The Pattern of Settlement in Central New York

Prior to European colonization, the area of Central New York that we consider was inhabited by an indigenous group that Europeans called the Iroquois. They referred to themselves as the Haudenosaunee, or “people of the longhouse.” Before the American Revolution, there were no European settlers in Central New York. The Haudenosaunee were very well organized militarily and resisted outside incursions. By a royal proclamation issued in 1763, King George III laid out a borderline for the 13 colonies that established a western limit for British settlement. The bulk of the Haudenosaunee lands of Central New York lay to the west of this line. This proclamation and the military power of the Haudenosaunee helped ensure that the land would remain unsettled by Europeans until after the Revolution.

At the outset of the American Revolution, the Haudenosaunee confederacy, after long hesitation, decided to side with the British Crown against the American colonists. After they had engaged in raids on American settlements, George Washington launched an expedition, led by General John Sullivan, to the heart of Haudenosaunee territory, located in the Finger Lakes region of New York (Graymont Reference Graymont1972, 220). Washington instructed Sullivan to engage in what in today’s terms would be called an ethnic cleansing operation.Footnote ¹⁴ The Haudenosaunee themselves commonly referred to Washington by the epithet “town destroyer.”Footnote ¹⁵ Though Sullivan’s army found Haudenosaunee settlements to be abandoned, the destruction of all crops grown in the area—principally maize—forced the Haudenosaunee to flee in the direction of Canada.

The result of the Sullivan Expedition and of the eventual American victory against the British was that by 1783 a large section of Central New York was almost completely depopulated. There were some white settlers who attempted to settle in the area in the decade afterwards, but they were extremely small in number. The area would not be settled more extensively until the establishment of the New York Military Tract.

The New York Military Tract

After the end of the Revolutionary War, the new State of New York faced the same problem as many other states: how to compensate military veterans for wages that had never been paid. One solution adopted was to offer military veterans—or their descendants—land in lieu of cash. In New York State, the way in which these lands were allocated forms the basis for the natural experiment that we use in this paper.

The general model for how the unoccupied lands in Central New York were to be allocated was provided by an act of the Confederation Congress passed in 1785. The Confederation Congress (the predecessor to the U.S. Congress) recommended how lands west of the Ohio river that had been “purchased” from Native Americans should be laid out. Each township, laid out in a square, should be subdivided into lots of 640 acres numbered from 1 to 36. Four of the 36 lots were to be reserved for the use of the United States Federal Government, and lot number 16 was to be reserved “for the maintenance of public schools.”Footnote ¹⁶

Perhaps inspired by this model, in 1789 the New York State Legislature passed an Act charging the surveyor general of the state with laying out a grid of towns and individual lots in the area of the west of the state previously occupied by the Onondaga and Cayuga Haudenosaunee groups who would now be “given” two reservations inside the tract. Each township in what came to be known as the “New York Military Tract” was to contain 100 square lots of 600 acres each. The tract was to contain as many townships of this size that would fit, and this number eventually came to 28. The individual tracts within the 28 townships were then to be allocated by lot to military veterans or their descendants. Finally, after 94 lots were allocated in this way in each town, the remaining six lots were to be retained for other purposes. Four of these additional lots were to be used for compensating commissioned officers as well as those whose assigned lots turned out to be underwater. One further lot was reserved for promoting literature.Footnote ¹⁷ One final lot—and this is the one of particular interest to us—was to be reserved for “promoting the gospel and a public school, or schools” (New York Legislature 1887). This would be the basis for the school funding, the effect of which we examine in this paper. The idea was that these lots could be used to provide income from rental or sale to support school expenditures.

The New York Military Tract was surveyed by a team led by Simeon Dewitt beginning in 1789, leading to the production of the map that can be seen in Figure 1. Figure 2 then shows a census map from 1850, the central year for our analysis of economic outcomes, with the dark black line distinguishing between towns inside and outside of the Military Tract. The lots for the Military Tract were assigned in the Summer of 1790. Lots for former soldiers from a given regiment were assigned randomly from areas across the whole tract and therefore not in the same locality.

Figure 1. Survey of the New York State Military Tract as Conducted by Simeon Dewitt Beginning in 1789

Note: The tract towns and boundaries inside towns are shown in the upper left part of the figure.

Figure 2. 1850 Census Map of New York State with Tract Boundary in Black

Source: New York Public Library.

Our research design depends on comparisons of economic and political outcomes inside and outside the Military Tract, with the objective of attributing any such differences to the greater availability of schooling funds inside the tract. Therefore, it is important for us to next consider patterns of initial settlement inside and outside the Military Tract. On the face of it, one might think that the pattern of initial settlement inside the Military Tract was very different from that outside the tract. Inside the tract, veterans of the Revolutionary War were given lots of 600 acres each selected by lottery. Outside of the tract, there was no such lottery or a formal plan for land allocation. In practice, the initial pattern of settlement in the towns that lay across the Military Tract boundary turned out to be very similar for three main reasons.

First, inside the tract boundary only 5.9% of the veterans who had initially been awarded the lots actually settled there.Footnote ¹⁸ What happened instead is that various individuals purchased lots, and often a significant number of lots, from the veterans and then resold these to settlers. This was the case with Jeremiah van Rensselaer, a prominent scion of the family that owned very substantial tracts of land in the Hudson Valley.

Second, as we will describe below, the pattern of settlement in the areas bordering the Military Tract often began with a small number of individuals purchasing a great tract of land that was then quickly divided through sale and resale into smaller tracts on which individual families settled.

Third, we also have evidence to suggest that the initial white settlers to the Military Tract and to the lands that bordered it tended to come from similar places and to have similar backgrounds. This area was settled during an era that one historian has called the “Yankee Invasion of New York,” where individuals and families from New England sought to settle on agricultural land that held great promise (Ellis Reference Ellis1951).

To see the similarities between the pattern of settlement inside and outside of the Military Tract, consider the cases of Caroline and Dryden: two adjacent towns with the former lying just outside the Military Tract and the latter lying just inside it. The initial division of lots in Dryden was laid out in the same pattern as for other towns in the tract: 100 lots of 600 acres each that were then successively sold and subdivided. A map of this initial configuration can be seen in Appendix H, Figure A.2.

The town of Caroline emerged through a different process, but it was one that resulted in similar settlement boundaries. The area that would eventually become the town of Caroline was initially part of a large parcel of land purchased by John Watkins and Royal Flint from New York State in 1794. The area was then surveyed and subdivided into a series of towns and individual lots. A mid-nineteenth-century map of Caroline shows a similar pattern of square lots (see Appendix H, Figure A.3).

Consider next the initial settlers in the towns of Caroline and Dryden. Town histories suggest a pattern often referred to in the region more generally.Footnote ¹⁹ Settlers tended to arrive either from the eastern portion of New York or especially from New England. At the time of its incorporation in 1811, the Town of Caroline had some 74 adult settlers, the first of whom arrived in 1794 but the great majority of which arrived in 1800 or after. Several of these are known to have previously exercised professions such as millwright, blacksmith, carpenter, or tailor. The first settler in what would become the town of Dryden—inside the Military Tract—arrived in 1797 (Goodrich Reference Goodrich1898). From that point it appears that settlers began to trickle into the community at about the same rate as in Caroline, again with some coming from eastern parts of New York and most coming from New England. As in Caroline, there is evidence that some of these individuals had already exercised a profession such as carpenter or millwright. In the case of some settlers, such as George Robertson in 1797, a lot was purchased directly from the Revolutionary War veteran who had first been awarded the lot. In other instances, lots were purchased from others who had already purchased the lot in question from a veteran.

The pattern of settlement in Caroline, and its resemblance to that of the Military Tract town of Dryden, was hardly exceptional. The early history of the town of Macedon, located to the north of the Military Tract, suggests a similar pattern of settlement (Eldridge Reference Eldridge1912). The territory that would become the town of Macedon was initially part of the Phelps and Gorham Purchase, a large tract purchased by a syndicate that subsequently resold and subdivided the land. An early map of Macedon shows a similar pattern of square lots seen in other towns both inside and outside of the Military Tract (see Appendix H, Figure A.4).

The territory that would become the town of Macedon was purchased by Nathan Comstock from Cummington, Massachusetts, in 1789 from Oliver Phelps and Nathaniel Gorham. He then sold thirty-one of the seventy-two lots in the township to Stephen Warner, who came from Massachusetts to settle in the area (Eldridge Reference Eldridge1912, 13). These two individuals then sold their lots to other arriving settlers. The pace of settlement, slow but steady for the next two decades, appears to have mirrored that in Caroline and Dryden with many, and probably most, settlers coming from New England. The attraction is described as being the “virgin soil” of the area, which was preferable to the “stony hillsides” of places like Berkshire and Litchfield Massachusetts (Eldridge Reference Eldridge1912, 35).

These case studies bolster our claim that the process of initial settlement inside and outside the Military Tract was similar and that what was different was that once common schools were set up, only towns in the Military Tract had access to the gospel and school lot funds. We will consider more formal tests of this fundamental assumption by presenting balance tests in presenting our research design below.

Before providing historical context for investment in education in the common school era, it is useful to consider how towns managed the gospel and school lot funding. In order to manage the funds, each town within the Military Tract elected three trustees who then decided how to manage the gospel and school lot and allocate the available funds. The process that led to different towns in the Military Tract having different school fund outcomes was most certainly an endogenous one. It could have been affected by the preferences of residents of different towns for investing in public education. It could have also depended on how well the lots and their finances were managed. For this reason, in all of our analyses in the sections that follow we will first report regressions where the level of local school funds is included as an (potentially endogenous) independent variable, and we will then report results where we instead use presence inside or outside of the Military Tract as an exogenous predictor of the level of school funding. Presence inside the Military Tract can then give us what is equivalent to an intention-to-treat estimate.

The final issue in understanding the potential effect of the Military Tract is that the funds were nominally for gospel and schools, but our argument is focused on the influence of school funding. Is it possible that the additional funds shaped outcomes through religion rather than education? This could have been the case if church participation and the encounter with the values imparted therein prompted people to be more active in democratic politics. It is known that Central New York during the period we consider experienced a great increase in religiosity as part of the Second Great Awakening. Due to an overall increase in religiosity within existing sects, as well as the proliferation of new sects including the Mormons, the Millerites, and the Oneida Society, Central New York during this era has often been referred to as the Burned Over District (Cross Reference Cross1950). There are two major reasons to believe that even if Central New York experienced an increase in religiosity during this period, our comparisons of towns inside and outside the Military Tract reflect the effect of school funding on economic outcomes and democratic participation.

The first reason is that those who have written most extensively about the Burned Over District, and in particular Cross (Reference Cross1950), have not described it as a phenomenon centered on the Military Tract. Evidence for this can be seen in the location of communities in New York that in 1831 requested the aid of the Reverend Charles Finney, one of the great leaders of the Second Great Awakening whose activities did much to create the idea of a Burned Over District. These locations can be seen in Appendix H, Figure A.5 (Cross Reference Cross1950, 157). The Military Tract does not seem to have had more revival activity than other areas. It is also the case that among the new sects that developed in New York State during this period, neither the Mormons, nor the Millerites, nor the Oneida Society originated within the towns of the Military Tract itself.

The second reason why we believe that our results are not biased by religion is that the existing evidence suggests that the funds for the gospel and school lots for towns within the tract went overwhelmingly to schools and not to churches. In some instances this was accomplished by an act of the New York State legislature. We have identified several cases in which the New York State Legislature weighed in on the question of how gospel and school lot funds should be disbursed. In all of these instances it was stated that the lot funds should be used exclusively for schools (The four towns in question are Geneva, Genoa, Hannibal, and Hector; Hobson Reference Hobson1918, 121; Keyes Reference Keyes1879). As discussed above, the towns themselves also decided how funds from the gospel and school lots should be allocated. We have a record of the town of Ithaca doing this at its first meeting, which was held on April 3, 1821. At the meeting it was resolved that one dollar of the funds should go to the gospel and the remainder should go to schools (Peirce and Hurd Reference Peirce and Hurd1879, 427). At a meeting in 1818 in the neighboring town of Dryden, it was decided that all but the nominal amount of six cents of the lot funds be devoted to schools (Peirce and Hurd Reference Peirce and Hurd1879, 481). We have not been able to identify any town in the Military Tract that devoted more than a nominal amount of gospel and school lot funds to the gospel.

Undergirding all of this, there appears to have been a general sentiment that using lot funds to support the gospel ran the risk of privileging one particular sect over others in an era where all sects were to be treated the same. The centennial history of the town of Dryden suggests this about the motivation for the town to devote all of its gospel and school lot funds to schools.

The deeper background to this statement was that prior to the Revolution, the Colony of New York had an “established” church, even if there was free exercise of religion. The established Anglican church received tax support. The first Constitution of New York State in 1777 did away with this arrangement (Esbeck Reference Esbeck2004). In leading to this outcome, the cause of the Anglican church was not helped by the fact that its members predominantly sided with the British Crown during the Revolution.

Common Schooling in New York State

During the Early Republic a number of states in the Northeastern United States passed legislation providing for the organization and funding of what were called “common schools.”Footnote ²⁰ Common schools were meant to have a curriculum that was to some extent supervised by the state government while also receiving funding from the state government. Although they were state subsidized, common schools were not free, nor at the time we are considering was attendance compulsory. The New York State Legislature passed a common school law in 1812, something that had been under consideration for some time without great success (Van Ness and Woodworth Reference Van Ness and Woodworth1813). The new state law of 1812 stipulated that each town should elect commissioners who would establish common schools, that trustees were to be elected to manage each school, and that a central state fund would allocate money to the towns according to their population, based on the census of the United States.

Data on Education

Our research design relies on comparisons between towns in Central New York in the first half of the nineteenth century to evaluate the effect of education on income, wealth, and political participation.

To measure the extent of education provision, we constructed three variables from the Annual Report of the Superintendent of Common Schools (1828). The variable Total School Funds is equal to the total public money for schools in each town in 1827. This includes both money distributed by New York State and any money from the gospel and school lots for the towns of the Military Tract.Footnote ²⁵ We focus on the year 1827 because it is the first year that this comprehensive source is available. We also include Length of School Year, which is equal to the average number of months that schools in a given town were open in 1827. Finally we include Number of Schools, which for each town represents the total number of establishments receiving common school funding.

The education data are summarized in Table 1. We report the means for towns inside and outside the Military Tract, with the restriction that the town be located within 30 kilometers of the tract boundary. As can be seen, towns inside the tract had significantly more funds, a greater number of schools, and also a longer school year on average.

Table 1. Descriptive Statistics of Educational Outcomes

Note: The table reports educational outcomes for 1827. The source is Annual Report of the Superintendent of Common Schools (1828). Note that there are 48 towns in the Military Tract in 1827. The original 28 towns inside the tract increased in number due to divisions from the time that the tract was originally established.

As a next step, we can use our education data to provide an initial check on whether the availability of gospel and school lot funding could have resulted in the differences in numbers of schools and length of the school year between towns inside and outside the Military Tract. We would expect this to be the mechanism associated with improved economic outcomes and greater participation in democratic politics. To begin with, we need estimates of how expensive it was for a town to add a school and also how costly it would be to extend the school year. To do this we can consult the records of Bethel Grove School in Tompkins County. These show that in 1837, during the period we consider, a teacher named W. Grant received $25.97 for three full months of teaching (Mitchell Reference Mitchell1960, 23). It is also the case that labor costs appear to have been by far the heaviest burden during this era when a one-room school building could be constructed for less than a hundred dollars.Footnote ²⁶

In addition to having an effect on the extensive margin of education provision—the number of schools within a town—our wage estimate also allows us to assess the intensive margin, how expensive it would be for a town to extend the length of its school year. In Antebellum New York State, the length of the school year was not standardized. Statute dictated that a town could only receive common school funds if its schools maintained an academic year of at least six months, but beyond that individual towns were free to choose. Once again, teacher wages would have been the dominant consideration here.

As a final step, we can look at whether the $240 difference in school funds between towns inside and outside the military tract (see Table 1 and recall that $170 is the average amount of funds from gospel and school lots) could plausibly explain the difference in schooling inputs on both the extensive and intensive margins. Towns inside the Military Tract and within 30 kilometers of the tract border had an average of 15.9 schools and an average school year of 7.8 months. Towns outside the Military Tract and within 30 kilometers of the border had an average of 12.9 schools and a school year of 7.1 months. Recalling that teachers’ wages in the 1830s were $8.6 dollars per month and that labor costs were by far the largest cost component for education, we can construct a back-of-the-envelope calculation of how much money it would have taken for a town with 12.9 schools and a school year of 7.1 months to transition to 15.9 schools with a school year of 7.8 months. The answer is $279. Given the back-of-the-envelope nature of this estimate, it seems quite close to the $170 average of gospel and school funds from which towns inside the Military Tract benefited.

Data on Labor Earnings and Wealth in 1850

Our analysis of economic outcomes includes measures of both earnings and wealth, as shown in Table 2. Our earnings measures are based on the occupation score index developed by the University of Minnesota Population Center as part of the Integrated Public Use Micro Data Series (IPUMS).Footnote ²⁷ This measure starts with 1950 occupation categories and assigns to each of them an income. The score is that occupation’s median income in hundreds of dollars in 1950. The 1850 census occupation categories have been standardized with the 1950 categories, and the 1850 occupation score indicates the 1950 median earnings of that category. It provides a rank ordering of 1850 income.Footnote ²⁸

Table 2. Descriptive Statistics for Main Outcomes

Notes: “Std.” signifies that the variable has been standardized; AA = African American.

We use the occupation scores to construct two measures for each town. Earnings is equal to the town’s median occupation score in 1850. Earnings Inequality is equal to the difference between the town’s mean and median occupation score in 1850. We use the difference between mean and median earnings rather than something like the share of total earnings going to the top 1% because the census income categories do a better job of describing variation in much of the distribution but not in the top end where income at this time would have come primarily from capital earnings. The same reason applies for not using the Gini index.

Our wealth measures also come from the 1850 census. The variable Wealth is equal to each town’s median value of wealth in terms of real property. The real property data for each individual in the 1850 Census is the contemporary dollar value of any real estate owned. This measure then comes with one caveat, as it included the full value of the property rather than the value net of liens and mortgages. Also, it does not include financial assets, but real estate assets dominated the composition of wealth during this period for this area of the country. Wealth Inequality is equal to the percentage of total wealth in the town owned by the wealthiest 1%.Footnote ²⁹

Data on Political Participation

Our empirical analysis of political outcomes focuses on turnout. We expect education to influence turnout through several different channels. The first would be if the curriculum content fosters a sense of civic duty, and we have already shown reasons why this may have applied in Antebellum New York State. The second would involve the effect of education on economic outcomes. Individuals with greater economic capacity might find it easier to vote. Related to this, it is possible that inequality of economic circumstances will lead to a decreased incentive for many people to vote (Leighley and Nagler Reference Leighley and Nagler2013).

Our study adds to recent efforts to understand how education influences political participation by looking at variations in educational duration within the same population (Berinsky and Lenz Reference Berinsky and Lenz2011; Henderson and Chatfield Reference Henderson and Chatfield2011; Wantchekon, Klašnja, and Novta Reference Wantchekon, Klašnja and Novta2015). These studies confirm that being excluded from education is associated with lower subsequent political participation (see also Persson Reference Persson2015). We contribute to this debate by demonstrating that places that differ in the intensive and extensive margins of education provision also differ in their levels of voting participation. This finding is likely more important for cases similar to ours where most of the population is constrained to only receive shorter spans of formal teaching.

We focus attention on three elections and one referendum in New York State during the mid-1840s: the 1842 Gubernatorial election, the 1844 Gubernatorial election, the 1844 Presidential election, and an 1846 referendum on African American suffrage as reported by Benson and Silbey (Reference Benson and Silbey2002). Their dataset provides us with a rare view of town-level measures of voter turnout, whereas most sources for this period only report county-level measures. For each election and the referendum, we create separate turnout measures equal to the number of voters as a percentage of the number of individuals eligible to vote. We then average these four measures and construct the variable Average Turnout, which is the focus of our analysis. Although our focus is on turnout of eligible voters, the value of voting and high turnout is connected to recognizing political equality, which suggests the expansion of eligibility in the context of nineteenth-century New York elections. As highlighted above, because the curriculum did not advance ideas related to greater inclusion, we do not expect that education in this setting would have such an effect, and this expectation indicates an important limit to the role that the state played in advancing a participatory democratic culture. To investigate this, we study differences across towns in voting in an 1846 referendum on abolishing the property requirement for African Americans. The variable 1846 Pro Suffrage is equal to the percentage of voters favoring extending suffrage by abolishing a property requirement for African Americans to vote. The source for these data is also Benson and Silbey (Reference Benson and Silbey2002).

Research Design

In order to establish the effect of the additional school funding on our outcomes of interest, we use a geographic regression discontinuity design. This is a well-established research strategy for analyzing the effects of policies that change discontinuously at a geographical boundary.Footnote ³⁰ The validity of the design hinges on observations being otherwise similar across the geographical boundary that decides what specific “treatment” applies.

Before discussing how well the assumptions of the design are met in our empirical setting, it is useful to review briefly the timeline for our analysis. The timeline is dictated both by data availability and by a desire to consider the consequences of education over a time horizon where we would expect it to matter. Though the legislation establishing the Military Tract dates from 1789, initial settlement in the tract as well as the areas bordering it did not occur for several years after this point. The communities in this area only began to approach a significant size around 1810, and the New York State Legislature did not pass a common school law until 1812, with the law becoming effective in 1813. There is little evidence of school development in the area prior to this date but much evidence afterwards. The earliest date for which we have school funding data is 1827; our political data are from the 1840s, and our economic data are from 1850.Footnote ³¹

Although the timing of our data is in part determined by data availability, the periods between when the differential availability of school funds could start to make a difference and our economic and political outcomes make for a compelling test of the influence of the state. If we are going to consider the consequences of schooling, we need to allow for the fact that from the moment a child entered primary school, there would inevitably be a time lag before they could become eligible to vote and enter the labor market.Footnote ³² We have access to voting data at the town level from the 1840s and economic data from 1850, which should allow for enough time after the establishment of schools to observe an effect of schooling.Footnote ³³

Our geographic regression discontinuity research design is based on the assumption that prior to receiving gospel and school lot funds, there was no other meaningful difference between towns located inside and outside the Military Tract. As long as the two types of towns were geographically close to each other, we think it is reasonable to believe that this assumption holds. We already established the fact that it was the general practice for towns in this area to be laid out in square lots. We also established that although lots inside the Military Tract were initially allocated to veterans—a potential source of difference—very few veterans ended up taking up residence inside the tract as a result of this allocation. Apart from the gospel and school lots, there were no other legal provisions regarding allocation of resources or governance that made the towns inside the Military Tract any different from towns outside.

For all our analyses, we will only study towns that are located within 30 kilometers of the Military Tract boundary. Furthermore, we include a robustness test that shows that our results hold at shorter distances. In all of our analyses, we include controls for border segment fixed effects. We also include controls for latitude and longitude.Footnote ³⁴

One principal way in which the towns inside and outside of the Military Tract might have differed prior to the treatment was in their potential economic viability, and at this time this would have primarily been a story of agricultural suitability. Europeans settling in the area followed the example of the Native Americans before them and concentrated on maize before shifting subsequently to wheat (Hedrick Reference Hedrick1933). Therefore, in our estimates we include measures of exogenous suitability for producing maize and wheat at the town level.Footnote ³⁵ In addition, we use elevation to proxy for the ruggedness of the terrain and access to water transport, which was critical during this period. A second way in which the towns inside the tract might have differed from those outside is if they had a higher proportion of veterans. We established above that few of the veterans initially awarded plots in the tract actually settled there. We also showed that the randomized process for allocating lots was not done in such a way as to settle veterans from the same regiment in the same town. Even so, some might say that the presence of even a small number of veterans in town could make a difference for its politics, say if they played leadership roles. We have information at the individual household level for areas inside and outside the tract reporting the number of veterans per household (almost invariably either zero or one). These are from 1840—and therefore posttreatment—but they can still be used to investigate the relative proportion of veterans in the population and compare this between towns inside and outside of the Military Tract.

A third way in which towns inside and outside of the Military Tract might have differed prior to treatment generalizes the concern about veterans. What if these areas were settled by different types of people and this initial sorting influenced the trajectory of subsequent economic and political outcomes? One reason not to think that individuals with different experiences, skills, preferences, or abilities might have decided to settle inside or outside the Military Tract is the common origin for many of the immigrants who arrived during what has been called the “Yankee Invasion of New York” (Ellis Reference Ellis1951). For the towns both inside and outside of the Military Tract in Central New York, we see an enormous number of references to individuals arriving from New England, some of whom exercised a trade while the great majority engaged exclusively in farming. It would be more problematic for our design if there was clear evidence that the individuals settling inside the Military Tract tended to come from different locations than did those settling outside the tract.

Even if most of the first white settlers in the Military Tract came from New England, there remains the possibility that among the individuals who came from this region, different types of people preferred to settle inside the Military Tract towns, possibly even because of the expectation of better education in the future because of the gospel and school funds. For example, it could have been the case that individuals who were more active democratic participants to begin with were more likely to settle inside the Military Tract because of the benefits that its schools could provide by encouraging civic behavior on the part of their children. This would lead to an upward bias in our voter-turnout estimates. The individuals who settled inside the Military Tract might also have been distinctive on other dimensions, perhaps leading to faster economic development and in turn higher rates of voter turnout several decades down the road. This would be a second source of upward bias in our estimates.

To deal with the first of these two possibilities, we identified a source for town-level voting data from the initial period of settlement. We have voting totals for elections to the New York State Assembly in 1813, a year in which the Military Tract and adjacent areas were first experiencing significant settlement. We then constructed a proxy for voter turnout by dividing these vote totals by the total population of each town, given that we lack the number of eligible voters (Spafford Reference Spafford1813). Below we report comparisons of this voter turnout proxy for towns inside and outside the Military Tract.

With respect to the second possibility, that early white settlers across the Military Tract boundary may have differed on other characteristics, perhaps leading to more economic dynamism inside the tract, in addition to the agricultural suitability measures discussed above, we can use population density: a common proxy measure in the economic history literature for economic development in an agricultural society. We used town population data and town area in 1813 to conduct a balance test of population density inside and outside the Military Tract at this initial moment of settlement.

In Table 3 we report summary statistics and balance tests for our measures of agricultural suitability, elevation, the fraction of households with a veteran, the turnout rate in 1813,Footnote ³⁶ and population density in 1813. For elevation we see essentially no difference. For maize suitability we see a small difference whereby potential agricultural output was about 6% higher for towns inside the tract, and this difference was statistically significant. For wheat suitability we see a very small difference (only about 1%) in favor of the towns inside the tract that is statistically significant at p = 0.09. Finally, for veterans we see no significant difference between towns inside and outside the tract.Footnote ³⁷ We find that turnout rates in 1813 were, on average, almost identical, and that the difference was not statistically significant. This result would seem to weigh against the possibility that our main estimates are biased upward due to sorting based on the initial proclivity to participate in democracy. The estimates show that although initial population density was, on average, slightly higher inside the Military Tract, this difference was not statistically significant.

Table 3. Descriptive Statistics of Control Variables

Note: Veteran fraction measures the fraction of households with at least one veteran.

In all the analyses to follow, we will report results based on two sets of estimating equations. The first is

(1) $$ {y}_i={\beta}_0+{\beta}_1{TotalSchoolFunds}_i+\boldsymbol{\gamma} {\boldsymbol{x}}_{\boldsymbol{i}}+{\alpha}_b+{\varepsilon}_i, $$

where $ i $ indexes towns and $ b $ indexes border segments; $ y $ is our various economic and political outcome variables; Total School Funds is our potentially endogenous measure of school funds per town; $ \boldsymbol{x} $ is a vector of geographic controls including latitude, longitude, wheat suitability, maize suitability, and elevation; $ {\alpha}_b $ are border segment fixed effects; $ \varepsilon $ is the error term; and $ {\beta}_0, $ $ {\beta}_1, $ and $ \boldsymbol{\gamma} $ are parameters to be estimated. We estimate this equation by ordinary least squares (OLS), with $ {\beta}_1 $ as our primary parameter of interest, and report robust standard errors.

The second equation is our geographic regression discontinuity specification:

(2) $$ {y}_i={\beta}_0+{\beta}_1{InTract}_i+\boldsymbol{\gamma} {\boldsymbol{x}}_{\boldsymbol{i}}+{\alpha}_b+{\varepsilon}_i, $$

which is analogous to the initial specification, but now $ {\beta}_1 $ measures the difference in outcomes between towns inside and outside of the Military Tract. As suggested before, presence inside the tract can be interpreted as an intention-to-treat estimate. There were exogenous reasons why some towns inside the Military Tract had more local school funds than others. This would have been linked to the value of their gospel and school lots as well as whether these funds were retained after towns subdivided. There would also have been endogenous reasons why some towns inside the tract had more funds. Those more disposed to aiding education could have done a better job of managing gospel and school lot funds. This second specification addresses these concerns by comparing outcomes between towns near the Military Tract border.

We further reestimate Equation 1 using the indicator variable InTract as an instrument for Total School Funds. This analysis provides a causal estimate of the effect of Total School Funds on our economic and political outcomes. The validity of this estimate requires first that being in the Military Tract increases Total School Funds. Previous discussion suggests the plausibility of this relationship, and we report an F statistic to assess the strength of the first stage in each table. Next, we need to assume that conditional on control variables, being in the Military Tract is as good as randomly assigned. This is what we have already argued is the case in presenting our geographic regression discontinuity design and estimating Equation 2—the towns inside and outside the tract are otherwise the same. Third, we need to assume that the instrument InTract satisfies the exclusion restriction in that its only effect on the outcome variables is through Total School Funds. We think the main two concerns here are the presence of veterans and spending on gospel as opposed to schooling. As we have already highlighted, veterans overwhelmingly sold their lots and, as of 1840, there were no differences in veteran presence inside and outside the Military Tract. Furthermore, as discussed above, our qualitative evidence suggests almost all of the Gospel and School Lot Funds were allocated to schools not churches.

Finally, it is important to clarify that our research design seeks to estimate the effect of additional schooling on various economic and political outcomes. The claim is that additional schooling from being inside the Military Tract put those towns on a different trajectory than towns outside the Military Tract. This includes both the effect of schooling on citizens raised in those towns and its effect on patterns of location among citizens and firms. Our analysis in this section suggests that there is no evidence that individuals sorted inside and outside of the Military Tract prior to towns setting up schools, but sorting remains a possibility after the common school era begins. We think that identifying the causal effect of education spending on the different trajectories of these towns is important because the effect of state policies more generally on economic and political outcomes is through their direct effects on its citizens and on patterns of entry and exit.