[w]here non-Indian settlers flooded into the opened portion of a reservation and the area has long since lost its Indian character, we have acknowledged that de facto, if not de jure, diminishment may have occurred.
[Solom v. Bartlett 465 U.S., at 471 (1984)]
INTRODUCTION
Homesteading in the United States was created through the 1862 Homestead Act and ended in 1934.Footnote 1 The Homestead Act is arguably the most significant land disposal legislation in U.S. history, resulting in 270 million acres of land transferring from the public domain to private parties.Footnote 2 Historians tend to divide the homesteading era in two: the first phase, 1862–c.1890; and the second, c.1890–1934 (e.g., Edwards, Friefeld, and Wingo Reference Edwards, Friefeld and Wingo2017, 12). Economists and political scientists have mostly focused on the first 30 years of homesteading.Footnote 3 This difference no doubt arises from an interest in methods of establishing federal de facto rights, and homesteading’s signature wealth dissipation caused by the rush to make a claim (Anderson and Hill Reference Anderson and Hill1990).
In fact, the bulk of homesteading took place between 1900 and 1930. Figure 1 shows the prevalence of homesteading over time. Panel (a) depicts acres of patented homesteads in the 16 most western states between 1862 and 1940, and shows there were actually three major periods of homesteading. Cash sales were the main alternative to homesteads in transferring title to private individuals, and panel (b) plots the ratio of the number of homestead patents to private land sale patents. Until c.1900, there was a cycle in the relative amounts: homesteading dominated first, followed by land sales in several waves until c.1900. Then, homesteading relative to cash sales took off, and by the early 1930s, there were more than 20 times the number of homesteads being patented relative to cash sales. We call the surge in absolute and relative homesteading that began c.1900, “late homesteading.”
Figure 1. Homestead Arrivals by Year
Note: Using data on individual land patents from Allen and Leonard (Reference Allen and Leonard2021), we count total acres patented as homesteads in panel (a) and the ratio of homestead patents to cash sale patents in panel (b).
Late homesteading presents an interesting political economy puzzle for three reasons. First, late homesteading was incongruent with the Progressive era, and the general sense that resources in the West had become too scarce to give away and the “open” frontier was supposedly “closed.”Footnote 4 By the 1890s, the western lands had mostly been inventoried through surveys, mappings, and measurement of natural resources. Concurrently, the federal government restricted open-ended privatization of the west: national parks, forests, and grasslands were created and a general philosophy of federal land management to avoid “corruption,” “exploitation,” and “speculation” was developed.Footnote 5
Second, late homesteading took place under and after revised homesteading laws. From 1890 to 1916, Congress passed a series of new homestead laws that expanded homesteading on some dimensions, but greatly constrained it on others. The systematic alteration of homesteading legislation indicates that late homesteading was the result of a deliberate choice by the federal state, not a matter of simple inattention to existing policy.Footnote 6
Finally, late homesteading seems inconsistent with the general explanations of preemption (De Soto Reference De Soto2000; Murtazashvili Reference Murtazashvili2013) and earlier homesteading (Allen Reference Allen1991; Reference Allen2019; Barzel Reference Barzel1997; Frymer Reference Frymer2014; Reference Frymer2017). These works argue that the state encouraged squatters and homesteaders to rapidly settle, organize, and develop territories claimed by others. Occupation through squatting, and then early homesteading, established meaningful and feasible property rights to the frontier for the United States and therefore encouraged U.S. land sales and other development. However, by 1890, the west was secure; the Confederacy, the British, and others were gone; and there was no serious threat of violence from Native tribes who had been decimated through military conflict, dispossession, and loss of numbers (Anderson and Mc Chesney Reference Anderson and Mc Chesney1994; Spirling Reference Spirling2012).
In light of the Progressive era, the legal revisions, and the emergence of the plenary state, why did the federal government continue the dissipating practice of homesteading? We develop a hypothesis that complements the earlier theories by drawing on the literature on extensive Native land dispossession starting c.1880. Whereas early homesteading took place mostly on lands ceded through treaties with Native tribes, we suggest late homesteading was a response to potential internal legal threats against the federal government that arose out of dubious Native land dispossessions after failed attempts to renegotiate earlier treaties.
Although the federal government entered into hundreds of treaties with tribes, over the course of the late nineteenth century many were signed under duress, others later unilaterally changed, and still others were simply never ratified and ignored by the federal government—all leading to lands effectively “taken” from the tribes (Clark Reference Clark1994; Pommersheim Reference Pommersheim2012; Spirling Reference Spirling2012).Footnote 7 Legal challenges and political opposition over these takings arose immediately, and the federal claims over the land—both in the short and long term—were uncertain (Clark Reference Clark1994). Allowing a flood of settlers who were required to occupy the lands in question lead to “de facto, if not de jure diminishment” of the Native rights to the land.Footnote 8 Hence, the distinguishing feature of homesteading—the requirement to occupy the claim—continued to serve the federal government’s strategic interests long after the threat of violent conflict had abated.
We are certainly not the first to note the connection between Westward Expansion, homesteading, and Native land dispossession.Footnote 9 Previous scholarship has identified the various homestead acts as part of a larger set of colonial policies designed to encourage settlement of the American frontier without particular regard to Native Americans’ prior use of land and other natural resources (Banner Reference Banner2007; Edwards, Friefeld, and Wingo Reference Edwards, Friefeld and Wingo2017; Frymer Reference Frymer2014; Reference Frymer2017). Our contribution is to provide new quantitative and qualitative evidence on the particular importance of late homesteading in the context of potential legal challenges to U.S. territorial claims, as opposed to a focus on the importance of early waves of homesteading in shaping more direct violent confrontations with tribes.
First, we review the stylized facts of homesteading and U.S. land disposal policy. Second, we show that changes to the internal structure of homesteading were consistent with our hypothesis. After reviewing the qualitative evidence, we use the universe of General Land Office homestead land patents and cash sales from 1862 to 1940, matched with land cessions, treaties, and reservations across 16 western states, to show that late homesteading primarily took place on lands taken and where the threat of legal challenges was greatest. Finally, we assess the plausibility of alternative explanations for late homesteading.
LATE HOMESTEADING FACTS
Brief History
Homesteading, created in the nadir of the Civil War and designed for those who never took up arms against the Union, very quickly moved from a policy of northern occupation of the West against the Confederacy, to white occupation on ceded lands to push Native tribes onto reserves. During this early era (1862–c.1890), the General Land Office was mostly concerned with occupation by an “actual settler,” even if proper protocols were not followed.Footnote 10 Indeed, every accommodation was made to squatters to become homesteaders without penalty, and all land, mineral, and timber rights on the plot of the settler’s choosing were transferred.
Late homesteading, by contrast, generally did not grant rights to minerals and timber (Gregg Reference Gregg2019, 259). Furthermore, the General Revision Act (1891) repealed the Preemption Act (1841), which removed squatting and forced late homesteaders to settle on only surveyed lands. Late homestead lands had to be unappropriated, not be mineral in character, and not be withdrawn, reserved, incorporated, or otherwise embraced for trade or business (General Land Office 1926, 2). This Act also changed the amount of time to commute a homestead from 6 to 14 months (Hibbard Reference Hibbard1939, 389), and granted the federal government the right to remove public lands from entry.Footnote 11
At the turn of the twentieth century, other federal acts placed more control over homesteads. The Kinkaid Homestead Act (1904), the Forest Homestead Act (1906), the Enlarged Homestead Act (1909), and the Stock-Raising Homestead Act (1916) all increased the amount of land that could be homesteaded, while reserving the mineral and timber rights to the federal state.Footnote 12 These larger plots, combined with irrigation projects stemming out of the Reclamation Act (1902), made dryland farming viable in areas previously constrained to 160-acre homesteads, but reduced the ability to commute and extract timber or minerals without a market purchase. The Three Year Homestead Act (1912) extended the commuting time to 3 years, and drastically reduced the ability to assemble large tracts of land quickly through homesteading.
During the early twentieth century, the Progressive Republican party dominated national politics, but the larger homesteading restrictions tended to occur when Democrats held greater power—as depicted in Supplementary Figure A1 and discussed in Appendix A of the Supplementary Material. Thus, both parties during the transition to late homesteading re-examined the policy and chose to redesign rather than terminate it. Throughout these revisions, what remained was the fundamental character of homesteading: land ownership through first occupancy by a settler. The amount of legislation, over a long period of time, suggests that late homesteading was not accidental, not an unintended consequence of some other land policy, and not driven by specific state interests.
Table 1 lists various totals of the number of different types of homesteads in our 16 state sample. Of the 1,271,372 total homesteads in our sample, 798,339 (63%) were after 1895. Although there were many types of different homesteads that corresponded to specific legislative acts, the vast majority of late homesteads were created through the original 1862 Act.
Table 1. Early versus Late Homesteading by State
Note: A late homestead is defined as being after 1895.
There was wide variation in the amount of late homesteading across the West, as reported in the lower panel of Table 1. Montana had the most late homesteading of any single state, both in terms of totals and percentages. Within states, there was also great variation in the amount of late homesteading. For Montana, the percentage of late homesteads out of all land patents varied from as low as 5% in Glacier county, to as high as 80% in Hill county. Figure 2 shows the share of late homesteads relative to all homesteads plus cash sales in each 6×6-mile township in our sample. The darker areas show more intensive late homesteading.
Figure 2. Geographic Distribution of Late Homesteading
Note: This figure depicts the intensity of late homesteading in each 6×6-mile Public Land Survey System township in the 16 western states. We define the share of late homesteads as Late Share = (Late Homesteads)/(All Homesteads + All Cash Sales), where a homestead is considered “late” if it has an arrival date after 1895. Data are described below.
Logistics of Land Disposal
Understanding late homesteading and how it competed with cash entries requires a knowledge of public land disposal during the era.Footnote 13 The process began with a land survey, after which lands that were not sold at auction or reserved by the government became “lands offered for private entry” and were available for both private purchase or homesteading. In each case, the land was claimed by the first person who could register at the local land office for a particular plot of land. For a cash entry, the patent was fee-simple immediately for $1.25 per acre. In contrast, a homestead patent was only granted after continuous 5-year occupation that involved land improvements and the construction of a home and out-buildings. After this “proving up,” homestead ownership was also fee-simple. Homesteaders were required to farm, and therefore homesteaders were only interested in farmland. Cash entrants potentially could have valued lands with commercial or future development potential.
Whether land was purchased or homesteaded depended in part on the net present value of the land, V, for each type of claimant. If offered lands had a negative value,
$ V<0 $
, then no homestead or cash claim would be made and the lands remained owned by the state. Lands that had a positive value for farming but were worth less than the minimum price,
$ 0<V<1.25 $
, were subject to racing only among homesteaders. However, if the value of offered lands became greater than the minimum price,
$ V>1.25 $
, then both cash and homestead entrants raced to claim. Since each type of claimant had different land valuations and costs for a given parcel, there is no reason a priori to think that one type of claimant would dominate the other when
$ V>1.25 $
.Footnote 14
There are two important implications. First, generally speaking, all lands offered for private entry were available to both cash and homestead entrants; lands were not typically “set aside” for just homesteading. Second, the minimum price for cash entry meant that for lands valued less than this price, homesteaders had no cash entrant competitors.
A THEORY OF LATE HOMESTEADING
Our theory is that late homesteading was the result of a time-consistency problem on the part of the U.S. federal state vis-à-vis its treaty relations with Native tribes, combined with the threat of future legal action over dispossessed lands.
The U.S. federal government had made treaties with Native tribes since the Revolution, and throughout the first half of the nineteenth century, treaties were made with western tribes for eastern tribe relocation; to secure trails to the west coast; and to establish forts, Indian reserve lands, and locations for white settlement. Prior to c.1870, these treaties were made because it was in the U.S. government’s interest to do so—it lacked the plenary power to confiscate the lands (Frymer Reference Frymer2014). Given their relative strengths at the time, several Great Plains tribes ended up with massive tracts of land in exchange for concessions to the United States.
After the Civil War, the situation changed rapidly. The relative military strengths continuously moved in favor of the federal government. Political forces in western states—driven by farming, ranching, and mining interests—clamored for more lands to be opened up for settlement, and lands set aside as Indian Territory were constantly encroached on (Carlson Reference Carlson1978). Furthermore, transcontinental railroads and their spur lines increased land values and induced westward migration (Allen Reference Allen2019). Within this context, attempts to renegotiate earlier treaties were made, but by c.1870 it became evident that this would not be possible, as tribes refused to engage.
These unexpected and dramatic changes meant that earlier treaties were no longer self-enforcing, and by the late nineteenth century, the federal state could, de facto, confiscate tribal lands on a wide scale, and it began to do so. There were hundreds of legislative and executive acts, with the series of allotment acts perhaps the largest and most infamous. There was only one thing that stood in the way: the rule of law. Lands questionably taken by the federal government could be reclaimed through the courts, creating significant legal uncertainty.
Late homesteading did not arise because lands were set aside for homesteading. In principle, most lands offered for private entry could be claimed by either homesteaders or cash purchasers. However, the questionable legal status of lands “taken” reduced their value, and made it more likely they were worth less than $1.25 per acre, and therefore unattractive for purchase. Thus, the legally questionable land confiscations that created large amounts of low valued land, combined with the logistics of land disposal, selected homesteaders over cash entrants.
But this begs the question: why would the state be interested in allowing homesteaders on these lands rather than cash entrants? An alternative policy might have been to hold the lands until land values increased to the point where cash entrants were willing to purchase them, and thus avoid the dissipation of rushing. Given that the federal government was reviewing and altering the homestead laws, and that the progressive Republican party dominated national politics during this period, it was feasible to eliminate homesteading altogether.
The answer is found in the signature characteristic of homesteading: occupation by actual settlers. Settler occupation disrupted tribal land uses, physical development, and infrastructure; it also created vested political interests in maintaining non-native settlement. These irreversible effects of settlement meant that even a future legal loss could only result in a payment to tribes, not the return of the land. This reduction of the tribal land base furthered federal efforts to continually diminish tribes’ sovereignty, which was inextricably linked to their ownership of the lands that comprised their territories (Carlos, Feir, and Redish Reference Carlos, Feir and Redish2022). By using homesteading to occupy these particular lands, any legal threats against dispossession became moot; any future court settlement effectively became a forced sale of the land.Footnote 15 Thus, the federal state strategically allowed homesteading to continue in order to solidify the transfer of lands away from tribes. This strategy complemented the various political forces that wanted lands to remain in the hands of non-native settlers.
Since homesteading dominated land sales when land values were low, we predict late homesteading arose where there was a possibility of dispossessed treaty lands returning to tribes. These lands were where homesteaders were likely to be most prevalent relative to cash purchases due to decreased land values, and they were precisely the lands where the government most wanted to ensure occupation. Thus, the probability of homesteading should increase with increases in the probability of losing a court challenge: the more legally dubious the land taking, the more likely the lands will be homesteaded rather than sold.
QUALITATIVE EVIDENCE FOR THE LATE HOMESTEADING HYPOTHESIS
The Evolution of Indian Land Policy
Until the early nineteenth century, tribes held an independent sovereignty, with usufruct and transfer rights to land, along with the ability to enter international treaties (Ball Reference Ball2016, 10; Fouberg Reference Fouberg2000, 3). In the 1820s–1830s, three Supreme court cases known as the “Marshall trilogy,” significantly redefined this understanding: no independent sovereignties could exist within the United States, despite historical treaties; tribes were ruled wards of the state and domestic dependent nations; and tribes were recognized as distinct nations with their own institutions and self-governance, but Congress was recognized as having a plenary right to protect tribes from state law. Thus, although Native American tribes occupied lands, and could enter treaties, the ultimate ownership rested in the federal state (Ball Reference Ball2016, 11, 20–1; Royce Reference Royce1899, 528), and therefore tribes did not have the right to sell their lands.Footnote 16
These three cases set the stage for the period of eastern tribal removal. The Indian Removal Act (1830) gave the federal government the authority to move eastern tribes west of the Mississippi in exchange for lands, annuities, and other support. The process of moving eastern tribes across the Mississippi required the arrangement of treaties with western tribes. Although these treaties transferred millions of acres of Indigenous lands to the federal government, they also established legal property rights to large amounts of land for both western and eastern tribes under U.S. law (Calloway Reference Calloway2013, 172–3). However, the arrival of eastern tribes, the steady flow of squatters and wagon trains, and the general absence of military enforcement led to a series of conflicts across the Great Basin, Southwest, and Great Plains. Earlier treaties were often considered void, and new treaties were then negotiated. Prior to 1870, even these later treaties still contained large tracts of Indigenous lands that reflected the substantial tribal power of the time. Appendix A of the Supplementary Material discusses the example of the 1851 Fort Laramie Treaty.
Immediately following the Civil War, the Indian Peace Commission set out to reconcile tribes to the coming American settlement.Footnote 17 The Medicine Lodge Treaties (1867) and the second Fort Laramie Treaty (1868) further reduced the land holdings of western tribes, but still left substantial areas under tribal control. Seldom were the terms agreed to by all parties, and terms negotiated by treaty agents were often altered unilaterally by the Senate. Furthermore, in practice and despite the treaties, the U.S. government often continued taking Indigenous lands. According to Calloway (Reference Calloway2013, 213) “Medicine Lodge was the last great treaty council held on the Southern Plains and it was one of the very last treaties.”
After 1870, the power balance between the western tribes and the U.S. shifted dramatically in the latter’s favor. Spirling (Reference Spirling2012) uses a textual analysis of all six hundred tribal treaties in the United States and finds that over time the treaties became harsher in terms of language and land loss, and that this was due to “the general trend of rising American government power …” (85).Footnote 18 The 1871 Indian Appropriation Act ended the acknowledgment of tribal nationhood and terminated the treaty process. The BIA ceased to recognize tribes’ “domestic dependent nation” status, and all Native Americans were held subject to the laws of the United States.
Native lands continued to be transferred to the federal government and subsequently placed into the public domain for settlement. These transfers came through agreements, legislation, and executive orders. Within the context of the Indian Wars of the time, if the federal terms were not met, lands were simply taken. Sometimes these takings were explicit, as in the 1877 extinguishment of Sioux off-reservation rights that shrank their lands from 134 million acres to just 15 million (Calloway Reference Calloway2013, 234). Other times, the lands were quietly expropriated.
Legal Challenges to Land Takings
While all Western lands were ultimately taken from tribes in various ways, our focus is on takings that were potentially dubious from the standpoint of U.S. law. Court challenges made by tribes over lost treaty rights are common, and challenges to lands taken in the late nineteenth century were immediate. Early cases include US v. Kagama (1886), Lone Wolf v. Hitchcock (1903), US v. Winans (1905), and US v. Winters (1908). Indeed, Lone Wolf provides an examplar of the legal threat to the practice of dispossession and the general public awareness of this threat.Footnote 19
In 1890–91, the Jerome Commission began negotiations for allotment with various tribes in Indian Territory, including the Kiowa who frustrated negotiation efforts by refusing to deal. Not deterred, the commissioners dictated an agreement to the secretary of the commission. Then,
[t]he Jerome Commissioners departed. Once safely in Washington, D.C., the commissioners switched versions of the agreement, substituting their altered copy for that which had been partially signed at the councils. The changed agreement in the commissioners’ hands, …ultimately contained only counterfeit signatures. …Even with the alterations to it, the Jerome Agreement was between 21 and 91 signatures short …. (Clark Reference Clark1994, 48–9)Footnote 20
Lone Wolf, a Kiowa band chief who had been coerced to take an allotment of land in the deal, went to court to halt the process. He was not alone in his opposition. The Office of Indian Affairs opposed ratification in 1892, and an independent agency called the Indian Rights Association (IRA) began to investigate and eventually helped fund the legal challenge.Footnote 21
The case went to the Supreme Court, which ruled in favor of the state in 1903, much to the surprise of those involved.Footnote 22
For our purpose, Lone Wolf demonstrates three key points: that the federal government breached good faith; many in government and in the public were aware of and opposed to this; and legal challenges were immediate.Footnote 23 Court cases continued throughout the twentieth century, and there have been at least eight cases involving the question of “reservation diminishment.” In Appendix A of the Supplementary Material, we discuss these cases in more detail and argue that they demonstrate: (i) how uncertain the claims of federal sovereignty are, (ii) that occupation by non-tribal residents is a factor in determining ultimate ownership, and (iii) that sometimes the tribes are victorious.Footnote 24
The Allotment Era
Federal policy regarding Indigenous lands during late homesteading was dominated by the General Allotment Act (1887)—and subsequent Acts and court decisions that amended it—which gave the federal government the power to unilaterally abrogate past treaties and divide tribal lands into individually owned parcels, varying in size from 80 to 320 acres (Ball Reference Ball2016, 24). The act gave the Secretary of the Interior the power to assign an allotment, even to tribal members who refused to participate. These lands were transferred to individual tribal members, who were intended to take up agriculture and assimilate.Footnote 25
The Allotment Act (aka the Dawes Act) was extended and supplemented over time. It was applied on a tribe-by-tribe basis, and during the nearly 50 years that the Act was in force, roughly half of western reservations were allotted (Wall Reference Wall2010). Tribes generally opposed these transfers, and litigation continues to this day over their legitimacy. The allotment era was a low point in the respecting of tribal rights by the federal government, even as laid out in Worcester, and ended in 1934 with the passage of the Indian Reorganization Act.
Once a reservation was selected for allotment, all tribal members on that reservation were given allotments. Any excess lands left over were declared “surplus” and were then taken and made available for non-Native settlement. In most cases, these lands were removed from the reservation and governance of the tribe (Dippel, Frye, and Leonard Reference Dippel, Frye and Leonard2022). In total, through the various allotment acts between 90 and 100 million acres of previously recognized Indigenous lands (amounting to two-thirds of the tribal land base) were removed from reservations (Ball Reference Ball2016, 21; Wall Reference Wall2010, 6).
Two features of allotment provide strong institutional support for our hypothesis. First, the specific terms under which surplus lands were made available to whites. By law, surplus lands under the Dawes Act were to be sold until the revenues provided enough compensation to the tribes as laid out in the relevant legislation, precluding the use of actual homesteading to settle these lands. However, lands sold under the Dawes Act required the same occupancy restrictions as homesteads (Edwards, Friefeld, and Wingo Reference Edwards, Friefeld and Wingo2017, 118). The fact that lands sold within the areas taken under the Dawes Act had homestead occupancy restrictions supports our hypothesis. The lands had to be sold to pay the tribes, but occupancy was intentionally added to mimic the one feature of homesteading that the state had maintained throughout the late homesteading era in order to avoid tribal repossession of land.
Second, allotment was abruptly ended in 1934, in large part due to concerns raised by the 1928 “Meriam Report.” Indian Agents were responsible for declaring allottees “competent” to hold fee-simple title, but significant top-down efforts by the federal government—such as the 1906 Burke Act and the “Competency Commission” in the 1910s led to massive conversion of land to fee-simple and subsequent sale to whites (Dippel, Frye, and Leonard Reference Dippel, Frye and Leonard2022). Broad criticism of these rapid and perfunctory transfers and their impact on tribes by Meriam (Reference Meriam1928) were partially responsible for the major federal shift away from allotment in 1934. This shift underscores the fact that the federal government was keenly aware of major criticisms leveled at its Indian Policy during the period we study, and that once the flow of low-valued allotted lands ceased, so did the surge in late homesteading.
The Enlarged Homestead Act
The Enlarged Homestead Act (EHA) of 1909 provides some additional evidence that federal officials were aware of the taken lands issue. The EHA increased the 160-acre restriction on homesteads to 320 acres on lands where irrigation was infeasible. Unlike the original Homestead Act that allowed settlers to select any lands offered for private entry, the Secretary of the Interior explicitly designated lands that could be settled under the Enlarged Homestead Act. If late homesteading was meant to encourage physical settlement of legally contestable lands, then we would expect these designated lands to be disproportionately located on taken lands in an attempt to direct homesteaders onto these lands.
We provide evidence consistent with this hypothesis using a map issued by the Department of the Interior for the state of Montana—where the majority of late homesteading occurred—in 1912.Footnote 26 We georeferenced this map to identify the relationship between lands designated under the EHA and those taken by the state. Supplementary Figure A4 depicts the original map and our georeferenced version overlaid with taken versus ceded lands. We estimate that 84% of land designated under the EHA was taken.Footnote 27 This provides additional evidence that federal officials at the time were actively seeking to promote homestead settlement on taken land. Next, we provide more formal quantitative tests of this hypothesis.
QUANTITATIVE EVIDENCE FOR THE LATE HOMESTEADING HYPOTHESIS
Identifying Taken Lands
To test our theory, an operational notion of “land taking” must be used.Footnote 28 Our major historical resource is Royce’s Reference Royce1899 compilation of Indigenous land cessions (Royce Reference Royce1899). We use this to determine which lands were ceded through something like a voluntary exchange, and which lands were taken using either illegal, illegitimate, or otherwise legally questionable means. Royce (Reference Royce1899) lists the dates, legal reference, and descriptions of the over seven hundred land cessions, and tracks their evolution as they were modified over time.Footnote 29 Most importantly, Royce often provides historical remarks that fill in context and details missing from the technical descriptions of the cessions. For example:
No treaty was ever made with these Indians for the extinguishment of their territorial rights. The U.S. simply took possession of their country, except such portions as have been set apart by executive order for their occupancy. (Royce Reference Royce1899, 857, in reference to Cession #532)
We use this textual information, along with various histories, to classify particular land cessions as either (legitimately) “ceded” or (illegitimately) “taken.” Although all Western lands were ultimately taken from tribes in one way or another, our interest is in distinguishing takings that were possibly subject to redress under U.S. law. We define a land cession as “ceded” if there is evidence it was voluntarily entered into by each side. That is, if there is evidence in the historical record of (i) an actual exchange of lands for some type of consideration, (ii) no indication of duress or ex post unilateral change of terms, and (iii) no evidence of outright Native opposition to the cession.
On the other hand, we define a land cession as “taken” if (i) there is evidence of a unilateral action by the federal government, (ii) there is no evidence of consideration, (iii) the treaty was not ratified, but the lands were taken anyway, and (iv) if there was open opposition by the tribes involved with the cession. We relied heavily on the comments in the Royce report. In that report, Royce describes many cessions as “taken possession of without formal purchase” (831); “no formal purchase, treaty not ratified” (825); “Country claimed by … was taken possession of by settlers …without purchase of title by the U.S.” (836); “U.S. took possession … without any treaty” (869).
Figure 3 depicts the borders of the various land Cessions from Royce (Reference Royce1899) along with our coding of which cessions were “taken.” The map shows that some cessions were enormous in size, while others were quite small. In total, there were around seven hundred cessions in the United States, with about four hundred taking place in the West—most of them prior to 1890.
Figure 3. Land Cessions and Lands Taken
Note: Lines show Cession borders, darker shaded areas indicate “Taken” cessions, based on Royce (Reference Royce1899).
Land Patent Data
Our data on homesteading and cash sales come from the digitized federal land patents originally issued by the General Land Office (GLO).Footnote 30 In addition to the issue date and the type of land claim, the BLM data report various characteristics of the patents themselves.Footnote 31 Each patent is geographically identified through its Public Land Survey System (PLSS) identifier, which is made up of an aliquot part, section, township, and range numbers. Within each township, there are 36 square-mile (640-acre) “sections” numbered 1 through 36. Each patent was georeferenced using federal government shapefiles of the PLSS to the section level, which means we locate every homestead and cash sale patented issued over 1860—1940 to within a square mile. We overlay the section grid with the land cessions to determine what land cession and treaty lies behind every land patent. This allows us to link lands taken or ceded with each land patent according to the coding of tribal lands from Royce (Reference Royce1899).
We also calculate a variety of geographic characteristics associated with each PLSS section including elevation, ruggedness, soil quality, and distance to streams and contemporary railroads. Additional details on variable construction are available in the Supplementary Material.
Empirical Strategy
To test the hypothesis that late homesteading is more likely to occur on tribal lands that were taken by the U.S. government, we construct three different sample comparisons: (i) homesteads after 1895 vs. homesteads before 1895 and all cash claims (late homesteads vs. all other claims), (ii) homesteads vs. cash sales after 1895 (late homesteads vs. late cash sales), and (iii) homesteads after 1895 vs. homesteads before 1895 (late homesteads vs. early homesteads).Footnote 32
The first comparison provides the broadest assessment of our main prediction. The next two comparisons break the sample into two parts for narrower tests. The second comparison narrowly tests whether late homesteading was more likely on taken lands by only considering land patents created after 1895. The final comparison exploits the fact that there were lands taken prior to 1895 as well, coupled with the fact that the differential impact of taken land on land values mostly occurred after the U.S. government’s plenary sovereignty emerged at the end of the nineteenth century. Taken together the results test whether late homesteading depended on the fact that lands were taken combined with the emergence of large amounts of taken land at the turn of the century.
For all three comparisons, we estimate a linear probability model of the form:
$$ \begin{array}{rl}1{(LateHomestead)}_i={\beta}_0+{\beta}_11{(Taken)}_i+\lambda {\overrightarrow{X}}_i+{\varepsilon}_i,& \end{array} $$
where
$ 1{(LateHomestead)}_i $
is an indicator that is equal to 1 if a plot of land was homesteaded after 1895, and 0 otherwise, and
$ 1{(Taken)}_i $
is an indicator equal to 1 if a plot of land falls within a territory that was taken by the federal government.
$ {\overrightarrow{X}}_i $
is a vector of control variables and
$ {\varepsilon}_i $
is a random error term.
Identification of
$ {\beta}_1 $
is primarily threatened by two factors. First, unobserved determinants of land takings may be correlated with homesteading. That is, something like good farmland may have been more likely taken, and may also have been more likely homesteaded. This is unlikely for two reasons. First, as noted, homesteading was more likely on marginal lands. Second, previous studies have found that the U.S. desire for lands were primarily driven by relative military capacity and politics (Anderson and Mc Chesney Reference Anderson and Mc Chesney1994; Spirling Reference Spirling2012), and not agricultural viability. Both suggest that “taken” territories were not selected on land characteristics. Still, we control flexibly for a variety of geographic characteristics to limit the possibility of omitted variable bias. Moreover, our comparisons of late homesteads to late cash sales are less likely to suffer from this problem, as cash sales and homesteads tended to have similar land characteristics within states (Allen and Leonard Reference Allen and Leonard2021).
The second threat to identification is reverse causality. Our interpretation of
$ {\beta}_1 $
would be compromised if late homesteaders encroached on tribal lands, prompting the government to take those lands rather than pursue treaties. Three facts make this unlikely. First, the federal government had actively sought out to renegotiate treaties in the 1870s and 1880s, ahead of late homesteading. Second, the General Revision Act (1891) removed the rights of squatters and forced homesteading on surveyed lands. Third, the timing is wrong; most land takings occurred prior to 1895, ruling out reverse causality.
Table 2 reports the results of estimating Equation 1 using ordinary least squares with a variety of specifications across the three different samples mentioned above. Column 1 includes state fixed effects but no controls. Column 2 adds linear controls for ruggedness, elevation, soil quality, distance to the nearest stream, distance to the nearest railroad, and the latitude and longitude of each plot’s centroid. Column 3 adds controls for the number of individuals associated with a patent and indicators for whether a patent included subsurface rights or was surveyed via metes and bounds. Column 4 uses semi-parametric, binned land quality controls by including indicator variables for each decile of each land quality variable (excluding the coordinate controls, which are still included linearly). Column 5 adds dummy variables for distance to the nearest reservation, in 50-mile increments from 0 to 200. Column 6 excludes plots associated with lands that were taken after 1895, ruling out reverse causality. We cluster standard errors by county in all cases.
Table 2. The Probability of Homesteading on Ceded versus Taken Land
Note: Results of estimating Equation 1. Land quality controls include elevation, ruggedness, soil quality, distance to nearest stream, distance to railroad, and the latitude and longitude of a section’s centroid (the latter is included linearly in specifications where we bin the other land quality controls. Patent controls include the number of names associated with a patent and indicator variables for whether a patent included mineral rights and whether it was part of an irregular survey. Early takings are those that occur before 1895. Supplementary Tables C1–C3 present coefficient estimates for all control variables. Standard errors are clustered by county and reported in parentheses. *
$ p<0.1 $
, **
$ p<0.05 $
, ***
$ p<0.01 $
.
Panel A of Table 2 compares late homesteads to all other claims in the sample: homesteads settled before 1895 and all cash sales. The impact of taken territory is positive and significant in all specifications. Hence, plots in taken territory were more likely to be late homesteads compared to any other type of claim in the sample. The column 6 estimates imply that taken plots were
$ 0.185/0.436=42.4\% $
more likely to be late homesteads.
In Panel B, the sample is restricted to homesteads and cash sales with an arrival date after 1895, so the comparison is between late homesteads and late cash sales. Across all six specifications, taken territories are associated with a statistically significant increase in the probability that a plot settled after 1895 was homesteaded rather than purchased. The coefficient estimates range from 0.045 to 0.1. Adding additional controls reduces the magnitude of the coefficient somewhat. In our preferred specification in column 6, taken lands are associated with a 8.4 percentage-point increase in the probability that a plot was homesteaded—a 10.7% increase relative to the baseline probability of 77.9%.
Panel C of Table 2 compares late homesteads to early homesteads, restricting the sample to only include homesteaded plots. Across all specifications, taken territory is associated with a statistically significant increase in the probability that a plot was homesteaded late, rather than early. Our preferred estimates in column 6 indicate that plots in taken territory prior to 1895 were
$ 0.210/0.684=30.7\% $
more likely to be homesteaded late, rather than prior to 1895.Footnote 33
Although we control for a variety of geographic and institutional factors, there may remain unobserved characteristics that are correlated with late homesteading and taken lands. To assess the extent to which unobserved confounders could be driving our results, we calculate Oster’s (Reference Oster2019)
$ \delta $
for each specification that contains controls and report the results in Table 2. Oster’s
$ \delta $
formalizes the logic of “coefficient stability” by comparing the change in the coefficient of interest to the corresponding change in the adjusted R-squared when controls are added to the model. The resulting parameter,
$ \delta $
, is a measure of how large selection on unobservables would have to be relative to selection on observed covariates to make the coefficient of interest (Land Taken) statistically insignificant. Our estimates of Oster’s
$ \delta $
range from 2.5 to 8.7, suggesting that unobserved covariates would have to matter 2.5–8.7 times more than the bundle of included controls to fully explain our estimated effect of Land Taken, which suggests that unobserved covariates are unlikely to undermine our core findings.
The Land Value Mechanism
Appendix B of the Supplementary Material provides suggestive evidence for the specific mechanism in our theoretical framework. Using county-level data on average land value per acre from the U.S. Census of Agriculture, we perform a simple mediation analysis and confirm that uncertainty about the legal title to taken lands reduced their value, making late homesteads relatively more likely. We relegate this exercise to the Supplementary Material because average farm values are only a proxy for the value of a specific land plot embodied in our theory, and because we lack the ability to interpret these results causally.
ALTERNATIVE HYPOTHESES
Here, we examine several alternative explanations for the practice of late homesteading. First, the “safety valve” hypothesis that westward expansion helped the United States absorb major waves of immigrants arriving during the late nineteenth and early twentieth centuries (Ferrie Reference Ferrie1997; Turner [1893] Reference Turner1966). Second, that late homesteading incentivized settlement in coordination with new irrigation projects being developed by the Bureau of Reclamation, which was established in 1902. Finally, we explore several other confounding factors that may be associated with late homesteading.
A Safety Valve?
It has long been argued that that homesteading was used to reduce the impact of large waves of immigrants on eastern settled states, and in 1920, close to 14 million Americans were immigrants.Footnote 34 Immigration at the turn of the twentieth century mostly included people from southern and eastern Europe. Thus, if late homesteading was the result of immigration pressures, the homesteaders should have been made up of people from these regions. One method to test for a difference is to simply look at the distribution of homesteader names before and after 1895, as reported in the GLO land patent data. Supplementary Figure B1 suggests there was little difference in surnames over time by comparing two density functions of the prevalence of last names for early and late homesteaders. There is no statistical difference in the distribution of names across the two periods.
We also use county-level census data to formally test the immigration hypothesis. The U.S. Census reports the total population as well as the “foreign-born” population for each county in every decade. We assemble a panel from 1870 to 1920 and calculate the change in the foreign-born population as a share of the total change in the population in each county in each decade to characterize the intensity of immigration at the county level, which we can then match with plots settled in each county in each decade. If homesteading was a mechanism for absorbing waves of immigration, then late homesteads should be more likely than cash sales over the period of 1895 to 1920 in counties where a larger share of the change in population is attributable to foreign-born individuals.
Panel A of Table 3 formally tests the immigration channel by regressing the late homestead indicator on the foreign share of county population change associated with the county and decade corresponding to the time and place a given plot was settled. The sample is homesteads and cash sales after 1895 only. The specifications mimic those found in columns 1–5 of Table 2. In column 6, we add the
$ Taken $
indicator to test whether the relationship between taken lands and late homesteading remains after controlling for foreign immigration. In all specifications, late homesteads are not more likely in counties with a larger share of foreign immigration in any given decade. Moreover, controlling for immigration actually increases the strength of the relationship between taken lands and homesteading in column 6 (compared to column 5 of Table 2).
Table 3. Alternative Explanations for Late Homesteading
Note: This table presents the results of tests for alternative mechanisms. The specifications mimic those in the first five columns of Table 2. In all cases, the sample consists of homesteads and cash sales after 1895. Individual patents are matched to county-level data based on the location of the patent and the arrival date. Supplementary Tables C4–C6 present coefficient estimates for included control variables. Standard errors are clustered by county and reported in parentheses. *
$ p<0.1 $
, **
$ p<0.05 $
, ***
$ p<0.01 $
.
In the Supplementary Material, we also show that, if anything, cash sales are more closely linked to aggregate U.S. immigration shocks (Supplementary Figure B2) and to county-level population increases than homesteading over 1890–1940.Footnote 35 This is consistent with our basic conceptual model of homesteading. Large population shocks would increase the demand for available land, raising its value, and reducing the total amount of land for which
$ 0<V<\$1.25 $
, where homesteaders held the advantage.
Reclamation and Homesteading
Congress created the Bureau of Reclamation (BOR) in 1902. The BOR used charges for water deliveries to help finance major dams and irrigation projects (Coman Reference Coman1911). To have any hope of solvency, a project needed farmers to work the land and pay for water. This, coupled with the desire to avoid speculation on future irrigated lands, led Congress to restrict settlement of BOR lands to homestead entries only (Coman Reference Coman1911). Perhaps late homesteading was driven by this channel.
To test this, we again use data from the 1910 Census of Agriculture, which reports the number of acres irrigated by BOR projects in each county. This figure was not reported regularly or systematically across censuses, so we focus on cross-sectional variation in BOR acreage. This test is sufficient because if the primary motivation for allowing homesteading was to settle lands serviced by the BOR, it seems unlikely that the government would have allowed homesteading to continue through the 1930s if no relationship between homesteading and BOR lands had emerged by 1910.
In Panel B of Table 3, we regress the late homestead indicator on BOR acreage in a county. As with immigration, the specifications mimic those in Table 2, and column 6 reintroduces the
$ Taken $
indicator. Across all specifications, the relationship between BOR acreage in a county in 1910 and the probability that a post-1895 plot was homesteaded rather than bought is small and statistically insignificant. Column 6 indicates that controlling for BOR acreage does not impact the relationship between taken lands and homesteading.
In Panel C, we include both potential mechanisms, following the same set of specifications and restricting the sample to plots settled after 1895, as in Panels A and B. Once again, there is no detectable relationship between late homestead and either immigration or BOR acreage, while the
$ Taken $
coefficient survives the inclusion of both measures. The
$ Taken $
coefficient is smaller in Panel C than in Panels A and B, and it is marginally significant, but this is primarily driven by changes in the sample associated with imperfect matches between the modern county boundaries reported in the patent data and the historical boundaries in the Census data.Footnote 36
Other Explanations
Finally, we explore the general relationship between late homesteading and frontier population growth, national politics, racial composition of settlement, and the transition to statehood for Western territories. Perhaps late homesteading was an important mechanism for pushing settlers into territories that remained sparsely populated in the 1890s, and had little to do with tribal land takings. Late homesteading may have been related to the dominance of the progressive Republican Party at the federal level, or to the desire of territories to transition to statehood at the local level.
Supplementary Table B8 presents the results of re-estimating our preferred specification from column 6 of Table 2 with a variety of additional controls. Column 1 reproduces column 6 of Table 2 for comparison. Column 2 adds a control for county-level population density in the year a plot was settled, and column 3 controls for Bazzi, Fiszbein, and Gebresilasse’s (Reference Bazzi, Fiszbein and Gebresilasse2020) “total frontier experience” through 1890. Column 4 adds indicators equal to one if Democrats controlled federal Congress, the presidency, or both. Column 5 adds a series of indicator variables for the decade in which the latest cession or taking associated with a plot occurred, to semi-parametrically control for the timing of Indigenous land transfers. Column 6 includes all of these variables together. The effect of land takings on late homesteading is robust across all columns and all panels in Supplementary Table B8, suggesting that these potentially confounding factors do not affect our core results.
Regarding statehood, Supplementary Figure B3 plots the total number of late homesteads against the date of statehood for all western states and demonstrates only four territories in the sample achieved statehood after 1895 during the “late” period. Moreover, as Supplementary Figure B4 depicts, the majority of late homesteading occurred in states that had already achieved statehood by 1895, such as Montana. Hence, it seems unlikely that territorial/state politics drove late homesteading.
Frymer (Reference Frymer2017) suggests that politicians had a strong interest in promoting white settlement of western lands, and perhaps homesteading was a way to accomplish this. In fact, there is little reason to think this would be the case, as homesteading had no specific racial restrictions associated with it. This is borne out by the data. Supplementary Figure B5 plots a binned scatter plot of the total number of late homesteads against the change in white population from 1900 to 1940 (both in logs), with no clear correlation or relationship emerging. Supplementary Table B7 tests this more formally, with a county-by-decade panel regression where we regress the decade-by-decade change in the white population share on various measures of homesteading. Once again, we find scant evidence to support this alternative explanation for homesteading.
CONCLUSION
During the early period of homesteading, large tribal territories were established, and homesteading (along with forced movement) was used to push tribes into these areas. By 1890, the entirety of U.S. policy to exercise sovereignty over the west had been successful. Railroad land grants ended, and the federal government pulled back the reins of transferring lands to private interests during the Progressive era. All land policy made an about-face and headed in the same direction. Except for homesteading.
We claim that the value of homesteading to the federal government always came from one key feature: homesteaders had to live on the land. When land was occupied, homes and barns were built, roads and stores arose, a certain type of development took place, and eventually population growth and cities made “going back” impossible. In the words of Justice Ginsburg, this would “…preclude the Tribe from rekindling embers of sovereignty that long ago grew cold.”Footnote 37
The Indian treaties of the mid-nineteenth century reflected the relative strength of western tribes at that time, and thus created a time consistency problem for the United States later in the century. The plenary power of the state allowed it to ultimately take back most of the tribal lands it had recognized in the earlier treaties, and late homesteading prevented the land from ever going back into tribal hands—even when the federal government’s actions were later ruled in violation. Consistent with this, we find that the location of late homesteading is well explained by the legal standing of the final land cessions that transferred Indigenous lands to the federal government. When those land cessions were more legally dubious, late homesteading took place. Late homesteading, therefore, was a major component of the dispossession of Indigenous lands during the early twentieth century. The large amount of late homesteading was a consequence of the enormous and questionable land transfers that took place in the latter half of the nineteenth century.
SUPPLEMENTARY MATERIAL
To view supplementary material for this article, please visit https://doi.org/10.1017/S0003055423001466.
DATA AVAILABILITY STATEMENT
Research documentation and data that support the findings of this study are openly available at the American Political Science Review Dataverse: https://doi.org/10.7910/DVN/C8V44A.
ACKNOWLEDGMENTS
Thanks to Jonathan Greenacre and Anji Redish for comments.
FUNDING STATEMENT
This research was funded by the National Science Foundation (Grant No. SES-2117660), and the Social Sciences and Humanities Research Council of Canada.
CONFLICT OF INTEREST
The authors declare no ethical issues or conflicts of interest in this research.
ETHICAL STANDARDS
The authors affirm this research did not involve human subjects.
Open access
