the markets for male and female labor should be
Testing for occupational barriers in agriculture 201

integrated, and their wages should be correlated. This test is analogous
to tests of geographical integration of markets which examine price
correlations. Hatton and Williamson use wage correlations to test for
integration between rural and urban markets by asking whether rural
wages responded to urban wages.26 They examine the gap between
urban and rural wages in the later nineteenth century and ¬nd that the
urban and rural labor markets were not segregated; rural wages responded
to changes in urban wages. Appendix 4.3 presents a mathematical model
demonstrating how the correlation in wages indicates an integrated
market; here I apply this test to ask whether men and women were
hired in segregated markets.
The correlation between male and female wages has also been
explored by E. H. Hunt, though for a different purpose. Hunt was trying
to explain the persistence of regional wage gaps and hypothesized that a
negative correlation between women™s wages and men™s wages would
equalize family income and discourage the migration necessary to
equalize male wages across regions. In his own words, “if women™s and
adolescents™ work was unobtainable or very badly paid in areas where
men™s wages were high then migration was less likely.”27 He ¬nds that
the evidence does not support this conclusion. Although he does not
speci¬cally calculate a correlation, he concludes, “regional variations in
the wages of women and young people failed to compensate for regional
variations in men™s wages. On the contrary, there was a generally positive
correlation between the two variables.”28 He does not, however, draw
the conclusion that the labor market was integrated by gender. On the
contrary, he claims, “the differentials between the earnings of men,
women, and young people were strongly in¬‚uenced by custom.
Women™s wages were determined, in large part, by consideration of what
most people believed they ought to earn and this was usually measured
as a customary proportion of the male ratio.”29 I will use wage correl-
ations to argue that, in fact, wages were set by the market, rather than by
custom.
I will use wage correlations to test for occupational segregation
constraints. Appendix 4.3 presents a mathematical model of this test.
The model predicts observable differences that will allow us to distin-
guish between integrated and segregated markets. If the market for male
and female labor is integrated, male and female wages will be correlated.
If there are occupational constraints, wages will not be correlated

26
Timothy Hatton and Jeffrey Williamson, “Integrated and Segmented Labor Markets:
Thinking in Two Sectors,” Journal of Economic History 51 (1991), pp. 413“25.
27
Hunt, Regional Wage Variations, p. 106. 28 Ibid., p. 117. 29 Ibid., pp. 117“18.
202 Gender, Work and Wages in Industrial Revolution Britain

because changes in the market for male labor would not translate into
changes in the market for female labor and vice versa.
In order for this test to work, there must be some variation in wages
that is not due to locational factors affecting all wages, such as the cost of
living. Evidence on wage gaps suggests that local labor markets were
indeed independent, and not part of a uni¬ed national labor market.
Both Hunt and Williamson found large wage gaps that persist even after
correcting wages for prices and compensating differentials.30 Looking at
the same data that I will use below, Cunningham noted that “What is
striking is how localized these employment markets were.”31 While there
was some labor mobility, there was not enough to equalize wages.
Persistent wage gaps are not surprising in light of the discouragements to
mobility. The Poor Laws played an important part in reducing labor
mobility, and thus preventing the equalization of wages. The poor could
only receive relief in the parish where they had a settlement, so the poor
were generally tied to the parish where they were born. New settlements
became dif¬cult to get because rate-payers did everything they could to
keep their poor law payments low. Before 1795, individuals could be
removed (sent back to their parish of settlement) if there was any sus-
picion that they might collect poor relief sometime in the future. Adam
Smith painted the effect of the settlement laws in extreme terms:
The very unequal price of labor which we frequently ¬nd in England in places at
no great distance from one another, is probably owing to the obstruction which
the law of settlements gives to a poor man who would carry his industry from one
parish to another . . . it is often more dif¬cult for a poor man to pass the arti¬cial
boundary of a parish, than an arm of the sea or a ridge of high mountains.32

Thus I feel con¬dent that there is some variation in real wages not due to
compensating factors.
Another problem, and one I cannot correct for, is that even if men and
women were always paid according to their productivity ratio, the cor-
relation would not be perfect if the productivity ratio varied from place
to place. Since different regions specialized in different industries and
different types of agriculture, they should have different productivity
ratios, and thus different wage ratios. If the market was truly integrated,
a low correlation might still result, owing to differences in the prod-
uctivity ratio. Wages in each parish would re¬‚ect the true productivity
ratio, but differences among parishes would lead to a less than perfect

30
Ibid., and Williamson, “Did English Factor Markets Fail.”
31
Hugh Cunningham, “The Employment and Unemployment of Children in England,”
Past and Present 126 (1990), p. 136.
32
Adam Smith, The Wealth of Nations (New York: Modern Library, [1776] 1965), p. 140.
Testing for occupational barriers in agriculture 203

correlation. For example, if the productivity ratio was one-half in parish
A and one-third in parish B, and male wages were 10 in parish A and 12
in parish B, female wages could be 5 in parish A and 4 in parish B under
perfect competition, which would give a negative correlation between
male and female wages. Thus, the correlation of wages might be low
even in an integrated market if there is variation in the productivity ratio.
This problem would lend a downward bias to the wage correlation, so if
we do ¬nd a strong positive correlation, we can still be con¬dent that the
markets for male and female labor were integrated.
The model presented in Appendix 4.3 implies an observable differ-
ence in the behavior of wages under the two different hypotheses and
will be the basis of the test which uses the correlation of wages. If the
labor market was competitive, male and female wages would be posi-
tively correlated. However, if there were occupational constraints, male
and female wages would not be correlated. Thus I will examine the
correlation of male and female wages to test for discriminatory con-
straints. Because this test is so simple, I am able to look at many different
data sets and track wage correlations through time. Most of the wages in
this section are for agricultural work, though the 1833 data also includes
other female occupations. This section shows that, for the agricultural
labor market, there is no indication that occupational segregation con-
straints appeared during the course of the Industrial Revolution.

A. 1768“70
In the previous section, I used data on employment and wages of farm
servants provided by Arthur Young. In his books, Young also provided
wages for day-laborers. Data on day wages for agricultural laborers are
provided for about seventy-¬ve towns. Fortunately, Young provides
multiple wages for each location, so that I can correct for the locational
¬xed effect. Unfortunately, for many locations no female winter wage is
reported (because women were less likely to be employed during the
winter), so that correlations including winter wages will have a smaller
sample size.
Table 4.7 presents the means of wages for men and women by season.
The wage ratios are about one half, which matches estimates of the wage
ratio presented in Chapter 2. The female“male wage ratio is 0.49 in
winter, 0.47 in hay harvest, and 0.56 in harvest. The fact that the ratio is
highest in harvest and lowest in hay harvest ¬ts with what we know
about the technologies used. Women were relatively effective with the
sickle, which was used to harvest grain, but were not as effective with
the scythe, which was used to cut the grasses for hay-making. Thus, the
204 Gender, Work and Wages in Industrial Revolution Britain

Table 4.7. Means of wages: 1770 (shillings per week)

Mean SD Min. Max. N

Men Winter 6.56 1.22 4.50 11.00 75
Hay 9.45 2.17 4.00 14.00 76
Harvest 11.27 2.85 6.00 21.00 76
Women Winter 3.20 0.93 2.00 5.50 47
Hay 4.47 1.53 2.00 11.50 77
Harvest 6.30 1.59 3.00 14.50 74

Source: Young, Northern Tour and Eastern Tour.


Table 4.8. Correlations of men™s and women™s wages: 1770

Correlation P-level N

Winter 0.44** 0.002 46
Hay 0.19* 0.096 75
Harvest 0.38** 0.001 74
Harvest “ hay 0.20* 0.086 73
Harvest “ winter 0.42** 0.004 46
Hay “ winter “0.01 0.947 45

* ¼ signi¬cantly different from zero at the 10% level
** ¼ signi¬cantly different from zero at the 5% level
Source: Young, Northern Tour and Eastern Tour.


wage ratio was highest when the productivity ratio was highest, and
lowest when the productivity ratio was lowest.
Table 4.8 presents the wage correlations. Correlations of male and
female wages in winter and harvest are signi¬cantly different from zero
at the 5 percent level, but the correlation of male and female hay-
making wages is lower and is only signi¬cantly different from zero at
the 10 percent level. This difference may re¬‚ect the fact that, because
the scythe required so much strength that it was never used by women,
hay-making offered fewer opportunities for substituting one sex for the
other.33 Because locational ¬xed effects may induce a spurious correl-
ation, I also present correlations of the difference in wages across sea-
sons. Two of these correlations are positive. The differences between
hay and winter wages is the exception, showing no correlation. Table 4.9
presents log regressions, which indicate the elasticity of the response.

33
Women never used the scythe because it required a great deal of strength. See Roberts,
“Sickles and Scythes.”
Testing for occupational barriers in agriculture 205
Table 4.9. Log-log regressions: 1770 (dependent variable ¼ women™s wage)

Winter Hay Harvest

Constant “0.290 0.949** 1.083**
(0.416) (0.323) (0.255)
Men™s wage 0.750** 0.227 0.307**
(0.221) (0.145) (0.107)
R2 0.207 0.032 0.103
N 46 76 74

Wages in logs. Standard errors in parentheses.
* ¼ signi¬cantly different from zero at the 10% level
** ¼ signi¬cantly different from zero at the 5% level
Source: Young, Northern Tour and Eastern Tour.


For winter wages, the male-wage elasticity of female wages is 0.75,
indicating a fairly strong response. For the other seasons, however, the
elasticity of the response is lower. Again, the relationship is not signi¬-
cant during hay-making, when the strength required to use a scythe
created a more rigid gender division of labor.
We saw in the previous section that English farmers in Young™s tours
were willing to substitute men and women farm servants. The wage
correlations just presented suggest that farmers also substituted male
and female day-laborers, though less so during hay-making. As we shall
see, wages from later dates show a positive correlation at least as large as
these, suggesting that the substitutability found in the 1770 agricultural
labor market persisted through the Industrial Revolution.

B. 1833
Wage data from the ¬rst half of the nineteenth century is provided by an
1833 survey conducted by the Poor Law Commissioners, just previous
to the passage of the New Poor Law. The survey contained a wide
variety of questions, and fortunately included questions on both male
and female wages.34 The survey was ¬lled out by local poor law

34
The questions on wages were: “8. Weekly wages, with and without Beer or Cyder, in
Summer and Winter?” and “12. What can Women and Children under 16, earn per
Week, in Summer, in Winter, and Harvest, and how employed?” Some wage information
was also obtained from question 13: “What in the while might a laborer™s Wife and Four
Children, aged 14, 11, 8 and 5 Years respectively (the eldest a boy), expect to earn in the
Year, obtaining, as in the former case, an average amount of Employment?” BPP 1834
(44) XXX. For a summary of what this survey tells us about the employment of women
and children, see Nicola Verdon, “The Rural Labour Market in the Early Nineteenth
206 Gender, Work and Wages in Industrial Revolution Britain

administrators at the parish level, so each observation corresponds to a
parish in England or Wales. The advantages of this data source are its
large size and broad coverage. The data cover all of England and Wales,
and there are 809 observations with summer wages for both men and
women. The data also have some disadvantages. The wages are sub-
jective reports by a parish of¬cial. They do not represent actual wages
paid to any individual, and the survey respondent may not have been as
conscientious an observer as Arthur Young. The data are from rural
areas, and most of the wages are for agricultural work. In spite of these
¬‚aws, the survey is valuable because it provides scarce data on wage rates
throughout England and Wales.
In constructing this data set, I added the value of in-kind payments in
the form of beer and food, which were sometimes given to workers, to
the cash wages. Sometimes the survey respondent would report wages
with and without beer; the difference is the correct valuation of this in-
kind payment. I used these valuations to estimate the value of beer in
parishes where beer was given but only the money wage was reported.
Where beer was given but I do not know its value, I used an imputed
value equal to the average value of beer in parishes reporting wages both
with and without beer, either in parishes of the same county, or, if none
of these existed, from surrounding counties. The value is usually about
1s. per week. On a few occasions, meals were also part of the wage. The
value of board was estimated the same way, but not for every county,
since it was used less frequently than beer. Since prices in Wales were
much different, the value of board in Wales was estimated separately
from England. I estimate the value of board for women by assuming it to
be worth three-fourths the value for men.35
The wages given for men are for agricultural work. Women™s wages,
however, are wages for a variety of unskilled occupations. Agriculture
was the most commonly cited employer of women, but some of the
wages in this data set are for weaving, lace-making, straw-plaiting, and
factory work. Thus, these data cover the unskilled labor market more
broadly than did Young™s purely agricultural data.
Women™s wages in this data set are biased for a number of reasons.
Wages are reported by the day, not by the hour, and since women often
worked fewer hours than men, this biases female wages down. In
domestic industry, women with children could work in their homes and

Century: Women™s and Children™s Employment, Family Income, and the 1834 Poor
Law Report,” Economic History Review 55 (2002), pp. 299“323.
35
The value of board for a man is assumed to be 4.8 shillings in England and 3.5 shillings
in Wales. I chose three-fourths as the sex ratio because adult women use about
73 percent as many calories as adult men. See Bekaert, “Caloric Consumption,” p. 638.
Testing for occupational barriers in agriculture 207
Table 4.10. Wages in 1833 (shillings per week)

Mean SD Min. Max. N

Men Summer 11.54 2.26 5.00 21.00 866
Winter 10.29 1.74 5.00 19.00 871
Harvest 17.63 5.30 5.00 34.38 182
Women Summer 4.37 1.28 1.00 9.60 563
(strict defn.) Winter 3.78 1.21 1.00 8.50 324
Harvest 7.87 3.28 1.50 18.00 380
Women Summer 4.27 1.37 1.00 9.60 834
(loose defn.) Winter 3.60 1.26 0.75 9.60 508
Harvest 7.52 3.23 1.00 18.00 595
Boys Summer 3.31 0.97 1.00 7.50 205
Winter 3.22 1.01 1.00 7.50 145
Harvest 5.08 2.01 1.00 15.00 66

Note: “Strict de¬nition” means that only those wages speci¬ed as being for women and not
children are included. “Loose de¬nition” means that ambiguous wages are also included.
Source: BPP 1834 (44) XXX.


thus save the child-care expense that would come with work outside the
home. The wage in domestic industry fails to capture this extra bene¬t,
and may lead to a wage that differs from the market wage for unskilled
women. Another important source of bias is the fact that the survey
includes only one question on both female and child wages. As a result,
the reply often does not distinguish between the two. I will use two
different de¬nitions of female wages. The “strict de¬nition” female wage
variable includes female wages only if the wages of women are clearly
distinguished from the wages of children. The “loose de¬nition” variable
includes also the ambiguous answers. Using the looser de¬nition will
result in a downward bias in the mean, but will not bias the correlations
as long as the measurement error is not correlated with the male wage.
The advantage of using the looser de¬nition is the larger sample size, which
will be especially useful when I examine differences across the seasons.
Table 4.10 provides descriptive statistics for these data. Ratio of
female to male wages is 0.38 in summer, 0.37 in winter, and 0.45 in
harvest (using the strict de¬nition). These ratios are lower than the ratios
from the 1770 data, which is consistent with other studies that have also
found a declining female“male wage ratio over this time period.36 As in
the 1770 data, the ratio is the highest in harvest. The simple correlation
between male and female summer wages, using the loose de¬nition, is

36
See Burnette, “Wages and Employment.”
208 Gender, Work and Wages in Industrial Revolution Britain
Table 4.11. Correlations of men™s and women™s wages: 1833

Summer Winter Harvest

A. Correlations
Women: strict de¬nition
Correlation 0.433** 0.359** 0.282**
N 540 317 82
P-level 0.000 0.000 0.010
95% con¬dence interval (0.36, 0.50) (0.26, 0.45) (0.07, 0.47)
Women: loose de¬nition
Correlation 0.416** 0.323** 0.379**
N 806 499 111
P-level 0.000 0.000 0.000
95% con¬dence interval (0.36, 0.47) (0.24, 0.40) (0.21, 0.53)
B. Rank correlations
Women: strict de¬nition
Correlation 0.463** 0.364** 0.258**
N 540 317 82
P-level 0.000 0.000 0.019
95% con¬dence interval (0.39, 0.53) (0.26, 0.46) (0.03, 0.46)
Women: loose de¬nition
Correlation 0.442** 0.334** 0.369**
N 806 499 111
P-level 0.000 0.000 0.000
95% con¬dence interval (0.38, 0.50) (0.25, 0.41) (0.20, 0.52)

Standard errors in parentheses.
* ¼ signi¬cantly different from zero at the 10% level
** ¼ signi¬cantly different from zero at the 5% level
Source: BPP 1834 (44) XXX.



0.416 and is signi¬cantly positive (see Table 4.11). The correlation
using the “loose de¬nition” is slightly lower than the correlation using
the “strict de¬nition,” probably because of the greater measurement
error. The size of the summer wage correlations is the same or higher
than the correlations from 1770 wages, which were around 0.4 in winter
and harvest, but lower in haytime. Correlations for winter and harvest
wages are lower than for summer wages, but all are signi¬cantly positive.
To check the robustness of the correlations, I also present rank correl-
ations, which are presented in panel B of Table 4.11. The rank correl-
ations give slightly higher correlations during the summer and winter, and
slightly lower correlations during harvest, but the differences are small.
The regressions presented in Table 4.12 indicate that the elasticity of
the effect was less than one, but still quite strong. For summer and
winter wages, a 10 percent increase in the male wage was associated with
an increase in the female wage of about 7 percent. The low R2™s,
Testing for occupational barriers in agriculture 209
Table 4.12. Log-log regressions: 1833 (dependent variable ¼ women™s wage)

Summer Winter Harvest

A. Women™s wage: strict de¬nition
Constant “0.226 “0.330 1.079**
(0.155) (0.258) (0.497)
Men™s wage 0.681** 0.696** 0.317*
(0.064) (0.111) (0.175)
R2 0.175 0.110 0.039
N 540 317 82
B. Women™s wage: loose de¬nition
Constant “0.269** “0.215 0.320
(0.136) (0.220) (0.400)
Men™s wage 0.684** 0.620** 0.558**
(0.056) (0.095) (0.142)
R2 0.157 0.079 0.124
N 806 499 111

Standard errors in parentheses.
* ¼ signi¬cantly different from zero at the 10% level
** ¼ signi¬cantly different from zero at the 5% level




however, show that differences in male wages explain only a small part of
the differences in female wages. To explain more of the variation in
wages, I add to the regressions dummy variables indicating in which
industry the women were working. Results are presented in Table 4.13.
Including industry effects reduces the coef¬cient on male wages only
slightly. The dummies have signi¬cant effects on women™s wages, and
the R2 increases, though the majority of the variation remains unex-
plained. Women™s wages in cottage industry were lower than in agri-
culture, while wages in washing were higher. These differences may
simply re¬‚ect the fact that women tended to work fewer hours in cottage
industry, and more hours in washing, or they may re¬‚ect productivity
differences arising from the selection of women into those industries.
Regressions in the second and fourth columns of Table 4.13 also include
regional dummies. Women™s wages were higher in the north, west, and
south-east than in the midlands. While adding the industry dummies does
not substantially change the coef¬cient on the men™s wage, adding regional
dummies does. This result makes sense because some of the covariance
between male and female wages is simply the result of different levels of
demand for labor in different regions. However, the coef¬cient on the male
wage is still signi¬cantly positive, indicating that even controlling for
industry and regional effects, male and female wages moved together.
210 Gender, Work and Wages in Industrial Revolution Britain


Table 4.13. Industry regressions: 1833 (dependent variable ¼ women™s wage)

Summer Winter

Constant “0.164 1.682** “0.111 1.342**
(0.125) (0.253) (0.206) (0.349)
Men™s wage 0.658** 0.206** 0.598** 0.204**
(0.051) (0.020) (0.089) (0.032)
Manufacture “0.118** “0.459** 0.020 0.022
(0.053) (0.230) (0.067) (0.245)
Washing 0.442** 2.843** 0.273 1.801**
(0.162) (0.641) (0.238) (0.780)
Straw “0.215** “0.229 “0.093 0.213
(0.071) (0.286) (0.088) (0.296)
Lace “0.585** “1.581** “0.558** “1.290**
(0.046) (0.195) (0.058) (0.208)
Other cottage “0.219** “0.526* “0.257** “0.566*
industry (0.073) (0.292) (0.088) (0.293)
Weaving “0.109 “0.686** “0.140 “0.771**
(0.079) (0.329) (0.100) (0.354)
North 0.969** 0.792**
(0.141) (0.183)
Industrial north 0.468** 0.552**
(0.158) (0.193)
West 0.395** 0.512**
(0.193) (0.219)
East “0.203 “0.078
(0.146) (0.194)
Lincoln 0.618* 0.472
(0.339) (0.406)
South-east 0.442** 0.584**
(0.131) (0.198)
South-west 0.100 0.052
(0.145) (0.169)
Wales 0.224 0.677**
(0.241) (0.319)
R2 0.321 0.343 0.241 0.279
N 806 806 498 498

All wages are in logs. Standard errors in parentheses.
* ¼ signi¬cantly different from zero at the 10% level
** ¼ signi¬cantly different from zero at the 5% level
The loose de¬nition is used for women™s wages. The omitted occupation is agriculture; the omitted
region is the midlands.
Source: BPP 1834 (44) XXX.
Testing for occupational barriers in agriculture 211

There are reasons the correlations measured here may be spurious.
The most important reason is location-speci¬c effects that will raise or
lower all wages in an area. Wage differentials that compensate for dis-
amenities would produce such location-speci¬c effects. Also, since all
wages for one parish were reported by the same person, the measurement
error may have a parish-speci¬c component. I correct for this problem by
subtracting winter from summer wages (and summer from harvest
wages), to eliminate the ¬xed effects.37 The question becomes: In areas
where the male wage increases substantially from winter to summer, are
female wages also more likely to increase? Unfortunately, using wage
differences will increase the relative size of the measurement error, and
thus should increase the bias due to measurement error. Since measure-
ment error introduces a downward bias, increasing the importance of the
measurement error will result in more of an underestimate.
Table 4.14 presents the correlations of wage differences. The summer“
winter wage differences have a correlation of 0.28, or 0.22 using the
strict de¬nition of women™s wages. As expected, the correlation of sea-
sonal wage differences is lower than the correlations of wage levels, but
there is still evidence that seasonal changes in male wages are positively
associated with seasonal changes in female wages. Most likely, the
original correlations contained some spurious correlation due to loca-
tion-speci¬c effects. After purging these effects, however, there is still
evidence of market integration: women™s wages increase where men™s
wages increase. Table 4.15 presents regressions of the wage differences.
The elasticities are much lower for the summer“winter differences, and
negative for the harvest“summer differences. Again, some of the cor-
relation seems to have been spurious. The summer“winter differences
indicate an integrated labor market. The harvest-summer differences do
not, but in this case the productivity ratio probably changed because of
the increased use of the scythe. The scythe required so much strength
that women never used it. The resulting inability to substitute women
for men in harvesting may have meant that in arable regions male wages
rose substantially during harvest but female wages did not.
While male and female wages were clearly not independent, as they
would be if there was complete occupational segregation, the relation-
ship may have been weaker than it should have been in a perfectly
competitive market. Women might have been at a disadvantage in their
competition with men if employers favored men and used women as

37
Assume that the parish-speci¬c component of the wage is additive, and is the same in
both seasons. The observed male summer wage is Wms ¼ wmsþDi, where Di is speci¬c to
parish i. Then the difference Wms À Wmw ¼ (wmsþD)À(wmwþD) ¼ wmsÀwmw.
212 Gender, Work and Wages in Industrial Revolution Britain

Table 4.14. Correlation of seasonal wage differences: 1833

Summer“winter Harvest“summer

A. Correlations
Women™s wage: strict de¬nition
Correlation 0.216** 0.229*
N 305 71
P-level 0.000 0.054
95% con¬dence interval (0.11, 0.32) (“0.005, 0.44)
Women™s wage: loose de¬nition
Correlation 0.283** 0.296**
N 483 102
P-level 0.000 0.003
95% con¬dence interval (0.20, 0.36) (0.11, 0.46)
B. Rank correlations
Women™s wage: strict de¬nition
Correlation 0.259** 0.228*
N 305 71
P-level 0.000 0.056
95% con¬dence interval (0.15, 0.36) (“0.006, 0.44)
Women™s wage: loose de¬nition
Correlation 0.337** 0.300**
N 483 102
P-level 0.000 0.002
95% con¬dence interval (0.26, 0.41) (0.11, 0.47)

Standard errors in parentheses.
* ¼ signi¬cantly different from zero at the 10% level
** ¼ signi¬cantly different from zero at the 5% level
Source: BPP 1834 (44) XXX.



marginal workers. Some of the survey responses suggested that women
were not hired if men were unemployed. This subordination of women™s
employment is suggested by such statements as, “Women . . . have but
little other work [besides harvest], there are so many Men and Lads out
of employ,”38 and “In consequence of the number of male hands,
Females are generally unemployed.”39 I look for this relationship in the
data by including in the regression male unemployment, as reported in
response to question 6: “Number of laborers generally out of employ-
ment, and how maintained in Summer and Winter?” The response to this
question is divided by the answer to question 5 “ “Number of Agricultural
laborers in your parish?” “ to produce an estimate of the unemployment
rate. Table 4.16 presents regressions including this variable. I ¬nd that

38
BPP 1834 (44) XXX, Bramshaw, Southampton, p. 423
39
BPP 1834 (44) XXX, Rother¬eld, Sussex, p. 529 .
Testing for occupational barriers in agriculture 213
Table 4.15. Difference-of-log regressions: 1833 (dependent variable ¼ difference
of log women™s wages)

Summer“winter Harvest“summer

A. Women™s wage: strict de¬nition
Constant 0.091** 0.501**
(0.015) (0.021)
Men™s log wage difference 0.229** “0.109
(0.088) (0.112)
R2 0.037 0.003
N 305 351
B. Women™s wage: loose de¬nition
Constant 0.107** 0.496**
(0.014) (0.016)
Men™s log wage difference 0.437** “0.095
(0.079) (0.082)
R2 0.060 0.002
N 483 573

All wages are in logs. Standard errors in parentheses.
* ¼ signi¬cantly different from zero at the 10% level
** ¼ signi¬cantly different from zero at the 5% level
Source: BPP 1834 (44) XXX.




Table 4.16. The effect of unemployment: 1833 (dependent variable
¼ women™s wage)

Summer Winter

Constant “0.131 “0.321
(0.176) (0.303)
Men™s wage 0.642** 0.664**
(0.071) (0.129)
Unemployment rate “0.769** “0.248
(0.159) (0.164)

R2 0.184 0.091
N 548 306

Standard errors in parentheses.
All wages are in logs. The loose de¬nition is used for the women™s wage.
* ¼ signi¬cantly different from zero at the 10% level
** ¼ signi¬cantly different from zero at the 5% level
Source: BPP 1834 (44) XXX.
214 Gender, Work and Wages in Industrial Revolution Britain
Table 4.17. Correlations with boys™ wages: 1833

Summer“ Harvest“
Summer Winter Harvest winter summer

Women™s wage, strict de¬nition
Correlation 0.303** 0.541** 0.642** 0.338** 0.477**
N 148 64 45 59 38
P-level 0.000 0.000 0.000 0.009 0.003
95% con¬dence (0.15, 0.44) (0.34, 0.69) (0.43, 0.79) (0.09, 0.55) (0.19, 0.69)
interval
Men™s wage
Correlation 0.295** 0.378** 0.434* 0.264** 0.579**
N 198 142 20 128 16
P-level 0.000 0.000 0.055 0.003 0.019
95% con¬dence (0.20, 0.42) (0.23, 0.51) (“0.01, 0.74) (0.09, 0.42) (0.012, 0.84)
interval

Standard errors in parentheses.
* ¼ signi¬cantly different from zero at the 10% level
** ¼ signi¬cantly different from zero at the 5% level
Source: BPP 1834 (44) XXX.


the unemployment rate reduced women™s summer wages, which suggests
that the demand for women workers went down when men were
unemployed. Such a relationship could be the result of gender discrim-
ination, or of incentives created by the poor law to employ male work-
ers.40 However, the negative effect of unemployment may also simply
re¬‚ect the fact that male wages were somewhat sticky, and thus do not
fully capture demand conditions in a region. If men were unemployed,
that suggests that the market-clearing male wage would be lower than the
prevailing market wage.41 If the unemployment rate is an indicator of
weak labor demand, then we would expect it to reduce women™s wages.
This data set also includes wages of boys, which can be used for
comparison. Examining correlations with boys™ wages may help us
determine if the correlation of men™s and women™s wages is large or
small. Table 4.17 shows that the correlation between women™s wages
and boys™ wages is 0.30 in the summer and 0.54 in the winter, using the
strict de¬nition of women™s wages. The correlation between men™s
wages and boys™ wage is 0.30 in the summer and 0.38 in the winter.

40
On the incentives created by the poor law system, see George Boyer, An Economic History of
the English Poor Law, 1750“1850 (Cambridge: Cambridge University Press, 1990).
41
Failure of male wages to fall to the market-clearing level may be explained either by the
incentives of the poor law system, or by ef¬ciency wage models. For a survey of ¬rms™
reasons for not lowering wages in a contemporary labor market, see Carl Campbell and
Kunal Kamlani, “The Reasons for Wage Rigidity: Evidence from a Survey of Firms,”
Quarterly Journal of Economics 112 (1997), pp. 759“89.
Testing for occupational barriers in agriculture 215
Table 4.18. Wages in England and Wales: 1860“1 (shillings per week)

Mean SD Min. Max. N

Men 1860:III 12.39 2.29 8.50 20.50 101
1860:IV 11.39 1.67 8.50 15.50 101
1861:I 11.48 1.92 6.33 18.50 100
1861:II 11.95 2.23 6.67 20.25 101
Women 1860:III 5.38 1.76 2.25 11.25 98
1860:IV 4.70 1.56 1.00 10.00 90
1861:I 4.46 1.41 1.00 11.00 87
1861:II 5.00 1.39 2.75 9.50 96
Children under 16 1860:III 4.14 1.33 2.00 8.00 94
1860:IV 3.60 1.08 1.50 7.00 92
1861:I 3.67 1.12 1.50 7.50 87
1861:II 3.79 1.04 1.50 6.50 92

Source: BPP 1861 (14) L.




These correlations are similar to the correlations between men™s and
women™s wages, and are also signi¬cantly positive. Women seem to have
been substitutable with men to the same extent that boys were.
For unskilled wages in 1833, I ¬nd a signi¬cantly positive correlation
between men™s wages and women™s wages that is robust. The simple
correlations seem to contain some location-speci¬c effects, and correcting
for these reduces the strength of the correlation. However, even after
correcting for ¬xed effects, there is still evidence that men and women
competed in the same labor market. The correlations were as strong in
1833 as correlations in 1770, so I ¬nd no evidence that the unskilled
labor market became more segregated between 1770 and 1833.

C. 1860
Data from the middle of the nineteenth century are provided by a wage
survey contained in the 1861 parliamentary returns. This survey covers
only agricultural wages and is smaller in size than the 1833 survey.
Fortunately, this data set contains different wages over the season,
allowing for a ¬xed-effect estimation. The survey also contains infor-
mation on in-kind payments. As in the 1833 data, internal evidence is
used to estimate the value of these in-kind payments where the money
value is not speci¬cally given.
Table 4.18 presents the descriptive statistics for this data set. Wages
are given by quarter, so that “Wages for the quarter ending Michaelmas,
1860” are recorded as wages for 1860:III, and similarly for the other
216 Gender, Work and Wages in Industrial Revolution Britain
Table 4.19. Correlations of men™s and women™s wages: 1860“1

1860:III 1860:IV 1861:I 1861:II

Men & women
Correlation 0.740** 0.508** 0.619** 0.694**
N 98 90 87 96
P-level 0.000 0.000 0.000 0.000
95% con¬dence (0.63, 0.82) (0.34, 0.65) (0.47, 0.73) (0.57, 0.79)
interval
Men & children
Correlation 0.503** 0.369** 0.552** 0.490**
N 94 92 87 92
P-level 0.000 0.000 0.000 0.000
95% con¬dence (0.33, 0.64) (0.18, 0.53) (0.39, 0.68) (0.32, 0.63)
interval
Women & children
Correlation 0.642** 0.474** 0.382** 0.551**
N 91 83 80 89
P-level 0.000 0.000 0.001 0.000
95% con¬dence (0.50, 0.75) (0.29, 0.64) (0.19, 0.55) (0.39, 0.68)
interval

Standard errors in parentheses.
* ¼ signi¬cantly different from zero at the 10% level
** ¼ signi¬cantly different from zero at the 5% level
Source: BPP 1861 (14) L.


quarters. The female“male wage ratios are not surprising: 0.43 for 1860:
III, 0.41 for 1860:IV, 0.39 for 1861:I, and 0.42 for 1861:II. Wages are
lowest in the ¬rst and fourth quarters (winter), and highest in the third
quarter, which includes harvest.
The simple correlations are presented in Table 4.19. The correlation
of men™s and women™s wages is higher than in the previous two data sets,
ranging from 0.51 to 0.74, and all the correlations are signi¬cantly
positive. Correlations with children™s wages are similar to the correl-
ations of men™s and women™s wages, but somewhat lower. This suggests
that the women were at least as good a substitute as children for men™s
labor. Table 4.20 shows the male wage elasticities of female wages.
Women™s wages appear to be very responsive to men™s wages; the elas-
ticities in Table 4.20 are higher than in previous data sets, and in three of
the four quarters exceed one. Do the higher correlations result from
larger location-speci¬c effects? Table 4.21 presents correlations of wage
differences, to correct for location-speci¬c effects. The correlations drop
substantially compared to the correlations in Table 4.19, suggesting that
some of the correlation was due to location-speci¬c effects. However,
even correcting for ¬xed effects, the correlations remain generally higher
Testing for occupational barriers in agriculture 217
Table 4.20. Log-log regressions: 1860“1 (dependent variable ¼ women™s wage)

1860:III 1860:IV 1861:I 1861:II

Constant “1.427** “1.562** “1.605** “0.691**
(0.307) (0.622) (0.540) (0.277)
Men™s wage 1.222** 1.262** 1.259** 0.919**
(0.122) (0.257) (0.223) (0.112)
R2 0.509 0.215 0.273 0.418
N 98 90 87 96

All wages in logs. Standard errors in parentheses.
* ¼ signi¬cantly different from zero at the 10% level
** ¼ signi¬cantly different from zero at the 5% level
Source: BPP 1861 (14) L.



Table 4.21. Correlations of wage differences: 1860“1

1860:III“ 1860:III“ 1860:III“ 1861:II“
1860:IV 1861:I 1861:II 1860:IV

Men“women
Correlation 0.489** 0.380** 0.436** 0.273**
N 89 86 96 88
P-level 0.000 0.000 0.000 0.010
95% con¬dence interval (0.31, 0.63) (0.18, 0.55) (0.26, 0.59) (0.07, 0.46)
Men“children
Correlation 0.370** 0.363** 0.184* 0.262**
N 89 84 89 87
P-level 0.000 0.001 0.084 0.014
95% con¬dence interval (0.18, 0.54) (0.16, 0.54) (“0.03, 0.38) (0.05, 0.45)
Women“children
Correlation 0.532** 0.374** 0.545** 0.446**
N 79 76 86 79
P-level 0.000 0.001 0.000 0.000
95% con¬dence interval (0.35, 0.70) (0.16, 0.55) (0.38, 0.68) (0.25, 0.61)

Standard errors in parentheses.
* ¼ signi¬cantly different from zero at the 10% level
** ¼ signi¬cantly different from zero at the 5% level
Source: BPP 1861 (14) L.


than those in Table 4.8 or Table 4.11. Seasonal variation in female
wages was not as responsive to seasonal variation in male wages as female
wage levels were to male wage levels, but there was still a signi¬cant
relationship. The elasticities in Table 4.22 are below one, ranging from
0.33 to 0.86. However, these elasticities are still relatively high
218 Gender, Work and Wages in Industrial Revolution Britain
Table 4.22. Difference-of-log regressions: 1860“1 (dependent variable ¼ difference
of the log of women™s wages)

1860:III“ 1860:III“ 1860:III“ 1861:II“
1860:IV 1861:I 1861:II 1860:IV

Constant 0.064* 0.127** 0.037 0.059
(0.038) (0.035) (0.024) (0.039)
Difference of 0.858** 0.586** 0.481** 0.328
men™s wages (0.234) (0.228) (0.152) (0.254)
R2 0.134 0.073 0.096 0.019
N 89 86 96 88

All wages are in logs. Standard errors in parentheses.
* ¼ signi¬cantly different from zero at the 10% level
** ¼ signi¬cantly different from zero at the 5% level
Source: BPP 1861 (14) L.


compared to those in the 1770 and 1833 data. The relationship between
male and female wages certainly did not erode over the course of the
Industrial Revolution; if anything it seems to be stronger in 1860 than it
was earlier. The evidence does not suggest that occupational segregation
constraints appeared in the agricultural labor market during the course
of the Industrial Revolution.

D. France in 1839
I have shown that there de¬nitely was a positive correlation between
male and female agricultural wages in England. However, it is more
dif¬cult to say whether the relationship should be considered strong or
weak. I have compared correlations over time, and I have compared the
correlations with children™s wages to those between male and female
adults, to get some idea of whether the correlations were large. Another
place to turn for comparison wages is another country. How did the
English labor market compare to the labor market in other countries?
Evidence on wages in France helps to put the English results in context.
The French wage data I will use come from a British parliamentary
report. The investigator, J. C. Symons, corresponded with government
of¬cials in France, who sent him wage returns. The wages seem to be
based on correspondence with employers in France.42 Table 4.23 pre-
sents descriptive statistics for these wage data. The wage ratio is 0.65,

42
Along with the list of agricultural wages, there are transcripts of letters from French
employers stating wages paid. BPP 1839 (159) XLII, pp. 137“49.
Testing for occupational barriers in agriculture 219

Table 4.23. Descriptive statistics: French agricultural day-laborers in 1839

Mean SD Min. Max. N

Male wage 0.992 0.28 0.33 1.63 63
Female wage 0.646 0.20 0.23 0.97 63

Source: BPP 1839 (159) XLII, pp. 147“9.



Table 4.24. French agricultural day-laborers: 1839

Wages Residuals§

Correlation 0.719 0.682
N 63 63
95% con¬dence interval (0.573, 0.821) (0.523, 0.795)
Regressions (dependent variable ¼ log female wage)
Constant “0.447** “0.000
(0.029) (0.027)
Log male wage 0.926** 1.030**
(0.091) (0.114)
R2 0.629 0.571
N 63 63

Standard errors in parentheses.
§ Residuals from regressions of each wage on dummies for province.
* ¼ signi¬cantly different from zero at the 10% level
** ¼ signi¬cantly different from zero at the 5% level
Source: BPP 1839 (159) XLII, pp. 147“9.


higher than in any of the measured ratios in English agriculture, but still
well within the range of gender wage ratios presented in Table 2.1. Table
4.24 presents wage correlations and regressions. The wage correlation is
high “ comparable to the correlations in the 1860 English data. Unfor-
tunately, seasonal wages are not given, so I cannot correct for ¬xed effects.
The best I can do is to correct for region-speci¬c effects. The second
column shows the correlation of wage residuals, after correcting for vari-
ation in wages across provinces. I ¬rst regress wages on dummies for eight
provinces, and then take the correlations of the residuals. The correlation
is still high, so it is possible that French agricultural labor markets were
more integrated by gender than English agricultural labor markets. It
remains possible, though, that the high elasticities measured in France
re¬‚ect my failure to correct for location-speci¬c effects.
220 Gender, Work and Wages in Industrial Revolution Britain

Conclusion
I have presented wage correlations from a variety of data sets. The
results consistently indicate a positive correlation between male and
female wages, although the correlations are in many cases small. Overall,
the evidence suggests that men and women did compete in the same
labor market. Most importantly, the results suggest that the integrated
labor market which we found in 1770 persisted throughout the whole
Industrial Revolution period. The correlations in 1860 were even
stronger than in 1770. French agricultural labor markets were also
integrated by gender, and may have been more integrated than English
labor markets.
This chapter has examined agricultural labor markets, and has shown
that men and women were considered substitutable by employers,
suggesting that employment was not determined by rigid gender roles. I
have tested only agricultural labor markets and, while I could expect
to ¬nd the same results in competitive portions of the labor market,
there were some sections of the labor market where men and women
were clearly not substitutable because monopolization allowed men to
construct barriers preventing women from entering the occupation.
Chapters 5 and 6 discuss those portions of the labor market.
5 Barriers to women™s employment




The pressure of male trade unions appears to be largely responsible for
that crowding of women into a comparatively few occupations, which is
universally recognized as a main factor in the depression of their wages.
Edgeworth, 19221

Having presented models of market-based occupational sorting, and
argued that in some portions of the labor market gender differences in
occupations and wages were the results of differences in strength, this
chapter turns to segments of the labor market where discriminatory
barriers, rather than comparative advantage, kept women out of the
best-paid occupations. I begin by examining cases where the predictions
of the sorting models do not hold, and then move on to examine possible
causes of the discriminatory barriers. I ¬nd that barriers were erected
where men could use their market power to reduce competition in order
to improve their own labor market outcomes. In this case occupational
sorting bene¬ted men and made women worse off, but increased com-
petition would have reduced occupational sorting. The conclusions of
this chapter thus support my general claims that the gender division of
labor was driven by economic motivations, and that women bene¬ted
from competition.


I. Occupational sorting not based on strength
While the absence of women from some occupations can be explained
by the strength requirements of the occupation, in many cases women
were absent from occupations not requiring strength, suggesting that
other forces must have been at work. Table 5.1 shows the prevalence of
women in certain occupations in the 1841 census. Section A of Table 5.1
shows the number of men and women in lower-skill wage-labor
occupations. In section B I have selected some white-collar, craft, and

1
F. Y. Edgeworth, “Equal Pay to Men and Women for Equal Work,” Economic Journal 32
(1922), p. 439.

221
Table 5.1. The percentage of women in selected occupations: the 1841 census
(Great Britain, persons 20 and over)

Occupation Men Women Percent women

A. Wage labor
Agricultural laborer 874,294 41,879 4.6
Laborer 333,786 12,474 3.6
Domestic servant 144,072 562,392 79.6
Textile manufacturea 268,557 177,251 39.8
B. Craft and professional
No strength required
Accountant 4684 0 0.0
Agent, factor 5365 61 1.1
Artist 3805 277 6.8
Attorney 13,918 0 0.0
Auctioneer, appraiser 3156 37 1.2
Banker 1791 8 0.4
Broker 2869 464 13.9
Clergyman, minister 23,496 0 0.0
Clerk 46,368 152 0.3
Dressmaker, milliner 436 84,064 99.5
Civil servant 15,853 617 3.7
Musician 3223 216 6.3
Pastry cook, confectioner 4241 1808 29.9
Tailor and 100,030 5339 5.1
breeches-maker
Teacher 25,207 31,557 55.6
Strength required
Anchor-smith and 1384 54 3.8
chain-maker
Blacksmith 80,543 512 0.6
Boat builder 24,149 142 0.6
Brass founder 4776 39 0.8
Brazier 5540 39 0.7
Bricklayer 36,049 107 0.3
Carpenter, joiner 141,750 452 0.3
Currier 9273 155 1.6
Engine and machine 5761 58 1.0
maker
Mason, pavior 72,934 184 0.3
Saddler 12,962 309 2.3
Sawyer 27,929 20 0.1
Wheelwright 22,537 147 0.6
Total employed in Britain 4,279,004 1,246,585 22.6

a Cotton manufacturer, ¬‚ax and linen manufacturer, silk manufacturer, weaver
(unspeci¬ed), woollen and cloth manufacturer, and worsted manufacturer.
Note: Total employed does not match Table 1.1 because these totals include only persons
over 20.
Source: BPP 1844 (587) XXVII.
Barriers to women™s employment 223
Table 5.2. Occupational sorting in skilled occupations: Manchester, 1846

Occupation Male Female Unknown Percent
female

Not requiring strength
Accountant 36 0 7 0.0
Agent 362 1 82 0.3
Attorney 168 1 59 0.6
Auctioneer and appraiser 45 0 3 0.0
Confectioner 59 28 7 32.2
Draper, mercer 167 21 25 11.2
Grocer and tea dealer 252 21 32 7.7
Hairdresser 116 0 2 0.0
Hosier, haberdasher 180 67 28 27.1
Librarian 27 13 0 32.5
Milliner, dressmaker 12 245 18 95.3
Pawnbroker 142 23 16 13.9
Publican 496 91 42 15.5
Schoolmaster/mistress 232 181 17 43.8
Shopkeeper 776 128 4 14.2
Tailor 239 1 18 0.4
These 16 occupations 3309 821 360 19.9
Requiring strength
Blacksmith and farrier 84 6 6 6.7
Brazier and coppersmith 116 8 33 6.5
Currier 96 8 4 7.7
Iron founder 29 0 16 0.0
Joiner and builder 187 0 33 0.0
Machine maker 117 5 33 4.1
Millwright 25 0 14 0.0
Saddler 34 1 0 2.9
Slater 14 1 0 6.7
Stone and marble mason 40 0 4 0.0
These 10 occupations 742 29 143 3.8

Source: Slater™s National Commercial Directory of Ireland, 1846.



professional occupations that I think would not require strength, and
some I think would require much strength. Differences in strength
explain some of the occupational sorting; women were generally less
likely to work in occupations requiring strength. They were more likely
to work as domestic servants than as agricultural laborers. Few women
worked as blacksmiths, masons, or sawyers. The same pattern can be
seen in Table 5.2, which presents employment by sex from the 1846
commercial directory for Manchester. While business owners in the
strength-intensive trades selected were only 4 percent female, owners in
224 Gender, Work and Wages in Industrial Revolution Britain
Table 5.3 The gender division of labor in staymaking

Trade Men Women Unknown Percent women

Shef¬eld, 1774 3 0 0 0.0
Manchester, 1788 10 2 2 16.7
Coventry, 1791 5 0 0 0.0
Manchester, 1824“5 9 2 1 18.2
Coventry, 1835 3 3 0 50.0
Manchester, 1846 14 11 0 44.0
Birmingham, 1850 6 59 6 90.8
Derby, 1850 4 12 0 75.0
Coventry, 1892 0 3 0 100.0

Sources: Sketchley™s Shef¬eld Directory, 1774; Lewis™s Manchester Directory for 1788; The
Universal British Directory, 1791; Pigot and Dean™s Directory for Manchester, 1825; Pigot &
Co.™s National Commercial Directory, 1835; Slater™s National Commercial Directory of Ireland,
1846; Slater™s Royal National and Commercial Directory, 1850.


the selected trades not requiring strength were 20 percent female.
Gender differences in strength had some impact on what trades women
were engaged in.
Changes in the amount of strength required may explain the change in
the prevalence of women in staymaking. In the eighteenth century,
staymakers were primarily male. Among the married couples Peter Earle
reports from London in 1695 to 1725 there are ¬ve male and two female
staymakers, and both the females had husbands who were staymakers.2
In Shef¬eld in 1774, and in Coventry in 1791, all of the staymakers were
male (see Table 5.3). Over the course of the Industrial Revolution,
women became more common in the trade, and by the middle of the
nineteenth century most staymakers were women. In the 1841 census 86
percent of staymakers were female.3 This shift seems to have resulted
from a change in how stays were constructed. In the ¬rst half of the
eighteenth century, only men made stays because it required strength. In
1747, R. Campbell commented on the prevalence of men in staymaking:
I am surprised the Ladies have not found out a Way to employ Women Stay-
Makers rather than trust our Sex with what should be kept as inviolably as Free-
Masonry; But the Work is too hard for Women, it requires more Strength than
they are capable of, to raise Walls of Defence about a Lady™s Shape . . . After the
Stays are stitched, and the Bone cut into thin Slices of equal Breadths and the


2
Earle, “The Female Labor Market in London,” pp. 348“52.

<<

. 8
( 14)



>>