Migration to the Agricultural Frontier and Wealth Accumulation, 1860-1870 August 2005

James I. Stewart August 2005 (forthcoming Explorations in Economic History)

_______________ * Assistant Professor, Economics Department, Reed College. Mailing: 3203 SE Woodstock Blvd., Portland, OR 97202. [email protected]. tel. (503)-517-7661. I am grateful to Joe Ferrie, Joel Mokyr, Chiaki Moriguchi, William Sundstrom, Alan Olmstead, Rob Vigfusson, John Woodruff, and Jim Sullivan for helpful comments or discussions about earlier drafts of this paper. In addition, participants in Northwestern University’s Economic History Workshop, and discussants and participants at the Economic History Association Meetings, the Cliometrics Conference, the All UC Conference on Economic History, and the Illinois Economic Association meetings provided valuable feedback. I would also like to thank two anonymous referees who contributed greatly to the improvement of this manuscript and the editor of this journal for his patience. Special thanks go to Joe Ferrie for making his sample of households linked between the 1860 and 1870 censuses available to me. Much of this research was made possible by a Northwestern University Dissertation Year Fellowship, which I gratefully acknowledge. An earlier version of this paper appeared under the title “Migration to the Agricultural Frontier and Economic Mobility, 18601880.” All errors are mine.

Abstract

I use a new data set of households linked between the 1860 and 1870 censuses to study frontier migration. Households that moved to the frontier to farm were more likely than non-migrants to have been poor, landless, and illiterate, and to have had young children. Also, after controlling for observable differences, migrants had below average abilities to accumulate wealth. These findings suggest fewer opportunities for migrants to accumulate wealth in non-frontier areas and a reason for their migration. Nonetheless, migrants fared well, accumulating wealth at high rates. The gains in wealth of migrants, especially those with long tenure on the frontier, suggest the extraordinary benefits of migration.

Keywords: migration; wealth accumulation; frontier; United States; nineteenth century; economic opportunity.

2

1. Introduction

A large body of research shows that between 1850 and 1900 inhabitants of the U.S. frontier enjoyed significant opportunity to accumulate wealth. For instance, in his path-breaking study, Merle Curti (1959, pp. 176-221) found low levels of initial wealth but also widespread land holding and high rates of wealth accumulation among the settlers of Trempealeau County, Wisconsin. concluded, “w[as] largely provided... by frontier conditions.”1

“...[T]hat opportunity,” he

Subsequently, other historians have

identified similar patterns of land holding and wealth accumulation in frontier areas of Utah (Kearl, Pope, and Wimmer, 1980), Arkansas and Texas (Schaefer, 1987), Iowa (Galenson and Pope, 1989), and Missouri (Gregson, 1996), giving rise to a set of facts about wealth in frontier areas and contributing to the perception of the frontier as a place of economic opportunity.2 While our understanding of wealth on the agricultural frontier is now considerable, our knowledge of who settled the frontier is less exhaustive. That is due in part to the effort required to trace households over time and between places in the census manuscript schedules.3 The recent contributions of Donald Schaefer (1985), Richard Steckel (1989), and Joseph Ferrie (1997) notwithstanding, important questions about the ability of households to migrate to the frontier remain. For instance, we do not know how many of the poorest Americans migrated to the frontier (Shannon, 1936; Danhof, 1941; Ankli, 1974; Klein, 1974; Atack and Bateman, 1987), nor whether, as Frederick Jackson Turner (1903) surmised,

1

Curti (1959, p. 446).

2

Roughly, those facts are: (1) rates of wealth accumulation on the frontier were high and often above those in non-

frontier areas (Curti, 1959; Galenson and Pope, 1989); (2) there was a gap between the wealth of early and late arriving settlers (Curti, 1959; Kearl, Pope, and Wimmer, 1980; Schaefer, 1987); (3) early arriving settlers accumulated wealth at higher rates than late arriving settlers (Gregson, 1996); and (4) inequality in wealth holding increased with time due to Riccardian rents and the accumulation of location specific knowledge (Kearl, Pope, and Wimmer, 1980; Schaefer, 1987; Gregson, 1996). 3

Until recently historians interested in frontier migration had to rely on anecdotal evidence to infer the

characteristics of migrants (Goodrich and Carter, 1936) or state-level census data on the places of birth of residents (Malin, 1935; Gallaway and Vedder, 1971; Steckel, 1983).

3

households in urban areas in the East availed themselves of the “free lands” of the West (Turner, 1893; Turner, 1903; Goodrich and Davidson, 1936; Shannon, 1945; Ferrie, 1997). This paper adds to our knowledge of nineteenth century frontier migration and wealth accumulation. I study households that migrated to the frontier—defined as the area within the presentday borders of Kansas, Nebraska, and South Dakota (then a part of the Dakota Territory)—and established farms between 1860 and 1870.4 I use two new longitudinal data sets constructed from the manuscript schedules of the federal censuses of population for this purpose. The first is a sample of 1571 male-headed households from the 1870 U.S. Census of Population of Kansas, Nebraska, and the Dakota Territory linked to the 1860 U.S. Census of Population.5 One thousand fifty-two of the 1571 households in the sample were farming on the frontier in 1870. Of these, 846 were living in non-frontier areas in 1860 and constitute the principal group of migrants to be analyzed in the paper.6 The data provide information about migrants in non-frontier areas in 1860 before their migration and in frontier areas a decade later after migration. The second data set is a sample of households from the 1870 U.S. Census 4

Although farming was just one of many occupations on the frontier, it was the most important. In the 1870 U.S.

Census Public Use Micro-sample for Kansas, Nebraska, and the Dakota Territory (Ruggles et. al., 2004), farmers constituted 45 percent of household heads over 20 years of age and 54 percent over 30 years of age. For the purposes of this paper, a farmer is a household head in the 1870 U.S. Census who reported his occupation as “farmer”, “farming”, “stockraiser”, etc. It is possible that some persons who reported their occupations as such were in fact not farmers but farm laborers. Census enumerators were, however, instructed to distinguish carefully between the two: “Be very particular to distinguish between farmers and farm laborers. In agricultural regions this should be one of the points to which the assistant marshal should especially direct his attention” (Instructions to Assistant Marshals, U.S. Census, 1870). This should have reduced the extent of misidentification, especially in comparison to earlier censuses. 5

I use the terms household and family interchangeably; however, the unit of analysis in this paper is technically the

family, defined by ties of blood or marriage or other relationships recognized under the law, not the household, defined by the Census as a group of persons occupying the same dwelling. 6

Other areas of the United States qualified as frontier in 1860, but Kansas, Nebraska, and the Dakota Territory,

along with California, attracted the largest numbers of migrants between 1860 and 1890 (U.S. Census of Population of 1890). For alternative definitions of the frontier, see Steckel (1989, p. 212), in 1850 as “Minnesota, Iowa, Kansas, Texas, and states farther west”; Schaefer (1985, p. 568), in 1850 as Arkansas and Texas; and Ferrie (1997, p. 6), in 1850 as places west of 90 degrees longitude.

4

Public Use Micro-sample linked to the manuscript schedules of the 1860 U.S. Census of Population (Ferrie, 2000). Eight hundred ninety-two households in this sample were living off the frontier in 1860 and 1870 and constitute the group of non-migrants to which the group of frontier migrants is compared. The Appendix briefly describes how the samples were constructed and assesses their representativeness. The migration of households to Kansas, Nebraska, and the Dakota Territory during the 1860s was part of the westward movement of population in the U.S., a process that unfolded over three centuries. But was this migration more than just a chapter in a well-scripted historical process? Beginning with Turner (1893; 1903), many historians have thought so. It is well known that after 1850 the economy of the United States underwent significant structural changes.

Population growth, urbanization,

technological change, the expansion of the scale and scope of industry, and the spread of long distance trade altered the kinds of goods and services being produced, the means of production, and the location of economic activity.

These changes ushered in new opportunities to accumulate wealth for some

households but also reduced or destroyed opportunities for many others (Atack, 1985; Libecap, 1992; Goldin, 1994). There is disagreement, however, over whether the frontier, despite its capacity for yielding wealth, afforded relief for households worst affected by these changes. For instance, historians have pointed to the high costs of farm-making (Danhof, 1941; Klein, 1974; Atack and Bateman, 1987) and the skills needed to farm on the frontier (Shannon, 1945) as barriers to the migration of blue collar workers in rural and urban areas. In absence of direct evidence on the pre-migration wealth, occupations, and locations of frontier households, however, historians have been unable to determine precisely who moved to the frontier (Curti, 1959; Galenson and Pope, 1989). Similarly, in longitudinal studies of wealth accumulation and occupational mobility in nineteenth century cities and rural areas, historians have been unable to observe the post-migration outcomes of non-persisters, preventing definitive judgments about economic opportunities in those places relative to economic opportunities elsewhere including the frontier (Bogue, 1963; Thernstrom, 1964).

5

The principal advantage of the linked sample is that both the pre-migration and post-migration characteristics of migrants and non-migrants are observable, permitting comparisons of their characteristics before migration and judgments about the effect of migration on the accumulation of wealth. In Section 2, I use information on wealth in the 1860 and 1870 censuses to study wealth accumulation in frontier and non-frontier areas. Sections 3 and 4 then tackle the central question of the paper: the factors underlying frontier migration. As part of the analysis, I estimate a switching regression model that accounts for the wealth maximizing behavior of households and the endogeneity of the migration decision and wealth accumulation. Section 5 then undertakes an analysis of the migration decisions of rural households. Section 6 concludes.

2. Wealth Accumulation in Kansas, Nebraska, and the Dakota Territory, 1860-1870

Between 1854 and 1890, the migration of thousands of American and immigrant households transformed Kansas, Nebraska, and the Dakota Territory from a wild and empty grassland into one of the nation’s most prosperous agricultural regions. The first settlers, however, did not go without hardship. As I show later in the paper, many started out poor, and most had to adapt to the region’s more arid and treeless growing environment (Webb, 1931, pp. 5-8; Malin, 1944). In addition, there were annual hazards. Nebraska, for instance, experienced droughts in 1860, 1863, and 1864 and grasshopper infestations in 1860, 1865, and 1866 (Olson and Naugle, 1997, p. 95). Not to be overlooked, between 1855 and 1865 settlers of Kansas also had to endure armed conflict over the issue of slavery (Miner, 2002, pp. 49-93). Despite these hardships, hazards, and perils, there were, however, strong incentives for households to migrate to the frontier: rates of wealth accumulation in Kansas, Nebraska, and the Dakota Territory between 1860 and 1870 were high and in excess of those in non-frontier areas.

2.1 Wealth in Frontier and Non-frontier Areas, 1860 and 1870

6

Panel A of Table 1 shows the wealth in 1860, the wealth in 1870, and the change in wealth over the decade of households not living on the frontier in 1860 and 1870 (“non-migrants”), households farming on the frontier in 1870 but living off the frontier in 1860 (“post-1860 migrants”), and households farming on the frontier in 1870 and living on the frontier in 1860 (“pre-1860 migrants”).7 In absolute and percentage terms, both groups of migrants outperformed non-migrants. For instance, migrants who moved to the frontier after 1860 increased their wealth on average by $167 or 11 percent over the decade. Those migrating to the frontier before 1860 did even better, however.8 Their wealth increased by $2307 or 189 percent over the same period. The large absolute and percentage gains in wealth of pre-1860 migrants likely reflect their low levels of initial wealth (Curti, 1959), capital gains on real estate due to population growth and transportation improvements (Kearl, Pope, and Wimmer, 1980; Schaefer, 1987; Gregson, 1996), and the accumulation of productivity-enhancing knowledge specific to the frontier (Gregson, 1996).9 In contrast, the average wealth of households that did not migrate to the frontier fell by $21 or 0.7 percent over the decade. The poor performance of non-migrants is consistent with the high rates of taxation in the Northeast (Soltow, 1975, p. 65), the slowing of the industrial sector of the economy, the destruction of capital and the loss of economies of scale in agriculture in the South, and the

7

All wealth data in Table 1 have been adjusted for differences in prices across time and regions using the price index

of Coelho and Shepherd (1974, p. 570, Table 3). Coelho and Shepherd show that the cost of living differed significantly by region of the country and that a large rise in the price level occurred during the 1860s. Both phenomena justify the use of the regional price index. 8

Unless they had been squatting illegally, households observed on the frontier in 1860 must have migrated there

after 1854. The federal government established the territories of Kansas and Nebraska in 1854 with the passage of the Kansas-Nebraska Act, opening the way for settlement of lands previously reserved for native-Americans. 9

The more modest gains in wealth of post-1860 migrants may be the consequence of their late arrival on the frontier

and recently-incurred costs of travel there. Schaefer (1987, p. 148) argues that the costs of travel to the Southern frontier (including search) were considerable and accounted for almost 40 percent of the difference in wealth in 1860 between early and late-arriving frontier migrants. The costs of travel to the frontier in the North a decade later were probably less, however, because the transportation network in the North was more extensive and northern migrants were not transporting slaves.

7

significant depreciation of the prices of farmland in the East, Midwest, and South (Lindert, 1988, p. 52) during the 1860s. While migrants, especially those with long tenure on the frontier, fared well in comparison to non-migrants, there may have been differences between migrants in their abilities to accumulate wealth. For instance, historians have long been interested in, though pessimistic about, the prospects of bluecollar workers who moved to the frontier to farm (Shannon, 1945, p. 54). To explore further the accumulation of wealth in frontier and non-frontier areas, Panels B and C of Table 1 display the wealth and changes in wealth of post-1860 migrants and non-migrants by several 1860 occupational categories and places of birth.10 Large differences existed between migrants and non-migrants with the same occupations and between migrants with different occupations in 1860. First, farmers, skilled laborers, and unskilled laborers in urban areas who moved to the frontier to farm made much greater absolute and percentage gains in wealth than their counterparts who did not move. For example, despite lacking immediate experience in agriculture, unskilled workers in urban areas who migrated to the frontier to farm increased their wealth eight-fold over the decade or twice as much as those who did not migrate.11 Additionally, among migrants, those in unskilled blue collar positions in 1860 tended to fare the best, increasing their wealth in absolute and percentage terms significantly more than those in farming or skilled blue-collar jobs. Historians have also long been interested in the extent to which the frontier sped up the process of assimilating immigrants (Malin, 1935; Curti, 1959; Kearl, Pope, and Wimmer, 1980; Galenson, 1991; 10

Based on the occupation reported in 1860, each household head was assigned to one of the following occupational

classes: professional white collar (lawyers, clergy, professors, etc.), clerical white collar (agents, sales workers, clerks, etc.), skilled blue collar (blacksmiths, coopers, tailors, etc.), unskilled blue collar (includes semi-skilled workers like operatives), or non-occupational response (student, retired, unemployed, etc.).

In assigning

occupations to classes, I followed the 1950 Census Bureau Occupational classification system, available at www.ipums.umn.edu. 11

A t-test of the hypothesis that the changes in wealth, 1860-1870 of migrant and non-migrant urban unskilled

workers are equal produces a test statistic of 1.57 and a p value of 0.123.

8

Conley and Galenson, 1998; Ferrie, 1999; Walker, 2000). One measure of assimilation is the extent to which immigrants shared in the rising incomes and wealth of the native-born. Panel C of Table 1 shows that compared to those born in the United States, the foreign-born did well between 1860 and 1870 in frontier areas. Both German and Irish migrants outperformed in absolute and percentage terms not only their native-born counterparts but also their countrymen who remained in non-frontier areas.

For

instance, the number of observations is small, but Irish-born migrants increased their wealth by $856 or 350 percent over the decade, whereas Irish-born non-migrants suffered a loss in wealth of $477 or 50 percent. Not surprisingly, the frontier had a significant leveling effect: Though large differences in premigration wealth existed between native-born and immigrant frontier households, by 1870 these differences, with the exception of the gap between the Irish and the U.S. born, had largely disappeared.

2.2 A Life Cycle Model of Wealth

Previous scholarship on the accumulation of wealth in the nineteenth century has shown that age-specific rates of wealth accumulation were generally higher in frontier than in non-frontier areas (Atack and Bateman, 1987; Galenson and Pope, 1989; Galenson, 1991).

To gauge better the opportunity to

accumulate wealth in Kansas, Nebraska, and South Dakota and to facilitate comparisons to previous works I estimate a life cycle model of wealth accumulation. I estimate the model using a sample selection technique to control for the potentially distorting effects on wealth accumulation of self-selection in migration to the frontier (Borjas, 1987).12 The dependent variable is the natural logarithm of wealth in 12

Selective migration amounts to sample truncation, in which presence on the frontier is not random and the only

observations of frontier wealth belong to the households that moved there. Let y be the natural logarithm of wealth on the frontier in 1870; X be a vector of 1870 characteristics affecting the level of wealth on the frontier in 1870; Z be an indicator variable, equaling 1 if the household migrated to the frontier between 1860 and 1870 and 0 otherwise; and W be a vector of 1860 characteristics affecting the decision to migrate. Then the sample selection model is: yi = β’Xi + εi

is observed only if Zi=1; and

9

1870 plus one dollar, while the independent variables are age in 1870, age in 1870 squared, household size, and dummy variables for literacy and nativity.13 So that I can control for the self-selection of migrants, the analysis is initially limited to families living in non-frontier areas in 1860, i.e., post-1860 migrants and non-migrants.14 Then I estimate the life cycle model by OLS using as observations families that migrated to the frontier before and after 1860. The regression results are in columns 1 and 2 of Table 2.15 According to the model, the wealth of migrant households farming in Kansas, Nebraska, and the Dakota Territory peaked at age 52. Controlling for literacy, nativity, and household size, the implied annual rates of wealth accumulation fall from 11.8 percent at age 20, to 8.1 percent at age 30, to 4.4 percent at age 40.16 These rates compare favorably to

Zi =

if γ’Wi + µi >= 0

1 0

otherwise 2

Also, (µ,ε) is distributed bi-variate normal(0,0,1,σε ,ρ), where the variance of µ is normalized to 1 and ρ is the correlation coefficient between µ and ε. See Greene (1997, pp. 974-982). Historians have recognized that ignoring the self-selection of households in migration may bias estimates of the true relationship between age and wealth. See Galenson and Pope (1989, p. 237), Gregson (1996, p. 533, f. 15), Conley and Galenson (1998, p. 489) and Walker (2000, p. 275). 13

Conley and Galenson (1998, pp. 472-475) and Walker (2000, pp. 264-265) identify three problems with OLS

estimation of semi-log wealth equations. First, the censoring of wealth in the census can generate bias in OLS regressions. Second, if the distribution of wealth is dominated by extreme values, OLS estimates may not be informative. Third, the practice of adding $1 to zero values of wealth in semi-log equation is not innocuous. None of these problems is likely to be serious in the sample of frontier migrants, however. First, 93 percent of migrants establishing farms reported positive wealth in 1870, lessening concerns about the first and third problems. Second, the distribution of 1870 wealth on the frontier contains few extreme values. All but 11 of the 846 migrants have wealth in 1870 below $10,000 and the maximum is $27,000. 14

The selection (migration) equation includes 1860 variables for wealth, age, household size, distance from the

frontier, urban residency, occupation, literacy, nativity, and a previous interstate move. The data are weighted to reflect actual migration rates. See the discussion of the mover-stayer model later in the paper for details. 15

The hypothesis of independence between the migration decision and the level of wealth in 1870 cannot be rejected,

as the estimate of the correlation coefficient ρ between the error terms in the wealth and sample selection equations is statistically indistinguishable from zero (Wald χ2(1) = .004, p =0.945). 16

The regression results also indicate that nativity had a significant effect on the level of wealth of farmers. Holding

other factors constant, the British possessed more wealth than the native-born in 1870, while the Irish held less.

10

those in other frontier areas and non-frontier areas in 1860 and 1870. The rates of wealth accumulation in Kansas, Nebraska, and the Dakota Territory were approximately equal to those in Iowa in 1870 (Galenson and Pope, 1989) but higher than those in Missouri in 1870 (Gregson, 1996) and in Utah in 1870 (Kearl, Pope, and Wimmer, 1980). The rates of wealth accumulation on the Kansas, Nebraska, and Dakota frontier were significantly higher than those in agricultural areas of the northern United States in 1860 (Atack and Bateman, 1987).17 When households that migrated to the frontier before 1860 are included in the regression analysis and age in 1870 is interacted with a dummy variable for presence on the frontier in 1860, the implied rates of wealth accumulation for post-1860 migrants do not change.18 The results in Tables 2 indicate, however, a significant premium for migrants reaching the frontier before 1860.

Rates of wealth

accumulation of pre-1860 migrants exceeded those of post-1860 migrants by approximately 2 percentage points over the life cycle.19 In her investigation of wealth in five Missouri townships between 1850 and 1870, Gregson (1996, p. 537) also found a premium for early settlers and concluded that approximately half of it was due to the accumulation of location-specific knowledge of farming and markets and the remainder to higher rates of capital gains on land. Whatever the sources of this premium in Kansas,

Kearl, Pope, and Wimmer (1980, p. 489) argue that earlier studies may have exaggerated differences in wealth between U.S.- and foreign-born settlers by ignoring the time of entry into the frontier economy. Time of entry is less of a concern in this case because all households migrated to the frontier between 1860 and 1870. 17

Among households in Appanoose County, Iowa in 1870 but not present in 1860, the implied rates of wealth

accumulation at ages 20, 30, and 40 were 14.1, 9.3, and 4.3 percent, respectively (Galenson and Pope, 1989, p. 653). Among farm households in Utah, Davis, and Weber counties, Utah in 1870 but not present in 1860, the implied rates of wealth accumulation at ages 20, 30, and 40 were 9.3, 3.6, and 1.5 percent, respectively (Kearl, Pope, and Wimmer, 1980, p. 489). Among farm households in Missouri in 1870 but not present in 1860, the implied rates of wealth accumulation at ages 20, 30, and 40 were 4.5, 3.5, and 2.5 percent, respectively (Gregson, 1996, p. 532). Among farm households in the northern United States in 1860, the implied rates of wealth accumulation at ages 20, 30, and 40 were 5.5, 3.5, and 1.5 percent, respectively (Atack and Bateman, 1987, p. 94). 18

This specification follows that of Gregson (1996, pp. 532-533).

19

The implied rates of wealth accumulation at ages 20, 30, and 40 for households migrating to Kansas, Nebraska, or

the Dakota Territory before 1860 and farming in 1870 are 13.9, 10.0, and 6.0, respectively.

11

Nebraska, and the Dakota Territory between 1860 and 1870, households with longer tenure on the frontier were rewarded with higher than average rates of wealth accumulation. In sum, the census data suggest that Kansas, Nebraska, and the Dakota Territory, like other frontier areas, were characterized by high rates of wealth accumulation. Frontier families often started with low levels of wealth, but accumulated wealth at rates that were high and in excess of those in nonfrontier areas. Also, unskilled workers in urban and rural areas in 1860 and immigrants able to reach the frontier prospered. There is also some evidence that the frontier had a leveling effect between native-born and foreign-born migrants.

Overall, the differences between frontier and non-frontier areas in the

accumulation of wealth suggest the extraordinary benefits of migration.

3. Who Migrated to the Frontier?

The preceding analysis raises the central question of this paper: who moved to the frontier to farm and to take advantage of the opportunity to accumulate wealth? In this section, I provide a preliminary answer to this question. According to economic theory, migration can be understood as an investment decision in which households change locations to earn higher wages and incomes provided the expected net benefits of migration are positive (Sjaastad, 1962; Mincer, 1978).20 The benefits and costs of migration will vary between households, however, and depend in large part on the differences (and similarities) between the sending and prospective receiving regions. In economic parlance, these are the reciprocal “push,” and “pull” forces of migration.21 20

The costs include the money costs of transportation, the opportunity cost of foregone earnings while traveling and

searching for work, and the psychic costs of leaving friends, family, and familiar surroundings (Sjaastad, 1962, pp. 83-84). 21

In 1860 most of the United States was still rural and agricultural, but long-distance migration was complicated by

increasing regional specialization in production (Atack, Bateman, and Parker, 2000, pp. 248-255), growing commercialization (Gates, 1960, pp. 398-420), and distinct growing environments arrayed along lines of latitude

12

3.1 The Characteristics of Frontier Migrants

The first two columns of Table 3 shows the mean characteristics of households that migrated to the frontier between 1860 and 1870 and were farming in 1870 as well as of households that did not migrate to the frontier.22 Compared to non-migrants, migrants were more likely to have been young, residing outside their states of birth, illiterate, living in rural areas, and farming in 1860. All of these differences are statistically significant and understandable in terms of the investment model. For example, the young were more likely to move to the frontier to farm because they had more time than the old to reap the benefits of investments in physical and human capital specific to the frontier. These investments were considerable because of environmental differences between frontier and non-frontier areas that rendered many eastern methods of farming ineffective (Webb, 1931, pp. 17-26; Malin, 1944, pp. 2-5). By a similar logic, previous long distance movers had lower costs and thus higher odds of migration because they had accumulated less capital specific to their 1860 locations than households that had not made such moves. Almost 70 percent of migrants were residing outside their state or country of birth in 1860, compared to just 42 percent of non-migrants. Very few households that migrated to the frontier to farm originated in urban areas, as well. Just 13 percent of migrants inhabited such places in 1860, whereas one-third of non-migrants did the same. The rural origins of migrants are expected and largely the consequence of the self-selection of migrants in (Steckel, 1983, pp. 20-27).

The investment model is particularly apposite because of these features and the

substantial investments in location-specific human capital (knowledge about markets and the cultivation of plants and animals) and physical capital (plows, livestock, etc.) required of farmers (Steckel, 1983; Schaefer, 1985; Atack and Bateman, 1987; Wright, 1987; Steckel, 1989). 22

A migrant is a household farming in Kansas, Nebraska, or the Dakota Territory in 1870 but living elsewhere in

1860. A non-migrant is a household living outside of Kansas, Nebraska, and the Dakota Territory in 1860 and 1870. Although there were changes to the boundaries of these states and this territory during the period of study, the sample of migrants contains only households residing in 1870 within the present-day borders of Kansas, Nebraska, and South Dakota.

13

agricultural occupations.23 Nearly 60 percent of migrants reported their occupation as farmer in 1860, whereas just 41 percent of non-migrants did the same. Like farmers, unskilled workers in 1860 were also well represented on the frontier a decade later. Twenty-one percent of migrants and non-migrants alike had been in unskilled positions in 1860. For unskilled workers, whose abilities to migrate to the frontier historians have long doubted, a frontier farm not only created new and improved opportunities to accumulate wealth but also constituted a step up the occupational ladder. While most characteristics of migrants and non-migrants can be understood in terms of differences and similarities between frontier and non-frontier areas, the origins of most migrants close to the frontier probably cannot. Table 3 shows that distance to the frontier—measured in a straight-line from the 1860 location of the household to the 95th meridian (the approximate eastern border of Kansas, Nebraska, and the Dakota Territory)—averaged 448 miles for migrants but 814 miles for non-migrants. Accordingly, 76 percent of migrants originated in the East North Central or West North Central regions, though just 28 percent of the 1860 U.S. population was living in these places. The proximity of migrants to the frontier likely reflected the money and time costs of travel or the costs of obtaining information about opportunities on the frontier or some combination thereof.24 Both types of costs increased with distance (Schwartz, 1973).25 The costs of travel were, however, steadily decreasing during this period, as

23

Steckel (1989, p. 212) finds that 9.5 percent of households migrating to the frontier between 1850 and 1860

originated in urban areas. Using a more representative sample, Ferrie (1997, p. 8) finds that 24 percent of frontier migrants between 1850 and 1860 were living in urban areas in 1850. 24

The origins of migrants in the Midwest may also reflect the value of latitude specific human capital, though

Steckel (1989, pp. 203-205) speculates that the more arid conditions of Kansas, Nebraska, and the Dakota Territory diminished the worth of such capital and thus its role in migration. 25

The role of information in migration is underscored by the virtual absence of migrants from New England during

the 1860s and the exceptionally large direct migration of households from Massachusetts and neighboring states during the 1850s under the auspices of the New England Emigrant Aid Company, which sought to populate Kansas and Nebraska with abolitionists and others opposed to the expansion of slavery (Miner, 2002, p. 39).

14

low cost transportation (primarily rail service) connecting the East with the frontier expanded (Taylor, 1957, pp. 141-144; Chandler, 1977, pp. 83-86).26

3.2 The Migration of Immigrants to the Frontier

Although most frontier migrants were natives of the United States, a significant number were immigrants.27 Because most immigrants in 1860 had recently made long distance moves to the United States, they were typically younger and had less human capital specific to U.S. job markets than the native-born. There should therefore have been fewer differences between immigrants who migrated to the frontier to farm and those who did not, and the investment model of migration should do less well at explaining their migration decisions. The right hand side of Table 3 shows this is the case. For instance, there are no statistically significant differences in the ages of migrants and non-migrants among household heads of British, German, and Irish birth.

In addition, I cannot reject the hypothesis of independence between the

migration decisions and the 1860 occupations of any immigrant group based on χ2(5) tests.28 There are, however, some significant differences between the foreign-born who migrated to the frontier and the native-born who did the same. For instance, 62 percent of Irish-born migrants and 30 percent of German-born migrants were living in urban areas in 1860, in comparison to just 11 percent of U.S.-born migrants. Also, 69 percent of Irish and 53 percent of German migrants were in laboring jobs, in contrast to just 27 percent of the native-born. These characteristics reflect the concentration of

26

In August of 1866, one could travel by rail as far west as Kearney, Nebraska (Olson and Naugle, 1997, p. 157).

To the south and four years later one could reach the Kansas-Colorado state line (Miner, 2002, p. 103). 27

In the 1870 U.S. Census PUMS, 24 percent of farm household heads in Kansas, Nebraska, and the Dakota

Territory were immigrants. See Table A.1 in the Appendix. 28

The sizes of some sub-samples of immigrants are small. This means that the large sample properties of the χ2

statistic under the null hypothesis of independence are not valid. When appropriate, I employed Fisher exact tests of independence (Agresti, 1990, pp. 59-62).

15

immigrants in cities and in skilled or unskilled blue-collar occupations.

Since the opportunity to

accumulate wealth in many urban areas of the East during the 1860s was limited (Coelho and Shepherd, 1976; Conley and Galenson, 1998), the frontier was, at least theoretically, of greater importance to immigrants than to the native-born.

3.3 The Pre-migration Wealth of Frontier Migrants and Non-migrants

Still other aspects of frontier migration can also be understood in terms of fundamental differences between the places of origin of migrants and the frontier. In Section 2, I showed that the frontier was characterized by high rates of wealth accumulation. Poor households, many of which toiled for low wages on farms or in factories in the East, would have found this feature of the frontier attractive. Yet could they afford to travel to the frontier and to establish farms there? Many historians have thought not. For example, Clarence Danhof (1941, p. 354) has estimated that a household in 1850 needed a minimum of $1000 to establish a farm of 40 acres. This sum, which included the costs of land, fences, barns, seed, draft animals, farm implements, shelter, and money for provisions until the first harvest, was substantial and beyond the means of many households. The costs were even higher in the more arid, treeless environments of Kansas, Nebraska, and the Dakota Territory (Webb, 1931). In fact, on the basis of the high costs of farm-making, Danhof (1941, p. 354) was led to conclude, “The West... cannot be regarded as offering great agricultural opportunities for the poor and poverty-stricken.” Table 4 shows various measures of the pre-migration wealth of migrants and non-migrants. In light of the costs of farm-making (Danhof, 1941; Klein, 1974; Atack and Bateman, 1987), the presumption must be that very few migrants could have been found in the bottom of the wealth distribution in 1860. The wealth averages in column 1 show that migrants had substantially less pre-migration wealth than non-migrants, with just 53 percent in 1860. Nonetheless, a frontier migrant with an amount equal to the average of $1512 was not poor, and, according to Danhof’s calculation, could afford to establish a

16

small farm. Similarly, as Panel B shows, in each region of the country except the West, migrants had substantially less wealth than non-migrants. Among the foreign-born, significant differences in wealth were largely absent.

Again, this equality likely reflects the leveling effects of immigration and

immigrants’ short durations in the United States. Averages, however, cannot reveal the extent to which those with wealth below and above the average migrated to the frontier, so columns 2-6 of Table 4 report the fractions of migrant and nonmigrant households reporting total estate less than or equal to $100, as well as the 25th, 50th, 75th, and 90th percentiles of wealth for migrants and non-migrants in 1860. The $100 benchmark corresponds to the poverty line of Soltow (1975, p. 24), who classified households with wealth in 1860 below this threshold as “very poor.” According to Soltow’s definition, a large proportion of households that migrated to the frontier to farm qualified as poor in 1860. The share was 30 percent—approximately equal to the share of nonmigrants who also reported wealth below this threshold. Panel B shows, however, there were significant differences between regions in the probability migrants fell into this category. Most interestingly, in the urban and industrial Northeast and the agricultural but also economically unequal South migrants were significantly more likely to be counted among the poor. For instance, in the South 38 percent of migrants reported wealth less than or equal to $100 in 1860, compared to the 25 percent of non-migrants who did similarly. The wealth percentiles in columns 3-6 show further that many migrants had little wealth in 1860 and were poorer than non-migrants. Half of all migrants had less than $500 in 1860. In the Midwest, the median wealth of migrants was $550, approximately half of that of non-migrants. The differences between migrants and non-migrants were even greater in the upper parts of the wealth distribution. In the South, for example, the 90th percentile of wealth among migrants was half of the 75th percentile of wealth of non-migrants. Examination of the wealth of migrants and comparison of their wealth with that of non-migrants show that many migrants had amounts in 1860 insufficient to cover the basic costs of farm-making.

17

According to estimates by Atack and Bateman (1987, pp. 128-145), the capital costs on owner-occupied farms in Kansas in 1860 ranged from $457 for a 40-acre farm to $2190 for a 160-acre farm. Thus, in 1860 a 160-acre farm was out of the reach of all but 21 percent of the households that migrated to the frontier.29 Just 49 percent could afford to purchase farms of the smallest size. How did households whose wealth in 1860 fell below the minimum required make ends meet? A definitive answer is beyond this paper, but some conjectures are possible. First, many households could have accumulated enough wealth between 1860 and the time of their migration to purchase farms. Alternatively, poor households could have rented farms (Bogue, 1963, pp. 56-66), obtained credit (Bogue, 1955, pp. 44-88), economized on the costs of farm-making after 1862 by acquiring land through the Homestead Act, or, as Robert Ankli (1974) has argued, deferred some costs of farm-making by improving land and building fences, barns, and other structures over the course of several years. Yet another possibility is that some poor households did not make ends meet, perhaps abandoning their farms shortly after being enumerated in the 1870 census.

4. A Model of Migration and Wealth Accumulation

Many of the hypotheses about the causes of migration raised so far cannot be definitively tested in a simple bi-variate framework. For instance, wealth and the likelihood of migration are closely related not just to each other but also to age. In this section, I estimate a “mover-stayer” model of migration and wealth accumulation to identify more clearly the causes of frontier migration between 1860 and 1870.

4.1 The Model

29

These estimates reflect the cost of capital (land, livestock, and machinery) on owner-operated farms in 1860 but

not expenses related to the purchase of seeds, hired labor, and provisions for living until the first harvest; hence, the proportions of migrants that could not afford farms of various sizes are lower bounds.

18

The mover-stayer model is a Type 5 Tobit or switching regression model (Amemiya, 1985, pp. 399402).30 There are equations describing wealth accumulation in frontier and non-frontier areas between 1860 and 1870 and an equation describing the migration decision. The model, described in more detail in a technical appendix available from the author, is:

Z =

∆ln(yi)F = βF’XFi + εFi

if Z=1

(1)

∆ln(yi)E = βE’XEi + εEi

if Z=0

(2)

1

if γ1’Wi + γ2 (∆ln(yi)F - ∆ln(yi)E) + µi >= 0

0

otherwise

(3)

The dependent variables in equations (1) and (2) are the changes in wealth (measured as the difference in the natural logarithms of wealth) between 1860 and 1870 for migrants and non-migrants, respectively. XF and XE are vectors of characteristics that affect wealth accumulation in frontier and non-frontier areas, respectively.31 Z is a binary variable indicating whether or not the household migrated to the frontier; W is a vector of characteristics affecting the decision to migrate;32 and ∆ln(yi)F - ∆ln(yi)E is the difference in expected wealth accumulation between frontier and non-frontier areas, which is not known and must be estimated. The structural parameter of the model is γ2, which should have a positive sign if households 30

Beside migration and economic mobility (Robinson and Tomes, 1982; Ferrie, 1997), the mover-stayer framework

has been used to study the effects of union membership on wages (Lee, 1978), the returns to schooling (Willis and Rosen, 1979), and the demand for residential housing (Lee and Trost, 1982). 31

XF and XE include variables for the following 1860 characteristics: age, occupation, literacy, nativity, household

size, the percentage of the household that is male and adult (nine years of age or older), urban residency, residency in the South, and interaction variables between unskilled laborer and urban residency and between farmer and the South. The definitions of urban and South correspond to the census definitions except that Delaware, the District of Columbia, and Maryland are included in the Middle Atlantic instead of the South. 32

W includes variables for the following 1860 characteristics: wealth, distance from the frontier, previous interstate

move, and the elements of XF except for the percentage of the household that is adult and male.

19

that expected larger gains in wealth were more likely to migrate. To capture the dependence between the migration decision and wealth accumulation in frontier and non-frontier areas (Borjas, 1987), the error terms are assumed to have a joint normal distribution with correlations ρ(µi, εiF)=ρF and ρ(µi, εiE)=ρE.

4.2 Results

Estimates of the parameters of the mover-stayer model were obtained via a multi-step procedure, which includes the joint estimation of equations (1), (2), and the reduced form of (3) by full information maximum likelihood, followed by the estimation of the structural migration equation (3) by probit maximum likelihood after having replaced the variable ∆ln(yi)F - ∆ln(yi)E with its predicted value.33

4.2.1 Wealth Accumulation in Frontier and Non-Frontier Areas

Columns 1 and 2 of Table 5 report the estimates of the parameters of the wealth accumulation equations. The results substantiate many of the impressions formed in Table 1 about the relationships between the 1860 occupations or places of birth of migrants and non-migrants and their accumulation of wealth over the decade. For instance, in comparison to farmers and white collar workers (the omitted category), unskilled workers in 1860 who subsequently migrated to the frontier fared well. A somewhat surprising result is that farmers in the South who migrated experienced significantly larger percentage gains in

33

The data were weighted so that the rate of migration in the model is 10 percent. Ferrie (1997, pp. 21-22) reports

frontier migration rates of 9.3 percent between 1850 and 1860 and 5.4 percent between 1860 and 1870. Also, the standard errors of the coefficients in the structural migration equation were obtained by bootstrapping. Five hundred samples were drawn with replacement from the data. See Johnston and DiNardo (1997, pp. 362-370). Last, some exclusion restrictions are imposed to ensure identification of the parameters of the model. To identify γ1 in equation (3), there must be at least one variable in XF or XE not in W. Likewise, to identify γ2, there must be at least one variable in W not in XF or XE. Both conditions are satisfied.

20

wealth than their counterparts in the North who also moved. This likely reflects, however, not the greater abilities of Southerners but rather their lower pre-migration wealth. Wealth accumulation also varied by one’s place of birth.

For instance, German and Irish

migrants made larger percentage gains in wealth over the decade than British or native-born migrants (the omitted category), according with growing evidence about the prosperity of foreign-born households in rural and urban areas of the frontier during the nineteenth century (Galenson, 1991; Ferrie, 1999, pp. 130155; Walker, 2000). Off the frontier, the effect of foreign-birth on wealth accumulation was less positive. The Irish, along with the foreign-born in the “other” category, did worse than the native-born. Household characteristics also affected the accumulation of wealth in frontier and non-frontier areas. Given the low ratio of labor to land and the high productivity of labor in frontier areas (Fogel and Rutner, 1972, pp. 402-403), large farm households should have had an advantage over small ones in accumulating wealth. This, however, was not the case. According to Table 5, household size was everywhere negatively correlated with the accumulation of wealth between 1860 and 1870. Instead, households with more adult males accumulated wealth faster, suggesting that what mattered was the productivity of each household member, which depended on the age and gender of the worker (Craig, 1991, p. 74). For instance, on frontier farms pre-adolescent and adolescent boys and girls made only modest contributions to output, because many tasks like digging wells and clearing land demanded adult strength. The percentage of the household that is male and adult had a smaller impact off the farming frontier than on, probably because male labor was relatively more valuable to farm than non-farm households and there were both farm as well as non-farm households in the group of non-migrants. Finally, the results show that households living in urban areas in 1860 that migrated to the frontier enjoyed large gains in wealth relative to those of households originating in rural areas. The change in wealth was 2.6 log points greater for urban households, indicating the substantial economic benefits for urban households that made it to the frontier.

4.2.2 Selection

21

A nice feature of the mover-stayer model is that it reveals the selection biases of migrants and nonmigrants, how, after controlling for differences in observable characteristics, migrants to the frontier and non-migrants would have fared (relative to one another) had they not made the decisions they did. Columns 1 and 2 of Table 5 report estimates of the correlation coefficients ρF and ρE between the error terms of the migration and wealth accumulation equations for movers and stayers, respectively. For movers to the farming frontier, ρF^ = -0.98, which means that migrants were negatively selected: had nonmigrants also moved to the frontier they would have experienced larger gains in wealth than migrants. For family heads staying off the frontier, ρE^ = -0.79, which means that non-migrants were positively selected: had migrants not moved to the frontier they would have made smaller gains in wealth than nonmigrants. Thus, after controlling for differences in observable characteristics, migrants would have done worse than non-migrants in terms of wealth accumulation between 1860 and 1870 in both frontier and non-frontier areas. This result suggests fewer opportunities for migrants to accumulate wealth.

4.2.3 Migration

What factors influenced the migration decision? The parameter estimates and the associated marginal effects in Columns 3 and 4 of Table 5 show that one such factor was the pre-migration occupation of the household head. The odds of moving to the frontier to farm between 1860 and 1870 were much higher for farmers and unskilled workers, a finding consistent with households’ past investments in occupationspecific human capital and the wider range of employment opportunities in non-frontier areas for white collar and skilled blue collar workers. Unskilled blue collar workers were, for instance, 3.6 percentage points or 15 percent more likely to migrate than white collar workers. In the South, there was no statistically significant difference in the likelihood of migration between farmers and persons in other occupations, although overall Southerners were substantially less likely to move to the frontier. The low

22

migration rates of Southern households in 1860 probably reflected regional differences in culture, the growing environment, and flows of information that made migration to the frontier more risky and less attractive. Another factor, whose role in frontier migration is in dispute, was the locations of households in rural or urban areas, as many historians have objected to Turner’s conception of the agricultural frontier as a safety-valve for urban blue collar workers. Fred Shannon (1945, p. 54), for one, has argued that urban blue-collar workers could neither afford to travel to the frontier nor prosper there because “what was the industrial worker going to substitute for experience when he turned to farming? It was hard enough for the Missouri farmer to get along in Kansas, or the Iowa farmer in Nebraska.” To be sure, farming was only one livelihood on the frontier, but Shannon and other critics were correct that few households in urban areas in 1860 subsequently established frontier farms. For instance, just 12 percent of unskilled workers who migrated to the frontier to farm had been living in urban areas in 1860, whereas 33 percent of unskilled workers who did not migrate had been living in such places. Nevertheless, the results in Table 5 show little support for the existence of barriers to migration specific to urban workers. The coefficient on urban residency is indistinguishable from zero and the associated marginal effect is negligible.

The interaction variable between urban residency and unskilled worker is similarly

insignificant. Skeptics of the safety-valve hypothesis have also argued that common laborers were too poor to afford the high costs of farm making and thus were largely absent from agricultural areas of the frontier (Danhof, 1941; Klein, 1974; Atack and Bateman, 1987). The results indicate, however, that being poor in 1860 was not an impediment to migration. In fact, the likelihood of moving to the frontier to farm was inversely related to wealth in 1860, a result consistent with the migration of households at the bottom of the wealth distribution in Table 4. The effect of wealth was large, as, for example, a family with $2875, the 75th percentile of wealth for non-migrants in 1860, was just 40 percent as likely to migrate as a family with $20, the 25th percentile of wealth for non-migrants.

23

If not their places of residency or wealth, what explains the absence of urban workers, especially those in blue-collar jobs, from agricultural areas of the frontier? Two factors stand out. First, only a small fraction of the urban workforce was employed in agriculture, the group of workers most at risk of migration. The second factor was the distance of most urban areas from the frontier. Distance had a strong and negative impact on migration. The model predicts, for example, a household in Philadelphia 1050 miles away was only 14 percent as likely to move to the frontier to farm as an otherwise identical household in Des Moines 177 miles away. Equally important, most urban households resided far from the frontier. In 1860, 81 percent of the urban population of the northern United States could be found in New England, the Middle Atlantic, Delaware, the District of Columbia, or Maryland (Historical Statistics of the United States, 1975, series A 172-194).34 Whether distance is capturing the money and time costs of travel or the costliness of obtaining information about opportunities on the frontier is unknown. Finally, several variables had little measurable impact on the probability of migration. For example, age had no bearing on the decision to migrate, a somewhat surprising though not unprecedented result (Schaefer, 1985; Steckel, 1989; Ferrie, 1997; Ferrie, 1999). Other variables such as wealth and occupation may be picking up the effects of age. Also, nativity had a large though not statistically significant effect. The structural parameter of the model has the expected sign but is not statistically significant at the conventional levels.

5. The Migration Decisions of Rural Households 34

To test this hypothesis further I grouped migrant (N=177) and non-migrant (N=195) unskilled workers in the

Northeast, Midwest, South, and West according to the size of their township of residence in 1860 (rural, <2500; small city, 2500-9999; or large city, >=10,000). In each region the resulting distributions of migrants and nonmigrants across township sizes were about the same and reflected the predominance of agriculture or industry in that region. In the Northeast, for instance, approximately 45 percent of unskilled migrants and non-migrants alike were living in urban areas. This compares to the 12 percent of migrants and the 14 percent of non-migrants in the Midwest who were similarly situated. It does appear, however, that in each region unskilled workers in the largest cities were underrepresented among migrants. Nonetheless, based on χ2(2) tests, I cannot reject the hypothesis of independence between the decision to migrate and the population of the township of residence in 1860 in any region.

24

While the opportunity to accumulate wealth on the frontier appealed to both rural and urban households, the specific factors bearing on their migration decisions were often different.

In rural areas the

opportunity to accumulate wealth in local agriculture relative to that on the frontier—a function of the availability of land, the prospects for capital gains on real estate, and access to markets in both places— would have been important.

In contrast, in urban areas labor market conditions—wages and

unemployment—would have been of greatest concern. In this section, I undertake a brief analysis of the migration decisions of rural dwellers, reserving an analysis of urban-to-frontier migration for future work.

5.1 Rural-to-Frontier Migration

In rural areas an important factor bearing on the accumulation of wealth (Soltow, 1975; Atack and Bateman, 1987, pp. 86-87) and therefore the incentives for migration (Gallaway and Vedder, 1971, p. 617; Atack and Bateman, 1987, pp. 76-77; Steckel, 1989, p. 212) was the opportunity to own land. The role of land in wealth accumulation is best understood from the perspective of the agricultural ladder, which posits a hierarchy of jobs in agriculture that workers ascended over their life cycles. Land enabled laborers and tenants, those on the lowest rungs, to become farm owners and to accumulate more wealth through rents and capital gains. The opportunity to own land and thus to climb the agricultural ladder was not equal across the United States, however. In 1860, the average price of a 40 acre farm in the Northeast or the North Central region equaled that of an 80 or 160 acre farm on the frontier (Atack and Bateman, 1987, p. 143).35 Within the East, the opportunity to own land also varied and was closely related to the local availability of land, a function of the quantity of unimproved farmland, the sizes of various age cohorts, and the mortality of farm operators (Leet, 1975; Atack and Bateman, 1987). Where there were

35

In addition to the availability of land, prices and rents reflected land productivity, the prevailing terms of trade,

and access to markets.

25

less farmland available, prices and rents would have been higher and the incentives for tenants, laborers, and others without land to migrate—to the frontier or elsewhere—would have been greater. Were households in densely settled areas more likely to move to the frontier? At first glance, it appears the answer is no. Table 6 reports averages of a measure of the availability of land in 1860 in the counties of residence of migrants and non-migrants. The availability of land is the excess demand or supply of farm sites in a county as a fraction of the total demand or supply of farm sites and takes on values between –1 and 1, with smaller negative numbers indicating greater excess supply of land and larger positive numbers indicating greater excess demand, and was constructed by following the method of Leet (1975, p. 147).36 Households in densely settled areas in 1860 were less, not more, likely to move to the frontier to farm. Factors possibly obscuring the true relationship between land availability and migration are the extent of land owning and distance from the frontier, both negative correlates of migration. Table 6 does show, however, that landless households—those reporting zero real estate wealth in 1860—were more likely to move to the frontier to farm. The share of migrants in this class was 47 percent and ranged from approximately 40 percent in New England and the Middle Atlantic states to almost 60 percent in the South. Overall, migrants were 15 percent more likely than non-migrants to have been landless in 1860. The difference between migrants and non-migrants in this respect was greatest in the Midwest and the South. In the Midwest—the biggest sending region—almost half of migrants were without land, whereas just one-third of non-migrants were similarly endowed.

36

The supply of farm sites in 1860 is the sum of farm sites that opened up between 1860 and 1870 because of the

deaths of farm operators and those that could have been developed from the supply of unimproved farmland in 1860. To estimate regional rural mortality rates for farm operators I used individual-level mortality data from the 1850 U.S. Census and the 1860 U.S. Census (Fogel et al., 2000). The demand for farm sites in 1860 was the number of males in the county between the ages of 20 and 29. The data on improved and unimproved farmland, population, and farm operators were obtained from the 1860 U.S. Census of Population, the 1860 U.S. Census of Agriculture, and the 1870 U.S. Census of Agriculture.

26

Another potentially important factor in rural-to-frontier migration was household fertility. Historians have conjectured that rural households with high fertility or prospective fertility sought less densely settled areas to reduce the costs of providing inheritances or inter vivos transfers of land to their children (Easterlin et al., 1978, pp. 69-73), although so far they have found little to substantiate this hypothesis (Steckel, 1989, pp. 212-214).37 Is there any evidence that rural households with high fertility were more likely to move to the frontier to farm? Column 3 of Table 6 shows that migrants had slightly larger households than non-migrants in 1860.38 This difference is, however, not conclusive, because large households also might have moved because of advantages in labor supply in frontier areas.

To

distinguish between these hypotheses, Table 6 also reports the average number of children between the ages of 0 and 8 and between the ages of 9 and 18 in migrant and non-migrant households. As columns 4 and 5 reveal, the difference in household size was largely driven by the migration of families with young children. Migrants and non-migrants had approximately the same number of children between the ages of 9 and 18, but migrants had significantly more below the age of 8. The migration of households with young, physically immature children is consistent with the bequest hypothesis and not that related to labor scarcity on the frontier, though selective migration of this sort may also reflect the independent effect of age on the likelihood of migration.

5.1.1 Regression Analysis of Rural to Frontier Migration

To explore further the factors bearing on the migration decisions of rural households, I estimate the mover-stayer model again, using only households in rural areas in 1860 and including variables for land 37

According to the target bequest model (Easterlin, 1976, pp. 63-70), parents want to leave their children with

inheritances at least as large as the ones they received. In rural areas, an inheritance often took the form of land for a farm. Historians have hypothesized that rural households in densely populated areas reduced their fertility in response to the increasing costs of providing such inheritances (Easterlin, 1976; Schapiro, 1982). 38

A t-test of the hypothesis that the sizes of migrant and non-migrant households were equal produces a test statistic

of 1.54 and a p value of 0.124.

27

ownership (a dummy variable indicating zero real estate wealth), land availability, the number of children in the household between the ages of 0 and 8 and the ages of 9 and 18, and interaction variables between not owning land and land availability and between the number of children of different ages and land availability.

To save space and because the variables affecting the wealth accumulation of rural

households in 1860 are essentially the same as those discussed earlier, only estimates of the coefficients and marginal effects of the variables in the structural migration equation are reported in Table 7. The results underscore the importance in migration of differences between frontier and nonfrontier areas in the opportunities to accumulate wealth and to own land.39 There is, for instance, clear evidence of wealth maximization as a motive for migration. The estimate of the structural parameter, γ2, indicates that households expecting to accumulate more wealth on the frontier areas were more likely to migrate there. An increase in the difference in expected wealth accumulation between frontier and nonfrontier areas of one log point doubled the likelihood of migration. Also, households reporting zero real estate wealth in 1860 were 34 percent more likely to move to the frontier to farm. Interestingly, the coefficient on wealth in 1860 is now small and statistically indistinguishable from zero, suggesting that whether the household owned land was the more important of these closely related factors. Still another variable bearing on rural-to-frontier migration was the number of children in the household and their ages. Each child between the ages of 9 and 18 reduced the probability a household would migrate to the frontier to farm by 1.4 percentage points, while the number of children between the ages of 0 and 8 had no effect. This implies a household with two teenage children was 27 percent less likely to migrate than an otherwise identical household with two children below the age of 8, contradicting the hypothesis about the advantages of households with older children in labor-scarce frontier areas. The model also predicts, however, no difference in the likelihood of migration between 39

While occupation, Southern residency, and distance from the frontier had effects on migration similar to those in

the original analysis, there are some differences from before. One is that in the South farmers were significantly less likely to migrate to the frontier to farm than persons in non-farm occupations. Another is place of birth, which now has a large and statistically significant effect. Immigrants of British or German birth were the most likely to move to frontier to farm.

28

households with young children and those without children, a result consistent with the migration of families with young children and of single males or married couples that did not have children but expected them. Both types would have moved to the frontier early in their life cycles to secure larger intergenerational transfers of wealth.40 Finally, while the prospective gains in wealth, real estate holdings, and the fertility of the household affected migration, the availability of land in the county of residence did not. The coefficient has the wrong sign and is statistically indistinguishable from zero. Moreover, neither households without land nor those with high fertility in densely settled areas were more or less likely to migrate to the frontier than otherwise identical households in sparsely settled areas. The insignificance of the land availability variables may indicate that differences between frontier and non-frontier areas in the availability of land were so great as to render those between non-frontier areas inconsequential.

6. Conclusion

Ample evidence now exists to demonstrate the agricultural frontier after 1850 was characterized by significant opportunity to accumulate wealth. Historians know less, however, about who migrated to the frontier. This paper has used a new sample of households linked between the 1860 and 1870 censuses to shed light on this question. Households migrating to Kansas, Nebraska, or the Dakota Territory to farm between 1860 and 1870 had below average abilities to accumulate wealth and were more likely than non-migrants to have been poor, landless, illiterate, and to have had high fertility in 1860. Interestingly, except for their rural locations and the larger sizes of their households, the pre-migration characteristics of frontier migrants

40

If families with young children or those anticipating children were more likely to migrate to the frontier, some of

the difference in fertility between frontier and non-frontier areas of the United States in the nineteenth century (Easterlin et al., 1978, pp. 35-40) may have been the consequence of out-migration by high fertility households.

29

resembled those of households departing Newburyport, Massachusetts between 1850 and 1860, who lacked “financial resources, occupational skill, [and] education.”41 Nonetheless, just as households leaving Newburyport for rural areas or large cities in the 1850s prospered (Herscovici, 1998, pp. 936-946), so did households migrating to the agricultural frontier in the 1860s. Despite being endowed with little wealth or human capital, frontier migrants accumulated wealth at rates that were high and in excess of those in non-frontier areas. Their achievements indicate the substantial economic benefits of frontier migration and provide additional evidence about the positive relationship between geographic mobility and economic opportunity in the mid-nineteenth century (Ferrie, 1997; Herscovici, 1998; Ferrie, 1999; Walker, 2000).

Appendix

Here I briefly describe how the sample of frontier households linked between the censuses was constructed. A more detailed description of the process is available from the author, but in general I followed the strategy outlined in Ferrie (1996). The sample of households linked from the 1870 U.S. Census Public Use Micro-sample to the 1860 U.S. Census (the sample of non-migrants) was constructed following almost identical procedures. First, 7287 household heads who were male and 26 years of age or older and had uncommon names were drawn from the manuscript schedules of the 1870 U.S. censuses of Kansas, Nebraska, and the Dakota Territory. A household head had an uncommon name if the combination of his first name and soundex-coded last name appeared at least once but fewer than three times in the index to the 1860 U.S. Census. (The index records the names and locations in the manuscript schedules of household heads and individuals whose surname differs from that of the head of the household in which they are found.) Then the 7,287 household heads were sought in the 1860 manuscript schedules. One thousand five hundred seventy-one household heads were positively matched in 1860 on the basis of their names, ages, and 41

Thernstrom (1964, p. 87).

30

birthplaces and those of their dependents. Allowances were made for small discrepancies between the censuses in the spelling of names and the recording of ages and birthplaces (for instance, Prussia in 1860 and Germany in 1870). How representative of frontier and non-frontier households are the linked samples? There are two potential sources of bias. The first is limiting the samples to households with uncommon names. This strategy has been used in the past, however, and has been shown to introduce no meaningful bias (Ferrie, 1996, pp. 13-15). The second and more serious potential source is the backward linkage of households. The linkage of households from 1870 to 1860 is likely to result in a sample in which young and immigrant households are underrepresented. Why? First, household heads who were young in 1870 were unlikely to have set up independent households a decade earlier and thus to appear in the index to the 1860 census. Similarly, only households residing in the United States in 1860 appear in the census for that year and thus are eligible to be included in the linked sample. Households immigrating to the U.S. after 1860 are necessarily excluded. The effect of these biases on our ability to infer the causes of migration to the agricultural frontier should not be great, however, because the biases are not large and both the migrant and non-migrants samples were linked backwards (from the 1870 to the 1860 census) using almost identical matching criteria. The two samples should have biases of similar magnitudes and directions. Table A.1 shows this is the case. The first and second columns compare the frontier farm households in the linked sample with frontier farm households in the 1870 U.S. Census Public Use Micro-Sample (PUMS). The third and fourth columns compare non-migrants in the linked sample with households residing outside of Kansas, Nebraska, and the Dakota Territory in the 1870 U.S. Census PUMS. Households in the linked samples are slightly older and wealthier and are more likely to have been born in the United States than those in the comparison samples. These differences are consistent with the backward linkage process.

31

References Agresti, A., 1990. Categorical Data Analysis. New York: John Wiley and Sons. Amemiya, T., 1985. Advanced Econometrics. Cambridge: Harvard University Press. Ankli, R., 1974. Farm Making Costs in the 1850s. Agricultural History 48, 51-70. Atack, J., 1985.

Industrial Structure and the Emergence of the Modern Industrial Corporation.

Explorations in Economic History 22, 29-52. Atack J., Bateman F., 1987. To Their Own Soil: Agriculture in the Antebellum North. Ames: Iowa State University Press. Atack, J., Bateman F., Parker W.N., 2000. The Farm, the Farmer, and the Market. In Engerman, S.L., Gallman, R.E. (Eds.), The Cambridge Economic History of the United States. New York: Cambridge University Press. Bogue, A., 1955. Money at Interest: The Farm Mortgage on the Middle Border. Ithaca: Cornell University Press. Bogue, A., 1963. From Prairie to Corn Belt: Farming on the Illinois and Iowa Prairies in the Nineteenth Century. Chicago: University of Chicago Press. Borjas, G., 1987. Self-Selection and the Earnings of Immigrants. American Economic Review 77, 531553. Chandler, A.D., 1977. The Visible Hand: The Managerial Revolution in American Business. Cambridge: Harvard University Press. Coelho, P.R.P., Shepherd, J.P., 1974. Differences in Regional Prices: The United States, 1851-1880. Journal of Economic History 34, 551-591. Coelho, P.R.P., Shepherd, J.P., 1976. Regional Differences in Real Wages: The United States, 18511880. Explorations in Economic History 13, 203-230. Conley, T., Galenson, D., 1998. Nativity and Wealth in Mid-Nineteenth Century Cities. Journal of Economic History 58, 468-493.

32

Craig, L., 1991. The Value of Household Labor in Antebellum Northern Agriculture.

Journal of

Economic History 51, 67-81. Curti, M., 1959. The Making of an American Community. Palo Alto: Stanford University Press. Danhof, C.H., 1941. Farm-Making Costs and The ‘Safety-Valve’: 1850-1860. Journal of Political Economy 49, 317-359. Easterlin, R.A., 1976. Population Change and Farm Settlement in the Northern United States. Journal of Economic History 36, 45-75. Easterlin, R.A., Alter G., Condran, G.A., 1978. Farms and Farm Families in Old and New Areas: The Northern States in 1860. In: Hareven, T.K., Vinovskis, M.A. (Eds.), Family and Population in Nineteenth Century America. Princeton: Princeton University Press. Ferrie, J.P., 1996. A New Sample of Americans Linked from the 1850 Public Use Micro-Sample of the Federal Census of Populations to the 1860 Federal Census Manuscripts. Historical Methods 39, 141-156. Ferrie, J.P., 1997. Migration to the Frontier in Mid-Nineteenth Century America: A Re-examination of Turner’s ‘Safety Valve’. unpublished Northwestern University paper. Ferrie, J.P., 1999. ‘Yankees Now’: Immigrants in the Antebellum U.S. 1840-1860. New York: Oxford University Press. Ferrie, J.P., 2000. Untitled Data Set (A Sample of Household Heads from the 1870 U.S. Census Public Use Micro-sample linked to the 1860 U.S. Census of Population Manuscript Schedules). Northwestern University. Fogel, R.W., Rutner, J.L., 1972. The Efficiency Effects of Federal Land Policy, 1850-1900: A Report of Some Provisional Findings.

In: Aydelotte, W.O., Bogue, A.G., Fogel, R.W. (Eds.) The

Dimensions of Quantitative Research in History. New Jersey: Princeton University, 390-418. Fogel, R.W., et al., 2000. U.S. Census Mortality: 1850, 1860, and 1870 (ICPSR Study 2526) Ann Arbor: ICPSR, University of Michigan.

33

Galenson, D.W., 1991. Economic Opportunity on the Urban Frontier: Nativity, Work, and Wealth in Early Chicago. Journal of Economic History 51, 581-603. Galenson, D.W., Pope, C.L., 1989.

Economic and Geographic Mobility on the Farming Frontier:

Evidence from Appanoose County, Iowa, 1850-1870. Journal of Economic History 49, 635-655. Gallaway, L.E. and Vedder, R.K., 1971. Mobility of Native Americans. Journal of Economic History 31, 613-649. Gates, P.W., 1960. The Farmer’s Age: Agriculture, 1815-1860. New York: Holt, Rinehart and Winston. Goldin, C., 1994. The Political Economy of Immigration Restrictions in the United States, 1890-1921. In: Goldin, C., Libecap, G.D. (Eds.), The Regulated Economy. Chicago: University of Chicago Press. Goodrich, C., Davidson, S., 1936. The Wage-Earner in the Westward Movement II. Political Science Quarterly 51, 61-116. Greene, W.H., 1997. Econometric Analysis. New Jersey: Prentice Hall. Gregson, M.E., 1996. Wealth Accumulation and Distribution in the Midwest in the Late Nineteenth Century. Explorations in Economic History 33, 524-538. Herscovici, S., 1998. Migration and Economic Mobility: Wealth Accumulation and Occupational Change Among Antebellum Migrants and Persisters. Journal of Economic History 58, 927-956. Johnston, J., DiNardo, J., 1997. Econometric Methods. New York: McGraw Hill. Kearl, J.R., Pope, C.L., Wimmer, L.T., 1980. Household Wealth in a Settlement Economy: Utah, 18501870. Journal of Economic History 40, 477-496. Klein, J., 1974. Farm-making Costs in the 1850s: A Comment. Agricultural History 48, 71-74. Lee, L-F., 1978. Unionism and Wage Rates: A Simultaneous Equation Model with Qualitative and Limited Dependent Variables. International Economic Review 19, 415-433. Lee, L-F., Trost, R.P., 1982. Estimation of Some Limited Dependent Variable Models with Application to Housing Demand. Journal of Econometrics 8, 357-382.

34

Leet, D.R., 1975. Human Fertility and Agricultural Opportunities in Ohio Counties: From Frontier to Maturity, 1810-1860. In: Klingaman, D.C., Vedder, R.K. (Eds.), Essays in Nineteenth Century Economic History. Athens: Ohio University Press. Libecap, G.D., 1992. The Rise of the Chicago Packers and the Origins of Meat Inspection and Antitrust. Economic Inquiry 30, 242-262. Lindert, P., 1988. Long Run Trends in American Farmland Values. Agricultural History 62, 45-85. Malin, J.C., 1935. The Turnover of Farm Population in Kansas. Kansas Historical Quarterly 4, 339-372. Malin, J.C., 1944. Winter Wheat in the Golden Belt of Kansas: A Study in the Adaptation to Subhumid Geographical Environment. Lawrence: University of Kansas Press. Mincer, J., 1978. Family Migration Decisions. Journal of Political Economy 86, 749-773. Miner, C., 2002. Kansas: The History of the Sunflower State, 1854-2000. Lawrence: University of Kansas Press. Olson, J.C., Naugle, R.C., 1997. History of Nebraska. Lincoln: University of Nebraska Press. Robinson C., Tomes, N., 1982. Self Selection and Interprovincial Migration in Canada. Canadian Journal of Economics 15, 474-502. Ruggles, S., Sobek, M., et. al., 2004. 1870 U.S. Census Integrated Public Use Microdata Series: Version 3.0, Historical Census Projects. Minneapolis: University of Minnesota. Schaefer, D.F., 1985.

A Statistical Profile of Frontier and New South Migration: 1850-1860.

Agricultural History 59, 563-577. Schaefer, D.F., 1987. A Model of Migration and Wealth Accumulation: Farmers at the Antebellum Southern Frontier. Explorations in Economic History 24, 130-157. Schapiro, M.O., 1982. Land Availability and Fertility in the United States, 1760-1870. Journal of Economic History 42, 577-600. Schwartz, A., 1973. Interpreting the Effect of Distance on Migration. Journal of Political Economy 81, 1153-1169.

35

Shannon, F.A., 1936. The Homestead Act and the Labor Surplus. American Historical Review 41, 637651. Shannon, F.A., 1945. The Farmer’s Last Frontier: Agriculture, 1860-1897. New York: Farrar and Rinehart. Sjaastad, L.A., 1962. The Costs and Returns of Human Migration. Journal of Political Economy 70, 8093. Soltow, L., 1975. Men and Wealth. New Haven: Yale University Press. Steckel, R.H., 1983. East-West Migration in America. Explorations in Economic History 20, 14-36. Steckel, R.H., 1989. Household Migration and Rural Settlement in the U.S., 1850-1860. Explorations in Economic History 26, 190-218. Taylor, G.R., 1957. The Transportation Revolution. New York: Rinehart and Company. Thernstrom, S., 1964. Poverty and Progress. Cambridge: Harvard University Press. Turner, F.J., 1893. The Significance of the Frontier in American History. In: The Frontier in American History, 1997. Tucson: University of Arizona Press. Turner, F.J., 1903. Contributions of the West to American Democracy. In: The Frontier in American History, 1997. Tucson: University of Arizona Press. Walker, T., 2000. Economic Opportunity on the Urban Frontier: Wealth and Nativity in Early San Francisco. Explorations in Economic History 37, 258-277. Webb, W.P., 1931. The Great Plains. New York: Grosset and Dunlap. Willis, R.J., Rosen, S., 1979. Education and Self-Selection. Journal of Political Economy 87, 7-36. Wright, G., 1987. The Economic Revolution in the American South. Journal of Economic Perspectives 1, 161-178.

36