Welcome to Wilma A. Dunaway's  Slavery and Emancipation in the Mountain South: Evidence, Sources, and Methods

The Supplementary Electronic Archive for  Slavery in the American Mountain South  (Cambridge University Press, 2003) and 

The African-American Family in Slavery and Emancipation (Cambridge University Press, 2003)

Site Map

Methodology & Sources

by

Wilma A. Dunaway

This is a copyrighted document from the electronic archive for Wilma A. Dunaway,  Slavery and Emancipation in the Mountain South: Evidence, Sources, and Methods, Virginia Tech Library. 

How to Cite   Contact the Author

Links to Topical Sections

• Regional Definition

• View a Map of  the Counties of the Mountain South

• Slavery: Primary and Archival Sources

• Access the List of Appalachian Slave Narratives

• Estimating Appalachian Slave Trading

• White Appalachian Labor Mechanisms

• Frontier Land Ownership Patterns: Sources and Methodology

• Mid-nineteenth century Land Ownership Patterns: Sources and Methodology

• Wealth Disparity and Class Differentiation: Methodology and Sources

• Measuring Subsistence Production

• Surplus Agricultural Production, Exchange and External Trading: Sources and Methodology

• Nonagricultural Economic Sectors: Sources and Methodology

Regional Definition

Appalachia is neither a political structure nor a clear cultural entity. Thus, all existing boundary definitions for this vast area are, at best, arbitrary. Long-running debates have emerged from several political "rediscoveries" and geographical redefinitions of the region. One group of scholars has argued that Appalachians share a homogeneous culture, common socioeconomic problems and similar political leanings. Conflicting writers contend that "Appalachia" is a regional myth that has been imposed from the outside to serve the vested interests of actors other than the area's residents. ⁽¹⁾

This research does not seek to devise yet another set of artificial boundary lines. Since many analyses have utilized topographical obstructions and physical isolation to explain the social problems of the region, this study has sought to keep terrain differences in central focus. For this reason, the current research relies on Campbell's broadly-inclusive delineation of the Southern Highlands as 219 counties in Alabama, Georgia, Kentucky, Maryland, North Carolina, South Carolina, Tennessee, Virginia and West Virginia. Because lowland counties have soil and climate characteristics that are more typical of the Deep South than of Appalachia, this research limits its regional delineation to those counties characterized by mountainous, hill/plateau and ridge/valley terrain. Campbell included four lowland counties of Alabama and South Carolina in his definition which have been excluded from this study. ⁽²⁾

View a Map  of  the Counties of the Mountain South:

Click Back to Return to this document

Slavery: Primary and Archival Sources

Unpublished theses and dissertations and collections at several regional archives were utilized. Many manuscript family paper collections at regional archives and the Coleman Papers on Slavery at the University of Kentucky provided many useful leads. In addition, the Civil War veteran questionnaires and published plantation journals provided considerable information about Southern Appalachian slavery. However, personal narratives offer the best source of detail from the viewpoint of the affected slaves. Appalachian slave narratives were selected from two manuscript collections at Fisk University and the University of Kentucky, from several obscure published primary sources, and from the national collection that was compiled by WPA workers in the 1930s. ⁽³⁾

Despite the controversy about sampling and methodological biases in these narratives, I am convinced they are just as reliable as the manuscript family papers of plantation owners or published elite journals and letters. Despite the advancing age of the elderly respondents, these narratives offer critical details about the labor mechanisms of slaveholders, about slaves' relationships with landless whites, about slave trading practices, about the work duties of slave women and children, about the techniques used by slaves to generate cash, and about slave breeding and sexual exploitation. ⁽⁴⁾

The selected Appalachian narratives offer several advantages. House servants, males and extremely young slaves were over-represented in the larger national collection. Among the Appalachians, however, two-thirds worked full-time or intermittently in the fields. In addition, nearly 45% of the Appalachian narratives were from female respondents-- a much higher proportion than the national collection. The WPA narratives have been criticized because so many of the respondents were too young during slavery to have been assigned major work duties or to remember significant details. In contrast, more than 75% of the Appalachian respondents were teenagers (slaves began work at age 10) or older at emancipation.

Finally, two-thirds of the Appalachian narratives were collected from respondents who had been slaves in the mountainous and hilly areas. This group of narratives provides insight into slavery as it operated in those counties of the South where such research has been neglected. My sample under-represents the ridge-valley counties of Virginia, but these areas are represented in other regional source materials. Furthermore, I also utilized archival (especially at the University of Virginia) and published sources to permit terrain comparisons.

I have grounded this study in analysis of indigenous transcripts of nearly 300 slaves and more than 400 white Civil War veterans. I spent many months locating Appalachian slave narratives within the Federal Writers Project, at regional archives, and among published personal histories. Beginning with Rawick's nineteen published volumes of the WPA slave narratives, I scrutinized every page for county of origin, for interregional sales or relocations that shifted slaves into the mountains for part of their lives, or for occurrences during the Civil War that shifted slaves into or out of the Mountain South. After that process, I identified other archival and published accounts, finding several narratives in unusual locations, including archives at Fisk University and the University of Kentucky. In this way, I did not ignore the life histories of slaves who were born outside the Mountain South then migrated there or those who were removed from families but whose narratives were collected by WPA writers in other regions. Ultimately, I located nearly 300 relevant transcripts to aggregate the first comprehensive list of Mountain South slave narratives.

C. Vann Woodward argues that historians should not consider slave narratives any no more biased than any other manuscripts:

Appalachian slave narratives are not handicapped by the kinds of shortcomings that plague the national WPA collection. Large plantations, males, and house servants are over-represented among the entire universe of respondents. In addition, two-fifths of the ex-slaves had experienced less than ten years of enslavement. The most serious distortions derived from the class and racial biases of whites who conducted the vast majority of the interviews. The Southern Mountain narratives were collected over a vast land area in nine states. Thus, the significant distances between respondents offer opportunities for comparison and for testing the widespread transmission of African-American culture. By checking the slave narratives against Census manuscripts and slave schedules, I established that the vast majority of the Appalachian narratives were collected from individuals who had been enslaved on plantations that held fewer than twenty slaves. Most of the Appalachian respondents had been field hands, and very few were employed full-time as artisans or domestic servants. In terms of gender differentiation, the Appalachian sample is almost evenly divided. In contrast to the entire WPA collection, three-quarters of the Appalachian ex-slaves were older than ten when freed. Indeed, when emancipated, one-third of the Southern Mountain respondents were sixteen or older, and 12 percent were 25 or older. Thus, the vast majority of the Appalachian ex-slaves had endured fifteen years or more of enslavement, and they were old enough to form their own memories and to retain family oral histories. Perhaps the greatest strength of the Appalachian collection has to do with the ethnicity of interviewers. More than two-fifths of the Appalachian narratives were written by the ex-slaves themselves or collected by black field workers, including many Tennessee and Georgia interviews that were conducted under the auspices of Fisk University and the Atlanta Urban league.

In my content analysis of slave narratives, I replicated the methodology of the Without Consent of Contract project so that I would be able compare small plantations and the Mountain South with national averages, with large Lower South plantations, and with slavery systems in other parts of the world. To insure an even-handed approach, I triangulated findings from slave narratives and from slaveholder manuscripts with statistical analyses of Censuses and tax lists and with other historical accounts. I have come away from this effort with a deep respect for the quality and the reliability of the indigenous narratives that I utilized. When I tested ex-slave claims against public records, I have found them to be much more accurate than most of the slaveholder manuscripts that I scrutinized, and quite often much less ideologically blinded than many of the scholarly works I have consulted about the subject.

Facts do not speak for themselves in any study, no matter what a writer might claim. Therefore, I made the conscious intellectual decision to engage in "the making of history in the final instance" by respecting the indigenous knowledge of the ex-slaves whose transcripts I analyzed. That means that I did not dismiss as "exaggeration" every slave voice that disagreed with intellectual fad or convention. In most instances, I triangulated the indigenous view against public records and found the slave's history to be more reliable than some recent scholarly representations. In other instances, I perceived that Appalachian slaves are a people without written history and that it is important to document the oral myths in which their community building was grounded. Because Appalachian slave narratives present a view of enslavement that is unconventional, I recognized that they and I were engaging in "the production of alternative narratives." When contacted by a Fisk University researcher in 1937, one Chattanooga ex-slave comprehended that he possessed a knowledge about slavery that was different from the social constructions of the African-American interviewer. "I don't care about telling about it [slavery] sometime, " he commented cynically, "because there is always somebody on the outside that knows more about it than I do, and I was right in it." Clearly this poorly educated man understood that historical facts are not created equal and that knowledge construction is biased by differential control of the means of historical production. I set myself the difficult goal of avoiding the kind of intellectual elitism the ex-slave feared while at the same time trying to avoid the pitfall of exaggeration.

Access the List of Appalachian Slave Narratives

Click Back to Return to this document

Quantitative Analysis of Slavery

The samples of 13,279 households drawn from frontier-era tax lists offer significant detail about the characteristics of slaveholders. In addition, the sample of 3,447 farm households selected from the 1860 Census manuscripts is the basis for examination of the extent of slavery among mid-1800s households and of the traits of slave owners. Finally, the published Censuses provide county totals for slaves and slaveholders to permit extensive comparative analysis. In addition, it is possible to do  some quantitative analysis of the Appalachian slave narratives.

Estimating Appalachian Slave Trading

Discussion of Sources

The most reliable sources for assessing the extent of slave trading in the Appalachian region are: extant slave narratives, Census manuscript samples, published Census statistics, records of legislative debates, antebellum newspapers, and comparison with other studies of Upper South slave trading. Unfortunately, I could not locate a manuscript collection for an Appalachian slave trading firm. Even though slaveholder letters and farm accounts proved helpful for analyzing the mechanisms for arranging slave hire-outs, such records are of little utility in evaluating the statistical extent of slave trading. The brief and scattered references to slave sales are far too inadequate to derive any overall picture of trading practices in this large region; and the existing Appalachian slaveholder manuscript collections disproportionately represent the elites of the region. Moreover, these sale notations almost never provide details about how a slave was sold, except to indicate a first or last name. Thus, it is almost impossible to determine whether each sale was truly local or out-of-state. Appalachian slave narratives make clear that slaveholders thought of themselves as having sold a slave "locally" when the transaction occurred on the auction block. In almost every instance, however, the Appalachian slave was bought locally by an out-of-state buyer or by a neighbor who quickly resold the slave to a distant buyer.

Most importantly, sole reliance upon slaveholder papers to analyze the extent of slave trading is the political equivalent of analyzing the public announcements and official memoranda of the Nazis to determine the number of Jews they killed in their concentration camps. Just as the Nazis camouflaged their activities in their internal records, so did Appalachian slaveholders deny slave-selling and define themselves as "reluctant sellers" forced to trade by the pressure of debts or because of the behavior of the slave. Slaveholder letters reflect the paternalism and cultural mythology of the times, and their statements about "not selling their people" just do not match casual entries of slave trading in their farm accounts. Consequently, these sources are highly-biased with respect to the statement of family practices in regard to selling slaves. I quickly discovered that Appalachian espousals about not selling slaves or breaking up slave families did not coincide with the statistical trends or with the overwhelming evidence to the contrary in Appalachian slave narratives. In contrast, the claims of Appalachian slaves are strongly supported by Census statistics and by parallel studies about slave trading in the rest of the Upper South. Thus, I followed the approach of beginning with the analysis of statistical trends, followed by analysis of slave narratives, legislative debates, newspaper advertisements, and slaveholder papers. Only on the subject of slave hire-outs did the slaveholder papers provide sufficient reliable information to be considered a worthwhile auxiliary source.

Methods for Estimating Intrastate Sales

Intrastate slave trading was estimated by using aggregated county totals from the published 1860 Census of Population. In line with the only estimates available, I assumed that 4.04% of the total slave population was involved each year in intrastate sales. It is likely that Southern Appalachians sold slaves more often than this rate, but I had no reliable method for arriving at any closer estimate. ⁽⁵⁾ For those slaves sold, I assumed the same sex/age distribution that is summarized below for interstate slave sales. To estimate gross intrastate sales proceeds, I assumed that slaves were valued at three-quarters of the interstate sales prices summarized below. All prices were converted to 1860 dollars to insure comparability over time.

Methods for Estimating Slave Hire-outs

Slave hire-outs were estimated by using aggregated county totals from the published 1860 Census of Population. As reported in the Appalachian slave narratives, I assumed that 13.8% of all slaves between the ages of fifteen and fifty-nine were hired out annually. By making minor adjustments of existing studies, I made the age/sex distribution of hired slaves, summarized in Table A.6. ⁽⁶⁾ I assumed a sliding scale of annual contract rates in which males aged 20-29, the prime marketable cohort, rented for $142. ⁽⁷⁾ The other age-sex cohorts were estimated at proportions of this prime rate, according to the ratio of their sales value to that of prime males as these prices were actually quoted to an Appalachian slaveholder by a Richmond trading firm. ⁽⁸⁾

Methods for Estimating Interstate Sales

To estimate the extent of interstate sales, it is necessary to complete two preliminary steps. First, I estimated the slave population that would have resided in the region if no outward or inward migration had occurred between 1840 and 1860. By applying the national rate of increase to the 1840 Appalachian slave population, I estimated the number of slaves that would have been in the region if no interregional transfer had occurred. Cross-comparisons were made with the 1840 and 1850 age cohorts to insure fertility rate accuracy. The total number of out-of-region slave migrations, then, equals the estimated population with natural increase minus the actual 1860 population. Second, I arrived at an estimate of what proportion of these interregional transfers occurred because slaves migrated out of Appalachia with their masters. The total number of interstate sales is equivalent to the total number of out-of-region slave transfers minus the number of migrations with masters.

In the final step, I assumed the age/sex distribution of Appalachian slaves sold through interstate transactions to have been like that obtained in other extensive demographic analyses of Upper South trading. To estimate the value of slave sales by age and sex, I drew on an archival letter to an Appalachian slaveholder in which a Richmond trader enumerates the prices of slaves by age and sex. Using this document, I was able to calculate the value of each age/sex cohort as a proportion of the value of prime field hands; then I applied those ratios to an existing study which documented the variations in the prices of prime field hands over the 1840 to 1860 period. The price estimates are summarized in Table A.8. All prices were converted to 1860 dollars to insure comparability over time. ⁽⁹⁾

To further illustrate the loss of slaves from the region between 1840 and 1860, I completed cohort analysis of the aggregated county slave populations in the 1840 and 1860 Censuses of Population. Adjustments were made for deaths, runaways and manumissions to arrive at an estimate of the 1840 cohort that should have remained in the region in 1860. Statistics for the 1840 cohort aged 0 to 9 were compared with the 1860 cohort aged 20 to 29. Statistics for the 1840 cohort aged 10 to 59 were compared with the 1860 cohort aged 30 to 79. ⁽¹⁰⁾

White Appalachian Labor Mechanisms

Antebellum Tenancy

Primary and archival sources. Even though regional archives have no collections of tenant manuscripts, I did locate several family papers and farm journals that include references to tenancy and share farming. I also found a few antebellum tenancy contracts and plat maps. Antebellum store accounts also provided useful details about regional debt practices. Few theses or dissertations address Appalachian tenancy directly; however, I did find scattered references in several works. The frontier tax lists offer detailed clues about landless farmers who paid the taxes on the land. In addition, discussion of tenancy occurs in the slave narratives and in the Civil War veteran questionnaires. Published travel diaries, family papers and farm journals also offer colorful insights about antebellum tenancy.

Analysis of five farm household types. The most significant source of information about antebellum tenancy is the sample of 3,447 farms drawn from the 1860 Census manuscripts. The enumerator manuscripts provide considerable descriptive detail about these households, including their accumulated wealth, their non-farm occupations, their crop production, and their ownership of farm implements and work stock. Using these details and information from other sources, it is possible to reconstruct the characteristics of the different household structures among antebellum landless farmers: cash renters, farm managers, tenant farmers, sharecroppers, and nonagricultural share contractors.

Use of Farm Labor Time

In order to compare the time allocation by farmers to subsistence and surplus production, I utilized the following estimates of labor hours required to produce crops in the antebellum era. ⁽¹¹⁾

Crop Cultivated Labor Hours Needed

Corn per bushel 0.969

Wheat per bushel 3.24

Barley per bushel 2.12

Potatoes per bushel 0.50

Sweet potatoes per bushel 3.02

Oats per bushel 1.67

Rye per bushel 2.52

Tobacco per pound 0.50

Cotton per bale 271.00

Antebellum Wage Earners and Artisans

Analysis of farm laborer households. Paid farm laborers constitute another antebellum labor mechanism in Southern Appalachian agriculture. When using details from the 1860 Census of Population manuscripts, it is possible to identify a sixth type of agricultural household that does not appear in the Census count of farms: those agricultural laborers who do not reside on farms. In addition, the 1860 Census of Population manuscripts enumerate some paid laborers with farm household members. These non-family wage-earners are identified in two ways. First, their surnames are different from that of the head of household, and no kinship tie is indicated. Second, their occupation is separately listed as "farm laborer" or "hired hand." ⁽¹²⁾

Other wage earners and artisans. The sample of households from the 1860 Census of Population provides extensive details about nonagricultural wage earners and artisans. In addition, much information was gleaned from archival store accounts and family business papers about arrangements with wage laborers and artisans. In addition, the Civil War veteran questionnaires and several archival collections provide details about antebellum wage earners.

Frontier Land Ownership Patterns: Sources and Methodology

The State of Public Land Records in Southern Appalachia

Complex and tedious methodology is required to research Southern Appalachian land ownership patterns over the period from 1790 to 1860 because there is no single source of land records. Deeds comprise the type of public record that is most consistently available. For a project of this scope, however, deeds are of little utility. Because they are recorded chronologically by parcel and because holdings are not aggregated for each owner, one must scrutinize the deeds over many years to ascertain, with certainty, whether any one owner held land. Most other early Southern Appalachian public records have been lost or destroyed. 1790 to 1800 Census records have been lost entirely for Virginia and Tennessee and partially for other counties in Southern Appalachia. ⁽¹³⁾


Consequently, it has been necessary to adapt data collection methods that permit the use of numerous state, county and national sources from multiple archives. Frontier county tax list manuscripts for Georgia, Kentucky, North Carolina, Tennessee, Virginia and West Virginia were rented on microfilm from the Family History Center, Church of Jesus Christ of the Latter Day Saints (Salt Lake City). In addition, frontier county tax list manuscripts were secured from the State Archives of Virginia and Maryland. Frontier land ownership patterns were also researched with the aid of Library of Congress maps. County tax lists offer the best archival source for researching frontier land distribution in Southern Appalachia. The lists were roughly alphabetized, and information was collected and recorded for households. In addition, many of these lists are accessible on microfilm or in published form. However, it is not possible to find in existence a complete set of records for any single year for all counties of Southern Appalachia. Moreover, each state enacted its own tax laws and collected tax information uniquely. ⁽¹⁴⁾

It is necessary, therefore, to devise distinct sampling techniques for each state subgrouping of records. Two separate samples are required for some states because they maintained two distinct sets of records. When utilizing these early tax lists, one must be careful to distinguish residents from absentee land owners. For this reason, it is essential either to cross-match the tax-list names with Census lists or to devise an alternate strategy for identifying residents. For this research, more than 13,000 frontier households were scrutinized in early nineteenth-century tax records. Table A.1 provides a summary of the frontier land record methodology and the tax-list sampling completed for frontier households. ⁽¹⁵⁾

Frontier Land Ownership Patterns in Eastern Kentucky

Kentucky resident ownership. Resident land ownership patterns were analyzed by using the manuscript tax lists for Floyd, Knox, Madison, Pulaski and Wayne counties, encompassing nearly three-fourths of the land area of Appalachian counties at that time. Beginning with a randomly-selected starting point, a systematic sample of 1,047 households was drawn by selecting every third taxpayer. Since the tax records were organized by districts, a single name might occur more than once in the county. Consequently, a search was made for each selected name in every district of the county to insure that landholdings were aggregated for each owner. At the 95% confidence level, this sample is large enough to obtain an estimate of a percentage that is within plus or minus less than three percentage points of the actual population value. ⁽¹⁶⁾

Kentucky absentee ownership. Beginning in 1800, the State Land Office maintained separate records of absentee taxpayers; thus, county tax lists did not include the names of nonresident land owners. To avoid this complication, the tax lists for Floyd (1793) and Madison (1792) counties were utilized to measure the extent of absentee land ownership. Information was collected about every land owner in these two tax lists. Since the tax records were organized by districts, a single name might occur more than once in the county. Consequently, a search was made for each selected name in every district of the county to insure that landholdings were aggregated for each owner. Then, the names of land owners were checked against the 1790 Census to identify residents. Absentee owners were assumed to be those individuals (a) who appear on the tax list, but not on the Census list and (b) who paid tax on no property other than land. At the 95% confidence level, this sample is large enough to obtain an estimate of a percentage that is within less than three percentage points of the actual population value. ⁽¹⁷⁾

Frontier Land Ownership Patterns in Western Maryland

To assess land ownership patterns in western Maryland, every household included in the 1804 tax list of Washington County was scrutinized. Since the tax records were organized by districts, a single name might occur more than once in the county. Consequently, a search was made for each selected name in every district of the county to insure that landholdings were aggregated for each owner. To determine residency of land owners, two methods were utilized. First, each name was cross-checked against the 1800 Census. Second, names not appearing on the Census list were scrutinized for ownership of other categories of wealth. Land owners who also paid tax on slaves, livestock, jewelry, household goods, or wagons were assumed to be residents. Essentially, absentees were assumed to be those taxpayers whose names did not appear on the Census list and who owned no property other than land. ⁽¹⁸⁾

Frontier Land Ownership Patterns in Western North Carolina

For western North Carolina, tax lists for Surry (1800), Burke (1805) and Ashe (1815) were the only extant usable records. Beginning with a random-number starting point, a systematic sample of 1,068 households was drawn by selecting every third taxpayer. Since the tax records were organized by districts, a single name might occur more than once in the county. Consequently, a search was made for each selected name in every district of the county to insure that landholdings were aggregated for each owner.

Absentee landholders were identified by following two methods. First, residents were legally required to pay poll taxes. Since women were exempted, female-headed households were counted as residents even when they paid no poll tax. Second, households not paying poll taxes were scrutinized for ownership of wealth other than land. Absentee owners were assumed to be those male taxpayers who paid no poll tax and who held no property other than land. At the 95% confidence level, this sample is large enough to obtain an estimate of a percentage that is within less than three percentage points of the actual population value. ⁽¹⁹⁾

Frontier Ownership Patterns in Appalachian South Carolina

To assess land ownership patterns in Appalachian South Carolina, all 1,242 households included in the 1790 Pendleton District Census were crossmatched with 1789-1792 deeds. Both lists were computerized, then sorted alphabetically to eliminate name duplications and to aggregate land and slave holding by each household. To evaluate the extent of absentee landholding, all deeds were crossmatched with the Census list to determine the residency of buyers. ⁽²⁰⁾

Frontier Land Ownership Patterns in Appalachian Tennessee

Because so many early Tennessee records have been lost or destroyed, it is not possible to secure tax lists for any single year. Instead, every available tax list in the period of 1790-1810 was identified and utilized. Eleven county tax lists, accounting for almost the entire Appalachian land area, were utilized from published sources. Beginning with a random number starting point, a systematic sample of 1,022 households was drawn by selecting every seventh household. Since the tax records were organized by districts, a single name might occur more than once in the county. Consequently, a search was made for each selected name in every district of the county to insure that landholdings were aggregated for each owner.

Absentee owners were identified by scrutinizing two elements of information on the tax lists. Absentees were assumed to be those land owners (a) who paid no poll tax and (b) who owned no property other than land. At the 95% confidence level, this sample is large enough to obtain an estimate of a percentage that is within less than three percentage points of the actual population value. ⁽²¹⁾

Frontier Land Ownership Patterns in Appalachian Virginia

Virginia resident ownership. Virginia maintained two separate tax lists during this era: personal tax and land books. Because of the broad scope of Virginia's tax laws, every resident household paid some form of tax. Since early Virginia Census records were destroyed, the Virginia State Archives recognizes these early personal tax lists as accurate population lists. Tax lists were utilized for every county for which both a personal tax list and a land book could be found for the same year. To assess the extent of land ownership by residents, two steps were completed. First, beginning with a random-number starting point, a systematic sample of 1,144 households was drawn by selecting every fourteenth taxpayer from the personal tax lists (published). Since the tax records were organized by districts, a single name might occur more than once in the county. Consequently, a search was made for each selected name in every district of the county to insure that landholdings were aggregated for each owner. Second, to check for land ownership by each taxpayer, the sample of names was cross-matched with the land books. At the 95% confidence level, this sample is large enough to obtain an estimate of a percentage that is within less than three percentage points of the actual population value. ⁽²²⁾

Virginia absentee ownership. A separate sample was required to assess the extent of absentee ownership. First, a systematic sample of 1,360 taxpayers was drawn, by selecting every eighth household from the land books. Since the tax records were organized by districts, a single name might occur more than once in the county. Consequently, a search was made for each selected name in every district of the county to insure that landholdings were aggregated for each owner. Then, the names were cross-matched with the personal tax lists to identify resident households. Absentees were assumed to be those land owners whose names did not appear on the personal tax list. At the 95% confidence level, this sample is large enough to obtain an estimate of a percentage that is within less than three percentage points of the actual population value.

Frontier Land Ownership Patterns in West Virginia

West Virginia resident ownership. Even though West Virginia was not yet formed as a state, several of its counties are identifiable in Virginia's early records. For 1800, personal tax lists and land books are extant for twelve of West Virginia's thirteen frontier counties. To assess the extent of land ownership by resident households, two steps were completed. First, a systematic sample of 1,052 taxpayers was drawn by selecting every ninth entry from the personal tax lists (published). Since the tax records were organized by districts, a single name might occur more than once in the county. Consequently, a search was made for each selected name in every district of the county to insure that landholdings were aggregated for each owner. Second, the sampled names were cross-matched with the land books (microfilm). At the 95% confidence level, this sample is large enough to obtain an estimate of a percentage that is within less than three percentage points of the actual population value. ⁽²³⁾

West Virginia absentee ownership. A separate sample is required to assess the extent of absentee ownership. First, a systematic sample of 1,675 households was drawn by selecting every fifth entry from the land books. Since the tax records were organized by districts, a single name might occur more than once in the county. Consequently, a search was made for each selected name in every district of the county to insure that landholdings were aggregated for each owner. Then, the sample of land owners was cross-matched with the personal tax lists to identify resident households. Absentees were assumed to be those land owners whose names did not appear on the personal tax lists. At the 95% confidence level, this sample is large enough to obtain an estimate of a percentage that is within less than three percentage points of the actual population value.

Frontier Land Ownership Patterns in Alabama and Georgia

Northern Alabama and northern Georgia remained Indian territory until after 1820. For Alabama, there were no public records before 1821; even those have been lost. For Appalachian Georgia, the only accessible records were 1818-1822 deeds, Cherokee Land Lottery Grant lists for 1821-1829, and the 1820 Census for Habersham County. To assess land ownership by residents, every name in the 1820 Census was cross-matched with the deeds. To analyze absentee ownership, every deed was cross-matched with the 1820 Census to check the residency of every buyer or seller of land. Cherokee Land Lottery grantee names were cross-matched with the 1820 and 1830 Censuses to determine residency. ⁽²⁴⁾

Difficulties in Using Early Records

Locating records. Using early nineteenth-century records poses several methodological dilemmas. First, many sources have been lost, making it impossible to rely on a single type of record or upon sources of the same year. When the target area is so large, one must be persistent in the aggregation of many different types of published and archival materials. Determining what public records still exist and where they may be utilized is a time-consuming process. For this research, locating records entailed borrowing obscure published lists from other libraries, travel to distant archives, and the rental of many reels of microfilm.

Matching lists. When using early nineteenth-century records for Southern Appalachia, one quickly notices that the names change radically from one year to the next. Over half of the households from one tax list will not appear on the next year's tax list. Contemporary research demonstrates that frontier Appalachian populations were highly mobile, particularly the landless households. For this reason, it is methodologically dangerous to rely on a sample derived by cross-matching a land or tax list of one year with a Census of a different year, even the year before or after. This researcher meticulously selected sources that permit-ted chronological matching of land or tax lists with Censuses. ⁽²⁵⁾

Cross-matching names. When cross-matching early nineteenth-century lists, one is handicapped by the use of varied phonetic spellings of the surnames. It is necessary, then, to match names by how they sound. In questionable instances, this researcher was greatly assisted by genealo-gical lists of name spellings. Cross-matching is also slowed by the lack of alphabetical order in early Census lists. Even though the land books and tax lists were nearly alphabetical, the information was collected by districts. As a result, a taxpayer's name could appear more than once; and one could not be sure that holdings were aggregated at any single entry. This researcher utilized two strategies to overcome these dilemmas. Some lists were computerized, then sorted alphabetically to eliminate duplications and aggregate holdings. Secondly, when using the nearly-alphabetical tax lists, one can aggregate holdings by checking for sample names in each of the two or three districts comprising the county. ⁽²⁶⁾

Determining residency of taxpayers. Several problems emerge when one attempts to determine taxpayers' residency by cross-matching lists. First, the early Censuses provide only the names of heads of household. However, the tax lists might reflect land for other family members. In the case of deceased male heads, the land might be listed in the name of the wife or several children. If land were deeded by a father to an unmarried son, the child's name would not appear on the Census list. Therefore, when checking residency, this researcher assumed that land owners were residents when they had a surname that recurred in the county Census list. This is somewhat problematic since it denies the possibility of migration by segments of the same family or the inheritance of estates by nonresident children. Given the nature of this project's exploration of absentee holdings, however, this researcher felt it essential to follow a methodology that might tend to over-count the residents, rather than the nonresidents.

Landless kin of owners. Nineteenth-century county tax lists collected information about the populace by situating the description of taxpayers into specific dwellings and districts. Enumerators proceeded geographically from one dwelling to the next, listing each household in the order they were encountered. When studying the extent of landless-ness, however, these lists pose a peculiar difficulty. Because of the legal requirement that all males older than twenty-one pay a poll tax, the tax assessors often enumerated unmarried sons or other males as a separate entry from the household in which they actually resided.

To avoid an undercount of owners, I scrutinized carefully each selected landless household. The enumeration order aids the identification of landless kin of owners since these persons are typically listed immediately before or after the land owner. When using published lists, I had the benefit of genealogical footnotes to link such persons to the correct households. When such landless kin were selected in the sampling process, I aggregated these persons with the rest of the appropriate owner household in which they actually dwelled. ⁽²⁷⁾

The enumeration order is studied carefully to identify extended families. Landless kin whose names appear at a great distance from the owner (often in a different district) in the enumeration order are counted as tenants. In those instances, the enumeration order seems to indicate they were residing in a dwelling located on non-family land and at a great geographical distance from the relatives. In short, these persons were providing their labor to farm on non-family land, and they were operating as households independent from their kin.

Later processing errors. Many early county records were reproduced manually by WPA workers during the 1930s, and the manuscripts were subsequently lost or destroyed. The most frequent error made during this later reproduction is the duplication of names. When using microfilm, one is also hampered by such duplication. Without any labelling to warn users, repeated camera shots of pages frequently occur on microfilm. To avoid sampling errors, this researcher followed the procedure of checking every reel of microfilm for such duplications, before collecting any data.

Mid-nineteenth Century Land Ownership Patterns: Sources and Methodology

Utilizing the 1860 Census Manuscripts

Because they provide extensive detail about land ownership, wealth accumulation, occupations, and agricultural production, the 1860 Census enumerator manuscripts are the best primary source for antebellum information on all Southern Appalachian counties. I drew two samples from the Census enumerator manuscripts.

In order to analyze landholding patterns among nonfarm households, I drew the first systematic sample of 3,056 households from Schedule I (population), by selecting every 100th entry. In order to analyze landholding patterns among farm households, I drew a second systematic sample of 3,447 farms from Schedule IV (Agriculture). To insure that the sample included enough farms to permit subregional comparisons, I selected 333 farms for Alabama (by drawing every 25th farm), 301 farms for Georgia (by drawing every 17th farm), 344 farms for Kentucky (by drawing every 7th mountains farm and every 23rd hill-plateau farm), 302 farms for Maryland (by drawing every 12th farm), 302 farms for North Carolina (by drawing every 21st farm), 325 farms for South Carolina (by drawing every 4th farm), 514 farms for Tennessee (by drawing every 28th farm), 517 farms for Virginia (by drawing every 25th farm), and 509 farms for West Virginia (by drawing every 24th farm). To calculate cumulative farm statistics for state or terrain subregions or for the region as a whole, I weighted each geographical category according to the actual proportion of total farms (aggregated from published county census data) located in each subregion. These 1860 samples represent nearly 1% of all the region's households and nearly 3% of all Southern Appalachian farms. At the 95% confidence level, these samples are large enough to obtain estimates of percentages that are within less than three percentage points of the actual population value. ⁽²⁸⁾

Methodology for Identifying Landless Farmers

Because the 1860 Census did not ask heads of household if they owned land, farm owners are not directly distinguishable from landless farmers. Consequently, it is necessary to determine land ownership by comparing several elements of information collected for each household on Schedules I and IV. To determine whether a particular farm operator owned land, it is necessary to cross-match farm sample names from Schedule IV with household names in Schedule I. Schedule IV contains all farm operators, reporting acres of improved land, acres of unimproved land, cash value of farm, value of farming implements, number and value of livestock, and quantity of agricultural commodities produced. Schedule I lists each household, providing details about value of real estate owned.

Several conventional techniques utilized by 1860 Census enumerators provide us with clues about land ownership. First, since "value of real estate" was reported on Schedule I, any farmer listed on that schedule without real property (i.e., "value of real estate" left blank) was landless. Enumerators frequently labelled landless farmers, using terms like "tenant," "renter," "cropper," or "farm laborer." In several Southern Appalachian states, enumerators omitted acreage and farm value (while reporting value of produced commodities) for tenants. ⁽²⁹⁾

The 1860 Census collected information about the populace by situating the description of residents into specific dwellings and districts. Enumerators proceeded geographically from one structure to the next, thereby treating "households" and residents of physical "dwellings" as synonymous entities. Because of this procedure, an extended family residing in separate dwellings upon the same farm were enumerated as distinct, unrelated households. The value of the farm and of real estate were designated for one dwelling while the other households were identified as owning no real estate.

As a result, the children, parents or other relatives who supplied labor on a family farm appear in the Census as landless. To avoid an undercount of owners, I scrutinized carefully each selected landless household. The enumeration order facilitates the identification of landless kin of owners since these households are typically enumerated immediately before or after the owners. When a selected landless household appeared to be part of an extended family residing in separate dwellings on a family farm, I aggregated characteristics for all pertinent households, treating then as a single "owner" household.

The enumeration order is studied carefully to identify extended families. Landless kin whose names appear at a great distance from the owner (often in a different district) in the enumeration order are counted as tenants. In those instances, the enumeration order seems to indicate they were residing in a dwelling located on non-family land and at a great geographical distance from the relatives. In short, these persons were providing their labor to farm on non-family land, and they were operating as households independent from their kin. ⁽³⁰⁾

Landholding by the Agricultural Elite

In their analyses of Southern Appalachia, scholars have ignored close analysis of the region's agricultural elite-- considering them so small in number they must have been politically and economically inconsequential. The 1860 published Census of Agriculture reports very few owners of 500 or more acres of farm land in the region. Believing this statistical count, I expected to find none of these largest holders among the farms selected from Schedule IV. Therefore, I conducted a separate sampling of the selected counties to collect information about the region's largest farmers. Using the 1860 Schedule IV manuscripts, I scrutinized every fifth entry to identify holders of 500 or more acres. Through this procedure, a sample of 2,505 of the largest farmers was drawn.

As is evident from Table A.3, the published 1860 Census reports a serious undercount of the region's largest farmers. Strangely, the published count of total number of farms corresponds closely with the Census manuscripts. However, there are very questionable flaws in the published Census distribution of farms, by size of landholding. The number of small Appalachian farms (less than 100 acres) is greatly over-counted while the number of large farms is under-reported. It is this invalid and unreliable statistical data to which scholars point when they depict Southern Appalachia as a region of smallholders! These published miscounts have led researchers to argue mistakenly that two-thirds of the region's antebellum farms were smaller than one hundred acres.

Wealth Disparity and Class Differentiation: Methodology and Sources

Frontier Wealth Disparity

It is possible to estimate frontier wealthholding patterns by drawing on part of the tax lists utilized for land analysis. This approach aggregates data for wealth in land, personal estate and slaves-- permitting comparisons with the 1860 Census sample which reported wealth in these same categories. The wealth of 10,264 frontier households was scrutinized for seven of the region's geographical zones.

No records are available for the Appalachian counties of Alabama or Georgia prior to 1830. For Maryland, analysis was completed for all 115 households included in the Washington County (1790-1800) inventory of estates. Use of this list necessitated the conversion of pounds to dollars, one pound being valued at $4.44. ⁽³¹⁾

Two early North Carolina tax lists were utilized to analyze frontier wealthholding patterns. These two lists were computerized, then sorted alphabetically to aggregate holdings for each owner. All 408 households were scrutinized in the 1777 Surry County tax list, which reported wealth. In addition, 573 households were analyzed in the 1815 Ashe County tax list, which reported the value of land held by each taxpayer. Slaves were estimated to be worth $200 each, following the methodology of Soltow. Total wealth for each taxpayer is the sum of land value (from the tax list) and estimated value of slaves owned. ⁽³²⁾

Pendleton District deeds (1789-1792) specify the value of every parcel of land transferred. To analyze wealthholding patterns, two steps were followed. From the deeds, the value of land was aggregated for each owner. The value of slaves (number listed in the 1790 Census) was estimated at $200 each. For each household, the total estimated wealth is the sum of the value of land and slaves. Use of these deeds necessitated the conversion of pounds to dollars, one pound being valued at $4.44. ⁽³³⁾

Frontier Virginia and West Virginia tax lists specify the value of land and town lots held by each household. Total wealth is the sum of the reported land values and the estimated value of enumerated slaves (at $200 each).

No land values or wealth are reported in the early Tennessee tax lists. Using the sample of 1,022 households from tax lists, wealth estimates were calculated in this fashion. Averages from the Ashe, North Carolina (1815) tax list were utilized to estimate the value of acres in the mountainous and hill-plateau counties. Averages from the Virginia (1800) tax lists were utilized to estimate the value of acres in the ridge valley counties. Slaves were valued at $200 each; and town lots were estimated at $47.40 each (the 1800 Virginia average). Total wealth for each household is the sum of the estimated values of land or town lots and slaves.

No land values or wealth are reported in the frontier Kentucky tax lists. For this reason, Ashe, North Carolina (1815) acreage values were utilized to estimate the land values of the two samples of 1,536 Kentucky households. Slaves were estimated at $200 each; and town lots were values at $47.40 each (the 1800 Virginia average). For each household, total estimated wealth is the sum of the values of land and slaves. ⁽³⁴⁾

Primary and Archival Sources

Between 1915-1922, two Tennessee state archivists devised a project to collect information from the state's Civil War veterans. 1,650 veterans responded to questionnaires about their antebellum lifestyles, their war experiences, and their post-war occupations. Except for one recent study, these questionnaires have been virtually neglected by scholars. Of the 1,650 veteran respondents, nearly one-third were Appalachians: 384 from Tennessee, 37 from Virginia, 24 from North Carolina, 19 from Georgia, nine from Alabama, and one from South Carolina. The mean 1860 age of the Appalachian veterans was 23.4 years, and two-thirds of them were nonslaveholders. I found these materials rich with insights about antebellum class perceptions, wealth disparity, land ownership, slaveholding and labor management techniques. ⁽³⁵⁾

A few other primary sources proved useful in the analysis of wealth disparity. The slave narratives provide comparative descriptions of poor whites and slaveholders. Diaries, journals, travel accounts, and other family manuscripts also provided colorful examples of antebellum class distinctions. ⁽³⁶⁾

Household Sample from 1860 Census Manuscripts

In order to analyze wealth-holding patterns by all Southern Appalachian households, I drew a systematic sample of 3,056 households from Schedule I of the 1860 Census enumerator manuscripts. Beginning with a different random-number starting point for each state subregion, I selected every 100th entry-- recording information about land ownership, occupation, value of real estate owned, and value of personal estate. For each household, total wealth is the sum of the values of real estate and personal estate. Table A.4 demonstrates that this sample is representative of the distributions of population, by states and by terrain type. At the 95% confidence level, this sample is large enough to obtain an estimate of a percentage that is within less than three percentage points of the actual population value. ⁽³⁷⁾

Measuring Subsistence Production

Identifying subsistence producers

Subsistence producers are defined as those farmers who:

(1) have limited market dependency. Subsistence households consume 80% or more of their food crops and livestock. In addition, they produce no major amounts of marketable staple crops (tobacco or cotton). ⁽³⁸⁾

(2) control the means of production. Subsistence farmers own no more than 99 acres. They are not landless, but they do not hold large surpluses of land. ⁽³⁹⁾

(3) are neither proletarians nor petty capitalists. Subsistence farmers report no external wage-earning, artisan or business source of nonagricultural income.

(4) do not utilize paid or coerced laborers. Subsistence farmers utilize no labor mechanism other than family members.

In short, a subsistence farmer is overwhelmingly dependent upon the land for survival and has few sources of money.

Measuring Antebellum Subsistence Production

Overview. While drawing the sample of 3,447 households from the 1860 Census of Agriculture manuscripts (Schedule IV), data was collected about the crop and livestock production of each farm. The quantitative methodology for distinguishing subsistence farms from surplus producers involves five steps.

(1) Land owners were distinguished from landless farmers.

(2) All crop and livestock production was converted into a single standardized measure (corn equivalencies).

(3) Subsistence consumption was estimated for each household.

(4) A subsistence score was calculated for each farm household. This score is a calculation of the proportion of total food crop production that was consumed for the household's subsistence.

(5) Determination was made whether each farm was a subsistence or surplus producer, based on the proportion of the total production remaining after subsistence requirements were met. I assumed that a subsistence producer consumed at least 80% of the total food crop and livestock production.

Estimating crop production. In order to determine what proportion of the farm production is consumed, it is first necessary to find a way to standardize the varied crops and livestock into a single measure. Earlier studies have devised methods for converting food crops and meat sources into their equivalent nutritional value in corn. Major food crops and livestock were converted into corn equivalencies, using these ratios: ⁽⁴⁰⁾

Food Source Equivalency in Corn

1 bushel wheat 1.30 bushel

1 bushel potatoes or yams 0.25 bushel

1 pound rice 0.01 bushel

1 beef 2.25 bushels

1 hog 5.00 bushels

1 sheep 0.50 bushel

1 ox 14.40 bushels

Overall, Southern farmers produced insignificant amounts of peas, beans, oats, rye, barley, buckwheat or hay. In contrast, many Southern Appalachian counties cultivated high levels of these crops to utilize as feed supplements for livestock. Since conventional methodology ignores these crops as nutritional resources for livestock, I converted these crops using these ratios from 1910. ⁽⁴¹⁾

Feed Source Equivalency in Corn

1 bushel buckwheat 0.90 bushel

1 bushel barley 0.90 bushel

1 Bushel rye 0.85 bushel

1 bushel oats 0.90 bushel

1 bushel peas or beans 1.20 bushels

1 ton hay 48.80 bushels

Estimating consumption. Earlier studies have provided extensive study data on human and animal consumption levels in the antebellum South. Based on these earlier sources, this research allocated three-and-one-half pounds of meat and one-and-one-quarter pecks of grain to each adult weekly, which converts into 24.8 annual corn equivalencies in grains, potatoes, peas or beans, and meat for each adult (aged 15 years or older). Half that amount was allowed to children younger than fifteen. I assumed that Appalachian farmers relied primarily on other meat sources so they could produce beef cattle for export. In addition to meat supplied from hogs, oxen were butchered at about eight years of age; thus, about one of every six oxen were utilized for food annually. Also about two of every five sheep were slaughtered each year for food. Thus, when I calculated the total corn equivalencies available for the food supply, I included one-sixth of the total number of oxen and two-fifths of the total number of sheep. In addition to food requirements, I allocated three gallons of whiskey annually per adult, allowing 1.5 bushels to produce each gallon. ⁽⁴²⁾

In addition to human consumption, one must take into account feeds for livestock. Antebellum southern livestock practices were characterized by neglect and near-subsistence feeding levels. Except during fattening, hogs, cattle, swine, and sheep were left to forage for their feed during most months of the year. During the winter months, they were supplemented with hay, coarse fodder, peas, beans, and pumpkins. Moreover, only 69.5% of all livestock were ever grain-fed; the rest consumed forage only. Livestock were estimated to have consumed the following annual levels of corn. ⁽⁴³⁾

Type of Farm Livestock Corn Equivalencies Yearly

cattle 2.25 bushels

hogs 5.00 bushels

horses 21.60 bushels

mules or oxen 14.40 bushels

sheep 0.50 bushel

In order to maintain conceptual precision, this research designated as subsistence requirements the grain feeds necessary to sustain consumed livestock and one or two work stock (horses, mules or oxen). Feed requirements above subsistence levels were considered crop reinvestments utilized to generate livestock surpluses.

Finally, additional amounts would need to be reserved for other exigencies. To cover the cost of mill or distillery tolls, I added 10% to the amounts of grain required for household subsistence. In addition, the subsistent household would need to reserve seed for the next year's essential food crops and to reserve part of the livestock for herd replacement. The following ratios were utilized to calculate the annual seed reservations required to reproduce the subsistence needs for the next year: ⁽⁴⁴⁾

Crop % of Crop Reserved as Seed

corn 5%

oats 10%

potatoes/yams 10%

rye 11%

wheat 13%

peas/beans 5%

To allow for the next year's production, one of every seven surplus swine was retained for herd replacement. In addition, 28.6% of the surplus cattle were retained for herd replacement. Since I assumed that sheep and oxen were consumed at low levels and were not exported, it was not necessary to allow for herd replacements. Even though existing studies ignore crop losses, I educted 15% of the corn and wheat production as having been lost due to spoilage or waste. ⁽⁴⁵⁾

Surplus Agricultural Production, Exchange and External Trading: Sources and Methodology

Measuring Farm Household Surpluses

Identifying the surplus producer. This research examines the surplus production of exportable grains, livestock, tobacco, cotton, and slaves for each farm household selected from the 1860 Census manuscripts. A surplus producer is a farm owner or cash renter that:

(1) consumes less than 80% of total crop production for household survival;

(2) produces more than minimal levels of staple crops (i.e., more than 20 pounds of tobacco or more than one-half bale of cotton);

(3) holds land surpluses in excess of subsistence acreage needs;

(4) makes use of paid laborers or tenants or owns slaves.

Surpluses used to generate profits. In order to make the concepts clear, it is helpful to state the distinction a little differently. Total agricultural production was utilized for three distinct functions: (a) to provide household subsistence needs, (b) to reinvest in the production of surpluses, and (c) to generate profits through exchange. Thus, real surplus production consists of the sum total of all commodities produced or accumulated by the household-- in excess of the needs for survival of the kinship unit.

The corn equivalencies used by a farm to generate surpluses are not vital to its household subsistence. A farm geared to subsistence production relies on family labor and does not have the means to afford external labor sources. Therefore, the food consumed by the non-family laborers who produce surpluses is not part of subsistence (i.e., the food consumption of slaves, tenants, or paid laborers). Similarly, the feeds for surplus livestock and seed reservations (stored to produce next year's surplus crops) are utilized in the production of marketable commodities-- not to sustain the household.

In simplest terms, then, total surplus production of a farm is the sum of:

(1) corn equivalencies reinvested to generate surpluses (consumption by paid laborers, slaves, tenants and surplus livestock and surplus seed reservations)

(2) excess corn, wheat, beef and pork available for marketing;

(3) staple crops (tobacco or cotton);

(4) slave laborers (who could be sold or hired out);

(5) land in excess of subsistence acreage requirements (that could be sold, rented or used for nonagricultural purposes).

Measuring External Trading by Counties

Using the antebellum published Censuses of Agriculture and Manufacturing, I calculated the surpluses of Southern Appalachian counties in seven steps. First, the county's food crops and livestock were converted into corn equivalencies. Second, survival consumption was deducted for all county residents (including slaves) and for the livestock. In order to estimate total livestock production and food crop consumption, it is necessary to take into account each county's nonagricultural livestock. Cattle, sheep and swine were produced off farms, contributing to each county's total food supply. In addition, horses and mules were employed in industry and commerce, requiring feed allocations before each county's grain surplus was calculated. Unfortunately, these animals were not enumerated by county in the published Census; however, summary tables by states were provided for nonfarm livestock in the 1860 census of Agriculture. These counts and the total counts of farm livestock were utilized to determine a ratio that could be utilized to estimate the number of nonagricultural live-stock for each county. The number of farm livestock was divided by the proportions in Table A.5 to estimate the total animal population in each county. ⁽⁴⁶⁾ Third, seed reservations and wastage allowances were calculated for all remaining crops. Fourth, food crop surpluses available for export were calculated by subtracting consumption and seed reservations from total production (all expressed in corn equivalencies). Fifth, the surplus corn equivalencies were reconverted into the original component amounts of corn, wheat, hogs, and beef. Since raw grains were not exported very often from the region, surplus corn and wheat were converted to meal, whiskey and flour, using information in the Census of Manufacturing enumerator manuscripts. Three gallons of whiskey were allocated to each resident adult, allowing 1.5 bushels to produce each gallon. ⁽⁴⁷⁾

Sixth, the monetary values of grain and livestock surpluses, of tobacco and of cotton were estimated using average annual antebellum commodity prices. Since no prices could be located for live swine and cattle, these livestock were converted to their estimated yields in dressed meat and priced as mess pork and mess beef (the cheapest cuts). Southern Appalachians also exported horses, mules and poultry to the Deep South. The earliest count of exported horses and mules is available in the 1910 published Census of Agriculture which enumerates the number of these animals sold. To estimate antebellum exports, I assumed that each county exported the same proportion of its horses and mules in the antebellum period as occurred in 1910-- a conserva-tive estimate since postbellum exports were lower. ⁽⁴⁸⁾ Seventh, the manufacturing and extractive surpluses were aggregated for each county from the published Censuses.

Nonagricultural Economic Sectors: Sources and Methodology

Assessing Antebellum Nonagricultural Economic Sectors

Occupational distribution. The best method for documenting antebellum economic sectors is to analyze the array of occupations in which households were engaged. The 1840 published Census provides the only antebellum count of occupations, by county-- designating several agricultural and nonagricultural sectors. In addition, the sample of 3,056 households drawn from the 1860 Census of Population manuscripts provides information about occupations declared by adults enumerated in each household. This detail permits a breakout of occupations by major economic sectors. ⁽⁴⁹⁾

The 1860 enumeration of occupations is also useful in the identification of households that report employment in more than one economic sector and those households that rely on income from labor that falls outside the formal marketplace (e.g., washwoman; prostitute; unemployed). Professionals, government employees and artisans are clearly identified in Schedule I, making possible a close analysis of these occupations. Because Schedule I also reports the accumulated wealth for each household, it is possible to investigate the degree of wealth concentration by certain economic sectors.

Antebellum commerce and trade. The major sources of statistical information on agricultural surpluses were the 1860 farm Census manuscripts and the 1840 and 1860 published Censuses. Several unpublished theses and dissertations were used for information on surpluses and external trading in specific Southern Appalachian counties. The Civil War veterans questionnaires, the slave narratives, and published personal narratives (especially diaries, journals and letters) were gleaned for unusual details about external trading and distant markets. Several reprinted antebellum sources provide details about East Tennessee trade linkages and exports. Manuscript collections were located for Appalachian plantations, and published plantation journals and letters were also utilized.

At regional archives, store accounts, travel diaries, farm journals and family business papers were searched for details about local marketing practices and external trade linkages. The economic and political significance of regional resorts were examined in several manuscripts and published sources. Dependence of these resorts upon Deep South planters was documented in several manuscript collections, including the Southern Historical Collection, Duke University and the University of Virginia. The 1840 published Census provides a count of commission merchants, retailers, wholesalers, persons involved in hauling and transporting goods, and persons engaged in commerce-related service occupations. The 1860 household sample also contributes to the description of antebellum commerce and trade.

Antebellum industrialization. At regional archives, I utilized many resources about antebellum artisans and about nineteenth-century industries. The 1810, 1840, and 1860 Censuses provide published information about manufacturing, mining and timbering-- including the value of the goods produced. In addition, numerous primary and secondary sources examine an array of antebellum manufacturing and extractive enterprises.

Notes

1. For overviews of the debates over the numerous definitions of Appalachia's regional boundaries, see: Shapiro, Appalachia on Our Mind; McKinney, "Political Uses;" Bruce Ergood, "Toward a Definition of Appalachia," in Appalachia: Social Context Past and Present, ed. B. Ergood and B.E. Kuhre (Dubuque, IA: Kendall-Hunt,1983): 31-40; Karl B.Raitz and Richard Ulack, Appalachia: A Regional Geography (Boulder: Westview Press, 1984).

For arguments that Appalachia is a homogeneous culture, see: Kephart, Southern Highlanders; Campbell, Southern Highlander; Eller, Miners, Millhands, Mountaineers.

For arguments that Appalachia is a regional myth, see: Robert F. Munn, "The Latest Rediscovery of Appalachia," Mountain Life and Work 40 (1965): 10-12; Shapiro, Appalachian on Our Minds; John Egerton, "Appalachia: The View from the Hills," in Appalachia: Social Context, 238-41.

2. See Campbell, Southern Highlander. All counties were categorized into predominant terrain types, utilizing guidelines of the Appalachian Regional Commission, Appalachia-- A Reference Book. The Shenandoah Valley counties are classified as ridge-valley terrain. The excluded counties are AL: Madison, Morgan and SC: Greenville, Spartanburg.

3. For the Fisk University collection, see: "Unwritten History of Slavery." The WPA collection is found in published form in 41 volumes collected in three series; see: Rawick, American Slave; Idem, American Slave: Supplement I; Idem, American Slave: Supplement II. Appalachian narratives were also selected from: Drew, North-side View; Bontemps, Great Slave Narratives; Killion and Waller, Slavery Times; Perdue, Barden and Phillips, Weevils in Wheat; and John W. Blassingame, ed., Slave Testimony: Two Centuries of Letters, Speeches, Interviews and Autobiographies (Baton Rouge: Louisiana State University Press, 1977). The following personal narratives were also utilized: Narrative of Bethany Veney; Louis Hughes, Thirty Years a Slave: From Bondage to Freedom (1897; reprint, New York: Negro Universities Press, 1969); and Charles Campbell, ed., Memoirs of a Monticello Slave (Charlottesville: University of Virginia Press, 1951).

4. Editing occurs in these instances, just as it did in the slave collection. Neither source is free from third-party biases, personal exaggerations or human forgetfulness. The slave narratives are no less reliable than the newspaper accounts, letters, political speeches, or

5. For estimates of intrastate sales, see Bancroft, Slave Trading in Old South, 382-406.

6. See Sutch, "Breeding of Slaves," Table 3; Tadman, Slaves and Speculators, 142.

7. Starobin, "Industrial Slavery," 422.

8. See later discussion of prices for interstate sales for sources.

9. For age/sex distribution, see Sutch, "Breeding of Slaves," Table 3. Price variations for prime field hands over the period 1840-1860 are available in Evans, "Some Economic Aspects," 329-37. The price ratio of all other age/sex cohorts to the value of prime field hands was deter-mined from a letter to an Appalachian slaveholder from a Richmond trader stating slave prices by age and sex; see document dated 13 November 1848 in Dickinson Family Papers, University of Virginia Manuscript Collections. For dollar value conversion rates, see David and Solar, "History of Cost of Living," Table 1.

10. For rates of mortality, manumission, and runaways, see Statistics of the United States in 1860, 286, 337-8.

11. For cotton and tobacco, see Pei-Kang Chang, Agriculture and Industrialization: The Adjustments that Take Place as an Agricultural Country Is Industrialized (Cambridge: Harvard University Press, 1949), 186. For other crops, see George K. Holmes, Supply of Farm Labor (Washington, DC: United States Department of Agriculture Bureau of Statistics Bulletin 94, 1912).

12. 1860 enumerators, typically, identified family members who had different surnames with labels like "son-in-law" in the occupation column.

13. For this entire period, deeds are recorded in nonalphabetical order in all states; and the availability of name indexes is sporadic. Thus, to aggregate land holdings and transfers for a single name, one would have to check every annual county list of deeds over a period of about twenty-five years.

14. To search for frontier county land ownership maps, begin with Richard W. Stephenson, Land Ownership Maps: A Checklist of 19th Century U.S. County Maps in the Library of Congress (Washington, DC: Government Printing Office, 1967). For assistance in borrowing from the Maryland State Archives, it is necessary to purchase county microform holdings guides; I used collection guides for Allegany, Frederick, Garret, and Washington counties. For assistance in identifying microfilm in the Salt Lake City collection, utilize microfiche locality guides (organized by states and counties) at your local Church of the Latter Day Saints.

15. Many studies have seriously distorted the picture of frontier land ownership by assuming that every listed taxpayer was a resident of the county; see Soltow, "Land Inequality on Frontier."

16. For a map of Kentucky during this time period, see Thorndale and Dollarhide, Map Guide, 123. Tax lists for Knox (1801), Madison (1800), Pulaski (1800) and Wayne (1801) were secured from the Family History Center, Church of Jesus Christ of the Latter Day Saints (Salt Lake City).

17. These tax lists were secured in published form; see: Betty Elswick, "Floyd County, Kentucky Tax List, 1793," Kentucky Genealogist 22 (1980): 43-55; Kentucky Department of State Archives, " Madison County Tax Lists, 1792," Register of the Kentucky State Historical Society 23 (1925): 115-41. To check residency, the published 1790 Census list was utilized. See Charles B. Heinemann, ed., First Census of Kentucky 1790 (Baltimore: Genealogical Publishing, 1965).

18. For an 1800 map of Maryland, see Thorndale and Dollarhide, Map Guide, 153. Records were not accessible for the other two Appalachian Maryland counties (Allegany and Frederick). The Washington (1804) tax list was obtained on microfilm from the Maryland Hall of Records. The 1800 Census was secured in published form; see "Washington County, Maryland 1800 Census," Maryland and Delaware Genealogist 1-2 (1959). The published Census was not alphabetized; to simplify cross-matching, the list was computerized and sorted alphabetically.

19. For an 1800 map of North Carolina, see Thorndale and Dollarhide, Map Guide, 246. I adapted the methodology of Soltow, "Land Inequality" in using poll tax to identify residents. Every male aged 21 or older was required to pay the poll tax; the poll tax was recorded by household of residence. Most of the few early North Carolina records still extant are illegible. Ashe (1815) and Surry (1800) tax lists were secured on microfilm from the Church of Jesus Christ of the Latter Day Saints (Salt Lake City). The Burke (1805) list was secured in published form; see Ransom McBride, "Burke County, North Carolina List of Taxables for 1805," North Carolina Genealogical Society Journal 8 (1982): 225-37.

20. For a 1790 map of South Carolina, see Thorndale and Dollarhide, Map Guide, 297. Pendleton District was subsequently divided into the counties of Pickens, Oconee, and Anderson. For the published 1790 Census, see U.S. Census Office, Heads of Families at First Census. For Pendleton District deeds, see Hendrix, Pendleton County Deed Books.

21. For 1800-1810 maps of Tennessee, see Thorndale and Dollarhide, Map Guide, 315-16.I adapted the methodology of Soltow, "Land Inequality," in using poll tax to identify residents. Every male aged 21 or older was required to pay the poll tax; the poll tax was recorded by household of residence. The following county tax lists were utilized: Anderson, 1802; Blount, 1801; Carter, 1796; Grainger, 1799; Hawkins, 1809; Jefferson, 1800; Knox, 1806; Rhea, 1808; Sullivan, 1796; Washington, 1789-90; and White, 1811. A microfilm copy of the White (1811) tax list was utilized at the East Tennessee Historical Society, Knoxville. Several county tax lists have been published; see: Pollyanna Creekmore, "Early East Tennessee Tax Lists," East Tennessee Historical Society Publications 23-32 (1951-1960). For Washington (1789-90), see Creekmore, "Early East Tennessee Taxpayers: Washington County, 1789-1790," Tennessee Ancestors 5 (1989): 73-87. For Rhea (1808), see Mary B. Curtis, Early East Tennessee Tax Lists (Ft. Worth: Arrow Printing, 1964).

22. For an 1800 map of Virginia, see Thorndale and Dollarhide, Map Guide, 350. The following personal tax lists and land books were utilized: Albermarle, 1800; Amherst, 1800; Augusta, 1800; Bath, 1800; Bedford, 1800; Botetourt, 1800; Fauquier, 1800; Franklin, 1800; Frederick, 1800; Russell, 1792; Wythe, 1800. Russell (1792) personal tax list and all county land books were secured on microfilm from the Family History Center, Church of Jesus Christ of the Latter Day Saints (Salt Lake City). Except for Russell County, personal tax lists were secured in published form. For Wythe County, see Netti Schreiner-Yantis, 1800 Tax Lists and Abstracts of Deeds, 1796-1800 of Wythe County, Virginia (Springfield, VA: By author, 1971). For all other counties see, The Virginia Genealogist, 3-9 (1959-80).

23. For an 1800 map of West Virginia, see Thorndale and Dollarhide, Map Guide, 368. Ohio County was not sampled because there is no extant personal tax list for this period. The following county tax lists and land books were utilized: Berkeley, 1800; Brooke, 1800; Greenbrier, 1800; Hampshire, 1800; Hardy, 1800; Harrison, 1800; Kanawha, 1800; Monongalia, 1800; Monroe, 1799; Pendleton, 1800; Randolph, 1805; and Wood, 1800. Use of the personal tax lists has been greatly simplified by the published compilation of all 1800 personal tax lists in one volume; see Steven A. Bridges, Virginians in 1800: Counties of West Virginia (Trumbull, CT: By author, 1987). The Randolph (1805) personal tax list and all county Land Books were secured on microfilm from the Family History Center, Church of Jesus Christ of the Latter Day Saints (Salt Lake City).

24. For 1790-1800 maps of Alabama and Georgia, see Thorndale and Dollarhide, Map Guide: 11, 80. The records needed for this analysis are available in one published source; see Herbert B. Kimsey, Early genealogical and Historical Records: Habersham County, Georgia (Athens, GA: By author, 1988).

25. For the most recent research on mobility, see David C. Hsiung, "Isolation and Integration in Upper East Tennessee, 1780-1860: The Historical Origins of Appalachian Characterizations" (Ph.D. diss., University of Michigan, 1991).

26. Legibility was rarely a problem in the lists utilized for this research. Typically, the handwriting was very precise.

27. I developed this methodological adaptation in response to critiques from some scholars who argued that the only landless people in Appalachia were the extended kin of land owners who were actually residing and working on family farms. In this way, I can demonstrate that landlessness occurred even when one accounts for extended households on family farms.

28. Microfilmed copies of Census manuscripts (Schedules I, II, IV) were secured from the Center for Research Libraries (Chicago), from the University of Tennessee, Special Collections, and were utilized on-site at several regional archives. Nine (9) farm households were dropped from the sample when they could not be crossmatched with Schedule I.

29. For full explication of these methodological conventions, see Frederick Bode and Donald E. Ginter, Farm tenancy and the Census in Antebellum Georgia (Athens, GA: University of Georgia Press, 1986), Ch. 2. 1860 Census enumerators were instructed to count as a farm operator any "person residing upon or having charge of the farm, whether as owner, agent, or tenant." The 1860 Census instructions gave enumerators only this broad distinction to follow: "The proprietor of a farm for the time being who pursues agriculture professionally or practically, is to be recorded as a farmer; the men who are employed for wages by him are to be termed farm laborers." See United States Census Office, Instructions to United States Marshals, 8th Census (Washington, DC: Government Printing Office, 1860). The following studies are earlier exemplars of this methodology: Frank L. Owsley, Plain Folk of the Old South (Baton Rouge: Louisiana State University Press, 1942); Blanche Clark, The Tennessee Yeomen, 1840-1860 (New York: Octagon Books, 1941); Herbert Weaver, Mississippi Farmers, 1850-1860 (Nashville: Vanderbilt University Press, 1945); Paul W. Gates, Frontier Landlords and Pioneer Tenants (Ithaca, NY: Cornell University Press, 1945).

30. Similar steps were followed in assessing landless households selected from the 1860 Census of Population.

31. The Washington estates inventory was secured in published form; see P.D. Shingleton, "Washington County, Maryland: Balance Books-- Estates," Maryland and delaware Genealogist 3, 5 (1960-1974). In 1797 dollars, 1 pound = $4.44; 1 shilling = 22 cents; 1 pence = 1.8 cents; see Arthur Nusbaum, A History of the Dollar (New York: Columbia University Press, 1967), 11.

32. For dollar conversions, see Nusbaum, History of Dollar, 11. The value of 1800 slaves is estimated at $200 each; see Lee Soltow, "Kentucky Wealth at the End of the 18th Century," Journal of Economic History 43 (1983).

33. For dollar conversions, see Nusbaum, History of Dollar, 11.

34. Soltow, "Kentucky Wealth," offers estimates land values for Kentucky that exceed the average values of Appalachian Virginia acreage. Since east Kentucky is comprised of mountainous and hill-plateau terrain, its lands would have sold in 1800 at a much cheaper rate than the acreage of the Blue Grass. Thus, the frontier land values of western North Carolina (of similar terrain) are more appropriate estimates for east Kentucky than those of Soltow. Since that county comprise mountainous and hill-plateau terrain, I applied average value per acre for land reported in the 1815 Ashe, North Carolina tax list.

35. The veterans questionnaires are available in published form; see Dyer and Moore, Tennessee Civil War Veterans Questionnaires.

36. For the published sources for the slave narratives, see the link to the list of Appalachian slave narartives above.

37. This sample was also utilized for analysis of landholding and occupational patterns.

38. For each farm household, consumption is subtracted from total crop production (both expressed in corn equivalencies) to determine how much of the total production remains as surplus.

39. A turn-of-the-century study determined that 100 acres were needed for a farmer to produce survival needs for the average mountain farm family in 1900. I followed this rule of thumb when assessing surplus land for 1860 farms. See United States Department of Agriculture, Agriculture in the Southern Appalachian and White Mountain Watersheds (Washington, DC: Government Printing Office, 1908.

40. For corn equivalency methodology, see: Raymond Battalio and John Kagel, "The Structure of Antebellum Southern Agriculture: South Carolina, A Case Study," in W.N. Parker, ed. The Structure of the Cotton Economy of the Antebellum South (Washington, DC: Agricultural History Society, 1970): 25-37. The count of beef cattle does not include milk cows. For rice conversion, see Hilliard, Hog Meat and Hoecake, 273n.

41. Feed equivalencies from Ralph D. Jennings, Consumption of Feed by Livestock, 1909-1956 (Washington, DC: United States Department of Agriculture Production Research Report 21, 1958), 74-5. I assumed that buckwheat was equivalent to barley in nutritional value.

42. An adult is defined to be a person 16 years of age or older. A household is defined as those persons residing in the same dwelling who are related by kinship ties or living as a family unit. Food requirements for persons residing in the household as hired laborers or slaves are not counted as part of subsistence needs. Human consumption levels are based on annual allowances per adult of 13 bushels corn, 2 bushels wheat, 2.2 swine, plus one beef per household; see Hilliard, Hog Meat and Hoecake. For food consumption of sheep and oxen, see Battalio and Kagel, "Structure of Antebellum Southern Agriculture," 36. Poultry were not allocated as part of regular food allowance since this was not customary in the antebellum period. Primary sources indicate that poultry were treated as a semi-luxury in the nineteenth-century Southern diet-- reserved for special occasions. Slaves usually sold their poultry to whites and seldom consumed the meat; see Hilliard, op.cit., 46, 65. Archival records indicate that poultry were exported from the region; however, the 1860 Census provides no count from which to make estimates. For whiskey allowances and production, see Gusfield, Symbolic Crusade, 75; Dabney, Mountain Spirits, 51n.

43. For livestock husbandry, see Gray, History of Southern Agriculture, ch. 35; Cornelius O. Cathey, "Agricultural Development in North Carolina, 1783-1860," James Sprunt Studies in History and Political Science 38 (1956), ch. 9. For the proportion of livestock fed on pasturage and forage only, see United States Department of Agriculture, Agricultural Statistics (Washington, DC: Government Printing Office, 1942). For livestock feed requirements, see Battalio and Kagel, "Structure of Antebellum Southern Agriculture," 28-30.

44. For information on seed reservations, see Battalio and Kagel, "Structure of Antebellum Southern Agriculture," 28n. In the actual calculation, I allowed an additional 3% to account for potential loss or wastage. This is a very conservative approach since archival sources indicate that most antebellum farmers producing surpluses purchased new seeds each year.

45. Regarding calculation of herd replacement ratios, see Battalio and Kagel, "Structure of Antebellum Southern Agriculture," 36.

46. In the early 1900s, 30.5% of all livestock consumed roughage only (no feed grains); this exclusion was also applied to the antebellum era; see Jennings, Consumption of Feed by Livestock, 79. To determine the number of nonagricultural livestock, the number of farm animals in each category was subtracted from these grand totals.

47. To reconvert corn equivalencies into their original crop components, one must work backwards. When I converted total crop production, I calculated what ratio each made up of the total corn equivalencies. To reconvert, I multiplied the number of surplus corn equivalencies by the appropriate ratios and then divided by the correct ratios. Surplus corn was converted to meal, using the ratio of 1.1 bushels raw grain to mill one bushel finished product. Bushels of meal were converted to barrels (the method by which they were exported) using these ratios: 1 bushel weighed 45 pounds; 1 barrel contained 196 pounds. For meal conversions, see Battalio and Kagel, "Structure of Antebellum Southern Agriculture," 27n; Cole, Wholesale Commodity Prices, x. One barrel of flour was produced from 4 and 1/2 bushels wheat; see Eugene Merritt, "International Trade in Farm and Forest Products, 1901-1910," United States Department of Agriculture Bureau of Statistics Bulletin 103 (1913), 52. Each gallon of whiskey required 1.5 bushels corn; see Dabney, Mountain Spirits, 51n.

48. Prices were obtained from Cole, Wholesale Commodity Prices. Each hog was estimated to yield 106.4 pounds dressed pork; each head of cattle 239.9 pounds dressed beef; see Battalio and Kagel, "Structure of Antebellum Southern Agriculture," 36-7. The 1910 published Census of Agriculture also provides average horse and mule prices.

49. The 1840 Census counted only the primary occupation of the head of household. Secondary occupations of the head and occupations of other household members were not counted. For this reason, the published count is seriously skewed-- particularly for nonagricultural occupations.