The Great Migration in Black and White

THE JOURNAL OF ECONOMIC HISTORY
VOLUME 75
DECEMBER 2015
NUMBER 4
The Great Migration in Black and White:
New Evidence on the Selection and
Sorting of Southern Migrants
WILLIAM J. COLLINS AND MARIANNE H. WANAMAKER
We construct datasets of linked census records to study internal migrants’ selection
DQGGHVWLQDWLRQFKRLFHVGXULQJWKH¿UVWGHFDGHVRIWKH³*UHDW0LJUDWLRQ´±
1930). We study both whites and blacks and intra- and inter-regional migration.
While there is some evidence of positive selection, the degree of selection was
small and participation in migration was widespread. Differences in background,
including initial location, cannot account for racial differences in destination
choices. Blacks and whites were similarly responsive to pre-existing migrant
stocks from their home state, but black men were more deterred by distance,
attracted to manufacturing, and responsive to labor demand.
A
t the turn of the twentieth century, real income per worker in the
South was less than one-half of that in the rest of the United States
(DVWHUOLQ0LWFKHQHUDQG0F/HDQ:KHQ::,OHGWRERWK
a labor demand boom in northern industrial centers and the interruption
The Journal of Economic History9RO1R'HFHPEHU7KH(FRQRPLF+LVWRU\
Association. All rights reserved. doi: 10.1017/S0022050715001527
:LOOLDP - &ROOLQV LV WKH 7HUHQFH ( $GGHUOH\ -U 3URIHVVRU RI (FRQRPLFV 'HSDUWPHQW RI
(FRQRPLFV %R[ % 9DQGHUELOW 8QLYHUVLW\ 1DVKYLOOH 71 (PDLO ZLOOLDP
FROOLQV#YDQGHUELOWHGX 0DULDQQH + :DQDPDNHU LV $VVLVWDQW 3URIHVVRU RI (FRQRPLFV 6WRNHO\ 0DQDJHPHQW &HQWHU 7KH 8QLYHUVLW\ RI 7HQQHVVHH .QR[YLOOH 71 (PDLO
[email protected].
7KHDXWKRUVDUHJUDWHIXOIRUVXJJHVWLRQVIURP-HUHP\$WDFN5LFKDUG%DNHU+R\W%OHDNOH\
-RKQ%URZQ%UDQW&DOODZD\.LWW&DUSHQWHU$QGUHZ'XVWDQ7D\ORU-DZRUNVL7RQJ/L5REHUW
0DUJR 3HWHU 0F+HQU\ $OHMDQGUR 0ROQDU 6XUHVK 1DLGX $ELJDLO :R]QLDN WZR DQRQ\PRXV
UHIHUHHVDQGVHPLQDUSDUWLFLSDQWVDWWKH8QLYHUVLW\RI*HRUJLD8QLYHUVLW\RI&DOLIRUQLDDW'DYLV
6WDQIRUG8QLYHUVLW\9DQGHUELOW8QLYHUVLW\WKH$66$0HHWLQJVWKH:RUOG&RQJUHVVRI
&OLRPHWULFV DQG WKH (FRQRPLF +LVWRU\ $VVRFLDWLRQ <H *X -XVWLQ 5RXVK DQG
6KHD 6DELQ SURYLGHG H[FHOOHQW UHVHDUFK DVVLVWDQFH 7KH DXWKRUV WKDQN 5LFKDUG +RUQEHFN DQG
'DYH 'RQDOGVRQ IRU VKDULQJ WUDQVSRUW FRVW GDWD 7KH *UH\ DQG 'RUQEXVK JLIWV DW 9DQGHUELOW
8QLYHUVLW\DQG2I¿FHRI5HVHDUFKDWWKH8QLYHUVLW\RI7HQQHVVHHKDYHSURYLGHGUHVHDUFKVXSSRUW
16)VXSSRUW6(6DQGLVJUDWHIXOO\DFNQRZOHGJHG
947
Downloaded from https:/www.cambridge.org/core. The University of Tennessee - Knoxville, on 17 Mar 2017 at 13:09:33, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/S0022050715001527
948
Collins and Wanamaker
of immigration from Europe, southern workers moved away from their
KRPHUHJLRQDWKLJKUDWHV7KLVZDVWKHVWDUWRIWKH³*UHDW0LJUDWLRQ´
ZKLFK ZDQHG GXULQJ WKH *UHDW 'HSUHVVLRQ EXW VXUJHG DJDLQ EHWZHHQ
DQG%\SHUFHQWRIVRXWKHUQERUQEODFNPHQDQG
SHUFHQWRIVRXWKHUQERUQZKLWHPHQDJHDQGRYHUZHUHOLYLQJRXWVLGH
of the South.
7KH *UHDW 0LJUDWLRQ ZDV DQ LPSRUWDQW HYHQW LQ WKH KLVWRU\ RI 86
ODERUPDUNHWLQWHJUDWLRQDQGLWKDVKDGORQJODVWLQJVRFLDOUDPL¿FDWLRQV
Social scientists have studied its causes and consequences for almost 100
\HDUVLQWHUDOLD6FURJJV86'HSDUWPHQWRI/DERU/HZLV
DQGUHFHQWO\+RUQEHFNDQG1DLGX&ROOLQVDQG:DQDPDNHU
%ODFNHWDO6FKRODUVKDYHIRFXVHGSULPDULO\RQWKHLQWHU
UHJLRQDO PRYHPHQW RI $IULFDQ $PHULFDQV ,QGHHG WKH WHUP ³*UHDW
0LJUDWLRQ´KDVWUDGLWLRQDOO\EHHQDSSOLHGVROHO\WREODFNPLJUDWLRQWRWKH
1RUWK,QWKLVDUWLFOHZHH[SDQGWKHIUDPHRILQYHVWLJDWLRQLQWZRZD\VWR
provide additional perspective. First, we examine the migration decisions
RIERWKZKLWHDQGEODFNPHQ7KHPDVVPLJUDWLRQRIZKLWHVRXWKHUQHUV
was important in its own right and provides a natural comparison for the
migration patterns of blacks. Second, our analysis includes both intraand inter-regional migrants, whereas much of the previous literature has
IRFXVHG HQWLUHO\ RQ WKH ODWWHU :H ¿QG WKDW GHVWLQDWLRQ FKRLFHV within
regions, including the South, provide valuable information and a more
complete picture of internal migration patterns during the early decades
RIWKH*UHDW0LJUDWLRQ
,QDGGLWLRQWRH[SDQGLQJWKHIUDPHRILQYHVWLJDWLRQZHGHYHORSQHZ
data that provide deeper insight than previously available into the careers
RI VRXWKHUQ PHQ EHWZHHQ DQG 7KH FURVVVHFWLRQDO GDWDVHWV
that inform most quantitative studies of U.S. internal migration in the
HDUO\WZHQWLHWKFHQWXU\KDYHDPDMRUOLPLWDWLRQUHVHDUFKHUVVLPSO\FDQQRW
observe the same person before and after migration.17KHDEVHQFHRIH[
ante information hinders the study of how individual and local characterLVWLFVLQÀXHQFHGERWKVHOHFWLRQLQWRPLJUDWLRQDQGWKHPLJUDQWV¶FKRLFHVRI
destination. Furthermore, using ex post measures of migrants’ outcomes
or human capital that are available in cross-sectional data sources may
JLYHPLVOHDGLQJLPSUHVVLRQVRIWKHLUSUHPLJUDWLRQVWDWXV7RRYHUFRPH
WKLVSUREOHPZHFUHDWHOLQNHGFHQVXVUHFRUGVIRUPRUHWKDQPHQ
providing a clearer view of the same men before and after the start of
1
6HH %RGQDU 6LPRQ DQG :HEHU &ROOLQV 0DORQH\ /RJDQ &ROOLQV DQG :DQDPDNHU DQG %ODFN HW DO IRU H[DPSOHV RI UHVHDUFK LQ ZKLFK
migrants are observed before and after moving.
Downloaded from https:/www.cambridge.org/core. The University of Tennessee - Knoxville, on 17 Mar 2017 at 13:09:33, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/S0022050715001527
Selection and Sorting of Southern Migrants
949
WKH*UHDW0LJUDWLRQ:HEXLOGWKHGDWDVHWE\VWDUWLQJZLWKDVDPSOHRI
southern-resident males, ages 0 to 40, in 1910. We then locate the same
men in the 1930 census manuscripts and transcribe data from the handZULWWHQGRFXPHQWV,QWKH\RXQJHUPDOHVLQWKHVDPSOHJHQHUDOO\
less than age 18) still lived with their parents and siblings, and the older
PDOHVZHUHDOUHDG\LQWKHVRXWKHUQODERUIRUFH,QERWKFDVHVIRU\RXQJHU
DQGROGHUPHQWKHGDWDVHWFRQWDLQVYDOXDEOHSUH*UHDW0LJUDWLRQLQIRUmation on personal, household, and local background.
For African Americans, the linked census records used here are the
VDPHDVLQ:LOOLDP-&ROOLQVDQG0DULDQQH+:DQDPDNHUZKLFK
focused on measuring black men’s income gains from inter-regional
PLJUDWLRQDIWHUDFFRXQWLQJIRUVHOHFWLRQ7KLVDUWLFOHJRHVEH\RQG&ROOLQV
DQG:DQDPDNHULQVHYHUDOZD\V,WVWXGLHVWKHVHOHFWLRQRIERWK
white and black migrants, which required the creation of a new set of
OLQNHGFHQVXVUHFRUGVIRUPRUHWKDQVRXWKHUQZKLWHPHQ,QDGGLtion, this article studies both intra-regional and inter-regional migration
SDWWHUQVZKHUHDV&ROOLQVDQG:DQDPDNHULJQRUHGLQWUDUHJLRQDO
PRYHPHQW ,QWUDUHJLRQDO ÀRZV ZHUH ODUJH LQ WKLV SHULRG DQG ZRUWK\
of scholars’ attention. Finally, as described later, much of this article is
dedicated to studying the migrants’ choices of destination and comparing
black and white migration patterns across potential destinations, a topic
WKDWLVQRWDGGUHVVHGLQ&ROOLQVDQG:DQDPDNHU
7KHOLQNHGFHQVXVUHFRUGVDOORZXVWRDGGUHVVVHYHUDOLPSRUWDQWWRSLFV
in the history of Americans’ internal migration. First, after documenting
the outstanding features of southern black and white migration patterns
and migrant characteristics, we investigate whether the migrants were
VWURQJO\ VHOHFWHG RQ WKH EDVLV RI WKHLU SUH*UHDW 0LJUDWLRQ FKDUDFWHUistics. Second, we examine how southern migrants sorted themselves
across potential destinations and the extent to which personal characteristics, such as place of origin and family background, account for blackZKLWHGLIIHUHQFHVLQPLJUDWLRQSDWWHUQV7KLUGZHHVWLPDWHWKHPLJUDQWV¶
responsiveness to variation in labor market opportunities and migration
costs across potential destinations, paying particular attention to racial
differences in behavior.
:H ¿QG WKDW VRXWKHUQ PHQ¶V SDUWLFLSDWLRQ LQ LQWHUVWDWH DQG LQWHU
regional migration was widespread in the sense that the migrants’ background characteristics were not much different from the non-migrants’
FKDUDFWHULVWLFV 7KHUH LV HYLGHQFH FRQVLVWHQW ZLWK D GHJUHH RI SRVLWLYH
selection into inter-state migration among both whites and blacks, as
PHDVXUHGE\LQGLFDWRUVRIMREVWDWXV,WLVDOVRFOHDUWKDWIDUPUHVLGHQWV
LQZHUHOHVVOLNHO\WRPRYHWKDQQRQIDUPUHVLGHQWV<HWRYHUDOOWKH
Downloaded from https:/www.cambridge.org/core. The University of Tennessee - Knoxville, on 17 Mar 2017 at 13:09:33, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/S0022050715001527
950
Collins and Wanamaker
differences between migrants and non-migrants were small within race
categories.
,Q VWXG\LQJ WKH SDWWHUQV RI LQWHUVWDWH PLJUDWLRQ ZH VHH WKDW WKHUH
was some overlap in the most popular destinations for white and black
migrants, but there were also notable differences. Approximately 28
percent of inter-state migrants would have to change their destination
WRHTXDOL]HWKHZKLWHDQGEODFNGLVWULEXWLRQVRYHUVWDWHV'LIIHUHQFHVLQ
the men’s background characteristics can account for surprisingly little
of the overall black-white differences in destination choice, which leads
us to study differences in responsiveness to economic variables across
SRWHQWLDOGHVWLQDWLRQV:H¿QGWKDWEODFNDQGZKLWHPHQZHUHVLPLODUO\
responsive to pre-existing distributions of state-to-state migrant stocks,
but that black men were more deterred by distance, more attracted to
manufacturing centers, and more responsive to cross-state variation in
DJJUHJDWH ODERU GHPDQG JURZWK FHWHULV SDULEXV $ WKHPH LQ WKH *UHDW
0LJUDWLRQOLWHUDWXUHHPSKDVL]HVUDFLDORSSUHVVLRQDVDVWURQJLQGHSHQGHQWPRWLYDWLRQIRUOHDYLQJWKH6RXWK:H¿QGRQO\PL[HGHYLGHQFHWKDW
black migrants were more inclined to leave the South than white migrants
and no evidence that black migrants moved to non-southern locations
more frequently than they did southern ones, conditional on the states’
HFRQRPLFFKDUDFWHULVWLFV7KHUHLVVWURQJHUHYLGHQFHKRZHYHUWKDWEODFN
migrants moved more frequently than southern white migrants to the
Northeast and Midwest, whereas southern whites moved more frequently
to the West, conditional on the states’ economic characteristics.
%$&.*5281'216287+(510,*5$7,21
'HVSLWH WKH VL]DEOH UHJLRQDO GLIIHUHQFHV LQ DYHUDJH LQFRPH FLWHG
HDUOLHU UHODWLYHO\ IHZ VRXWKHUQERUQ PHQ OHIW WKH UHJLRQ EHIRUH ::,
)RUSHUVSHFWLYHWKHFKDUWVDWWKHWRSRI)LJXUHXVHWKH,3806FURVV
VHFWLRQDOGDWDIRUHDFKFHQVXVIURPWR5XJJOHVHWDO
to depict the cumulative inter-regional migration rate for men born in the
6RXWKEHWZHHQDQGLHWKHVKDUHRIVRXWKHUQERUQPHQLQ
each ten-year birth cohort who resided outside the South at each census
GDWH 7KLV VSDQV DOO WKH ELUWK FRKRUWV RI PHQ LQ RXU VDPSOH RI OLQNHG
FHQVXV PDQXVFULSWV WR ZKLFK LV GHVFULEHG LQ GHWDLO LQ WKH
article’s next section, as well as those born up to ten years before and
DIWHUWKHPHQLQWKHOLQNHGVDPSOH7KHERWWRPSDQHORI)LJXUHSORWV
cumulative interstate migration withinWKH6RXWKWKHVKDUHRIVRXWKHUQ
ERUQPHQUHVLGLQJRXWVLGHWKHLUVWDWHRIELUWKEXWLQWKH6RXWK7KHUH
LV QR ,3806 VDPSOH IRU DQG ZH VLPSO\ FRQQHFW WR Downloaded from https:/www.cambridge.org/core. The University of Tennessee - Knoxville, on 17 Mar 2017 at 13:09:33, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/S0022050715001527
Selection and Sorting of Southern Migrants
951
White, Out of South
Black, Out of South
0
1880 1890 1900 1910 1920 1930 1940 1950 1960
1880 1890 1900 1910 1920 1930 1940 1950 1960
.4
.3
.2
.1
Black, Within South
White, Within South
.4
.3
.2
.1
0
1880 1890 1900 1910 1920 1930 1940 1950 1960
Birth cohorts
1860s
1890s
1880 1890 1900 1910 1920 1930 1940 1950 1960
1870s
1900s
1880s
1910s
FIGURE 1
,17(5$1',175$5(*,21$/0,*5$7,212)6287+(51%2510(1%<5$&(
$1'%,57+&2+25772
Notes7KH¿JXUHSORWVWKHVKDUHRIVRXWKHUQERUQPHQLQHDFKFRKRUWZKRUHVLGHRXWVLGHWKHLU
VWDWHRIELUWKHLWKHU³2XWRI6RXWK´RU³:LWKLQ6RXWK´7KH6RXWKLVGH¿QHGDVHOVHZKHUHLQWKH
DUWLFOHH[FOXGLQJ'HODZDUH:DVKLQJWRQ'&DQG0DU\ODQG&RKRUWVDUHREVHUYHGDWPRVWVL[
WLPHVFRUUHVSRQGLQJWRDJHV±±DQGVRRQ<HDURIELUWKLVGH¿QHGDV\HDURIREVHUYDWLRQ
PLQXVDJH7KHVELUWKFRKRUWLV¿UVWSORWWHGIRUDWDJHV±9HUWLFDOOLQHVDW
1910 and 1930 indicate the timeframe examined using the linked dataset.
Sources 3ORWV DUH EDVHG RQ VWDWHRIELUWK DQG VWDWHRIUHVLGHQFH IURP WKH ± ,3806
FURVVVHFWLRQV5XJJOHVHWDO7KHUHLVQRVDPSOHIRU
Downloaded from https:/www.cambridge.org/core. The University of Tennessee - Knoxville, on 17 Mar 2017 at 13:09:33, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/S0022050715001527
952
Collins and Wanamaker
IRU WKH UHOHYDQW FRKRUWV 7KH YHUWLFDO OLQHV DW DQG FRUUHVSRQG WR WKLV DUWLFOH¶V PDLQ ZLQGRZ RI DQDO\VLV UHÀHFWLQJ WKH VWUXFWXUHRIWKHGDWDVHWRIOLQNHGFHQVXVUHFRUGVDQGWKH¿UVWGHFDGHVRIWKH
*UHDW 0LJUDWLRQ 7KURXJKRXW WKH DUWLFOH WKH ³VRXWKHUQ´ VWDWHVLQFOXGH
$ODEDPD$UNDQVDV)ORULGD*HRUJLD.HQWXFN\/RXLVLDQD0LVVLVVLSSL
1RUWK&DUROLQD2NODKRPD6RXWK&DUROLQD7HQQHVVHH7H[DV9LUJLQLD
DQG:HVW9LUJLQLD:HFDWHJRUL]H'HODZDUH0DU\ODQGDQGWKH'LVWULFW
RI&ROXPELDDV³QRQVRXWKHUQ´
,QSULRUWRWKHRQVHWRIWKH*UHDW0LJUDWLRQRQO\SHUFHQWRI
EODFNDQGZKLWHVRXWKHUQERUQPDOHVLQWKHLUVWKHVELUWKFRKRUW
resided outside the South. For whites, this is approximately the same level
RIFXPXODWLYHLQWHUUHJLRQDOPLJUDWLRQDVREVHUYHGWHQ\HDUVHDUOLHULQ
IRUPHQDWDVLPLODUSRLQWLQWKHOLIHF\FOHWKHVELUWKFRKRUW
Later cohorts of whites undertook substantially more inter-regional
migration: 15 percent of the 1890s birth cohort had left the South by
DQGFORVHUWRSHUFHQWKDGOHIWDPRQJWKRVHH[SRVHGWRWKH::,,
ERRPREVHUYHGLQ)RUEODFNVWKHVELUWKFRKRUWKDGDVOLJKWO\
KLJKHUUDWHRIFXPXODWLYHLQWHUUHJLRQDOPLJUDWLRQE\WKDQWKHV
FRKRUWREVHUYHGLQFRPSDUHGWRSHUFHQW7KDWGLIIHUHQFHLV
dwarfed by subsequent changes in black inter-regional migration rates,
with nearly one-quarter of the 1890s birth cohort leaving the South by
DQGHYHQKLJKHUUDWHVIRUFRKRUWVH[SRVHGWR::,,HJSHUFHQW
of the 1910s birth cohort observed in 1950).
7KH FRPSDUDWLYHO\ ORZ UDWH RI LQWHUUHJLRQDO PLJUDWLRQ EHIRUH does not imply that the southern labor force was stationary, as there is
considerable evidence of mobility withinWKH6RXWK:ULJKW7KH
lower panels of Figure 1 indicate that 15 percent of blacks and nearly
20 percent of whites in the 1870s birth cohort had moved away from
WKHLUELUWKVWDWHE\EXWVWD\HGZLWKLQWKH6RXWK,QWKLVVHQVHLQWUD
regional mobility was more common than inter-regional mobility prior
WR,QSDUWLFXODUWKHUHZHUHVL]DEOHÀRZVRIVRXWKHUQERUQZKLWHV
LQWR7H[DVDQG2NODKRPDDQGVL]DEOHÀRZVRIVRXWKHUQERUQEODFNVLQWR
Florida, Arkansas, and Mississippi.
$IWHU ÀRZV ZLWKLQ WKH 6RXWK FRQWLQXHG HVSHFLDOO\ WR 7H[DV
2NODKRPD DQG )ORULGD EXW GHVWLQDWLRQV OLNH &DOLIRUQLD 2KLR,OOLQRLV
3HQQV\OYDQLDDQG1HZ<RUNEHFDPHPXFKPRUHSURPLQHQWWKDQSUHYLously. Between 1910 and 1930, Figure 1 shows slight declines in within6RXWK PLJUDWLRQ EXW VLJQL¿FDQW LQFUHDVHV LQ LQWHUUHJLRQDO PLJUDWLRQ
UHÀHFWLQJDVKLIWLQWKHUHODWLYHDWWUDFWLRQRIQRQVRXWKHUQUHVLGHQFHIRU
both blacks and whites.
Downloaded from https:/www.cambridge.org/core. The University of Tennessee - Knoxville, on 17 Mar 2017 at 13:09:33, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/S0022050715001527
Selection and Sorting of Southern Migrants
953
Scholars have suggested several reasons for the relatively low rates of
$IULFDQ $PHULFDQ PLJUDWLRQ IURP WKH 6RXWK SULRU WR ::, ORZ OHYHOV
RIKXPDQDQG¿QDQFLDOFDSLWDOZKLFKKDYHEHHQIRXQGWRLPSHGHORQJ
GLVWDQFH PLJUDWLRQ LQ WKLV DQG RWKHU VHWWLQJV /RJDQ +DWWRQ DQG
:LOOLDPVRQ GLVFULPLQDWLRQ E\ QRUWKHUQ HPSOR\HUV ZKR UHDGLO\
KLUHGLPPLJUDQWODERUHUVIURP(XURSH0\UGDO&ROOLQVDQG
weak integration of northern and southern labor markets compared to the
strong ties between northern and European labor markets, a legacy of mass
PLJUDWLRQIURP(XURSHWKDWEHJDQLQWKHV7KRPDV:ULJKW
+DWWRQDQG:LOOLDPVRQ5RVHQEORRP,PSHGLPHQWVWR
WKHPRELOLW\RIVRXWKHUQDJULFXOWXUDOZRUNHUVDOVRPD\KDYHEHHQVLJQL¿FDQW5DQVRPDQG6XWFK*URVVPDQ1DLGX
'HVSLWH VL]DEOH LQWHUVWDWH ÀRZV RI VRXWKHUQ ZKLWHV WKH KLVWRU\ RI
their internal migration has been far less explored than that of African
$PHULFDQV$NHUV.LOOLDQ%HUU\*UHJRU\2I
course, some of the same factors that inhibited blacks’ movement prior
WR::,DOVRPD\KDYHDIIHFWHGVRXWKHUQZKLWHV%XWLWVHHPVFOHDUWKDW
the black and white stories differ in important ways. Most obviously,
whites were not recently removed from slavery, were less concentrated
than blacks in the Cotton Belt, were more likely than blacks to have
acquired some wealth and literacy, and likely faced less discrimination
in distant labor markets. For perspective, in 1870 nearly 40 percent of
ZKLWHVRXWKHUQPHQDJHWRRZQHGVRPHDPRXQWRIUHDOSURSHUW\
compared to less than 5 percent of blacks. Approximately 74 percent
RI VRXWKHUQ ZKLWHV RYHU DJH FRXOG UHDG DQG ZULWH FRPSDUHG WR SHUFHQWRIEODFNV5XJJOHVHWDO+RZHYHUZKLOHVRXWKHUQZKLWHV
might have found it easier to afford long-distance moves than blacks
RQDYHUDJHWKH\DOVRPLJKWKDYHIRXQGLWOHVVDWWUDFWLYH)RULQVWDQFH
opportunities to advance in southern labor markets, whether by ascending
the agricultural ladder or moving into skilled non-agricultural work, may
KDYHVHHPHGPRUHSOHQWLIXOWRZKLWHVWKDQWREODFNV,WLVDOVRSRVVLEOH
WKDWOLQJHULQJKRVWLOLW\IURPWKH&LYLO:DULQÀXHQFHGZKLWHVRXWKHUQHUV¶
perceptions of the North.
,I LJQRUDQFH DQG SRYHUW\ LQKLELWHG ORQJGLVWDQFH PLJUDWLRQ E\ EODFN
and white southerners, this constraint may have receded with each generation’s educational and economic advances in the late nineteenth and
HDUO\WZHQWLHWKFHQWXULHV+LJJV0DUJR&ROOLQVDQG
0DUJR,QDGGLWLRQLPSURYLQJWUDQVSRUWDWLRQDQGFRPPXQLFDWLRQ
networks in the South may have facilitated migration by lowering the
associated costs and uncertainties. Surfaced roads more than doubled
Downloaded from https:/www.cambridge.org/core. The University of Tennessee - Knoxville, on 17 Mar 2017 at 13:09:33, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/S0022050715001527
954
Collins and Wanamaker
EHWZHHQ DQG LQ WKH 6RXWK 86 'HSDUWPHQW RI $JULFXOWXUH
1917). Nearly all southerners lived in counties with railroad access by
$WDFN DQG SHUVRQDO FRPPXQLFDWLRQ 7KH FLUFXODWLRQ RI
northern newspapers, such as the Chicago Defender, increased in the
6RXWK GXULQJ WKH V *URVVPDQ (QYLURQPHQWDO VKRFNV DOVR
PD\ KDYH GULYHQ VRPH PLJUDWLRQ +LJJV /DQJH 2OPVWHDG DQG
5KRGH %RXVWDQ .DKQ DQG 5KRGH +RUQEHFN DQG 1DLGX
0RVW SURPLQHQWO\ WKH EROO ZHHYLO VSUHDG IURP 7H[DV ZKHUH LW
gained a foothold in the 1890s, to Mississippi by 1907 and North Carolina
E\ 7KH EROO ZHHYLO GLVUXSWHG FRWWRQ SURGXFWLRQ DQG LPSDUWHG D
long-lasting negative productivity shock, perhaps making southern agriculture a less attractive option than before, at least within the Cotton Belt
+LJJV/DQJH2OPVWHDGDQG5KRGH%DNHU)LQDOO\
VSHFL¿FDOO\ IRU $IULFDQ $PHULFDQV SROLWLFDO GLVHQIUDQFKLVHPHQW PRE
violence, de jure segregation, and, in general, the ascendance of the Jim
Crow regime may have provided a strong incentive to leave the region
0\UGDO:RRGZDUG.RXVVHU7ROQD\DQG%HFN
0DUJR7DNHQWRJHWKHUWKHVHWUHQGVPD\KDYH\LHOGHGDVRXWKHUQ
labor force circa 1910 that was more able and more inclined to migrate
long distances than ever before.
$JDLQVWWKLVEDFNGURSWKHH[RJHQRXVVKRFNRI::,FUHDWHGERWKD
PDMRU ODERU GHPDQG ERRP LQ LQGXVWULDO FHQWHUV SUHGRPLQDQWO\ ORFDWHG
in the North, and a temporary halt to European immigration, which was
later reinforced by immigration restrictions. Many industrial employers
UHFUXLWHGVRXWKHUQPLJUDQWVIRUWKH¿UVWWLPHJDLQHGH[SHULHQFHLQKLULQJ
training, and evaluating them, and established networks to draw on the
VRXWKHUQODERUVXSSO\:KDWOH\%HUU\)RRWH:KDWOH\DQG
Wright 2003). As is commonly found in studies of migration, networks
of previous migrants helped perpetuate migration patterns in and from
WKH6RXWK&DUULQJWRQ'HWUDJLDFKHDQG9LVKZDQDWK+DWWRQDQG
:LOOLDPVRQ 0F.HQ]LH DQG 5DSRSRUW &RQVHTXHQWO\ KLJK
UDWHVRIPLJUDWLRQFRQWLQXHGORQJDIWHUWKHLPSHWXVRI::,ZLWKPDMRU
repercussions for American economic and social history.
Within this historical setting, our thinking about migration and location
FKRLFHLVJXLGHGE\WKHRUHWLFDODQGHPSLULFDOZRUNE\/DUU\$6MDDVWDG
-HQQLIHU5REDFN*HRUJH-%RUMDV6WHSKHQ*%URQDUV
DQG 6WHSKHQ - 7UHMR -HIIUH\ *URJJHU DQG *RUGRQ + +DQVRQ
DQG(QULFR0RUHWWL:HVXSSRVHWKDWDQLQGLYLGXDO¶VORFDtion decision depends on expected income, amenities, and relocation
FRVWV EURDGO\ GH¿QHG ZKLOH UHFRJQL]LQJ WKDW WKHVH H[SHFWDWLRQV PD\
vary across types of workers, by race, skill, or other initial conditions.
Downloaded from https:/www.cambridge.org/core. The University of Tennessee - Knoxville, on 17 Mar 2017 at 13:09:33, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/S0022050715001527
Selection and Sorting of Southern Migrants
955
7KHVHGLIIHUHQFHVDFURVVZRUNHUVPD\JLYHULVHWRLQWHUHVWLQJSDWWHUQVRI
selection into inter-state migration and of sorting across potential destinations, which we explore later. For both selection and sorting, the dataset
of linked census records provides new opportunities for analysis.
1(:'$7$/,1.,1*&(16865(&25'6±
7RFRQVWUXFWWKHOLQNHGGDWDVHWZHEHJDQZLWKWKH,3806VDPSOHRI
manuscript data from the 1910 Census of Population and selected all
VRXWKHUQUHVLGHQW PDOHV EHWZHHQ WKH DJHV RI DQG 5XJJOHV HW DO
:HWKHQDWWHPSWHGWROLQNWKHVHPHQ³IRUZDUGÉ\ORFDWLQJWKHPLQ
the handwritten manuscripts of the 1930 Census of Population. We used
each individual’s name, place of birth, and age from the 1910 records as
search criteria for location in the 1930 records.2 From an initial sample of
LQGLYLGXDOVWKHOLQNLQJSURFHVVVXFFHVVIXOO\ORFDWHGLQGLviduals, a 24 percent match rate.37KH¿QDOVDPSOHLVSHUFHQWEODFN
Additional details on the linking process and the variables available in
each census year are provided in the Appendix.
As mentioned above, the linked data offer several advantages relative to the state-level aggregates or micro-level cross-sections that have
W\SLFDOO\VXSSRUWHGTXDQWLWDWLYHVWXGLHVRIWKH*UHDW0LJUDWLRQ7KHNH\
distinction is that we observe the same person both before and after the
RQVHWRIWKH*UHDW0LJUDWLRQ¶V¿UVWELJZDYH,QZHVHHWKH\RXQJHU
PHQLQRXUVDPSOHZKHQWKH\VWLOOUHVLGHGZLWKWKHLUIDPLO\7KHUHIRUH
we observe many characteristics about the household in which they grew
up—what their parents did for a living, where they were located, whether
WKH\DWWHQGHGVFKRROWKHLUOLWHUDF\LIROGHUWKDQZKHWKHUWKH\OLYHG
RQDIDUPDQGVRRQ7KHROGHUPHQLQRXUVDPSOHDQGROGHULQ
are observed after they have left their parents, but also after they have
HQWHUHGWKHVRXWKHUQODERUIRUFH7KHUHIRUHZHNQRZZKDWNLQGRIMRE
they held in 1910 in addition to whether they were literate, where they
OLYHGZKHWKHUWKH\RZQHGDKRPHDQGVRRQ,QZHVHHZKHWKHUWKH
PHQKDYHPRYHGVLQFHDQGLIVRWRZKHUH*LYHQWKHFRXQW\
2
2XUVHDUFKFULWHULDLQFOXGHD6281'(;YHUVLRQRIWKHODVWQDPH¿UVWWKUHHOHWWHUVRIWKH
¿UVWQDPHVWDWHRIELUWKDQGELUWK\HDUZLWKLQWZR\HDUV6281'(;LVDQDOJRULWKPXVHGWR
generate alternative spellings of a surname. SOUNDEX matches include the exact last name and
reasonably close approximations to that name.
3
2XUOLQNDJHUDWHVDUHVLPLODUWRWKRVHLQ/RQJDQG)HUULHDQG$EUDPLW]N\%RXVWDQ
DQG(ULNVVRQ:HGHOHWHGFDVHVZKHUHPRUHWKDQRQHLQGLYLGXDOLQPDWFKHGWRWKH
VDPHLQGLYLGXDOLQ0DWFKUDWHVIRUEODFNVSHUFHQWDQGZKLWHVSHUFHQWGLIIHUZKLFK
ZHEHOLHYHLVSDUWLDOO\DWWULEXWDEOHWRDGYDQFHVLQVHDUFKWHFKQRORJ\RYHUWLPH7KHVDPSOHVZHUH
matched sequentially.
Downloaded from https:/www.cambridge.org/core. The University of Tennessee - Knoxville, on 17 Mar 2017 at 13:09:33, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/S0022050715001527
Collins and Wanamaker
DQGVWDWHLGHQWL¿HUVLWLVVWUDLJKWIRUZDUGWRPHUJHWKHOLQNHGGDWDVHWZLWK
FRXQW\ DQG VWDWHOHYHO LQIRUPDWLRQ IURP 0LFKDHO +DLQHV ZKLFK KHOSV FKDUDFWHUL]H HDFK LQGLYLGXDO¶V ORFDO HFRQRPLF DQG VRFLDO
environment.
$PDMRUFRQFHUQIRUDQ\VXFKGDWDVHWLVWKDWWKHOLQNHGVDPSOHPLJKW
exhibit selection bias that interferes with subsequent analyses and interpretations. Fortunately, the men in the linked sample are similar to
WKRVH LQ WKH UDQGRPO\ VHOHFWHG EDVH VDPSOH IURP 7DEOH VHSDrately reports the summary statistics for the linked and base samples
of blacks and whites. Although there are some small differences, their
VWDWLVWLFDO DQG HFRQRPLF VLJQL¿FDQFH LV OLPLWHG 6LPLODU FRQFOXVLRQV
follow when we estimate the probability of being in the linked sample
as a function of observable characteristics, conditional on being in the
EDVH,3806VDPSOH)URPWKDWH[HUFLVHGHWDLOVRIZKLFKDUHGLVFXVVHG
in the Appendix), we conclude that literacy, farming occupations, and
:HVW 9LUJLQLD UHVLGHQFH DUH VWDWLVWLFDOO\ VLJQL¿FDQW SUHGLFWRUV RI EHLQJ
IRXQGLQWKHFHQVXVPDQXVFULSWVEXWWKHSUDFWLFDOVLJQL¿FDQFHRI
these increased probabilities is small, generally less than a 3 percent
increase.4
$ VHSDUDWH FKHFN ZLWK WKH ,3806 FURVVVHFWLRQDO VDPSOH RI
VRXWKHUQERUQEODFNPHQDJHWRWRFRUUHVSRQGWRWKRVHWR
in 1910), reveals that 22 percent resided outside the South at the time of
WKHFHQVXV7KLVLVFORVHWRWKHSHUFHQWRIRXUPDWFKHGVDPSOH
who resided in the South in 1910 but not in 1930. We do not expect
these numbers to be exactly the same because of interregional migration
RI PHQ SULRU WR LQ WKH ,3806 FURVVVHFWLRQ ZKHUH PLJUDQWV DUH
GH¿QHG XVLQJ SODFH RI ELUWK LQIRUPDWLRQ DQG VDPSOLQJ YDULDWLRQ 7KH
corresponding numbers for southern-born whites observed outside the
South in 1930 are 15 percent and 17 percent. Across all states, the distribution of inter-state migrants in the linked sample is highly correlated
ZLWKLQWHUVWDWHPLJUDQWVLQWKH,3806VDPSOHIRUERWKWKH
white and black samples), and rarely deviates by more than 1 percentage
SRLQW 7KH FRUUHODWLRQ LV HVSHFLDOO\ KLJK IRU QRQVRXWKHUQ GHVWLQDWLRQV
where pre-1910 migration is less likely to confound the comparisons
$SSHQGL[7DEOH2YHUDOOWKHUHVXOWVGRQRWVXJJHVWWKDWWKHOLQNHG
sample is biased in a way that will confound our analyses, and we take
the linked sample to be fairly representative.
We evaluated the results’ sensitivity to using stricter match criteria by restricting the sample
WRH[DFWPDWFKHVRQODVWQDPHDQGFORVHPDWFKHVRQ¿UVWQDPH-DUR:LQNOHUGLVWDQFH7KH
main results are unchanged. See the Online Appendix.
4
Downloaded from https:/www.cambridge.org/core. The University of Tennessee - Knoxville, on 17 Mar 2017 at 13:09:33, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/S0022050715001527
957
Selection and Sorting of Southern Migrants
TABLE 1
&203$5,6212)/,1.('$1'%$6(6$03/(&+$5$&7(5,67,&6
6287+(510$/(6
White Males
Matched
Sample
Full
,3806
Sample
Black Males
p-value
of Diff.
Matched
Sample
Full
,3806
Sample
p-value
of Diff.
10.3
0.23
3DQHO$'LVWULEXWLRQRIVWDWHRIUHVLGHQFHSHUFHQWDJHV
Alabama
Arkansas
1.00
9.8
0.09
5.0
4.9
0.61
Florida
2.1
2.3
0.09
4.0
3.8
0.43
*HRUJLD
7.5
7.5
0.99
13.5
.HQWXFN\
10.8
10.4
0.14
3.2
3.0
0.44
Louisiana
4.7
5.2
0.01
8.2
8.7
0.26
Mississippi
4.2
4.1
0.74
12.9
North Carolina
7.9
0.14
8.9
Oklahoma
8.0
8.1
0.84
1.5
South Carolina
3.5
0.28
11.0
13.9
12.2
8.0
1.7
10.2
0.48
0.14
0.03
0.31
0.10
7HQQHVVHH
0.86
5.3
5.5
0.54
7H[DV
0.02
9.3
9.0
0.60
Virginia
7.0
0.57
7.9
0.01
West Virginia
0.00
0.7
1.0
0.04
3DQHO%3HUVRQDOFKDUDFWHULVWLFVSHUFHQWH[FHSWDJHYDULDEOHV
$WWHQGLQJVFKRRODJH±
82.9
81.8
0.04
0.94
,QRZQHURFFXSLHGKRXVLQJ
52.4
48.8
0.00
22.4
23.9
0.05
/LWHUDWHDJH±
91.8
90.1
0.00
0.86
/LWHUDWHDJH±
91.2
0.00
0.91
)DWKHULVIDUPHUDJH±
59.0
0.00
0.30
1910 city population
Not in city
&LW\SRS City pop. >25,000
70.0
0.09
74.5
73.2
0.05
20.4
20.8
0.25
17.3
0.07
9.8
0.31
0.53
9.2
9.5
Min age
0
0
—
0
0
—
Max age
40
40
—
40
40
—
Median age
15
0.00
15
0.08
Mean age
17.1
0.01
17.0
0.06
St. dev.
11.1
11.5
0.00
11.2
11.3
0.50
Notes: A variance-ratio test is used to compare sample standard deviations and a Wilcoxon matched-pairs
signed-rank test is used to compare sample medians. All others comparison of means are done with standard
WWHVWV7KHPDWFKHGVDPSOHVFRQWDLQZKLWHPHQDQGEODFNPHQ
Sources7KHOLQNHGVDPSOHLVFUHDWHGE\WDNLQJWKH,3806VDPSOHRIZKLWHDQGEODFNPDOHVDJH±
ZKRUHVLGHLQWKH6RXWKDQGVHDUFKLQJIRUWKHVHPHQLQWKHFHQVXVPDQXVFULSWV7KHWH[WDQG$SSHQGL[
GHVFULEHVDPSOHFRQVWUXFWLRQLQGHWDLO7KH,3806GDWDDUHIURP5XJJOHVHWDO
Downloaded from https:/www.cambridge.org/core. The University of Tennessee - Knoxville, on 17 Mar 2017 at 13:09:33, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/S0022050715001527
958
Collins and Wanamaker
0,*5$176(/(&7,21
3DWWHUQV RI PLJUDQW VHOHFWLRQ DUH LPSRUWDQW IRU ERWK VHQGLQJ UHJLRQV
HJ ZKHWKHU WKH UHJLRQ WHQGV WR ORVH KLJKO\ VNLOOHG ZRUNHUV DQG
UHFHLYLQJUHJLRQVHJZKHWKHUPLJUDQWVDUHOLNHO\WRDVVLPLODWHTXLFNO\
and whether they are substitutes or complements for the area’s native
ZRUNHUV 7KH OLQNHG GDWDVHW LV HVSHFLDOO\ ZHOO VXLWHG WR FKDUDFWHUL]H
selection into migration because it has such detailed background inforPDWLRQRQWKHPHQLQ7KHVHFKDUDFWHUL]DWLRQVLQWXUQSURYLGHD
better basis than previously available for understanding the origins and
FRQVHTXHQFHV RI WKH *UHDW 0LJUDWLRQ HYHQ WKRXJK DVVHVVLQJ WKH IXOO
range of possible implications from selection, such as the wage impact
RQVHQGLQJDQGUHFHLYLQJDUHDV%RXVWDQLVEH\RQGWKLVDUWLFOH¶V
scope.
,QWKLVVHFWLRQZHIRFXVRQFKDUDFWHUL]LQJZKRPLJUDWHGDQGZKHWKHU
there are clear differences between migrants and non-migrants in terms
of their background characteristics. We start by classifying all men in the
OLQNHGVDPSOHLQWRWKUHHPXWXDOO\H[FOXVLYHFDWHJRULHV³QRQPLJUDQWV´
ZKRUHVLGHGLQWKHVDPHVWDWHLQDQG³ZLWKLQ6RXWKPLJUDQWV´
ZKRFKDQJHGVWDWHRIUHVLGHQFHZLWKLQWKHUHJLRQDQG³LQWHUUHJLRQDO
PLJUDQWV´ %HWZHHQ DQG D ODUJH IUDFWLRQ RI WKH PHQ LQ WKH
OLQNHGVDPSOHSHUFHQWRIZKLWHVDQGSHUFHQWRIEODFNVKDGOHIW
WKHLUVWDWHRIUHVLGHQFHZLWKURXJKO\HYHQVSOLWVEHWZHHQ³ZLWKLQ
6RXWK´DQG³LQWHUUHJLRQDO´PLJUDWLRQ
7DEOHFROXPQUHSRUWVWKHZKLWHPHQ¶VDYHUDJHFKDUDFWHULVWLFVLQ
EHIRUHWKHVWDUWRIWKH*UHDW0LJUDWLRQLQFOXGLQJWKHLUOLWHUDF\UDWHV
school attendance, occupational income and education scores, farm status
DQGKRPHRZQHUVKLSUDWHV7KH¿JXUHVDUHWDEXODWHGVHSDUDWHO\E\
PLJUDWLRQFDWHJRU\7KHVHFRQGFROXPQUHSRUWVdifferences in characteristics between the migrant categories and the non-migrant category, and
the third column reports the differences that remain after controlling for
DJHDQGFRXQW\RIRULJLQ¿[HGHIIHFWV5,QSUDFWLFHWKHFRXQW\OHYHO¿[HG
effects in column 3 absorb local push factors, such as boll weevil destruction, and control for selection that derives from differences in place-ofRULJLQ FKDUDFWHULVWLFV ZKLFK LV LQFOXGHG LQ WKH VLPSOH FRPSDULVRQ LQ
5
³7KH ³DGMXVWHG GLIIHUHQFHV´ DUH FRHI¿FLHQWV RQ PLJUDWLRQ FDWHJRU\ GXPPLHV H[SUHVVHG
UHODWLYHWRWKHQRQPLJUDQWFDWHJRU\IURPUHJUHVVLRQVZLWKDJHȖiDQGFRXQW\RIRULJLQșj¿[HG
effects: Yi ĮıSMiȡRMiȖișjei, where SM is a dummy for within-South migrants and
RM is a dummy for inter-regional migrants. YYDULDEOHVDUHWKHFKDUDFWHULVWLFVHJOLWHUDF\
occupation score, etc.).
Downloaded from https:/www.cambridge.org/core. The University of Tennessee - Knoxville, on 17 Mar 2017 at 13:09:33, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/S0022050715001527
Selection and Sorting of Southern Migrants
959
column 2).7KHQH[WWKUHHFROXPQVUHSHDWWKHWDEXODWLRQVIRUWKHVDPSOH
of black men. Although this is very basic information about the migrants
and non-migrants, none of it can be inferred from census cross-sections,
and in this respect the linked manuscript data are crucial.
,IDELOLW\RUSURGXFWLYLW\ZHUHSRVLWLYHO\FRUUHODWHGZLWKPLJUDWLRQWKHQ
we would expect migrants to have better outcomes than non-migrants in
terms of human capital, occupational status, or family background before
leaving the South. While discrimination in the South slowed black men’s
economic and educational progress, there was considerable variation in
literacy, education, occupation, property ownership, and other measures
in 1910. For whites and blacks, there is some evidence of positive selecWLRQLQWRLQWHUUHJLRQDOPLJUDWLRQRQWKHEDVLVRIOLWHUDF\SDQHO$
but the differences are quantitatively small at 1 to 4 percentage points.
7KH\DUHVWDWLVWLFDOO\LQVLJQL¿FDQWDIWHUFRQWUROOLQJIRUDJHDQGSODFHRI
RULJLQ7KLVFRQWUDVWVZLWKVRPHRIWKHSUHYLRXVOLWHUDWXUH¶VVXJJHVWLRQRI
strongly positive migrant selection in the early twentieth century based
RQ FHQVXV FURVVVHFWLRQV WKDW REVHUYH PLJUDQWV H[ SRVW HJ 7ROQD\
7KHGLVFUHSDQF\VXJJHVWVWKDWPLJUDQWVPLJKWKDYHLPSURYHGWKHLU
literacy after leaving the South, or that census enumeration of literacy
might have been regionally biased. More generally, this raises concern
regarding the practice of using ex post migrant characteristics from crosssections to make inferences about selection into migration. As linked
historical datasets become more common, scholars may be able to avoid
this measurement problem.
For an alternative view of educational background, we examined
school attendance in 1910 and found small differences in attendance
UDWHVEHWZHHQWKHPLJUDQWFDWHJRULHVSDQHO%ZLWKRUZLWKRXWFRQWUROV
From this perspective, migrant selection on the basis of formal educaWLRQVHHPVUDWKHUZHDNLQWKHHDUO\GHFDGHVRIWKH*UHDW0LJUDWLRQ:H
note, however, that in comparison to studies of migration in more recent
GHFDGHV %RUMDV %URQDUV DQG 7UHMR 9LJGRU *URJJHU DQG
+DQVRQZKLFKDUHRIWHQPRWLYDWHGE\WKH5R\PRGHORXU
metric for educational attainment is fairly crude. We cannot follow such
VWXGLHV ZLWKRXW EHWWHU GDWD IRU TXDOLW\DGMXVWHG HGXFDWLRQDO DWWDLQPHQW
DQG YDULDWLRQ LQ UHWXUQV WR HGXFDWLRQ E\ UDFH DFURVV ORFDWLRQV LQ WKH
early twentieth century, which to our knowledge does not exist.
,I LQGLYLGXDOV WHQG WR PLJUDWH IURP SDUWLFXODUO\ DIÀXHQW ORFDWLRQV IRU H[DPSOH ZH ZRXOG
infer positive selection in column 2 but not necessarily in column 3. Both perspectives are useful.
&KDQJHVLQFRHI¿FLHQWVEHWZHHQFROXPQVDQGDUHGULYHQSULPDULO\E\WKHLQFOXVLRQRIFRXQW\
¿[HGHIIHFWVQRWDJHHIIHFWV
Downloaded from https:/www.cambridge.org/core. The University of Tennessee - Knoxville, on 17 Mar 2017 at 13:09:33, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/S0022050715001527
White Males
Sample Average
,QWHUUHJLRQDOPLJUDQWV
%6FKRRODWWHQGDQFHUDWHDJHWRN Non-migrants
0.828
Within-South migrants
0.823
,QWHUUHJLRQDOPLJUDQWV
0.838
Black Males
Difference vs.
Non-migrants
$GMXVWHG
Difference
Sample Average
Difference vs.
non-Migrants
$GMXVWHG
Difference
—
0.001
0.013**
—
±
±
—
—
—
0.003
0.035*
—
±
0.008
—
±
0.010
—
±
±
—
—
—
±
0.043
—
±
0.035
—
0.80***
0.81***
12.3
13.4
—
13.2
—
—
1.07***
0.85**
—
0.53*
&2FFXSDWLRQLQFRPHVFRUHLQKXQGUHGVRIGROODUVN Non-migrants
—
Within-South migrants
17.3
1.00***
,QWHUUHJLRQDOPLJUDQWV
18.2
Collins and Wanamaker
$/LWHUDF\UDWHN ZKLWHDQGEODFN
Non-migrants
0.924
Within-South migrants
0.925
Downloaded from https:/www.cambridge.org/core. The University of Tennessee - Knoxville, on 17 Mar 2017 at 13:09:33, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/S0022050715001527
TABLE 2
6(/(&7,21,172±0,*5$7,2121%$6,62)&+$5$&7(5,67,&6
,QWHUUHJLRQDOPLJUDQWV
10.92
()DUPRULJLQN Non-migrants
Within-South migrants
0.530
,QWHUUHJLRQDOPLJUDQWV
0.500
)+RPHRZQHUVKLSUDWHN Non-migrants
0.543
Within-South migrants
0.472
,QWHUUHJLRQDOPLJUDQWV
0.505
—
0.104**
0.148***
—
0.079
7.90
8.04
—
8.02
—
—
0.131**
0.120***
—
0.082
0.025
—
±
±
—
±
±
0.592
0.521
—
0.502
—
—
±
±
—
±
±
—
±
±
—
±
±
0.222
0.200
—
0.251
—
—
±
0.030**
—
±
0.007
6LJQL¿FDQWDWWKHSHUFHQWOHYHO
6LJQL¿FDQWDWWKHSHUFHQWOHYHO
6LJQL¿FDQWDWWKHSHUFHQWOHYHO
Notes³$GMXVWHGGLIIHUHQFHV´DUHUHJUHVVLRQFRHI¿FLHQWVWKDWPHDVXUHGLIIHUHQFHVLQFKDUDFWHULVWLFVDPRQJLQWHUUHJLRQDOLQWUDUHJLRQDQGQRQPLJUDQW
FDWHJRULHVZKHUHQRQPLJUDQWVDUHWKHEDVHFDWHJRU\FRQWUROOLQJIRUDJHDQGFRXQW\RIRULJLQ¿[HGHIIHFWV1RQPLJUDQWVDUHGH¿QHGDVUHVLGLQJLQWKHVDPH
VWDWHLQDQGZLWKLQ6RXWKPLJUDQWVOHIWWKHLUVWDWHRIRULJLQEXWUHPDLQHGLQWKH6RXWK/LWHUDF\LVUHFRUGHGIRUWKRVHZKRDUHDJHDQGRYHUVFKRRO
DWWHQGDQFHLVH[DPLQHGIRUWR\HDUROGV2FFXSDWLRQDOLQFRPHDQGHGXFDWLRQVFRUHVDUHFRQGLWLRQDORQODERUIRUFHSDUWLFLSDWLRQLQ7KHVWDQGDUG
GHYLDWLRQIRURFFXSDWLRQLQFRPHVFRUHRFFXSDWLRQHGXFDWLRQVFRUHLVIRUZKLWHVDQGIRUEODFNV6WDQGDUGHUURUVFOXVWHUHGE\FRXQW\RI
origin, are in parentheses.
Sources: Data are from the sample of linked census records, as described in the text and Appendix.
Selection and Sorting of Southern Migrants
Downloaded from https:/www.cambridge.org/core. The University of Tennessee - Knoxville, on 17 Mar 2017 at 13:09:33, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/S0022050715001527
'2FFXSDWLRQHGXFDWLRQVFRUHN Non-migrants
10.77
Within-South migrants
10.87
Collins and Wanamaker
$PRQJWKRVHROGHQRXJKWRUHSRUWRFFXSDWLRQVLQSDQHOV&DQG
D), there is somewhat stronger evidence of positive selection on the basis
RISUH*UHDW0LJUDWLRQODERUPDUNHWRXWFRPHV7RWKHH[WHQWWKDWEHWWHU
skills, ability, or motivation translated into better occupational standing
EHIRUH::,WKHRFFXSDWLRQDOGDWDSURYLGHDPRUHVHQVLWLYHLQGLFDWRURI
VHOHFWLRQWKDQOLWHUDF\DORQH,QGLYLGXDOLQFRPHZDVQRWUHSRUWHGLQWKH
FHQVXVDQGVRZH¿UVWH[DPLQHDPRGL¿HGYHUVLRQRIWKH³RFFXSDWLRQDOLQFRPHVFRUH´SDQHO&77KLVLVDFRPPRQO\XVHG,3806YDULable that assigns an income to each detailed occupation category based
on the median income observed in that occupation in the 1950 census
5XJJOHV HW DO 3UHZDU RFFXSDWLRQDO VFRUHV IRU ³ZLWKLQ6RXWK´
DQG³LQWHUUHJLRQDO´ZKLWHPLJUDQWVDUHWRSHUFHQWKLJKHUWKDQIRU
non-migrants. For blacks, most group differences are slightly smaller in
magnitude than for whites, but the point estimates are consistent with
positive selection.8
,QSDQHO'ZHDVVLJQDUDFHVSHFL¿F³HGXFDWLRQVFRUH´WRHDFKGHWDLOHG
occupation in 1910 based on the average educational attainment of
southern workers in the corresponding occupation categories in the 1940
FHQVXV5XJJOHVHWDOZKLFKZDVWKH¿UVWFHQVXVWRLQTXLUHDERXW
HGXFDWLRQDODWWDLQPHQW7KHJRDOLVVLPSO\WRSURYLGHDQDOWHUQDWLYHFKDUDFWHUL]DWLRQRIRFFXSDWLRQDOVWDWXVWKDWLVFRQQHFWHGWRIRUPDOHGXFDWLRQDO
attainment rather than income. Again, there is some evidence of positive selection into migration, approximately one-tenth of a grade, but the
GLIIHUHQFHV DUH VPDOO DQG VWDWLVWLFDOO\ LQVLJQL¿FDQW DIWHU FRQWUROOLQJ IRU
age and initial location.
,QVXPPDU\RXUPHDVXUHVRIVNLOO²ZKHWKHUEDVHGRQRFFXSDWLRQDORU
educational information—are generally consistent with a limited degree
of positive selection for both whites and blacks in the early decades of
WKH*UHDW0LJUDWLRQ7KHGLIIHUHQFHVDFURVVPLJUDQWJURXSVDUHTXDQWLWDtively small, however, and thus the degree of migrant selection from this
perspective seems rather weak.
7
*LYHQWKHGLIIHUHQFHLQVWDWXVIRUWHQDQWIDUPHUVUHODWLYHWRIDUPRZQHUVZHIXUWKHUDGMXVWHG
WKHRFFXSDWLRQVFRUHIRUIDUPHUVEDVHGRQWKHLUKRPHRZQHUVKLS+RPHRZQHUVKLSLVQRWUHSRUWHG
LQWKHPLFURGDWDVDPSOHWKH\HDURQZKLFKRFFXSDWLRQVFRUHVDUHRULJLQDOO\EDVHG7KHUHIRUH
IRU IDUPHUV ZKR UHSRUWHG RZQLQJ UHQWLQJ WKHLU KRPHV ZH PXOWLSOLHG PHGLDQ IDUPHU LQFRPH
LQ WKH ,3806GH¿QHG occscore) by the ratio of median income among farmer-owners
IDUPHUUHQWHUVLQWKH,3806VDPSOHWRPHGLDQLQFRPHRIDOOIDUPHUVLQWKHVDPSOH
UHVWULFWLQJWKHVDPSOHWRVRXWKHUQUHVLGHQWV7KLVDGMXVWHGoccscoreLVDQDO\]HGLQWKHWKLUG
SDQHORI7DEOH
8
For reference, a standard deviation in the occupation income score variable is 12.0 for whites
DQGIRUEODFNVSDQHO&RI7DEOH$VWDQGDUGGHYLDWLRQLQWKHRFFXSDWLRQHGXFDWLRQVFRUH
YDULDEOHLVIRUZKLWHVDQGIRUEODFNVSDQHO'
Downloaded from https:/www.cambridge.org/core. The University of Tennessee - Knoxville, on 17 Mar 2017 at 13:09:33, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/S0022050715001527
Selection and Sorting of Southern Migrants
7DEOH SDQHO ( H[DPLQHV VHOHFWLRQ RQ WKH EDVLV RI IDUP VWDWXV LQ
1910. From this perspective, there are notable differences. For both
whites and blacks, those who lived on farms in 1910 were less likely
WRPLJUDWHRXWRIWKHVWDWHWKDQRWKHUV,QWKH³XQDGMXVWHG´FROXPQVWKH
farm-status difference between the migrant groups and the non-migrant
JURXS LV EHWZHHQ DQG SHUFHQWDJH SRLQWV ZKLWHV DQG EODFNV 7KH
addition of age and county-of-origin controls reduces the gap relative to
non-migrants for white inter-regional migrants and for both categories
of African Americans migrants, but non-trivial differences remain, espeFLDOO\IRUZKLWHV7KXVLWDSSHDUVWKDWDQDJULFXOWXUDOEDFNJURXQGWHQGHG
to hinder long-distance migration, even when comparisons are based on
within-county variation.
2ZQHURFFXSLHGKRXVLQJVWDWXVZKLFKLVH[DPLQHGLQ7DEOH¶VODVW
panel, is of particular interest because it is the only census variable in 1910
WKDW UHÀHFWV KRXVHKROG DVVHWV RZQHUVKLS RI UHDO SURSHUW\ ,Q JHQHUDO
household wealth may facilitate long-distance migration, but in this
historical context homeownership may also indicate a prior decision to
VHWWOHGRZQLQDSDUWLFXODUDUHDRUIRU\RXQJPHQLQRXUVDPSOHUHVLGLQJ
ZLWKWKHLUSDUHQWVLQRZQHURFFXSLHGKRXVLQJLWPD\UHÀHFWDQH[SHFWDWLRQWRLQKHULWRUUHFHLYHDSDUHQWDOJLIWRIORFDOSURSHUW\$EUDPLW]N\
Boustan, and Eriksson 2013). Among whites, residing in owner-occupied housing in 1910 is associated with substantially less long-distance
PLJUDWLRQEHWZHHQ DQG SHUFHQWDJHSRLQWV HYHQ DIWHUDGMXVWPHQWV
for age and county of residence. African Americans were far less likely
than whites to own their homes, and the pattern with respect to migration
LVGLIIHUHQW,QWHUUHJLRQDOEODFNPLJUDQWVZHUHVRPHZKDWPRUHOLNHO\WR
UHVLGH LQ RZQHURFFXSLHG KRXVLQJ LQ ,I WKH QHJDWLYH FRHI¿FLHQWV
for whites are interpreted as indicators of relatively strong local attachments among property owners and their children, then it would appear
that black property owners and their children did not share such strong
attachments.
Overall, migration in the linked dataset does not conform to a simple
FKDUDFWHUL]DWLRQ LQ WHUPV RI QHJDWLYH RU SRVLWLYH VHOHFWLRQ 0LJUDQWV
KHOG MREV LQ ZLWK KLJKHU RFFXSDWLRQ VFRUHV WKDQ QRQPLJUDQWV RQ
average, which is consistent with positive selection. But these differences
were not large. Moreover, differences across migrant groups in terms of
literacy, school attendance, or occupational-education scores were small
RUQRQH[LVWHQW)DUPUHVLGHQFHZDVWKHPRVWUREXVWQHJDWLYHSUHGLFWRU
RIVXEVHTXHQWPLJUDWLRQIRUZKLWHVDQGEODFNVD¿QGLQJWKDWLVFRQVLVtent with our expectations but novel in the sense that, to our knowledge,
Downloaded from https:/www.cambridge.org/core. The University of Tennessee - Knoxville, on 17 Mar 2017 at 13:09:33, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/S0022050715001527
Collins and Wanamaker
no previously constructed dataset could observe the pre-migration farm
status of individual men in this period. We interpret the overall results as
DUHÀHFWLRQRIUHPDUNDEO\EURDGSDUWLFLSDWLRQLQLQWHUQDOPLJUDWLRQ7KH
differences between migrants and non-migrants in the early decades of
WKH *UHDW 0LJUDWLRQ DSSHDU OHVV VDOLHQW WKDQ WKH LQFUHDVHG YROXPH DQG
QHZGLUHFWLRQVRIPLJUDWLRQ7KLVOHDGVXVWRLQYHVWLJDWHIXUWKHUwhere the
migrants moved and why they decided to move there.
29(59,(:2)0,*5$7,213$77(516$1'6257,1*
7KH OLQNHG GDWDVHW SURYLGHV QHZ RSSRUWXQLWLHV IRU VWXG\LQJ DQG
comparing migrant sorting patterns, both as a function of individual and
place-of-origin characteristics and as a function of labor market condiWLRQVDQGPLJUDWLRQFRVWVDFURVVVSDFH,QDGGLWLRQWKHOLQNHGGDWDDOORZ
XV WR EH PRUH VSHFL¿F DERXW ZKHQ SHRSOH PRYHG EHWZHHQ DQG
1930) than is possible with census cross-sections, where prior location is
known only at the time of birth and migration could have occurred at any
time afterwards.
7DEOH SURYLGHV VXPPDU\ VWDWLVWLFV RI WKH PLJUDWLRQ SDWWHUQV LQ WKH
OLQNHGGDWDVHWLQFOXGLQJWKHSURSHQVLW\WRPLJUDWHSDQHO$DQGGLVWDQFH
DQG GLUHFWLRQ RI PLJUDWLRQ SDQHO % 7KH WDEOH¶V WKLUG FROXPQ UHSRUWV
VLPSOH EODFNZKLWH GLIIHUHQFHV 7KH IRXUWK FROXPQ UHSRUWV FRHI¿FLHQWV
on a black indicator variable from regressions that control for state-ofRULJLQ¿[HGHIIHFWV$VPHQWLRQHGDERYHEHWZHHQDQGDODUJH
fraction of the men in the linked sample left their 1910 state of resiGHQFH ZLWK DSSUR[LPDWHO\ DQ HYHQ VSOLW RI PLJUDQWV EHWZHHQ ³ZLWKLQ
6RXWK´ DQG ³RXWVLGH 6RXWK´ GHVWLQDWLRQV $PRQJ WKRVH ZKR OHIW WKH
South, the Midwest was the most common destination for both whites
and blacks, but black inter-regional migrants moved relatively strongly
LQWR WKH 1RUWKHDVW FRPSDUHG WR ZKLWHV ZKHUHDV ZKLWHV PRYHG UHODtively strongly into the West.9 Detailed state-to-state migration patterns
are reported in the Online Appendix.
3DQHO%RI7DEOHUHSRUWVPHDVXUHVRIGLVWDQFHWUDYHOOHGEDVHGRQWKH
latitude and longitude of the center of each individual’s 1910 and 1930
9
7KHPRVWFRPPRQGHVWLQDWLRQFRXQWLHVIRULQWHUUHJLRQDOEODFNPLJUDQWVZHUH&RRN&RXQW\
&KLFDJR,/3KLODGHOSKLD3$:D\QH'HWURLW0,DQG1HZ<RUN1<:KLWHLQWHUUHJLRQDO
PLJUDQWVPRVWIUHTXHQWO\UHORFDWHGWR/RV$QJHOHV&$&RRN&KLFDJR,/:D\QH'HWURLW
0,DQG6XPPLW$NURQ2+,QWUDUHJLRQDOEODFNPLJUDQWVPRYHGPRVWIUHTXHQWO\WR6KHOE\
&RXQW\0HPSKLV71'XYDO-DFNVRQYLOOH)/+DPLOWRQ&KDWWDQRRJD71DQG&KDWKDP
6DYDQQDK*$:KLWHLQWHUUHJLRQDOPLJUDQWVUHORFDWHGPRVWIUHTXHQWO\WR2NODKRPD&LW\2.
6KHOE\0HPSKLV71-HIIHUVRQ%LUPLQJKDP$/DQG'DOODV7;
Downloaded from https:/www.cambridge.org/core. The University of Tennessee - Knoxville, on 17 Mar 2017 at 13:09:33, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/S0022050715001527
Selection and Sorting of Southern Migrants
TABLE 3
0,*5$7,213$77(516800$5<67$7,67,&6%<5$&(±
White
Black
Unconditional Black- Conditional BlackWhite Difference White Difference
3DQHO$0LJUDWLRQSURSHQVLWLHVVKDUHV
No state-to-state migration
Within-South migrants
0.18
0.19
,QWHUUHJLRQDOPLJUDQWV
0.17
0.20
South-to-Northeast migrants
0.03
0.08
South-to-Midwest migrants
0.09
0.11
South-to-West migrants
0.05
0.01
±
0.0088
0.028***
0.050***
0.016***
–0.037***
±
±
0.075***
0.050***
0.049***
–0.024***
±
±
±
±
±
2.23***
0.23***
3DQHO%*HRJUDSKLFFKDUDFWHULVWLFVRIPLJUDWLRQ
'LVWDQFHWUDYHOOHGPLOHV
Full sample
219.8
210.7
Within-South migrants
321.8
Inter-regional migrants
&KDQJHLQODWLWXGHKRPHWRGHVWLQDWLRQGHJUHHV
Full sample
1.38
Within-South migrants
±
0.47
Inter-regional migrants
&KDQJHLQORQJLWXGHKRPHWRGHVWLQDWLRQGHJUHHV
±
0.30
Full sample
1.21***
1.09***
±
0.20
0.71***
1.34***
Within-South migrants
±
1.42
Inter-regional migrants
Notes7KHODVWFROXPQUHSRUWVUHJUHVVLRQVFRHI¿FLHQWVWKDWFRPSDUHEODFNDQGZKLWHPLJUDWLRQ
SDWWHUQV FRQWUROOLQJ IRU VWDWHRIRULJLQ ¿[HG HIIHFWV 7KH ³IXOO VDPSOH´ LQFOXGHV QRQPLJUDQWV
GH¿QHGDVWKRVHZKRGRQRWOHDYHWKHVWDWHRIRULJLQ6RXWKWR1RUWKHDVWLQFOXGHVUHVLGHQWV
of the Northeast census regions and also Delaware, Maryland, and Washington, DC. Latitude
differences are positive for south-to-north migration. Longitude differences are positive for westto-east migration. Standard errors, clustered by county of origin, are in parentheses.
Sources: Data are from the linked sample of census records, as described in the text and Appendix.
Downloaded from https:/www.cambridge.org/core. The University of Tennessee - Knoxville, on 17 Mar 2017 at 13:09:33, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/S0022050715001527
Collins and Wanamaker
FRXQW\ RI UHVLGHQFH ,Q WKH IXOO VDPSOH DYHUDJH PLJUDWLRQ GLVWDQFHV
LQFOXGLQJ]HURVIRUWKRVHZKRGLGQRWFKDQJHFRXQWLHVDUHVLPLODUDFURVV
the black and white samples, with or without controls for state-of-origin.
,WLVFOHDUKRZHYHUWKDWFRQGLWLRQDORQPLJUDWLQJZKLWHVPRYHGIDUWKHU
WKDQ EODFNV RQ DYHUDJH ERWK ZLWKLQ WKH 6RXWK YHUVXV PLOHV
DQG ZKHQ OHDYLQJ WKH 6RXWK YHUVXV PLOHV DOVR VHH 7ROQD\
HW DO &RQWUROOLQJ IRU VWDWHRIRULJLQ DPSOL¿HV WKH ODUJH GLIIHUHQFH LQ GLVWDQFH WUDYHOOHG E\ LQWHUUHJLRQDO PLJUDQWV FROXPQ 7KH
DYHUDJHEODFNPDOHPRYHGQRUWKZDUGDQGHDVWZDUGSRVLWLYHFKDQJHVLQ
latitude and longitude), whereas the average white male moved northward and westward, though not as far north as blacks. Among white
LQWHUUHJLRQDO PRYHUV WKH PLJUDQWV WR WKH :HVW VWURQJO\ LQÀXHQFH
the average change of longitude, and black-white differences in eastwest mobility patterns are striking, even when controlling for state of
origin.10
For the sake of concise description and to facilitate discrete-choice
PRGHOLQJZHIRFXVRQWKHPLJUDQWV¶FKRLFHRIVWDWH,QRXUVDPSOH
percent of inter-state migrants chose non-urban locations, and therefore focusing solely on migrants to cities would omit a large share of
the sample, distort the ex ante set of destination choices, and generDOO\PLVFKDUDFWHUL]HWKHSHULRG¶VLQWHUQDOPLJUDWLRQSDWWHUQV0RUHRYHU
because the multinomial logit model, described below, estimates a large
QXPEHU RI FRHI¿FLHQWV IRU HDFK SRWHQWLDO GHVWLQDWLRQ ZRUNLQJ DW WKH
county- or city-level would be computationally prohibitive.
Figure 2 maps the distribution of inter-state southern migrants across
destinations between 1910 and 1930. Continuing pre-1910 migration
WUHQGVPLJUDQWVZLWKLQWKH6RXWKWHQGHGWRIDYRU7H[DV2NODKRPDDQG
)ORULGDZLWK7H[DVDQG2NODKRPDEHLQJSDUWLFXODUO\LPSRUWDQWGHVWLQDWLRQVIRUZKLWHV,QDGGLWLRQDQGEUHDNLQJZLWKSUHSDWWHUQVEODFN
PLJUDQWVIUHTXHQWO\FKRVHORFDWLRQVLQWKHLQGXVWULDO1RUWK3HQQV\OYDQLD
,OOLQRLV2KLRDQG1HZ<RUNDOOUHFHLYHGPRUHVRXWKHUQEODFNPLJUDQWV
DIWHUWKDQGLGDQ\VRXWKHUQVWDWH3HQQV\OYDQLDZDVWKHPRVWFRPPRQ
GHVWLQDWLRQIRUEODFNVDQGLWGUHZKHDYLO\IURP9LUJLQLD*HRUJLDDQG
South Carolina. Southern white migrants were also drawn strongly to
2KLR DQG WR D OHVVHU H[WHQW,OOLQRLV %XW &DOLIRUQLDDOVR HPHUJHGDVD
PDMRUGHVWLQDWLRQIRUZKLWHVGUDZLQJKHDYLO\IURPWKHUHVLGHQWVRI
7H[DV2NODKRPD$UNDQVDVDQG.HQWXFN\
10
Appendix Figure 1 in the Online Appendix graphs scatterplots of changes in latitude and
ORQJLWXGHVHSDUDWHO\IRUVRXWKHUQZKLWHVDQGEODFNV7KHPDLQYLVXDOLPSUHVVLRQLVRQHRIVWURQJ
northward movement for blacks but a much more diffuse pattern for whites.
Downloaded from https:/www.cambridge.org/core. The University of Tennessee - Knoxville, on 17 Mar 2017 at 13:09:33, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/S0022050715001527
Selection and Sorting of Southern Migrants
FIGURE 2
',675,%87,212),17(567$7(0,*5$176,1/,1.('6$03/(±
Notes7KHPDSVVKRZWKHVKDUHRILQWHUVWDWHPLJUDQWVFKRRVLQJHDFKGHVWLQDWLRQVWDWHVHSDUDWHO\
E\UDFH7KHVKDUHVDGGWRSHUFHQWIRUHDFKUDFHFDWHJRU\
Sources: Data are from the linked sample of census records, as described in the text and
Appendix.
Although there was a substantial degree of overlap in black and
ZKLWH PLJUDWLRQ ÀRZV WKHUH ZHUH DOVR QRWDEOH GLIIHUHQFHV DPRQJ WKH
PRVWIUHTXHQWO\FKRVHQORFDOHV2QO\2KLRDQG,OOLQRLVUDQNDPRQJWKH
WRS¿YHGHVWLQDWLRQVWDWHVIRUERWKJURXSVDQGLQVRPHFDVHVWKHEODFN
ZKLWHGLIIHUHQFHVDUHTXLWHODUJHHJ&DOLIRUQLDDQG3HQQV\OYDQLD$
Downloaded from https:/www.cambridge.org/core. The University of Tennessee - Knoxville, on 17 Mar 2017 at 13:09:33, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/S0022050715001527
Collins and Wanamaker
GLVVLPLODULW\LQGH[SURYLGHVDVLPSOHZD\WRVXPPDUL]HWKHPDJQLWXGH
RIEODFNZKLWHGLIIHUHQFHVLQWKHPLJUDQWV¶ORFDWLRQFKRLFHV,WLQGLFDWHV
the share of migrants that would have to choose a different location for
WKH GLVWULEXWLRQ RI FKRLFHV WR EH WKH VDPH DFURVV UDFH FDWHJRULHV 7KH
index is calculated as
1 N bi wi where i denotes a state, b w ) is
Σ
−
i
i
2 i=1 B W
WKHQXPEHURIEODFNZKLWHPHQPRYLQJWRVWDWHi and BW ) is the total
QXPEHURIEODFNZKLWHPLJUDQWVLQWKHVDPSOH)URPWKLVSHUVSHFWLYH
DSSUR[LPDWHO\SHUFHQWLQGH[YDOXHRIRIEODFNRUZKLWHLQWHU
state migrants would have to choose a different destination for the black
and white post-migration distributions to be equivalent.
%$&.*5281'&+$5$&7(5,67,&6$1'',))(5(1&(6
,1%/$&.$1':+,7(0,*5$7,213$77(516
Because black and white men differed in their observable characteristics and starting locations circa 1910, it is natural to ask whether such
differences can account for black-white differences in migration patterns.
We take two different approaches to this question. First, looking deeper
LQWRWKHVXPPDU\VWDWLVWLFVUHSRUWHGLQ7DEOHZHHVWLPDWHDVHULHVRI
RUGLQDU\ OHDVW VTXDUHV 2/6 UHJUHVVLRQV WKDW LQFOXGH D UDFH LQGLFDWRU
variable and a rich set of background variables, such as age, literacy,
LQGXVWU\RIHPSOR\PHQWRUIDWKHU¶VLQGXVWU\RIHPSOR\PHQWSODFHRI
RULJLQDQGVRRQ&RPSDULQJ³XQDGMXVWHGÉODFNZKLWHPLJUDWLRQGLIIHUHQFHVWRWKRVH³DGMXVWHGÍRUEDFNJURXQGFKDUDFWHULVWLFVZH¿QGWKDWWKH
background characteristics generally cannot account for the differences
LQPLJUDWLRQRXWFRPHV,QIDFWDGGLQJFRQWUROYDULDEOHVWHQGVWRZLGHQ
not narrow, some black-white differences in migration choices. For the
sake of brevity, these results and additional details are provided in the
Online Appendix.
Second, similar in spirit to the above but with a sharper focus on the
actual choice of destination, we estimate multinomial logit models to
FKDUDFWHUL]HWKHORFDWLRQFKRLFHVRILQWHUVWDWHPLJUDQWVDVDIXQFWLRQRI
UDFH DQG EDFNJURXQG FKDUDFWHULVWLFV 7KH PXOWLQRPLDO IUDPHZRUNWUHDWV
each state as a potential destination, with the caveat that we combined
some less populous states to facilitate estimation. For the subsample of
individuals age 17 and under in 1910, the model includes indicator variables for race, father’s literacy and industry of employment, own school
attendance, place in the family’s birth order, owner-occupied housing
VWDWXVVWDWHRIRULJLQDQGFLW\VL]H)RUWKRVHDQGROGHULQWKHOLVW
Downloaded from https:/www.cambridge.org/core. The University of Tennessee - Knoxville, on 17 Mar 2017 at 13:09:33, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/S0022050715001527
Selection and Sorting of Southern Migrants
of independent variables is similar, but includes own literacy and industry
RIHPSOR\PHQWUDWKHUWKDQIDWKHU¶VDQGPDULWDOVWDWXVEXWRPLWVVFKRRO
DWWHQGDQFHDQGELUWKRUGHU7KH2QOLQH$SSHQGL[SURYLGHVPRUHGHWDLOV
Using the model’s parameter estimates, the importance of black-white
disparities in observable characteristics is revealed by comparing two
counterfactual migration distributions in which men have the same charDFWHULVWLFVH[FHSWIRUUDFH6SHFL¿FDOO\ZHWDNHWKHIXOOVDPSOHRIZKLWH
and black men and predict destination choices when all are assigned
³EODFN´VWDWXVDQGWKHQDJDLQZKHQDOODUHDVVLJQHG³ZKLWH´VWDWXVVXFK
that the differences across the two sets of predictions are attributable only
to differences in race as all other personal attributes are equivalent across
FRXQWHUIDFWXDOV 6WDWD&RUS )RU HDFK PDQ WKLV SURFHGXUH HVWLmates probabilities of choosing each state under white and black model
SDUDPHWHUVWKHSUREDELOLWLHVDUHDYHUDJHGWRJHWWKHSUHGLFWHG³DOOZKLWH´
RU ³DOO EODFN´ FRXQWHUIDFWXDO GLVWULEXWLRQV RI PLJUDQWV DFURVV VWDWHV ,I
differences in background characteristics largely explained black-white
differences in destination choice, then the value of the dissimilarity index
between the all black and all white counterfactual distributions would be
VXEVWDQWLDOO\OHVVWKDQWKHLQGH[YDOXHEDVHGRQWKHDFWXDOGLVWULEXWLRQV,Q
SUDFWLFHKRZHYHUUHFDOFXODWLQJWKHGLVVLPLODULW\LQGH[EHWZHHQWKH³DOO
EODFN´DQG³DOOZKLWH´FRXQWHUIDFWXDOGLVWULEXWLRQV\LHOGVDQLQGH[YDOXH
of 0.27. From this perspective, only a small portion of black-white differences in destination choice, on net, can be accounted for by the background characteristics available in the census. Underlying the dissimilarity index results, we see that background characteristics are helpful in
QDUURZLQJWKHEODFNZKLWHGLIIHUHQFHLQPLJUDWLRQWRVRPHVWDWHVPRVW
notably California), but they widen the black-white difference in other
VWDWHVVXFKDV,OOLQRLVDQG0LVVRXULVXFKWKDWWKHRYHUDOOGLVVLPLODULW\
index changes little.11
,QVXPZH¿QGWKDWWKHUHZDVVXEVWDQWLDORYHUODSLQEODFNDQGZKLWH
migration patterns, but black-white differences in migration patterns
ZHUHHFRQRPLFDOO\VLJQL¿FDQWDQGJHQHUDOO\ZHUHQRWDUHÀHFWLRQRIWKH
migrants’ background characteristics. Observationally similar southern
men circa 1910 tended to make different migration decisions depending
RQWKHLUUDFH7KLV¿QGLQJOHDGVXVWRIRFXVRQKRZVWURQJO\DQGKRZ
differently) white and black migrants responded to variation in the costs
DQGEHQH¿WVRIUHORFDWLQJWRHDFKSRWHQWLDOGHVWLQDWLRQ&RQGLWLRQDOORJLW
models are particularly useful for studying such issues.
11
7KHVWDWHVSHFL¿FUHVXOWVDUHLOOXVWUDWHGLQWKH2QOLQH$SSHQGL[)LJXUH
Downloaded from https:/www.cambridge.org/core. The University of Tennessee - Knoxville, on 17 Mar 2017 at 13:09:33, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/S0022050715001527
970
Collins and Wanamaker
&21',7,21$//2*,702'(/2)'(67,1$7,21&+2,&(
Suppose that individual i chooses to migrate to state j if
U ( X ij ) > U ( X ik ), ∀ k ∈ { ,
, J}
d j ≠ k,
ZKHUH8UHSUHVHQWVXWLOLW\RYHUDYHFWRUXik, which contains variables
WKDW UHÀHFW WKH H[SHFWHG LQFRPH DPHQLWLHV DQG PLJUDWLRQ FRVWV IRU
individual i in potential destination k ,I WKH XWLOLW\ RYHU HDFK VWDWH
FKRLFH LQFOXGHV D UDQGRP FRPSRQHQW ZLWK DQ ([WUHPH 9DOXH 7\SH ,
distribution, the probability of choosing any particular state is represented by:
P( Di = j ) =
exp(γ X ij )
Σ kJ =1 exp (γ X ik )
, ∀j ∈ [1
J]
where Di is the location choice of individual i7KLVLVWKHIDPLOLDUFRQGLtional logit framework for discrete choice described in Daniel McFadden
DSSOLHG WR D VHWWLQJ LQ ZKLFK PLJUDQWV FKRRVH GHVWLQDWLRQV
'DYLHV *UHHQZRRG DQG /L 9LJGRU :R]QLDN $V
cited earlier in the article, our interpretation of the model in this setting is
VWURQJO\LQÀXHQFHGE\SUHYLRXVUHVHDUFKRQWKHHFRQRPLFVRIPLJUDWLRQ
DQGORFDWLRQFKRLFH6MDDVWDG5REDFN%RUMDV%URQDUVDQG
7UHMR*URJJHUDQG+DQVRQ0RUHWWL
0D[LPXPOLNHOLKRRG HVWLPDWHV RI Ȗ GHVFULEH KRZ GLIIHUHQFHV LQ
economic characteristics across potential destination states are correlated
with the choices of inter-state migrants. Note that in conditional logit
models, any variable that does not vary across potential destinations for
individual i VXFK DV DQ LQGLYLGXDOVSHFL¿F YDULDEOH RU SODFHRIRULJLQ
VSHFL¿FYDULDEOHIDOOVRXWRIWKHHVWLPDWLQJHTXDWLRQDQGLVQRWLGHQWL¿HG
&RHI¿FLHQWV RQ GHVWLQDWLRQVSHFL¿F FKDUDFWHULVWLFV DQG RQ LQWHUDFWLRQV
EHWZHHQGHVWLQDWLRQFKDUDFWHULVWLFVDQGLQGLYLGXDOVSHFL¿FDWWULEXWHVFDQ
EHLGHQWL¿HGHJLQWHUDFWLRQVRIODERUGHPDQGDQGUDFH
We estimate the conditional logit model with the sample of inter-state
migrants. We do not include non-migrants in the analysis because doing
VRSRVHVFRQFHSWXDOSUREOHPVZLWKNH\YDULDEOHVVXFKDV³GLVWDQFH´DQG
³PLJUDQWVWRFN´DQGZLWKWKHWUHDWPHQWRI³KRPH´DVDSRWHQWLDOGHVWLnation choice. Since this section of the article is primarily concerned
with describing the migrants’ choices of destinations rather than their
selection into migration, we believe that concentrating on the migrants
Downloaded from https:/www.cambridge.org/core. The University of Tennessee - Knoxville, on 17 Mar 2017 at 13:09:33, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/S0022050715001527
Selection and Sorting of Southern Migrants
971
VLPSOL¿HVWKHDQDO\VLVDQGWKHLQWHUSUHWDWLRQRIYDULDEOHVLQDZD\WKDWLV
helpful.12
Several variables comprise XijVWDUWLQJZLWKWKRVHWKDWPD\KDYHLQÀXenced expected employment opportunities and earnings. We construct
D PHDVXUH RI SRVW ODERU GHPDQG JURZWK Bj IROORZLQJ 7LPRWK\
%DUWLN,Q%DUWLN¶VVSHFL¿FDWLRQHPSOR\PHQWJURZWKLQHDFKVWDWH
LVSUHGLFWHGE\PXOWLSO\LQJWKHVL]HRIWKHODERUIRUFHLQHDFKLQGXVWU\
LQHDFKVWDWHHjlE\D±QDWLRQZLGHHPSOR\PHQWJURZWKUDWHLQ
WKDWLQGXVWU\gl), and then summing across industries within states.
B j = Σ lL=1e jl × gl .
7KH%DUWLNVW\OHPHDVXUHJDXJHVWRWDOHPSOR\PHQWJURZWKLQHDFKVWDWH
ZLWKRXWXWLOL]LQJH[SRVWVWDWHVSHFL¿FODERUIRUFHJURZWKUDWHVZKLFKDUH
HQGRJHQRXVWRPLJUDWLRQ:HDOVRHVWLPDWHVSHFL¿FDWLRQVWKDWLQFOXGHWKH
percentages of the 1910 labor force employed in agriculture and manufacturing separately rather than the Bartik measure, which combines
LQIRUPDWLRQRYHUDOOLQGXVWULHV,QWKHVSHFL¿FDWLRQVZLWKDJULFXOWXUHDQG
manufacturing employment variables, we also include a control variable
IRUVWDWHSRSXODWLRQVL]HEHFDXVHODUJHUVWDWHVDUHOLNHO\WRDWWUDFWPRUH
migrants all else equal.
7KH YHFWRU Xij DOVR LQFOXGHV UDFHVSHFL¿F LQFRPH HVWLPDWHV IRU HDFK
VWDWH FLUFD 7KHVH PHDVXUHV FRPELQH VWDWHOHYHO HVWLPDWHV RI UHDO
LQFRPH SHU ZRUNHU IURP .ULV -DPHV 0LWFKHQHU DQG ,DQ : 0F/HDQ
ZLWKUDFHVSHFL¿FDGMXVWPHQWIDFWRUVIRUHDFKVWDWHWKDWDUHFDOFXODWHGIURPWKHFHQVXVPLFURGDWDWKH¿UVWFHQVXV\HDUZLWKZDJH
data).13,QHVVHQFHWKHRXWSXWSHUZRUNHUEHQFKPDUNLQHDFKVWDWHFLUFD
1910 is scaled up or down depending on the ratio of black or white men’s
ZDJHVWRDOOZDJHVLQWKDWVWDWHLQWKHFHQVXV7KH2QOLQH$SSHQGL[
provides more detail.
12
:HXVHWKHPXOWLQRPLDOORJLWZKHUHWKHHPSKDVLVLVRQLQGLYLGXDOVSHFL¿FYDULDEOHVDQGWKH
FRQGLWLRQDOORJLWZKHUHWKHHPSKDVLVLVRQGHVWLQDWLRQVSHFL¿FYDULDEOHV$PRGHOWKDWFRPELQHV
LQGLYLGXDODQGGHVWLQDWLRQVSHFL¿FYDULDEOHVLQWKHVDPHIUDPHZRUNDVLPSOHPHQWHGLQ6WDWD¶V
asclogit command, can be estimated successfully only after stripping out nearly all the personal
background variables. We have also explored nested logit models in which the census regions
and the home state serve as nests.
13
,PSOLFLWO\ZHDVVXPHWKDWEODFNDQGZKLWHPHQ¶VZDJHVUHODWLYHWRVWDWHDYHUDJHVZHUHQRW
VWURQJO\ DIIHFWHG E\ WKH PLJUDWLRQ LWVHOI 7KH EHVW UHOHYDQW HYLGHQFH LV IURP %RXVWDQ ZKLFK¿QGVDPRGHVWQHJDWLYHHIIHFWDSSUR[LPDWHO\SHUFHQWRIVRXWKHUQPLJUDQWVRQEODFN
wages in the North between 1940 and 1970, when the volume of migration was larger than before
1930. On this basis, we expect that migration between 1910 and 1930 had relatively small effects
on overall wage levels.
Downloaded from https:/www.cambridge.org/core. The University of Tennessee - Knoxville, on 17 Mar 2017 at 13:09:33, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/S0022050715001527
972
Collins and Wanamaker
Variables related to the cost of migration are also in Xij. We calculate
the log distance from each individual’s county of residence in 1910 to
each potential destination state to capture relocation costs that are proporWLRQDOWRORJGLVWDQFH,QDGGLWLRQZHFDOFXODWHWKHVKDUHRIDOOSHRSOH
born in person i’s home state who resided in state j in 1910, separately by
UDFHXVLQJWKH,38065XJJOHVHWDO7KHVHSUHPLJUDWLRQUDWHVKHOSDFFRXQWIRUWKHLQÀXHQFHRISUHH[LVWLQJDQGUDFHVSHFL¿F
relationships between states including, but not limited to, networks that
facilitate migration by providing a cultural home and assistance with
¿QGLQJHPSOR\PHQWDQGKRXVLQJ
$UHJLRQDOGXPP\YDULDEOHIRU³1RQ6RXWK´FDSWXUHVWKHLQÀXHQFHRI
UHJLRQVSHFL¿FDPHQLWLHVVRPHRIZKLFKPD\KDYHGLIIHUHQWLDOO\LQÀXenced decisions of black and white migrants, such as more secure civil
ULJKWVRUOHVVULJLGVRFLDOVHJUHJDWLRQ7KHFRHI¿FLHQWUHÀHFWVPLJUDWLRQ
above or below what would be expected on the basis of the economic
variables included in Xij7KLVLVDVLPSOHZD\WRWHVWZKHWKHUDUHDVRXWVLGH
the South were especially attractive to southern migrants, conditional on
other X variables. Of course, identifying this border effect relies upon the
sample’s inclusion of both intra-regional and inter-regional migrants, and
LGHQWLI\LQJ D UDFHVSHFL¿F ERUGHU HIIHFW UHOLHV XSRQ KDYLQJ ERWK ZKLWH
DQG EODFN PLJUDQWV 7KH WKHPH RI HVFDSLQJ WKH 6RXWK LV SURPLQHQW LQ
many narrative descriptions of African Americans’ motives for interUHJLRQDO PLJUDWLRQ HJ :LONHUVRQ EXW RWKHUV KDYH QRWHG WKDW
UDFLDORSSUHVVLRQORQJSUHFHGHGWKHRQVHWRIWKH*UHDW0LJUDWLRQ+LJJV
9LFNHU\:KHWKHUDUHDVRXWVLGHWKH6RXWKZHUHHVSHFLDOO\
DWWUDFWLYHWREODFNPLJUDQWVDQGZKHWKHUWKDWLQÀXHQFHFDQEHGHWHFWHG
DIWHUFRQWUROOLQJIRUHFRQRPLFFRQGLWLRQVDUHRSHQHPSLULFDOTXHVWLRQV,Q
VXEVHTXHQWDQDO\VLVZHHVWLPDWHVHSDUDWHFRHI¿FLHQWVIRUWKH1RUWKHDVW
0LGZHVWDQG:HVWUDWKHUWKDQWUHDWLQJWKH1RQ6RXWKDVDVLQJOHUHJLRQ
RUZLWKDIXOOVHWRIGHVWLQDWLRQ¿[HGHIIHFWV
7KHEDVHOLQHDQDO\VLVDOVRLQFOXGHVDYDULDEOHIRUHDFKVWDWH¶VOHYHORI
XUEDQL]DWLRQWRVHHZKHWKHUPLJUDQWVZHUHGUDZQWRKLJKO\XUEDQL]HGVWDWHV
for reasons that are not captured by other independent variables. Robustness
FKHFNVZLWKGLIIHUHQWVSHFL¿FDWLRQVDQGIXQFWLRQDOIRUPDVVXPSWLRQVDUH
discussed later and in further detail in the Online Appendix.
&21',7,21$//2*,75(68/76
7KHEDVHOLQHUHVXOWVDUHSUHVHQWHGLQ7DEOH&ROXPQVWRSHUWDLQ
WRWKHVDPSOHRIZKLWHPHQFROXPQVWRSHUWDLQWREODFNPHQDQG
FROXPQVWRSRROWKHVDPSOHVDQGUHSRUWFRHI¿FLHQWVRQX variables
Downloaded from https:/www.cambridge.org/core. The University of Tennessee - Knoxville, on 17 Mar 2017 at 13:09:33, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/S0022050715001527
Selection and Sorting of Southern Migrants
973
LQWHUDFWHGZLWKDUDFHGXPP\YDULDEOHEODFN WRKLJKOLJKWGLIIHUHQFHV
0RGHOVLQFROXPQVDQGLQFOXGHGHVWLQDWLRQVWDWH¿[HGHIIHFWV7KH
SUHVHQFHRIVWDWH¿[HGHIIHFWVOLPLWVWKHVHWRIFRHI¿FLHQWVZHFDQLGHQWLI\
to those that vary across individuals within potential destinations, which
means we cannot identify most of the X YDULDEOHV¶ FRHI¿FLHQWV LQ WKLV
VSHFL¿FDWLRQ1RQHWKHOHVVEHFDXVHGLVWDQFHDQGPLJUDQWVWRFNYDULDEOHV
GRYDU\ZLWKLQGHVWLQDWLRQFDWHJRULHVDFURVVLQGLYLGXDOVWKHLUFRHI¿FLHQWV
DUHLGHQWL¿HGDQGWKHVSHFL¿FDWLRQSURYLGHVDXVHIXOUREXVWQHVVFKHFN
$SRVLWLYHORJLWFRHI¿FLHQWLQGLFDWHVWKDWDQLQFUHDVHLQWKDWYDULDEOH
for a particular state is associated with an increase the probability of
PLJUDWLRQ WR WKDW VWDWH ,QWHUSUHWLQJ WKH PDJQLWXGH RI ORJLW FRHI¿FLHQWV
is not straightforward, and therefore we present some counterfactuals to
illustrate the results. For reference, marginal effects for each variable for
each destination state and race are reported in the Online Appendix.
$FURVVDOOVSHFL¿FDWLRQVDQGERWKUDFHVPLJUDQWVUHVSRQGHGQHJDWLYHO\
to distance and positively to pre-existing stocks of migrants from the
same state. Black migrants appear to have been more strongly deterred by
GLVWDQFHWKDQZKLWHVDQGWKLVGLIIHUHQFHLVVWDWLVWLFDOO\VLJQL¿FDQWHYHQ
ZKHQ FRQWUROOLQJ IRU SUHH[LVWLQJ PLJUDQW VWRFNV DQG GHVWLQDWLRQ ¿[HG
HIIHFWV7KHUHODWLYHO\VWURQJUHVSRQVHWRGLVWDQFHPD\UHÀHFWVRXWKHUQ
blacks’ lower average levels of wealth and educational attainment, which
could affect their access to information about distant opportunities and
WKHLU DELOLW\ WR ¿QDQFH ORQJGLVWDQFH WUDYHO ,W LV QRWHZRUWK\ KRZHYHU
WKDWVSOLWWLQJWKHEODFNVDPSOHLQWROLWHUDWHDQGLOOLWHUDWHPHQLQ
GRHV QRW UHYHDO D VWDWLVWLFDOO\ VLJQL¿FDQW GLIIHUHQFH LQ WKHLU PLJUDWLRQ
EHKDYLRUZLWKUHVSHFWWRGLVWDQFH2QOLQH$SSHQGL[7DEOHQRUGRHV
splitting the sample by 1910 homeownership status. For perspective on
WKHPDJQLWXGHRIWKHGLVWDQFHFRHI¿FLHQWVLQFROXPQVDQGRI7DEOH
the results suggest that if southern migrants had been located one stanGDUGGHYLDWLRQIDUWKHUDZD\IURP2KLRDOOHOVHKHOGFRQVWDQWWKHQWKH
share of migrants going to Ohio would have declined by 3 percentage
points for whites and 4 percentage points for blacks, relative to a base of
8 and 7 percentage points respectively.147KHDYHUDJH³PDUJLQDOHIIHFW´
RYHUDOOGHVWLQDWLRQVWDWHVHTXDOO\ZHLJKWHGLV±SHUFHQWDJHSRLQWV
IRU ZKLWHV DQG ± SRLQWV IRU EODFNV IRU D RQHVWDQGDUG GHYLDWLRQ
FKDQJH LQ ORJ GLVWDQFH 7KH HPSLULFDO LPSRUWDQFH RI GLVWDQFH VXJJHVWV
14
7KHVH ¿JXUHV DUH FDOFXODWHG DV WKH GLIIHUHQFH EHWZHHQ WKH SUHGLFWHG VKDUH RI PLJUDQWV
choosing Ohio in our base regression and, given the parameter estimates, the predicted share
FKRRVLQJ2KLRZKHQGLVWDQFHWR2KLRLVLQFUHDVHGE\RQHVWDQGDUGGHYLDWLRQLQORJSRLQWV
Since the model is non-linear, the effect is different for every state.
Downloaded from https:/www.cambridge.org/core. The University of Tennessee - Knoxville, on 17 Mar 2017 at 13:09:33, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/S0022050715001527
3RROHG
&RHI¿FLHQWV
on Race
,QWHUDFWLRQ
%
3RROHG
&RHI¿FLHQWV
on Race
,QWHUDFWLRQ
&
±
±
±
±
±
0.193***
0.198***
0.157***
±
±
±
—
—
0.0478
—
—
±
±
—
—
—
—
—
—
0.789***
—
—
—
—
0.352***
—
—
—
—
—
—
±
—
—
—
—
0.0101**
—
—
—
—
—
—
—
—
±
—
—
—
—
±
—
—
—
—
±
—
—
White
$
White
%
White
&
Black
$
Black
%
±
±
±
±
0.193***
0.201***
0.182***
1.40***
1.10***
3HUFHQWPDQXIDFWXULQJ
3HUFHQWDJULFXOWXUH
Migrant stock
Labor market variables
Log average income
Log labor demand
Collins and Wanamaker
Black
&
3RROHG
&RHI¿FLHQWV
on Race
,QWHUDFWLRQ
$
Cost variables
Log distance
974
Downloaded from https:/www.cambridge.org/core. The University of Tennessee - Knoxville, on 17 Mar 2017 at 13:09:33, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/S0022050715001527
TABLE 4
0,*5$176257,1*&21',7,21$//2*,7&2()),&,(176
Urban
Log population
3VHXGRR2
N
±
—
—
±
—
—
0.121
—
—
±
±
—
—
0.00587*** ±
—
—
0.0143***
0.00404
—
—
—
—
0.503***
—
—
—
—
0.879***
—
—
—
—
—
—
0.21
7,498
0.21
7,498
7,498
2,114
2,114
0.29
2,114
0.22
0.22
0.27
±
6LJQL¿FDQWDWWKHSHUFHQWOHYHO
6LJQL¿FDQWDWWKHSHUFHQWOHYHO
6LJQL¿FDQWDWWKHSHUFHQWOHYHO
Notes7KHVDPSOHLQFOXGHVLQWHUVWDWHPLJUDQWVIURPWKHOLQNHGGDWDVHWRIFHQVXVUHFRUGV&ROXPQV±LQFOXGHZKLWHPHQ&ROXPQV±LQFOXGHEODFNPHQ
&ROXPQV±UHSRUWLQWHUDFWLRQWHUPVIURPDSRROHGUHJUHVVLRQZLWKHDFKUHJUHVVRULQWHUDFWHGZLWKDUDFHGXPP\ EODFN6SHFL¿FDWLRQV$DQG%GLIIHU
LQWKHLQFOXVLRQH[FOXVLRQRIODERUGHPDQGORJSRSXODWLRQSHUFHQWPDQXIDFWXULQJDQGSHUFHQWDJULFXOWXUHYDULDEOHVRQO\6SHFL¿FDWLRQ&LQFOXGHVSRWHQWLDO
GHVWLQDWLRQVWDWH¿[HGHIIHFWV'LVWDQFHUHIHUVWRFRXQW\RIUHVLGHQWLQWRSRSXODWLRQZHLJKWHGFHQWHURIHDFKSRWHQWLDOGHVWLQDWLRQVWDWH0LJUDQWVWRFN
refers to the share of persons born in state i who are residing in state j LQFDOFXODWHGZLWK,38065XJJOHVHWDO7KHODERUGHPDQGYDULDEOHIROORZV
%DUWLNDQGXVHVLQGXVWU\FRPSRVLWLRQDWWKHVWDWHOHYHOWRIRUPDZHLJKWHGDYHUDJHRIQDWLRQDOOHYHOLQGXVWU\VSHFL¿FJURZWKUDWHV/RJDYHUDJH
LQFRPHFRPELQHVLQIRUPDWLRQIURP0LWFKHQHUDQG0F/HDQDQGWKHFHQVXVPLFURGDWD5XJJOHVHWDODVGHVFULEHGLQWKHWH[WDQG$SSHQGL[
1RQ6RXWKHTXDOVRQHIRUGHVWLQDWLRQVWDWHVRXWVLGHWKH6RXWKLQFOXGLQJ0DU\ODQG'HODZDUHDQG:DVKLQJWRQ'&3HUFHQWRIODERUIRUFHLQPDQXIDFWXULQJDQG
DJULFXOWXUHSHUFHQWXUEDQDQGWRWDOSRSXODWLRQLQDUHIURP+DLQHV6WDQGDUGHUURUVFOXVWHUHGDWWKHFRXQW\RIRULJLQDUHLQSDUHQWKHVHV
Sources: Data are from the linked sample of census records, as described in the text and Appendix.
Selection and Sorting of Southern Migrants
975
Downloaded from https:/www.cambridge.org/core. The University of Tennessee - Knoxville, on 17 Mar 2017 at 13:09:33, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/S0022050715001527
Region and other control variables
Non-South
±
Collins and Wanamaker
that improvements in transportation and information networks played an
important role in facilitating internal migration and integrating U.S. labor
markets.
7KH EODFN DQG ZKLWH FRHI¿FLHQWV RQ WKH SUHH[LVWLQJ PLJUDQW VWRFN
YDULDEOH DUH SRVLWLYH VLPLODU DFURVV VSHFL¿FDWLRQV DQG QRW VWDWLVWLFDOO\
GLIIHUHQW IURP RQH DQRWKHU LQ FROXPQV WR 7KXV UHVSRQVLYHQHVV WR
migrant networks measured in this manner did not distinguish blacks’
PLJUDWLRQEHKDYLRUIURPWKDWRIZKLWHV+RZHYHUEODFNVDQGZKLWHV¶SUH
H[LVWLQJPLJUDQWVWRFNVZHUHGLVWULEXWHGGLIIHUHQWO\7KHUHIRUHQHWZRUNV
PLJKWVWLOOKDYHEHHQDQLPSRUWDQWGHWHUPLQDQWRIGLIIHUHQFHVLQÀRZV15
7KHUHVXOWVDUHIDLUO\VLPLODULQVSHFL¿FDWLRQVZLWKGHVWLQDWLRQ¿[HGHIIHFWV
FROXPQVDQGVXJJHVWLQJWKDWWKHPLJUDQWVWRFNFRHI¿FLHQWVLQRWKHU
VSHFL¿FDWLRQV DUH QRW VLPSO\ SLFNLQJ XS GHVWLQDWLRQVSHFL¿F XQREVHUYables that had drawn previous migrants. For some states, especially those
that had relatively small stocks of previous southern migrants, plausibly
VL]HGFKDQJHVLQWKHPLJUDQWVWRFNYDULDEOHGRQRWLPSO\ODUJHFKDQJHV
LQWKHSUHGLFWHGÀRZRIPLJUDQWVEHWZHHQDQG,Q&DOLIRUQLD
for example, increasing the black migrant-stock variable’s value to equal
WKDWRIZKLWHVZLWKLQHDFKVWDWHRIRULJLQZRXOGUDLVH&DOLIRUQLD¶VVKDUH
RIEODFNPLJUDQWVIURPWRSHUFHQW7KLVDFFRXQWVIRUOLWWOHRIWKH
VL]DEOHUDFLDOGLIIHUHQFHLQPLJUDWLRQWR&DOLIRUQLDSHUFHQWDJHSRLQWV
+RZHYHULQVRPHRWKHUVWDWHVWKHLPSOLFDWLRQVRIVFDOLQJXSWKHVWRFN
RISUHYLRXVPLJUDQWVDUHODUJHU,QFUHDVLQJWKHEODFNYDULDEOH¶VYDOXHWR
HTXDOWKDWRIZKLWHVIRU7H[DVZRXOGUDLVH7H[DV¶VVKDUHRIEODFNPLJUDQWV
IURPWRSHUFHQWFRPSDUHGWRDVKDUHRISHUFHQWIRUZKLWHV
7KHDYHUDJHPDUJLQDOHIIHFWRIDRQHVWDQGDUGGHYLDWLRQFKDQJHLQSUH
existing migrant stock over all states is 0.45 percentage points for whites
and 0.33 points for blacks.
7KHUHVXOWVLQFROXPQVDQGRI7DEOHLQGLFDWHWKDWZKLWHDQGEODFN
migration was strongly correlated with exogenous variation in aggregate
ODERUGHPDQGJURZWK7KHSRVLWLYHUHVSRQVHWRWKLVYDULDEOHLVQRJUHDW
surprise because booming labor markets are commonly cited as motivation
15
6WXDUWDQG7D\ORUXVHGDWDRQSODFHRIELUWK±ELUWKFRKRUWVDQGSODFHRI
UHVLGHQFHDIWHUIURP0HGLFDUHDQG6RFLDO6HFXULW\UHFRUGVWRH[DPLQHWKHUROHRIORFDO
QHWZRUNVLQEODFNDQGZKLWHLQWHUUHJLRQDOPLJUDWLRQSULPDULO\LQWKHSRVWSHULRG7KH\¿QG
that networks were more important for black inter-regional migrants from the Deep South than
IRUZKLWHLQWHUUHJLRQDOPLJUDQWVIURPWKH*UHDW3ODLQVRUWKH'HHS6RXWK$OWKRXJKDOOUHVXOWV
point towards networks being important for blacks, their results contrast with ours regarding the
comparison of blacks and whites. Whether the contrast is due to our relatively rough proxy for
QHWZRUNV RU GLIIHUHQFHV LQ HFRQRPHWULF PHWKRGV VDPSOH FRYHUDJH GH¿QLWLRQV RI PLJUDQWV RU
FKDQJHVRYHUWLPHZLOOUHTXLUHPRUHUHVHDUFK$OVRVHH&KD\DQG0XQVKLRQPLJUDWLRQ
networks.
Downloaded from https:/www.cambridge.org/core. The University of Tennessee - Knoxville, on 17 Mar 2017 at 13:09:33, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/S0022050715001527
Selection and Sorting of Southern Migrants
977
for long-distance migration and, in general, large states tend to have large
changes in employment and, therefore, attract large shares of migrants.
,W LV LQWHUHVWLQJ WKDW EODFN PHQ DSSHDU WR KDYH EHHQ KLJKO\ UHVSRQVLYH
WR WKLV VLJQDO UHODWLYH WR ZKLWH PHQ FROXPQ SHUKDSV UHÀHFWLQJ WKH
UHODWLYHLPSRUWDQFHRI¿QGLQJHPSOR\PHQWTXLFNO\XSRQDUULYDORUH[WUD
weight given to this signal as a correlate of changing employment opporWXQLWLHV IRU EODFNV VSHFL¿FDOO\ LH UHODWLYH WR SUHYLRXV GLVFULPLQDWRU\
practices in northern and western labor markets). For perspective on the
FRHI¿FLHQWLI7HQQHVVHH¶VODERUGHPDQGKDGJURZQRQHVWDQGDUGGHYLDtion faster, the estimates suggest that its share of black migrants would
KDYHLQFUHDVHGE\SHUFHQWDJHSRLQWVUHODWLYHWRDEDVHRISHUFHQWIRU
whites, the predicted increase is 2 percentage points relative to a base
RI SHUFHQW *LYHQ D RQHVWDQGDUGGHYLDWLRQ FKDQJH LQ ODERU GHPDQG
the average marginal effect over all potential destination states is 1.10
percentage points for whites and 2.27 points for blacks.
,QFROXPQVDQGZHHQWHUVHSDUDWHYDULDEOHVIRUPDQXIDFWXULQJ
and agricultural employment shares in 1910, rather than aggregate
employment growth, to provide a different perspective on the economic
FKDUDFWHULVWLFV RI VWDWHV WKDW PD\ KDYH LQÀXHQFHG PLJUDWLRQ SDWWHUQV
Black men were inclined to select states with high levels of manufacturing employment, but disinclined to select agricultural states, all else
WKHVDPH,WLVQRWDEOHWKDWWKHFRHI¿FLHQWRQPDQXIDFWXULQJLVSRVLWLYH
HYHQ ZKHQ D ³1RQ6RXWK´ GXPP\ YDULDEOH XUEDQ VKDUH DQG DYHUDJH
LQFRPHDUHLQFOXGHGDVLQGHSHQGHQWYDULDEOHVLQWKHVSHFL¿FDWLRQ,QRWKHU
ZRUGVWKHSRVLWLYHPDQXIDFWXULQJFRHI¿FLHQWIRUEODFNVVXJJHVWVWKDWWKH
sector itself was an important determinant of black migration patterns.
)RUZKLWHPHQERWKWKHPDQXIDFWXULQJDQGDJULFXOWXUDOFRHI¿FLHQWVDUH
negative in column 2, suggesting that they tended to seek residence in
VWDWHV ZLWK KLJKHU OHYHOV RI HPSOR\PHQW LQ ³RWKHU´ LQGXVWULHV VXFK DV
construction, mining, trade, transportation, and services.
We also estimate models that distinguish among parts of the South
that were differentially affected by the spread of the boll weevil. We
GH¿QHGWKUHHJURXSVRIVWDWHVGHSHQGLQJRQWKHLUFRWWRQLQWHQVLW\DQGWKH
EROOZHHYLO¶VWLPHRIDUULYDO:HFRQVLGHUHGVWDWHV³FRWWRQLQWHQVLYH´LI
at least 20 percent of total crop value came from cotton production in
+DLQHV %DVHG RQ WKH PDS IURP :DOWHU +XQWHU DQG %5
&RDGZHFRGHG7H[DV$UNDQVDV/RXLVLDQDDQG0LVVLVVLSSLDV
³HDUO\EROOZHHYLO´VWDWHV7KH\ZHUHFRWWRQLQWHQVLYHDQGUHFHLYHGWKH
EROOZHHYLOSULRUWRRXUZLQGRZRIREVHUYDWLRQEHIRUH:HFRGHG
$ODEDPD *HRUJLD 6RXWK &DUROLQD 1RUWK &DUROLQD DQG 2NODKRPD DV
³ODWHEROOZHHYLO´VWDWHV7KH\ZHUHFRWWRQLQWHQVLYHDQGUHFHLYHGWKHEROO
Downloaded from https:/www.cambridge.org/core. The University of Tennessee - Knoxville, on 17 Mar 2017 at 13:09:33, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/S0022050715001527
978
Collins and Wanamaker
ZHHYLODIWHUZLWKLQRXUZLQGRZRIREVHUYDWLRQ7KHWKLUGJURXSLV
FRPSULVHGRI7HQQHVVHH.HQWXFN\9LUJLQLD:HVW9LUJLQLDDQG)ORULGD
which were not cotton intensive.
,IWKHEROOZHHYLOLPSDUWHGDORQJODVWLQJQHJDWLYHSURGXFWLYLW\VKRFN
WKDWLVQRWFDSWXUHGE\FRQWUROYDULDEOHVDQGZDVQRWQHXWUDOL]HGE\FRWWRQ
price changes or the redistribution of economic activity to other crops or
sectors, then we would expect to see less migration to cotton-intensive
SODFHVLQRXUVDPSOHFHWHULVSDULEXV:H¿QGWKDWUHODWLYHWRVWDWHVZKHUH
FRWWRQZDVOHVVLPSRUWDQWWKHRPLWWHGFDWHJRU\EODFNVDQGZKLWHVZHUH
ERWK OHVV OLNHO\ WR PRYH WR FRWWRQLQWHQVLYH VWDWHV ZLWK WKH H[FHSWLRQ
RI EODFNV LQ VSHFL¿FDWLRQ % FRQVLVWHQW ZLWK D QHJDWLYH DQG SHUVLVWHQW
productivity shock across cotton-producing regions.17 &RHI¿FLHQWV DUH
PRUHQHJDWLYHIRUWKH³ODWHÉROOZHHYLOVWDWHVWKDQIRUWKH³HDUO\ÉROO
ZHHYLOVWDWHVEXWWKHGLIIHUHQFHVEHWZHHQ³ODWH´DQG³HDUO\´VWDWHJURXSV
DUHQRWVWDWLVWLFDOO\VLJQL¿FDQW
,Q 7DEOH WKH FRHI¿FLHQWV RQ WKH DYHUDJH LQFRPH YDULDEOH ZKLFK
varies by destination state and race, are positive but somewhat weaker
LQ VSHFL¿FDWLRQV WKDW LQFOXGH FRQWUROV IRU LQGXVWULDO FRPSRVLWLRQ 7KH
FRHI¿FLHQWV LQ FROXPQV DQG LPSO\ WKDW D RQHVWDQGDUGGHYLDWLRQ
LQFUHDVHLQLQFRPHLQ3HQQV\OYDQLDZRXOGKDYHUDLVHGLWVVKDUHRIZKLWH
PLJUDQWVE\DERXWSHUFHQWDJHSRLQWVDQGLWVVKDUHRIEODFNPLJUDQWV
by about 1.7 percentage points, relative to base shares of 3 and 8 percent,
respectively. Columns 7 and 8 suggest that white migrants were more
UHVSRQVLYHWKDQEODFNVWRYDULDWLRQLQSUHZDULQFRPH,WLVSRVVLEOHWKDW
whites, perhaps because they had higher levels of education or access
to more information, were better informed than blacks about wages in
distant states and, therefore, more responsive to the existing variation.
But again, there is no evidence that literate blacks were more responsive
WKDQ LOOLWHUDWH EODFNV WR YDULDWLRQ LQ DYHUDJH LQFRPH 2QOLQH $SSHQGL[
7DEOH$QRWKHUSRVVLELOLW\LVWKDWWKHHVWLPDWHVRIUDFHVSHFL¿FLQFRPH
circa 1910 are a better proxy for whites’ expected earnings opportunities
after 1910 than for blacks.
)RUERWKZKLWHDQGEODFNPHQLQFROXPQVDQGRI7DEOHWKH
FRHI¿FLHQWRQWKH1RQ6RXWKYDULDEOHLVQHJDWLYH7DNHQDWIDFHYDOXH
5HVXOWV FKDQJH RQO\ VOLJKWO\ LI ZH VKLIW 7HQQHVVHH ZKHUH FRWWRQ ZDV SHUFHQW RI FURS
YDOXHLQWRWKH³ODWHEROOZHHYLO´FDWHJRU\1RWHWKDWVWDWHVZHUHQRWQHDWO\GLYLGHGDFFRUGLQJWR
the boll weevil’s presence in 1910. Our coding is approximate and based on visual inspection of
+XQWHUDQG&RDG¶VPDS
17
5HVXOWV DUH UHSRUWHG LQ WKH 2QOLQH $SSHQGL[ 7DEOH :KHQ WKH GDWD DUH SRROHG EODFN
ZKLWHGLIIHUHQFHVLQDWWUDFWLRQWRWKHHDUO\DQGODWHEROOZHHYLOVWDWHVDUHVWDWLVWLFDOO\LQVLJQL¿FDQW
5HVXOWVDUHTXDOLWDWLYHO\VLPLODUZKHQVDPSOHVDUHUHVWULFWHGWRIDUPUHVLGHQWVQRWVKRZQ
Downloaded from https:/www.cambridge.org/core. The University of Tennessee - Knoxville, on 17 Mar 2017 at 13:09:33, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/S0022050715001527
Selection and Sorting of Southern Migrants
979
neither white nor black southern migrants were especially attracted to
the Non-South, conditional on other X variables.187KHUHLVVRPHFRQFHUQ
that controlling for pre-existing migrant stocks may absorb some of
the attraction of regional amenities to the extent that previous migrants
UHVSRQGHGWRWKRVHVDPHDPHQLWLHV+RZHYHURPLWWLQJWKHPLJUDQWVWRFN
YDULDEOHIURPWKHVSHFL¿FDWLRQVLQFROXPQVDQGKDVOLWWOHHIIHFWRQWKH
1RQ6RXWKFRHI¿FLHQWIRUEODFNPLJUDQWVQRWVKRZQ,QFROXPQVDQG
WKHFRHI¿FLHQWVRQ1RQ6RXWKLQWHUDFWHGZLWKUDFHEODFN DUHQRW
VWDWLVWLFDOO\VLJQL¿FDQWZKLFKDJDLQUXQVFRXQWHUWRVRPHRIWKHOLWHUDture’s emphasis on the idea that black migrants were especially motiYDWHGWRHVFDSHWKH6RXWKEH\RQGWKDWZKLFKLVHPEHGGHGLQWKHRWKHUX
variables).19
$PRUHQXDQFHGSLFWXUHHPHUJHVE\VHSDUDWHO\HVWLPDWLQJFRHI¿FLHQWV
IRU1RUWKHDVW0LGZHVWDQG:HVWH[SUHVVHGUHODWLYHWR6RXWKLQVWHDG
RIMXVW³1RQ6RXWK´(VWLPDWHVRIWKHVHFRHI¿FLHQWVDUHORFDWHGLQ7DEOH
%ODFNPLJUDQWVZHUHVLJQL¿FDQWO\PRUHDWWUDFWHGWRWKH1RUWKHDVWDQG
Midwest than whites conditional on the other X variables, but they were
VLJQL¿FDQWO\ OHVV OLNHO\ WR PRYH WR WKH :HVW 7KXV WKH ³1RQ6RXWK´
YDULDEOH LQ 7DEOH WHQGV WR PDVN VXEVWDQWLDO XQH[SODLQHG GLIIHUHQFHV
LQ EODFNV¶ DQG ZKLWHV¶ FKRLFHV RI VXEUHJLRQ RXWVLGH WKH 6RXWK 7KHVH
UHVLGXDO GLIIHUHQFHV LQ PLJUDWLRQ PLJKW UHÀHFW GLIIHUHQFHV LQ ODERU
market discrimination, civil liberties, or social norms across northern and
ZHVWHUQORFDWLRQVDIWHUDK\SRWKHVLVWKDWPHULWVIXUWKHUUHVHDUFK,Q
this interpretation, white migrants serve as a control group to capture the
LQÀXHQFHRIXQREVHUYHGUDFHQHXWUDOVXEUHJLRQDOFKDUDFWHULVWLFVVRWKDW
black-white differences might then be interpreted as evidence of raceVSHFL¿FIDFWRUVWKDWYDU\RYHUVXEUHJLRQV1RWHWKDWIRUEODFNVFROXPQV
DQG WKH FRHI¿FLHQWV IRU 0LGZHVW DQG 1RUWKHDVW DUH VWLOO QHJDWLYH
H[SUHVVHGUHODWLYHWRWKH6RXWKEXWWKH\DUHVPDOOHULQPDJQLWXGHWKDQ
WKHQHJDWLYHFRHI¿FLHQWVIRUZKLWHV
5HWXUQLQJWR7DEOH¶VUHVXOWVLQFROXPQVDQGWKHFRHI¿FLHQWRQ
³XUEDQ´LVQHJDWLYHIRUZKLWHPHQEXWSRVLWLYHIRUEODFNPHQ7KHFRHI¿FLHQWVVXJJHVWVPDOOHIIHFWVRQPLJUDWLRQSDWWHUQV5DLVLQJ0LVVLVVLSSL¶V
18
9LJGRU¿QGVWKDWEODFNPLJUDQWVWRFLWLHVZHUHPRUHOLNHO\WRFKRRVHQRQVRXWKHUQ
ORFDWLRQVLQWKHFHQVXVFURVVVHFWLRQ+LVFRQGLWLRQDOORJLWIUDPHZRUNGLIIHUVIURPRXUVLQ
VHYHUDOZD\V+HXVHVFHQVXVFURVVVHFWLRQVLQDQGDGLIIHUHQWVHWRIFRYDULDWHVFLWLHV
DVGHVWLQDWLRQFKRLFHVDQGDVDPSOHRILQWHUVWDWHPLJUDQWVZKRUHVLGHGLQFLWLHVLQRPLWWLQJ
interstate migrants who do not reside in cities).
19
,Q D VSHFL¿FDWLRQ WKDW H[SUHVVHV YDULDEOHV LQ OHYHOV UDWKHU WKDQ ORJV WKH FRHI¿FLHQWV RQ
³1RQ6RXWK´ DUH FRQVLVWHQW ZLWK WKH K\SRWKHVLV WKDW EODFNV ZHUH PRUH DWWUDFWHG WR WKH 1RUWK
WKDQZKLWHVEXWWKHFRHI¿FLHQWIRUEODFNVLVVWLOOQRQSRVLWLYHWKHZKLWHFRHI¿FLHQWLVODUJHULQ
PDJQLWXGHDQGQHJDWLYH6HHWKH2QOLQH$SSHQGL[7DEOH
Downloaded from https:/www.cambridge.org/core. The University of Tennessee - Knoxville, on 17 Mar 2017 at 13:09:33, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/S0022050715001527
980
Collins and Wanamaker
TABLE 5
0,*5$176257,1*&21',7,21$//2*,7&2()),&,(176215(*,21$/
,1',&$7256
White
$
White
%
Black
$
Black
%
3RROHG
3RROHG
&RHI¿FLHQWV &RHI¿FLHQWV
on Race
on Race
,QWHUDFWLRQ
,QWHUDFWLRQ
%
$
5HJLRQDO,QGLFDWRUV6RXWKLVRPLWWHG
Northeast
±
±
±
±
0.733***
Midwest
±
±
± ±
0.318**
0.472***
West
1.054***
1.224***
±
0.381*
±
±
3VHXGRR2
0.24
0.24
0.25
0.25
N
7,498
7,498
2,114
2,114
Notes $OO UHJUHVVLRQV DOVR LQFOXGH FRUUHVSRQGLQJ FRYDULDWHV UHSRUWHG LQ 7DEOH ZLWK WKH
H[FHSWLRQRIWKH³1RQ6RXWK´LQGLFDWRU5HJLRQDOGH¿QLWLRQVIROORZVWDQGDUGFHQVXVUHJLRQVZLWK
WKH H[FHSWLRQ RI 'HODZDUH 0DU\ODQG DQG :DVKLQJWRQ '& ZKLFK ZH FODVVLI\ DV ³1RUWKHDVW´
UDWKHUWKDQ³6RXWK´6HHQRWHVWR7DEOHIRUPRUHGHWDLOV
Sources: Data are from the linked sample of census records, as described in the text and Appendix.
XUEDQVKDUHE\RQHVWDQGDUGGHYLDWLRQSHUFHQWDJHSRLQWVZRXOGUDLVH
its share of black migrants by only 0.5 percentage points, relative to a
EDVHRISHUFHQW%HFDXVHXUEDQL]DWLRQLVKLJKO\FRUUHODWHGZLWKPDQXIDFWXULQJDQGDJULFXOWXUDOHPSOR\PHQWZHGRQRWHPSKDVL]HWKHFRHI¿cient on urban in columns 2 and 5.
,QVXPWKHDUWLFOH¶VHDUOLHUDQDO\VLVIRXQGWKDWLWZDVQRWSRVVLEOHWR
account for differences in white and black migration patterns with the
LQGLYLGXDOV¶ EDFNJURXQG FKDUDFWHULVWLFV ,QVWHDG WKH GLIIHUHQW SDWWHUQV
must have resulted from differences in perceived opportunities and
PLJUDWLRQFRVWVDFURVVVSDFHX) and differences in how men responded
WRWKRVHRSSRUWXQLWLHVDQGFRVWVȖ7RJDXJHKRZPXFKRIWKHEODFN
white differences in destination choice can be accounted for by distance
and migrant stock variables, we apply the parameters of the black condiWLRQDOORJLWPRGHOZLWKGHVWLQDWLRQ¿[HGHIIHFWVWRDOOPHQLQWKHVDPSOH
EODFNDQGZKLWHDQGWKHQDSSO\WKHSDUDPHWHUVRIWKHZKLWHPRGHOWRDOO
Downloaded from https:/www.cambridge.org/core. The University of Tennessee - Knoxville, on 17 Mar 2017 at 13:09:33, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/S0022050715001527
Selection and Sorting of Southern Migrants
981
men.207KLV\LHOGVWZRFRXQWHUIDFWXDOGLVWULEXWLRQVRIPLJUDQWVZLWKWKH
same underlying XYDOXHV7KHGLVVLPLODULW\LQGH[RIWKHWZRFRXQWHUIDFtual distributions is approximately 0.20, substantially less than the unadMXVWHGLQGH[YDOXHRI)URPWKLVSHUVSHFWLYHHTXDOL]LQJWKHPLJUDQW
stock and distance variables across black and white migrants reduces the
GLVVLPLODULW\ LQGH[ E\ DERXW RQHIRXUWK 7KH UHPDLQLQJ GLIIHUHQFHV LQ
PLJUDWLRQSDWWHUQVUHÀHFWGLIIHUHQFHVLQUDFHVSHFL¿FZDJHVDFURVVVWDWHV
DEVRUEHG E\ UDFHVSHFL¿F GHVWLQDWLRQ ¿[HG HIIHFWV DQG PRGHO FRHI¿FLHQWVWKHȖ¶VZKLFKDUHGLI¿FXOWWRLQWHUSUHWGH¿QLWLYHO\1RQHWKHOHVV
ZHEHOLHYHWKDWDGGLWLRQDOUHVHDUFKFRXOGLVRODWHDQGWHVWVSHFL¿FK\SRWKHVHVDQGLQWHUSUHWDWLRQVHJZKHWKHUWKHSDWWHUQRIEODFNZKLWHGLIIHUHQFHVLQGHVWLQDWLRQ¿[HGHIIHFWVFRUUHODWHVZLWKDPHDVXUHRIORFDOODERU
market discrimination, perhaps following Sundstrom 2007).
&21&/86,216
7R VWXG\ PLJUDWLRQ SDWWHUQV GXULQJ WKH ¿UVW GHFDGHV RI WKH *UHDW
0LJUDWLRQ ZH EXLOW D QHZ GDWDVHW RI PRUH WKDQ VRXWKHUQ PHQ
WKDWZHOLQNHGIURPWKHWRWKHFHQVXV7KHGDWDVHWLVQRYHO
LQWKDWLWREVHUYHVWKHVDPHPHQEHIRUHDQGDIWHUWKHVWDUWRIWKH*UHDW
0LJUDWLRQ7KHVFRSHRIDQDO\VLVLVQRYHOLQWKDWLWLQFOXGHVERWKZKLWH
and black southerners and it incorporates information about those who
moved within the South as well as those who left the region.
First, we study selection into inter-state and inter-regional migraWLRQ7KHUHLVVRPHHYLGHQFHRISRVLWLYHVHOHFWLRQLQWRPLJUDWLRQDPRQJ
both whites and blacks in terms of occupational status, and it is clear
WKDWQRQIDUPUHVLGHQWVLQZHUHPRUHOLNHO\WRPRYHDFURVVVWDWH
and regional lines than farm residents. Overall, however, the differences
between migrants and non-migrants were small within race categories,
HYHQ IRU WKRVH PRYLQJ LQWHUUHJLRQDOO\ ,Q WKLV VHQVH SDUWLFLSDWLRQ LQ
internal migration by southern men was remarkably widespread after
1910.
Second, we examine migration patterns between origins and destinations and ask whether individual and local background characteristics
account for differences in black and white migration choices. Although
Note that distance and migrant stock variables differ by race across destinations, and their
FRHI¿FLHQWV DUH LGHQWL¿HG LQ WKH FRQGLWLRQDO ORJLW PRGHO ZLWK GHVWLQDWLRQ ¿[HG HIIHFWV 7KH
³DOO ZKLWH´ DQG ³DOO EODFN´ FRXQWHUIDFWXDOV DUH HVWLPDWHG ZLWKD SRROHG VDPSOH WKXV WKH VDPH
underlying distributions of distance and migrant stock variables.
20
Downloaded from https:/www.cambridge.org/core. The University of Tennessee - Knoxville, on 17 Mar 2017 at 13:09:33, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/S0022050715001527
982
Collins and Wanamaker
there was substantial overlap in black and white migrants’ choices of
GHVWLQDWLRQV WKHUH ZHUH DOVR QRWDEOH GLIIHUHQFHV ,Q 2/6 DQG PXOWLQRPLDOORJLWIUDPHZRUNVZH¿QGWKDWRQQHWDVXUSULVLQJO\VPDOODPRXQW
of the black-white differences in migration patterns can be accounted for
E\GLIIHUHQFHVLQEDFNJURXQGFKDUDFWHULVWLFV,QRWKHUZRUGVWRWKHH[WHQW
that black and white migration patterns differed, those differences were
QRWLQJHQHUDODVLPSOHUHÀHFWLRQRIZKHUHWKHPHQVWDUWHGJHRJUDSKLcally or their personal circumstances circa 1910. Rather, observationally
similar men made different location choices depending on their race.
7KLUGJLYHQWKH¿QGLQJVDERYHZHPHDVXUHEODFNDQGZKLWHPLJUDQWV¶
responsiveness to variation in the characteristics of potential destinations.
Black and white men were similarly responsive to pre-existing migrant
stocks, but black men were more deterred by distance than whites,
more attracted to manufacturing centers, and more responsive to variation in labor demand growth. Conditional on the potential destination
VWDWHV¶FKDUDFWHULVWLFVZH¿QGRQO\PL[HGHYLGHQFHWKDWEODFNPLJUDQWV
ZHUH PRUH OLNHO\ WKDQ ZKLWHV WR OHDYH WKH 6RXWK 7KHUH LV KRZHYHU
interesting variation across areas outside the South, with blacks sorting
more strongly than whites into the Midwest and Northeast and whites
sorting more strongly into the West, conditional on state characteristics.
Variation in the characteristics of potential destinations, such as distance
and pre-existing migrant stocks, can account for a non-trivial portion of
the black-white dissimilarity index in destinations, but a larger portion is
associated with racial differences in responsiveness to the destinations’
FKDUDFWHULVWLFVDQGLQGHVWLQDWLRQ¿[HGHIIHFWV7KHVHGLIIHUHQFHVLQYLWH
further research.
7KH DUWLFOH¶V ¿QGLQJV UHO\ KHDYLO\ RQ WKH OLQNHG GDWDVHW RI FHQVXV
records for southern males from 1910 to 1930, and it is our hope that
RWKHU VFKRODUV ZLOO ¿QG WKLV GDWDVHW WR EH XVHIXO :H KDYH HQGHDYRUHG
to keep the analyses focused on a set of fundamental questions about
WKH QDWXUH RI LQWHUQDO PLJUDWLRQ LQ WKLV SHULRG EXW LW LV QRW GLI¿FXOW WR
LPDJLQHH[WHQVLRQVWKDWORRNDW¿QHUOHYHOVRIJHRJUDSKLFGHWDLOH[SORLW
more information on the migrants’ outcomes, or link to other historical or
administrative datasets. More generally, it is clear that this kind of dataset
KROGV JUHDW SURPLVH IRU KLVWRULFDO UHVHDUFK RQ PLJUDWLRQ HJ )HUULH
$EUDPLW]N\%RXVWDQDQG(ULNVVRQDQGQHZWHFKQRORJLHV
and initiatives are likely to make them far more common and accessible
LQWKHIXWXUH/\LQJMXVWEH\RQGWKHVFRSHRIWKLVDUWLFOHWKHUHLVPXFK
WKDWUHPDLQVWREHOHDUQHGDERXWPLJUDWLRQGXULQJWKH*UHDW'HSUHVVLRQ
WKHJURZWKRIEODFNJKHWWRVDQGVXEXUEDQL]DWLRQRIZKLWHVWKHLQWHUDFWLRQ
Downloaded from https:/www.cambridge.org/core. The University of Tennessee - Knoxville, on 17 Mar 2017 at 13:09:33, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/S0022050715001527
Selection and Sorting of Southern Migrants
983
of migration and intergenerational mobility, and much more. All these
topics are ripe for reassessment as new datasets that follow individuals
over time are brought to light.
Appendix
)RUHDFKPDOHDJHWRLQWKH,3806VDPSOH5XJJOHVHWDOZHUDQ
two searches in the 1930 census manuscripts: one with exact last name and one with a
6281'(;YHUVLRQRIWKHODVWQDPH,QERWKFDVHVZHFRQGLWLRQHGRQWKH¿UVWWKUHHOHWWHUV
RIWKH¿UVWQDPHRUOHVVLIVKRUWHUWKDQWKUHHOHWWHUVLQOHQJWKH[DFWVWDWHRIELUWKDJHZLWKLQ
two years, race, and gender. We counted any individual with a unique match in the exact
last name or SOUNDEX search as a successful match. We then eliminated all duplicate
PDWFKHVLHWZRGLIIHUHQWLQGLYLGXDOVLQOLQNHGWRWKHVDPHLQGLYLGXDOLQWKH
census).
From the 1930 census, we extracted detailed location of residence as well as inforPDWLRQ RQ OLWHUDF\ HPSOR\PHQW RFFXSDWLRQ LQGXVWU\ DQG YHWHUDQ VWDWXV ,QGLYLGXDO
and household-level variables available in the linked sample are detailed in Appendix
7DEOH7KHDYDLODELOLW\RIVRPHYDULDEOHVGHSHQGVRQDQLQGLYLGXDO¶VDJHDWWKHWLPH
RI HQXPHUDWLRQ $QG DOWKRXJK VRPH YDULDEOHV DUH WHFKQLFDOO\ ³DYDLODEOH´ IRU DOO DJH
JURXSVWKH\DUHSUDFWLFDOO\PLVVLQJHJoccupationIRU±\HDUROGV
7KH FKDUDFWHULVWLFV RI PDOHV LQ WKH PDWFKHG VDPSOH DUH FRPSDUHG WR WKRVH RI WKH
HQWLUH,3806VDPSOHLQ7DEOHRIWKHPDLQWH[W7KHUHZHFRQFOXGHGWKDWWKH
GLIIHUHQFHVDUHQRWHFRQRPLFDOO\VLJQL¿FDQWDQGDUHXQOLNHO\WRELDVWKHVWXG\RIPLJUDWLRQ SDWWHUQV +HUH ZH DOVR SURYLGH HVWLPDWHV RI D OLQHDU SUREDELOLW\ PRGHO RI EHLQJ
IRXQGLQRXWRIWKRVHLQWKHRULJLQDO,3806VDPSOHFRQGLWLRQDORQREVHUYDEOH
FKDUDFWHULVWLFV7KHUHVXOWVDUHORFDWHGLQ$SSHQGL[7DEOHDQGODUJHO\HFKRWKH
FRQFOXVLRQV IURP 7DEOH LQ WKH PDLQ WH[W 6SHFL¿FDOO\ ZH HVWLPDWH WKH SUREDELOLW\
of being located in the 1930 manuscripts separately by race and three age categories:
± ± DQG ± 7KH DJH JURXSLQJV FRUUHVSRQG WR WKH DYDLODELOLW\ RI VFKRRO
attendance, literacy, and farming occupation information in the 1910 data. We observe
a slightly increased probability of being found for literate individuals, farmers, and resiGHQWVRI:HVW9LUJLQLDFRQFOXVLRQVKLJKOLJKWHGLQGLVFXVVLRQRI7DEOHDVZHOO2QWKH
RWKHUKDQGKRPHRZQHUVKLSLVDVWDWLVWLFDOO\LQVLJQL¿FDQWSUHGLFWRURIVDPSOHLQFOXVLRQ
FRQGLWLRQDORQRWKHUIDFWRUVREVHUYDEOHLQ,QHDFKFDVHWKHHQKDQFHGSUREDELOLW\
RI EHLQJ ORFDWHG LQ WKH PDQXVFULSWV LV VPDOO ± SHUFHQWDJH SRLQWV LQ WKH
FDVHRIOLWHUDF\DQG±SHUFHQWDJHSRLQWVIRUIDUPLQJRFFXSDWLRQV5HVLGHQFHLQ
West Virginia in 1910 raises the probability of being located in 1930 by as much as 5.5
SHUFHQWDJHSRLQWVEXWIRUDUHODWLYHO\VPDOOQXPEHURIPHQSHUFHQWRIZKLWHPDOHV
in the sample lived in West Virginia in 1910).
,Q$SSHQGL[7DEOHZHFRPSDUHWKHFKDUDFWHULVWLFVRILQWHUUHJLRQDOPLJUDQWV
LQWKHOLQNHGVDPSOHWRWKRVHLQWKH,3806VDPSOH7KHPDWFKHGVDPSOH¶VLQWHU
regional migrants are individuals with a southern state of residence in 1910 and a nonVRXWKHUQUHVLGHQFHLQ,3806PLJUDQWVRQWKHRWKHUKDQGDUHLQGLYLGXDOVZLWK
a southern place of birthDQGDQRQVRXWKHUQUHVLGHQFHLQ7KXVLQGLYLGXDOVZKR
Downloaded from https:/www.cambridge.org/core. The University of Tennessee - Knoxville, on 17 Mar 2017 at 13:09:33, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/S0022050715001527
984
Collins and Wanamaker
APPENDIX TABLE 1
9$5,$%/($9$,/$%,/,7<±/,1.('6$03/(2)&(16865(&25'6
1910
Race
Owner-occupied housing status
Value of home/rent paid
Farm residence status
School attendance
Literacy
Marital status
City population
,QGXVWU\RIHPSOR\PHQW
Father’s industry of employment
Occupation
Father’s occupation
Employment status
Father’s employment status
State of residence
County of residence
State of birth
Age
Veteran status
1930
$JHV±
$JHV±
$JHV±
All Ages
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
Notes:7KHWDEOHUHSRUWVGDWDDYDLODELOLW\IRUDQGREVHUYDWLRQVLQWKHOLQNHGVDPSOHRI
FHQVXVUHFRUGV6RPHYDULDEOHVHJIDWKHU¶VLQGXVWU\RIHPSOR\PHQWLQDUHRFFDVLRQDOO\
DYDLODEOHIRUXQFKHFNHGDJHJURXSVEXWXVXDOO\VXFKLQIRUPDWLRQLVPLVVLQJ2EVHUYLQJIDWKHU¶V
LQGXVWU\ ZRXOG UHTXLUH FRUHVLGHQFH ZLWK IDWKHU LQ ,Q JHQHUDO YDULDEOH GH¿QLWLRQV DUH
FRQVLVWHQWDFURVV\HDUV:HXVHWKH,3806YDULDEOHocc1950 and ind1950 for 1910 and coded the
1930 variables into the same categories. *For marital status in this column, availability pertains
WRDJHV±UDWKHUWKDQ±
Source: Linked sample of census records, as described in the text.
migrated out of the South prior to 1910 will not be included in our sample but will
EH LQFOXGHG LQ WKH ,3806 VDPSOH :H FDQQRW FRUUHFW IRU WKHVH IXQGDPHQWDO VDPSOH
differences.21 Nonetheless, the differences across samples are small even when they
DUHVWDWLVWLFDOO\VLJQL¿FDQW7KHPRVWQRWDEOHH[FHSWLRQLVWKHVKDUHRIZKLWHPLJUDQWV
LQRZQHURFFXSLHGKRXVLQJLQWKHOLQNHGVDPSOHSHUFHQWUHODWLYHWRWKHFURVV
VHFWLRQDOVDPSOHSHUFHQW$VPHQWLRQHGLQWKHPDLQWH[WZKHQFRPSDULQJGHVWLQDWLRQ VKDUHV LQ WKH OLQNHG DQG FURVVVHFWLRQDO GDWDVHWV VHSDUDWHO\ IRU EODFN DQG ZKLWH
inter-state migrants), the correlation in migration patterns is high. With inter-regional
GHVWLQDWLRQVWDWHVJURXSHGDVLQ$SSHQGL[7DEOHWKHFRUUHODWLRQLVPRUHWKDQIRU
both whites and blacks.
21
We focus on inter-regional migrants for comparison because the relatively high rate of withinSouth migration prior to 1910 is likely to confound comparisons for within-South migrants across
the datasets.
Downloaded from https:/www.cambridge.org/core. The University of Tennessee - Knoxville, on 17 Mar 2017 at 13:09:33, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/S0022050715001527
985
Selection and Sorting of Southern Migrants
APPENDIX TABLE 2
(67,0$7('352%$%,/,7<2),1&/86,21,10$7&+('6$03/(
%<$*(*5283$1'5$&(
White Males
Age 0 to 9
+RPHRZQHUVKLS
Black Males
10 to 18
19 to 40
±
±
±
0.0012
0.0044
0.001
School attendance
Literacy
Age 0 to 9
10 to 18
19 to 40
—
0.0143
—
—
±
—
—
—
—
—
—
0.024***
0.017***
—
0.0078*
0.0081***
—
—
0.0094
±
0.0028
±
0.0041
0.0077
0.0019
0.0050
0.018
0.020***
0.010
0.029***
0.0044
±
0.029***
0.011
0.014
±
±
±
0.015
0.0017
±
±
±
±
0.028
±
±
±
±
0.017
±
±
±
±
.HQWXFN\ ±
±
±
±
0.0025
±
0.025
±
0.005
±
City population
&LW\SRS City pop. >25,000
Father is farmer/
individual is farmer†
6WDWHRIUHVLGHQFH
Alabama
Arkansas
Florida
*HRUJLD ±
Louisiana ±
Mississippi
0.021
0.014
0.031
0.037
±
North Carolina ±
0.0011
0.010
0.0019
±
±
0.011
±
±
±
±
Oklahoma
Downloaded from https:/www.cambridge.org/core. The University of Tennessee - Knoxville, on 17 Mar 2017 at 13:09:33, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/S0022050715001527
Collins and Wanamaker
APPENDIX TABLE 2 (CONTINUED)
(67,0$7('352%$%,/,7<2),1&/86,21,10$7&+('6$03/(
%<$*(*5283$1'5$&(
White Males
Age 0 to 9
Black Males
10 to 18
19 to 40
Age 0 to 9
10 to 18
19 to 40
0.011
0.0072
0.040
0.029
±
0.019
0.017
±
±
±
6WDWHRIUHVLGHQFH
South Carolina
7HQQHVVHH
7H[DV
West Virginia
±
±
±
0.028
0.055***
0.041***
±
±
N
25,791
9,214
7,331
$GMXVWHGR2
0.0032
0.0031
0.0017
0.00050
0.0025
6LJQL¿FDQWDWWKHSHUFHQWOHYHO
6LJQL¿FDQWDWWKHSHUFHQWOHYHO
6LJQL¿FDQWDWWKHSHUFHQWOHYHO
Notes2PLWWHGFDWHJRULHVLQWKHUHJUHVVLRQVDUH1R&LW\&LW\3RSXODWLRQDQG
9LUJLQLD6WDWHRI5HVLGHQFH7KHPRGHODOVRLQFOXGHVDJH¿[HGHIIHFWVQRWVKRZQ
†
,QGLFDWRUIRUIDWKHU¶VIDUPLQJRFFXSDWLRQDJHV±RUVHOIDJHV
Sources: 7KHOLQNHGVDPSOHLVFUHDWHGE\WDNLQJWKH,3806VDPSOHRIZKLWHDQG
EODFNPHQDJH±ZKRUHVLGHLQWKH6RXWKDQGVHDUFKLQJIRUWKHVHPHQLQWKH
FHQVXVPDQXVFULSWV7KHWH[WFRQWDLQVGHWDLOVRQVDPSOHFRQVWUXFWLRQ7KH,3806
GDWDDUHIURP5XJJOHVHWDO
Downloaded from https:/www.cambridge.org/core. The University of Tennessee - Knoxville, on 17 Mar 2017 at 13:09:33, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/S0022050715001527
Black Migrants
Matched Sample
of Migrants
5HVLGHGLQ6RXWK
in 1910)
,38066DPSOH
of Migrants
%RUQLQ6RXWK
p-value of
Difference
0.01
0.03
0.00
0.01
0.07
0.14
0.04
0.01
0.02
0.00
0.10
0.08
0.00
0.04
0.01
0.08
0.15
0.04
0.00
0.02
0.00
0.08
0.00
0.77
0.81
0.05
0.57
0.93
0.33
0.42
0.75
0.13
0.88
—
0.08
0.10
0.40
0.03
0.53
0.78
0.02
Matched Sample
of Migrants
5HVLGHGLQ6RXWK
in 1910)
0.02
0.14
0.02
0.05
0.05
0.09
0.01
0.05
0.01
0.01
0.08
0.01
0.01
,38066DPSOHRI
Migrants
ERUQLQ6RXWK
0.03
0.15
0.02
0.05
0.08
0.07
0.01
0.03
0.01
0.07
0.01
0.01
0.01
p-value of
Difference
0.12
0.38
0.79
0.08
0.23
0.71
0.00
0.00
0.03
0.51
0.30
0.04
0.07
0.31
0.33
987
'LVWULEXWLRQRIVWDWHRIUHVLGHQFH
$UL]RQD
California
Colorado
&RQQHFWLFXW5KRGH,VODQG
Delaware
D.C./Maryland
,GDKR2UHJRQ:DVKLQJWRQ
,OOLQRLV
,QGLDQD
,RZD
.DQVDV
0DLQH1+9HUPRQW
Massachusetts
Michigan
Minnesota
Missouri
Montana/Wyoming
Nebraska
Nevada/Utah
White Migrants
Selection and Sorting of Southern Migrants
Downloaded from https:/www.cambridge.org/core. The University of Tennessee - Knoxville, on 17 Mar 2017 at 13:09:33, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/S0022050715001527
APPENDIX TABLE 3
&203$5,6212)/,1.('$1')8//6$03/(66287+(510,*5$176
727+(1216287+
Black Migrants
White Migrants
Linked Sample
of Migrants
5HVLGHGLQ6RXWK
in 1910)
Linked Sample
of Migrants
5HVLGHGLQ6RXWK
in 1910)
,38066DPSOH
of Migrants
%RUQLQ6RXWK
p-value of
Difference
0.12
0.00
0.14
0.15
0.01
0.08
0.13
0.12
0.17
0.10
0.32
0.45
0.14
0.08
0.25
0.02
0.03
0.05
0.17
0.01
0.02
0.03
0.04
0.17
0.05
0.01
0.41
0.24
0.95
0.54
0.12
35.7
0.18
0.95
0.13
0.18
0.95
0.11
0.05
0.71
0.94
0.12
0.38
0.99
0.18
0.33
0.99
0.18
0.00
0.17
0.44
0.01
0.84
0.01
0.82
0.09
0.91
0.09
0.90
0.10
0.10
,38066DPSOHRI
Migrants
ERUQLQ6RXWK
p-value of
Difference
Notes7KH,3806VDPSOHLQFOXGHVDOOVRXWKHUQERUQEODFNPDOHVREVHUYHGRXWVLGHRIWKH6RXWKLQWKHFHQVXV7KHWH[WFRQWDLQVPRUHGHWDLOVRQWKH
FRQVWUXFWLRQRIWKHOLQNHGVDPSOHRIFHQVXVUHFRUGV9DOXHVRI³´LQWKHDERYHWDEOHUHSUHVHQWWUXH]HURHV9DOXHVEHWZHHQDQGDUHUHSUHVHQWHGDV
³´6DPSOHVL]HVIRUEODFNPLJUDQWVLQRXUPDWFKHGVDPSOHDQGLQWKH,3806VDPSOHDUHDQGUHVSHFWLYHO\)RUZKLWHPLJUDQWVVDPSOHVL]HV
DUHDQGUHVSHFWLYHO\
Sources'DWDDUHIURP5XJJOHVHWDODQGWKHOLQNHGGDWDVHWRIFHQVXVUHFRUGVGHVFULEHGLQWKHWH[W
Collins and Wanamaker
'LVWULEXWLRQRIVWDWHRIUHVLGHQFH
New Jersey
New Mexico
1HZ<RUN
North/South Dakota
Ohio
3HQQV\OYDQLD
Wisconsin
3HUVRQDOFKDUDFWHULVWLFV
Mean age
,QRZQHURFFXSLHGKRXVLQJ
Literate
Veteran
-REFKDUDFWHULVWLFV
Farmer
Employed
988
Downloaded from https:/www.cambridge.org/core. The University of Tennessee - Knoxville, on 17 Mar 2017 at 13:09:33, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/S0022050715001527
APPENDIX TABLE 3 (CONTINUED)
&203$5,6212)/,1.('$1')8//6$03/(66287+(510,*5$176
727+(1216287+
Selection and Sorting of Southern Migrants
989
REFERENCES
$EUDPLW]N\5DQ/HDK3ODWW%RXVWDQDQG.DWKHULQH(ULNVVRQ³(XURSH¶V7LUHG3RRU
+XGGOHG 0DVVHV 6HOI6HOHFWLRQ DQG (FRQRPLF 2XWFRPHV LQ WKH $JH RI 0DVV
0LJUDWLRQÁmerican Economic ReviewQR±
$EUDPLW]N\5DQ/HDK3ODWW%RXVWDQDQG.DWKHULQH(ULNVVRQ³+DYHWKH3RRU$OZD\V
%HHQ/HVV/LNHO\WR0LJUDWH"(YLGHQFHIURP,QKHULWDQFH3UDFWLFHVGXULQJWKH$JH
RI0DVV0LJUDWLRQ´Journal of Development Economics±
Akers, Elmer. Southern Whites in Detroit$QQ$UERU0,0DQXVFULSW
$WDFN-HUHP\³2QWKH8VHRI*HRJUDSKLF,QIRUPDWLRQ6\VWHPVLQ(FRQRPLF+LVWRU\
7KH$PHULFDQ7UDQVSRUWDWLRQ5HYROXWLRQ5HYLVLWHG´Journal of Economic History
QR±
%DUWLN7LPRWK\-Who Bene¿ts from State and Local Economic Development Policies"
.DODPD]RR:(8SMRKQ,QVWLWXWH
Berry, Chad. Southern Migrants, Northern Exiles 8UEDQD ,/ 8QLYHUVLW\ RI ,OOLQRLV
3UHVV
%DNHU 5LFKDUG % ³)URP WKH )LHOG WR WKH &ODVVURRP 7KH %ROO :HHYLO¶V ,PSDFW RQ
(GXFDWLRQLQ5XUDO*HRUJLD´8QSXEOLVKHG:RUNLQJ3DSHU
%ODFN'DQ$6HWK*6DQGHUV(YDQ-7D\ORUHWDO³7KH,PSDFWRIWKH*UHDW0LJUDWLRQ
RQ 0RUWDOLW\ RI $IULFDQ$PHULFDQV (YLGHQFH IURP WKH 'HHS 6RXWK´ American
Economic ReviewQR±
%RGQDU-RKQ5RJHU6LPRQDQG0LFKDHO3:HEHULives of Their 2wn: Black, Italians,
and Poles in Pittsburgh, 1900–1960&KDPSDLJQ,/8QLYHUVLW\RI,OOLQRLV3UHVV
1982.
%RUMDV *HRUJH - 6WHSKHQ * %URQDUV DQG 6WHSKHQ - 7UHMR ³6HOI 6HOHFWLRQ DQG
,QWHUQDO0LJUDWLRQLQWKH8QLWHG6WDWHV´Journal of Urban Economics 32, no. 2
±
%RXVWDQ/HDK3ODWW³&RPSHWLWLRQLQWKH3URPLVHG/DQG%ODFN0LJUDWLRQDQG5DFLDO
:DJH&RQYHUJHQFHLQWKH1RUWK±´Journal of Economic History
QR±
%RXVWDQ/HDK3ODWW0DWWKHZ(.DKQDQG3DXO:5KRGH³0RYLQJWR+LJKHU*URXQG
0LJUDWLRQ 5HVSRQVHV WR 1DWXUDO 'LVDVWHUV LQ WKH (DUO\ 7ZHQWLHWK &HQWXU\´
American Economic Review: Papers and Proceedings QR ±
44.
&DUULQJWRQ :LOOLDP - (QULFD 'HWUDJLDFKH DQG 7DUD 9LVKZDQDWK ³0LJUDWLRQ ZLWK
(QGRJHQRXV0RYLQJ&RVWVÁmerican Economic ReviewQR±
&KD\.HQQHWKDQG.DLYDQ0XQVKL³%ODFN1HWZRUNVDIWHU(PDQFLSDWLRQ(YLGHQFH
IURP 5HFURQVWUXFWLRQ DQG WKH *UHDW 0LJUDWLRQ´ 8QSXEOLVKHG :RUNLQJ 3DSHU
2012.
&ROOLQV:LOOLDP-³:KHQWKH7LGH7XUQHG,PPLJUDWLRQDQGWKH'HOD\RIWKH*UHDW
%ODFN0LJUDWLRQ´Journal of Economic HistoryQR±
&ROOLQV :LOOLDP - ³$IULFDQ$PHULFDQ (FRQRPLF 0RELOLW\ LQ WKH V $ 3RUWUDLW
IURPWKH3DOPHU6XUYH\´Journal of Economic HistoryQR±
&ROOLQV:LOOLDP-DQG5REHUW$0DUJR³+LVWRULFDO3HUVSHFWLYHVRQ5DFLDO'LIIHUHQFHV
LQ6FKRROLQJLQWKH8QLWHG6WDWHV´,QHandbook of the Economics of Education:
Volume 1HGLWHGE\(+DQXVKHNDQG):HOFK±1HZ<RUN1<1RUWK
+ROODQG
Downloaded from https:/www.cambridge.org/core. The University of Tennessee - Knoxville, on 17 Mar 2017 at 13:09:33, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/S0022050715001527
990
Collins and Wanamaker
&ROOLQV:LOOLDP-DQG0DULDQQH+:DQDPDNHU³6HOHFWLRQDQG,QFRPH*DLQVLQWKH
*UHDW0LJUDWLRQRI$IULFDQ$PHULFDQVÁmerican Economic Journal: Applied
QR±
'DYLHV 3DXO 6 0LFKDHO - *UHHQZRRG DQG +DL]KHQJ /L ³$ &RQGLWLRQDO /RJLW
$SSURDFKWR866WDWHWR6WDWH0LJUDWLRQ´Journal of Regional Science 41, no.
±
(DVWHUOLQ5LFKDUG$³,QWHUUHJLRQDO'LIIHUHQFHVLQ3HU&DSLWD,QFRPH3RSXODWLRQDQG
7RWDO,QFRPH±´,QTrends in the American Economy in the Nineteenth
Century±1DWLRQDO%XUHDXRI(FRQRPLF5HVHDUFK&RQIHUHQFHRQ5HVHDUFK
LQ,QFRPHDQG:HDOWK3ULQFHWRQ1-3ULQFHWRQ8QLYHUVLW\3UHVV
)HUULH-RVHSK3³+LVWRU\/HVVRQV7KH(QGRI$PHULFDQ([FHSWLRQDOLVP"0RELOLW\LQ
WKH8QLWHG6WDWHVVLQFH´Journal of Economic PerspectivesQR
±
)RRWH &KULVWRSKHU / :DUUHQ & :KDWOH\ DQG *DYLQ :ULJKW ³$UELWUDJLQJ D
'LVFULPLQDWRU\/DERU0DUNHW%ODFN:RUNHUVDWWKH)RUG0RWRU&RPSDQ\±
´Journal of Labor EconomicsQR±
*UHJRU\-DPHV1The Southern Diaspora: How the Great Migrations of Black and
White Southerners Transformed America &KDSHO +LOO 8QLYHUVLW\ RI 1RUWK
&DUROLQD3UHVV
*URJJHU-HIIUH\DQG*RUGRQ++DQVRQ³,QFRPH0D[LPL]DWLRQDQGWKH6HOHFWLRQDQG
6RUWLQJRI,QWHUQDWLRQDO0LJUDQWV´Journal of Development Economics 95, no. 1
±
*URVVPDQ -DPHV 5 Land of Hope: Chicago, Black Southerners, and the Great
Migration&KLFDJR8QLYHUVLW\RI&KLFDJR3UHVV
+DLQHV0LFKDHO5DQG,QWHUXQLYHUVLW\&RQVRUWLXPIRU3ROLWLFDODQG6RFLDO5HVHDUFK
Historical, Demographic, Economic, and Social Data: The United States, 1790–
2002>FRPSXWHU¿OH@,&365Y$QQ$UERU0,,&365>GLVWULEXWRU@
+DWWRQ7LPRWK\-DQG-HIIUH\*:LOOLDPVRQThe Age of Mass Migration: Causes and
Economic Impact1HZ<RUNDQG2[IRUG2[IRUG8QLYHUVLW\3UHVV
+DWWRQ 7LPRWK\ - DQG -HIIUH\ * :LOOLDPVRQ ³:KDW )XQGDPHQWDOV 'ULYH :RUOG
0LJUDWLRQ"´1%(5:RUNLQJ3DSHU&DPEULGJH0$
+LJJV 5REHUW ³7KH %ROO :HHYLO WKH &RWWRQ (FRQRP\ DQG %ODFN 0LJUDWLRQ ±
Ágricultural History QR±
²²²³$FFXPXODWLRQRI3URSHUW\E\6RXWKHUQ%ODFNV%HIRUH:RUOG:DU,Ámerican
Economic Review QR±
+RUQEHFN5LFKDUGDQG6XUHVK1DLGX³:KHQWKH/HYHH%UHDNV%ODFN0LJUDWLRQDQG
(FRQRPLF 'HYHORSPHQW LQ WKH 6RXWK´ American Economic Review 104, no. 3
±
+XQWHU:DOWHU'DYLGDQG%5&RDGThe Boll Weevil Problem: Farmers’ Bulletin No.
1329. Washington, D.C.: U.S. Department of Agriculture, 1923.
.LOOLDQ /HZLV 0 ³7KH $GMXVWPHQW RI 6RXWKHUQ :KLWH 0LJUDQWV WR 1RUWKHUQ 8UEDQ
1RUPV´Social Forces QR±
.RXVVHU - 0RUJDQ The Shaping of Southern Politics: Suffrage Restriction and the
Establishment of the 2ne-party South, 1880–19101HZ+DYHQ<DOH8QLYHUVLW\
3UHVV
/DQJH)DELDQ$ODQ/2OPVWHDGDQG3DXO:5KRGH³7KH,PSDFWRIWKH%ROO:HHYLO
±´Journal of Economic HistoryQR±
Downloaded from https:/www.cambridge.org/core. The University of Tennessee - Knoxville, on 17 Mar 2017 at 13:09:33, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/S0022050715001527
Selection and Sorting of Southern Migrants
991
Lewis, Edward E. The Mobility of the Negro: A Study in the American Labor Supply.
1HZ<RUN&ROXPELD8QLYHUVLW\3UHVV
/RJDQ7UHYRQ³+HDOWK+XPDQ&DSLWDODQG$IULFDQ$PHULFDQ0LJUDWLRQEHIRUH´
Explorations in Economic HistoryQR±
/RQJ-DVRQDQG-RVHSK)HUULH³,QWHUJHQHUDWLRQDO2FFXSDWLRQDO0RELOLW\LQ%ULWDLQDQG
WKH866LQFHÁmerican Economic ReviewQR±
0DORQH\ 7KRPDV 1 ³0LJUDWLRQ DQG (FRQRPLF 2SSRUWXQLW\ LQ WKH V 1HZ
(YLGHQFHRQ$IULFDQ$PHULFDQ2FFXSDWLRQDO0RELOLW\LQWKH1RUWKÉxplorations
in Economic History QR±
0DUJR5REHUW$³$FFXPXODWLRQRI3URSHUW\E\6RXWKHUQ%ODFNV%HIRUH:RUOG:DU
,&RPPHQWDQG)XUWKHU(YLGHQFHÁmerican Economic ReviewQR
±
——— . Race and Schooling in the South 1880–1950: An Economic History. Chicago:
8QLYHUVLW\RI&KLFDJR3UHVV
0F)DGGHQ 'DQLHO ³&RQGLWLRQDO /RJLW $QDO\VLV RI 4XDOLWDWLYH &KRLFH %HKDYLRU´ ,Q
Frontiers in EconometricsHGLWHGE\3=DUHPEND±1HZ<RUN$FDGHPLF
3UHVV
0F.HQ]LH'DYLGDQG+LOOHO5DSRSRUW³1HWZRUN(IIHFWVDQGWKH'\QDPLFVRI0LJUDWLRQ
DQG ,QHTXDOLW\ 7KHRU\ DQG (YLGHQFH IURP 0H[LFR´ Journal of Development
EconomicsQR±
0LWFKHQHU.ULV-DPHVDQG,DQ:0F/HDQ³865HJLRQDO*URZWKDQG&RQYHUJHQFH
±´Journal of Economic HistoryQR±
0RUHWWL (QULFR ³/RFDO /DERU 0DUNHWV´ ,Q Handbook of Labor Economics, Vol.
%HGLWHGE\2UOH\$VKHQIHOWHUDQG'DYLG&DUG±$PVWHUGDP1RUWK
+ROODQG
0\UGDO*XQQDUAn American Dilemma: The Negro Problem and Modern Democracy.
1HZ<RUN+DUSHU%URWKHUV3XEOLVKHUV
1DLGX 6XUHVK ³5HFUXLWPHQW 5HVWULFWLRQV DQG /DERU 0DUNHWV (YLGHQFH IURP WKH
3RVWEHOOXP866RXWK´Journal of Labor EconomicsQR±
Ransom, Roger L., and Richard Sutch. 2ne Kind of Freedom: The Economic
Consequences of Emancipation6HFRQG(GLWLRQ1HZ<RUN&DPEULGJH8QLYHUVLW\
3UHVV>@
5REDFN-HQQLIHU³:DJHV5HQWVDQGWKH4XDOLW\RI/LIH´Journal of Political Economy
QR±
Rosenbloom, Joshua L. Looking for Work, Searching for Workers: American Labor
Markets during Industrialization1HZ<RUN&DPEULGJH8QLYHUVLW\3UHVV
5R\$'³6RPH7KRXJKWVRQWKH'LVWULEXWLRQRI(DUQLQJV´2xford Economic Papers
QR±
5XJJOHV 6WHYHQ - 7UHQW $OH[DQGHU .DWLH *HQDGHN HW DO Integrated Public Use
Microdata Series: Version 5.0 >0DFKLQHUHDGDEOH GDWDEDVH@ 0LQQHDSROLV
University of Minnesota, 2010.
6FURJJV :P 2 ³,QWHUVWDWH 0LJUDWLRQ RI 1HJUR 3RSXODWLRQ´ Journal of Political
EconomyQR±
6MDDVWDG/DUU\$³7KH&RVWVDQG5HWXUQVRI+XPDQ0LJUDWLRQ´Journal of Political
Economy2FWREHU±
6WDWD&RUS³3RVW(VWLPDWLRQ&RPPDQGVIRU0XOWLQRPLDO/RJLW´Stata Base Reference
Manual: Release 11&ROOHJH6WDWLRQ7;6WDWD&RUS/3
Downloaded from https:/www.cambridge.org/core. The University of Tennessee - Knoxville, on 17 Mar 2017 at 13:09:33, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/S0022050715001527
992
Collins and Wanamaker
6WXDUW %U\DQ $ DQG (YDQ - 7D\ORU ³6RFLDO ,QWHUDFWLRQV DQG /RFDWLRQ 'HFLVLRQV
(YLGHQFHIURP860DVV0LJUDWLRQ´8QSXEOLVKHG:RUNLQJ3DSHU
6XQGVWURP:LOOLDP$³7KH*HRJUDSK\RI:DJH'LVFULPLQDWLRQLQWKHLQWKH3UH&LYLO
5LJKWV6RXWK´Journal of Economic HistoryQR±
7KRPDV %ULQOH\ Migration and Economic Growth 1HZ <RUN 1< &DPEULGJH
8QLYHUVLW\3UHVV
7ROQD\6WHZDUW(³(GXFDWLRQDO6HOHFWLRQLQWKH0LJUDWLRQRI6RXWKHUQ%ODFNV±
´Social ForcesQR±
7ROQD\ 6WHZDUW ( DQG (0 %HFN ³%ODFN )OLJKW /HWKDO 9LROHQFH DQG WKH *UHDW
0LJUDWLRQWR´Social Science HistoryQR±
7ROQD\6WHZDUW(.DWKHULQH&XUWLV:KLWH.\OH'&URZGHUHWDO³'LVWDQFHV7UDYHOHG
'XULQJ WKH µ*UHDW 0LJUDWLRQ¶ $Q $QDO\VLV RI 5DFLDO 'LIIHUHQFHV $PRQJ 0DOH
0LJUDQWV´Social Science History QR±
U.S. Department of Agriculture. Public Roads Mileage and Revenues in the Southern
States, 1914 86'$ %XOOHWLQ 1R 2I¿FH RI 3XEOLF 5RDGV DQG 5XUDO
Engineering. Washington, DC, 1917.
U.S. Department of Labor. Negro Migration in 1916–17:DVKLQJWRQ'&*32
Vickery, William E. The Economics of Negro Migration, 1900–19601HZ<RUN1<
$UQR3UHVV
9LJGRU-DFRE/³7KH3XUVXLWRI2SSRUWXQLW\([SODLQLQJ6HOHFWLYH%ODFN0LJUDWLRQ´
Journal of Urban Economics QR±
:KDWOH\:DUUHQ³*HWWLQJD)RRWLQWKH'RRUµ/HDUQLQJ¶6WDWH'HSHQGHQFHDQGWKH
5DFLDO,QWHJUDWLRQRI)LUPV´Journal of Economic HistoryQR±
:LONHUVRQ ,VDEHO The Warmth of 2ther Suns: The Epic Story of America’s Great
Migration1HZ<RUN1<5DQGRP+RXVH
Woodward, C. Vann. The Strange Career of Jim Crow7KLUG5HYLVHG(GLWLRQ1HZ
<RUN2[IRUG8QLYHUVLW\3UHVV2ULJLQDOZRUNSXEOLVKHG
:R]QLDN$ELJDLO³$UH&ROOHJH*UDGXDWHV0RUH5HVSRQVLYHWR'LVWDQW/DERU0DUNHW
2SSRUWXQLWLHV"´Journal of Human ResourcesQR±
:ULJKW*DYLQ2ld South, New South: Revolutions in the Southern Economy since the
Civil War1HZ<RUN1<%DVLF%RRNV
Downloaded from https:/www.cambridge.org/core. The University of Tennessee - Knoxville, on 17 Mar 2017 at 13:09:33, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/S0022050715001527

Download Report

The Great Migration in Black and White

Paperzz.com

Your Paperzz