Transcript Module 12 An Econometric Model of Sex

An Econometric Model of Sex-Specific
U.S.
Immigration before the Civil War
Michael J. Greenwood
University of Colorado at Boulder
Donna Gabaccia: “Historical studies of international female migration
scarcely exist” (1996, p. 91).
A study by the United Nations (1994) indicates that little is known about
women’s migration, either internal or international, and laments “the
neglect of research on women’s migration” (p. xv).
Why is the sex composition of international migration important?
1. Females were less likely to be “economic migrants” than males.
Economic migrants are motivated by their own economic
advantages and costs. Thus, females were less likely to participate
in formal labor markets upon their arrival in the U.S.
2. The child-bearing capacity of females increases the potential for
growth of the second generation immigrant population.
The United States began collecting immigration data, by sex, in 1820. This
is by far the earliest date that any “New World” country began systematically
recording information on immigrants, by country of birth. These data are not
without problems, but at least they provide a resource with which to work.
Later I will discuss some of these problems, and I also will discuss my
efforts to “correct” or “adjust” the data to account for the shortcomings.
Table 1. Sex-Specific U.S. Immigration, by Source Country, 1820-1855
Country
Males
Females
Sex-ratio
Belgium
Denmark
France
Germany
Ireland
Italy
Netherlands
Sweden/Norway
Portugal
Spain
3,996
2,028
123,252
752,431
1,013,566
6,782
10,679
17,592
2,876
9,536
2,995
1,028
65,540
487,864
817,106
1,444
6,902
11,722
686
1,988
133.4
197.3
188.1
154.2
124.0
469.7
154.7
150.1
419.2
479.7
United Kingdoma
311,607
198,918
156.7
a
Less Ireland.
Ireland
United Kingdom (less Ireland)
50.0%
50.0%
40.0%
40.0%
30.0%
30.0%
20.0%
20.0%
10.0%
10.0%
0.0%
0.0%
1820
1825
1830
1835
1840
1845
1850
1855
1820
1825
1830
Germany
1835
1840
1845
1850
1855
Netherlands
50.0%
50.0%
40.0%
40.0%
30.0%
30.0%
20.0%
20.0%
10.0%
10.0%
0.0%
0.0%
1820
1825
1830
1835
1840
1845
1850
1855
1820
1825
1830
Sweden and Norway
1835
1840
1845
1850
1855
1840
1845
1850
1855
Denmark
50.0%
60.0%
50.0%
40.0%
40.0%
30.0%
30.0%
20.0%
20.0%
10.0%
10.0%
0.0%
0.0%
1820
1825
1830
1835
1840
1845
1850
1855
1820
1825
1830
1835
Figure 1. Percentage of US Immigrants Who Were Female by Country of Origin, 1820-1856
France
Belgium
50.0%
60.0%
50.0%
40.0%
40.0%
30.0%
30.0%
20.0%
20.0%
10.0%
10.0%
0.0%
0.0%
1820
1825
1830
1835
1840
1845
1850
1855
1820
1825
1830
Italy
1835
1840
1845
1850
1855
1840
1845
1850
1855
Spain
50.0%
50.0%
40.0%
40.0%
30.0%
30.0%
20.0%
20.0%
10.0%
10.0%
0.0%
0.0%
1820
1825
1830
1835
1840
1845
1850
1855
1840
1845
1850
1855
1820
1825
1830
1835
Portugal
50.0%
40.0%
30.0%
20.0%
10.0%
0.0%
1820
1825
1830
1835
Figure 1 (Cont.). Percentage of US Immigrants Who Were Female by Country of Origin, 1820-1855
Table 2. Sex Composition of U.S. Immigration, 1821-1990
Period
Sex Ratio
a
1821-30
222.3
a
1831-40
179.8
a
1841-50
146.3
1851-60a
137.1
a
1861-70
153.3
1871-80a
159.1
a
1881-90
156.6
a
1891-1900
163.7
a
1901-10
230.2
Period
1911-20a
1921-30b
1931-40b
1941-50b
1951-60b
1961-70b
1971-80bc
1981-90def
1991-2000ef
Sex Ratio
171.1
125.1
76.6
67.7
85.0
81.1
88.4
99.3
85.2
Background Research on Sex-Specific Migration
Ravenstein’s (1885) Seventh Law of Migration: “Females are more migratory
than males.”
D.S.Thomas (1938) devotes a chapter tp “sex differentials” and there she
poses the following question: “What opportunities, social and
economic, are offered young men and women migrants in what types
of cities ? … How far is migration an adjustment to these
opportunities?” (pgs. 68-69).
Hatton and Williamson (1998), in their book on the era of mass migration,
devote a little attention to female migration.
Donato, Katherine M. “Understanding U.S. Immigration: Why Some Countries
Send Women and Others Send Men.” In Seeking Common Ground, edited
by Donna Gabaccia, 159-184. Connecticut: Praeger, 1992.
Donato, Katherine M. and Andrea Tyree. “Family Reunification, Health
Professionals, and the Sex Composition of Immigrants to the United
States.” In Sociology and Social Research 70, no. 3, edited by Marcus
Freeman, 226-230. Los Angeles: University of Southern California, 1986.
Greenwood, Michael J. and John M. McDowell, Legal U.S. Immigration:
Influences on Gender, Age, and Skill Composition. Kalamazoo:
W.E.UPJOHN Institute for Employment Research, 1999.
Greenwood, Michael J., Family and Sex-Specific U.S. Immigration from
Europe, 1870-1910: A Panel Data Study of Rates and Composition,
Explorations in Economic History, 45 (3), Sept. 2008, 356-382.
The Model
I distinguish two types of migrants:
1. Independent migrants, who presumably entered the U.S. with
economic incentives primarily in mind, and
2. Accompanying migrants, who entered with an independent migrant or
who followed one, but whose own economic incentives were not the
primary force behind the move.
Given that independent economic migrants were primarily male and
accompanying migrants in immigrant family units, apart from children,
were primarily female, variations in the sex composition of the total
flows from the various source countries should be a function of the
relative incentives for single-unit independent migration versus familyunit migration. Economic migrants should be particularly responsive to
differential advantages between the U.S. and their source countries and
to the relative costs of migrating, although the costs of migrating are
relevant for accompanying migrants as well.
Let Mijt represent migration from country i to the United States (j) in year t:
Mijt = zijt + ijt ,
(1)
where  is a vector of unknown parameters, zijt is a vector of independent
variables with zijt = [xijt1, xijt2, xijt3], and ijt is random errors. Mijt may
represent the migration rate from country i to the U.S. in year t or,
alternatively, the female share of migrants from i to the U.S. in t. The x’s
represent, respectively, variables, relating to differential economic
opportunities between source countries and the U.S.(xijt1), variables relating
to the costs of migrating to the U.S.(xijt2), and control variables (xijt3).
Econometric Procedures
The Hausman-Taylor instrumental variable estimator (1981) is used to address
the econometric problems encountered. This approach is briefly described
below:
mist = αi + δt + βxit + γzi + єist,
i=1,…,11; t=1,…,36,
(2)
where mist is migration of sex s from country i in year t. Again mist could refer
either a migration rate or to sex composition. The notation refers to 18201855. Next we partition the set of explanatory variables into two groups,
paying close attention to the model’s time-invariant variables, such as
distance and the dummy variable for English-speaking countries: [ xit | zi ]
where xit is a Kx1 vector of variables that measures characteristics of
country i in period t and zi is a Gx1 vector of time-invariant variables. (I drop
the s subscript because each sex class has the same set of right-hand side
variables.) In (2), β and γ are vectors of unknown parameters, єist is a
random disturbance, and the vectors αi and δt are unobserved countryspecific and time-specific variables, respectively.
The error terms єist are assumed to have zero mean and to be
independent across countries. The model allows different crosssectional and time-specific intercepts, and various assumptions
about these are discussed in some detail in Greenwood, et al.
(1996). For present purposes, only those assumptions relevant to the
Hausman-Taylor approach are considered.
COMPOSITIONAL METHODOLOGY
The methodology used here to examine immigrant sex composition long has
been employed by economists to analyze systems of demand or
expenditure equations (e.g., Leser, 1961; Pollak and Wales, 1969; Parks,
1969; and Barten, 1977) and systems of cost-share equations (e.g., Berndt
and Wood, 1975). However, except for a few recent exceptions, no
previous attempt to use such an approach to study the composition of
migration appears to have been made. In the present context, the idea is to
estimate a two-equation model, with one equation for males and the second
for females, that satisfies an adding-up condition. The adding-up condition
is simply that the fractions of male and female U.S. immigration from
country i during year t must sum to one.
Let i represent source country, j independent variable, t year, and s sex. The
share of U.S. immigration of a given sex is the dependent variable and may
be expressed in the following way, where SHAREist represents the share of
U.S. immigration from country i, of sex s, during year t:
SHAREist = βs∑2s=1SHAREist + ∑nj=1 βsjXisjt + єist,
(3)
where
∑2s=1 βs = 1, and
∑2s=1βsj = 0, V βj .
The conditions state that the coefficients on the constant term (the share of
immigration of sex s from i during t) must sum to one across the two sex
equations. Moreover, if the data are otherwise the same on the right-hand
side of each equation, the coefficients on each independent variable must
sum to zero. Thus, if each independent variable were set at zero the sex
shares would sum to one, as they must logically. Furthermore, any change
in an independent variable that increases (decreases) one share must
correspondingly decrease (increase) the other share, so that the shares
continue to sum to one. The coefficients on the independent variables are
interpreted as the percentage point change in the share of immigrants of
sex s due to an incremental change in the independent variable.
Examples of earlier work using this methodology:
Greenwood, Michael J. and John M. McDowell. Legal U.S. Immigration:
Influences on Gender, Age, and Skill Composition. Kalamazoo:
W.E.UPJOHN Institute for Employment Research, 1999.
Greenwood, Michael J. “Family and Sex-Specific U.S. Immigration from
Europe, 1870-1910: A Panel Data Study of Rates and Composition,”
Explorations in Economic History 45 (3), Sept. 2008, 356-382.
Greenwood, Michael J. “Modeling the Age and Age Composition of Late 19th
Century U.S. Immigrants from Europe,” Explorations in Economic History
44, no. 2, 2007, 255-269.
Greenwood, Michael J., John M. McDowell, and Donald M. Waldman. “A Model
of the Skill Composition of US Immigration,” Applied Economics 28, 1996,
299-308.
Greenwood, Michael J., John M. McDowell and Matt Wierman. “SourceCountry Social Programs and the Age Composition of Legal US
Immigrants,” Journal of Public Economics 87, no. 3-4, 2003, 739-771.
The Data
As indicated earlier, the data cover 1820-1855.
The data end in 1855 because the sex of U.S. immigrants is unknown for
1856-1868.
The Dependent Variables
The data are characterized by several shortcomings:
I. The periods represented by “years” are not uniform (Davis, 1931). Prior to
adopting a regular fiscal year (ending June 30) beginning in 1869, the State
Department reports refer to various different “years”: (1) 1820-1832, 12
months ending September 30;
(2) 1833-1843, calendar year, but the last quarter of 1832 was apparently
lost in the change from one type of period to the other; (3) 1844, nine
months ending September 30;
(4) 1845-1849, 12 months ending September 30;
(5) 1850, 15 months ending December 31; and
(6) 1851-1855, calendar year. In this study, these different periods are taken
into account through the econometric approach discussed above, but the
two nine-month periods are multiplied by a factor of 1.33 and the 15-month
period is multiplied by 0.8 so that each “year” refers to a 12-month period.
II. For certain countries and for specific years, relatively small numbers
of immigrants did not report their sex. For the country and year in
question, these numbers were distributed to the sexes in proportion
to the known sex distribution.
III. Another data issue for this period concerns the U.K. for all years.
During the entire period, large numbers of immigrants (775,158
males, or 58.5 percent of all males, and 572,956 females, or 56.4
percent of all females) from the United Kingdom did not specify
whether they come from England, Ireland, Scotland, or Wales.
Since the United Kingdom was an extremely important source of
immigrants during this period, and since Ireland must be separated
form the other three, those not specified, by sex, were distributed in the
same proportions as those whose origin was specified.
IV. Since immigration data do not exist prior to 1820 and a five-year lag is used
as an independent variable, where necessary the 1820 value is applied to
earlier years.
The data set ultimately employed utilizes 12 source countries commonly used
to study historical U.S. immigration:
1. Belgium,
2. Denmark,
3. France,
4. Germany,
5. Ireland,
6. Italy,
7. The Netherlands,
8. Norway,
9. Portugal,
10. Spain,
11. Sweden, and
12. the United Kingdom (England, Scotland, and Wales).
However, for 1820-1855, reported data on U.S. immigration combine Norway
and Sweden, so only 11 panels are available for study. These 11 countries
or country groups provide 396 observations for the 36-year period under
study.
Independent Variables
The number of countries in the data set is critical because it imposes a
degrees-of-freedom constraint on the estimated model or models. The
number of independent variables included in the models may not exceed n1, where n is the number of countries or country groups (or panels). This
restriction demands that as many countries as possible be included in the
analysis, because no matter which countries are included, the number is not
great. The first consideration in selecting countries of origin was the
availability of data on the dependent variable or variables (i.e., sex of
immigrants). The second was data on the independent variables. Because
a panel data technique is used in this study, annual time-series data had to
be available back to 1820 (and in the case of the birth and death rates, to
1800) for each country. Certain variables (such as sex-specific population
and, in some cases, per capita GDP) had to be interpolated to form a
“synthetic” series. Other data (such as birth and death rates for some
countries) had to be extrapolated.
The econometric modeling approach employed in this study requires a
continuous time series for each time-varying variable included in the model.
Other variables that do not vary over time, such as distance to the United
States, are specifically treated in the econometric approach, but
nonetheless must be assigned a value for each year. Consequently, any
temporal “gaps” in the data must be filled for the approach to work.
Measures for certain of the independent variables employed in this study are
not available on an annual basis. This condition is especially true for
census-based measures, such as population. Such measures were
interpolated or extrapolated to yield continuous a 36 (1820-1855) year time
series for each variable. Variables such as those noted above change
sufficiently slowly, and a sufficient number of data points is available, that I
feel that I have not done great injustice to the temporal aspect of the data.
Moreover, the cross-sectional aspect of the data for years for which the
gaps were filled align closely with those observations that center around
census years.
Another data problem for 1820-1855 is that measures that are available or can
be created for later periods are simply not available for this period.
Whereas certain measures can be synthesized, for others no reasonable
basis exits to generate annual observations because no data points are
available for any country for any year. Thus, for 1820-1855 no information
is available on sex-specific origin populations and none exists for the
economically active populations. However, birth and death rates for certain
countries were generated back 1800. In the sex-specific migration-rate
regressions that are reported below, half the source country population is
used to normalize the sex-specific flows.
Table 3. Means and Standard Deviations
Variable
Mean
Std.
dev.
Male immigrants (103)
Female immigrants (103)
3
Total immigrants (10 )
Single male immigrants (103)
Male immigration rate (per '000)
Female immigration rate (per '000)
Single male immigration rate (per '000)
US immigration from i, t-1 to t-5
2
Female share (10 )
Total population (106)
Birthrate (per '000)
Death rate (per '000)
Rate of natural increase, t-20
3
Per capita GDP, country i, t-1 (10 )
Per capita GDP, U.S., t-1 (103)
Relative per capita GDP, t-1
Per capita GDP difference, US-i (103)
3
Distance (10 )
English
Eng * Relative per capita GDP
Eng * Absolute per capita GDP Diff (103)
5.656
4.002
9.663
1.655
0.479
0.364
0.230
37.071
28.759
13.652
33.390
23.725
8.680
1.345
1.537
1.328
0.192
3.675
0.182
0.358
0.048
16.298
12.561
28.818
4.173
1.887
1.569
0.681
109.609
13.120
11.813
4.551
5.586
7.365
0.450
0.208
0.812
0.457
0.318
0.386
1.040
0.337
Table 4. The Rate of Emigration of Males and Females from Europe to the U.S.,1820-1855:
Hausman-Taylor Instrumental Variable Estimates and Absolute t-Ratios
Variable
Males
1
2
3
4
Differential Economic Opportunities
Relative per capita GDP, t-1 (x 1)
:
t:
1.403
(8.422)
0.070
(0.165)
Per capita GDP, US - i, t-1 (x 1)
Total migration prior 5 yrs (x 2)
Birthrate in i (x 1)
Distance from i to U.S. (z1)
Control Variables
Population of i * 0.5 (x 1)
Constant
Test for the exogeneity of the
HT instruments
0.142
(0.352)
4
1.778
(5.717)
-0.028
(0.654)
-0.015
(1.189)
-0.014
(0.333)
-0.006
(0.529)
1.416
(3.105)
-0.047
(1.047)
-0.006
(0.418)
0.431
(0.280)
0.006
(6.743)
0.027
(1.026)
0.044
(0.030)
-2.067
(1.993)
0.006
(6.682)
0.015
(0.700)
-0.536
(0.760)
8.742
(1.038)
0.010
(16.208)
0.017
(0.528)
6.665
(0.902)
8.550
(0.995)
0.010
(16.188)
0.015
(0.441)
6.524
(0.868)
0.612
(0.326)
0.005
(6.617)
0.026
(1.126)
0.380
(0.214)
-1.791
(1.301)
0.005
(6.139)
0.015
(0.715)
0.360
(0.323)
8.553
(1.048)
0.009
(16.899)
0.009
(0.322)
6.767
(0.945)
8.501
(1.015)
0.009
(16.879)
0.008
(0.276)
6.732
(0.919)
-0.027
(0.638)
0.443
(0.062)
-0.006
(0.301)
3.000
(0.584)
-0.365
(4.481)
-19.602
(0.678)
-0.357
(3.747)
-19.387
(0.651)
-0.056
(1.177)
-0.671
(0.087)
-0.007
(0.204)
2.405
(0.428)
-0.371
(5.432)
-20.128
(0.722)
-0.369
(4.586)
-20.381
(0.705)
0.624
0.582
0.177
0.201
0.856
0.964
0.140
0.160
Relative per capita GDP * English (x 2)
Costs of Migrating
English spoken in i (z2)
1.240
(8.851)
3
-2.170
(4.674)
2.097
(5.032)
-0.047
(1.044)
-0.005
(0.408)
U.S. per capita GDP, t-1 (x 1)
Natural increase, t-20 (x 1)
2
2.125
(5.628)
Source cntry per capita GDP, t-1 (x 1)
Relative growth, t-1 to t-3 (x 1)
Females
1
-0.026
(0.750)
-0.010
(0.932)
-0.016
(0.446)
-0.004
(0.379)
1.202
(2.888)
-0.042
(1.146)
-0.002
(0.180)
-1.794
(4.677)
1.769
(5.156)
-0.042
(1.142)
-0.002
(0.175)
Table 5. The Sex Composition of U.S. Immigration from Europe, 1820-1855: Hausman-Taylor Instrumental
Variable Estimates and Absolute t-Ratios for the Female Share
Variable
1
2
3
4
5
Differential Economic Opportunities
Relative per capita GDP, t-1 (x 1)
:
t:
6.213
(3.809)
23.913
(5.003)
Per capita GDP, US - i, t-1 (x 1)
16.937
(5.174)
20.612
(4.950)
Source cntry per capita GDP, t-1 (x 1)
U.S. per capita GDP, t-1 (x 1)
Relative growth, t-1 to t-3 (x 1)
Natural increase, t-20 (x 1)
-0.098
(0.248)
0.290
(2.485)
Relative per capita GDP * English (x 2)
-0.259
(0.662)
0.210
(1.799)
-19.664
(3.975)
-0.272
(0.696)
0.237
(2.047)
Per capita GDP, US - i, t-1 (x 1) * English (x 2)
Costs of Migrating
English spoken in i (z2)
Total migration prior 5 yrs (x 2)
Birthrate in i (x 1)
Distance from i to U.S. (z1)
Control Variables
Population of i (x 1)
Constant
Test for the exogeneity of the
HT instruments
-0.314
(0.803)
0.235
(2.030)
-11.108
(2.834)
20.170
(5.635)
-0.262
(0.674)
0.262
(1.958)
-9.339
(1.491)
-39.710
(0.679)
-0.017
(1.964)
-1.243
(4.321)
-38.568
(0.693)
2.662
(0.054)
-0.009
(0.986)
-1.082
(3.799)
-23.592
(0.515)
-47.349
(0.749)
-0.006
(1.010)
-1.165
(4.123)
-44.771
(0.807)
-29.083
(0.519)
-0.000
(0.068)
-1.167
(4.155)
-31.585
(0.647)
-19.191
(0.401)
-0.005
(0.952)
-0.975
(3.289)
-23.169
(0.556)
3.896
(4.940)
192.578
(0.888)
2.864
(3.700)
131.928
(0.734)
2.618
(3.793)
245.754
(1.128)
2.047
(2.795)
201.892
(1.050)
1.452
(1.986)
150.011
(0.888)
0.387
0.562
0.252
0.370
0.555
Table 6. The Rate of Emigration of Single Males from Europe to the U.S.,1820-1855:
Hausman-Taylor Instrumental Variable Estimates and Absolute t-Ratios
Variable
1
2
3
Differential Economic Opportunities
Relative per capita GDP, t-1 (x 1)
:
t:
0.344
(4.683)
4
0.223
(0.995)
Per capita GDP, US - i, t-1 (x 1)
0.557
(4.213)
0.167
(0.871)
Source cntry per capita GDP, t-1 (x 1)
U.S. per capita GDP, t-1 (x 1)
Relative growth, t-1 to t-3 (x 1)
Natural increase, t-20 (x 1)
0.003
(0.183)
-0.005
(0.976)
Relative per capita GDP * English (x 2)
0.005
(0.246)
-0.005
(0.827)
0.136
(5.127)
-0.008
(0.408)
-0.007
(1.240)
Per capita GDP, US - i, t-1 (x 1) * English (x 2)
Costs of Migrating
English spoken in i (z2)
Total migration prior 5 yrs (x 2)
Birthrate in i (x 1)
Distance from i to U.S. (z1)
Control Variables
Population of i * 0.5 (x 1)
Constant
Test for the exogeneity of the
HT instruments
5
0.001
(0.033)
-0.006
(1.069)
-0.510
(3.694)
0.584
(3.619)
-0.008
(0.447)
-0.007
(1.296)
1.220
(4.163)
0.158
(0.173)
0.002
(5.432)
0.004
(0.353)
-0.182
(0.209)
-0.139
(0.125)
0.002
(5.127)
0.004
(0.289)
-0.291
(0.298)
0.767
(1.041)
0.003
(11.992)
0.008
(0.676)
0.195
(0.309)
-1.065
(0.733)
0.002
(7.855)
0.007
(0.522)
-1.038
(0.834)
0.844
(1.199)
0.003
(11.827)
0.010
(0.781)
0.236
(0.400)
0.022
(0.924)
-0.264
(0.069)
0.030
(1.130)
0.144
(0.034)
-0.012
(0.797)
-0.179
(0.056)
0.036
(1.340)
3.583
(0.695)
-0.014
(0.975)
-0.481
(0.154)
1.145
1.068
0.664
0.154
1.054
Summary of Empirical Findings
1. Two types of variables stand out, relative per capita GDP and migration
over the prior five years. Both male and female U.S. immigrants tended to
come from relatively low income countries. With respect to the female
regressions, these findings are interpreted to mean that relatively much
migration of intact families originated in low-income countries. Respective
elasticities estimated at the means are 3.90 for males and 4.52 for females
(families).
2. The variable for the per capita GDP gap also is positive and significant.
3. When separate variables are introduced for origin countries and the U.S.,
both coefficients have the expected sign and are highly significant. Higher
U.S. per capita GDP attracted immigrants, whereas higher source-country
per-capita GDP discouraged U.S. immigration.
4. The close association between the findings for males and females is likely
the result of many males and females moving together in families and thus
responding to the same basic stimuli. Nevertheless, the absolute values on
the respective male coefficients are greater than their counterparts for
females, which is as expected since males were more likely to be economic
migrants. The interaction terms are significant, which suggests that migrants
from the U.K. were more responsive to wage gaps or to opportunities in the
U.S., which presumably reflects their ability to transfer accumulated
education and occupational skills to the U.S. labor market.
5.
6.
7.
Current immigrants had a strong tendency to follow past
immigrants. About 6 current male immigrants came to the U.S. per
1,000 persons who migrated over the past five years, whereas
about 5 females migrated per 1,000 earlier migrants.
Other than the source-country population in the last regression for
each sex, which is negative and significant, no other variable is
statistically significant in any regression of Table 4.
Malthusian population pressures do not appear to have
encouraged more emigration of either males or females during the
early years of the nineteenth century, although they were to
become more important later in the century. Higher source-country
birth rates and the associated increased costs associated with larger
family size do not appear to have discouraged family migration.
8.
9.
10.
During 1820-1855, the sex composition of U.S. immigration was
significantly less oriented toward females if the source countries
had relatively high per capita GDP compared to the U.S. In other
words, compared to males more females and thus presumably
more family migrants originated in relatively low-income source
countries.
Table 5 also provides estimates for per capita GDP broken down
by source country and the United States. These results suggest
that relative to males, females (families) were attracted by better
economic opportunities in the U.S.
The interaction term between relative per capita GDP and English
is negative and significant. Thus, relative to males, females from
Ireland and the rest of the U.K. were less responsive to wage
gaps. In the presence of the interaction term, the relative per
capita GDP variable is positive and significant, which suggests
that females and thus families from the rest of Europe were more
responsive to economic incentives than females from the U.K.
11. Higher birth rates in source countries discouraged female migration
relative to that of males, as anticipated, but current labor force
pressure at entry ages (as reflected in the rate of natural increase
lagged 20 years) encouraged more female migration. Thus, families
relative to lone males were more influenced by population pressure
during the early 19th century.
12. With the exception of the first regression of Table 5 (in which the
coefficient on lagged migration is negative and significant), past
migrants do not appear to have differentially influenced the current
movement of females relative to males.
13. English and distance are never significant.
14. Single males moved to the U.S. from relatively high-income European
source countries. During 1870-1910 single males, especially those from
Ireland and Scandinavia, were relatively quite responsive to per capita GDP
differences between the U.S. and their source countries (Greenwood,
2008). So why should single males have behaved differently during the
early 19th century? One possibility is that England, Scotland, and Wales
sent to the U.S. the most single males per million population (518.2), higher
even than Ireland (507.5), and yet had the second highest per capita GDP
gap favoring a European source country (-$435.31) after the Netherlands
(-610.64). Of course, another possibility is that single males were drawn
from the lower tail of the income distribution in their relatively high-income
source countries.