"The Laborer Is Worthy of His Hire":
Shaker Religious Communes in External Labor Markets

John E. Murray, University of Toledo

I. Introduction
This essay considers the extent of markets for labor in 19th century America, with special attention to New York state in the late antebellum period. I employ records of Shaker apprentices and hired hands. The Shakers were a Christian communal society in which property was owned jointly and members received a constant return to their labor: room, board, and the collective religious experience. This paper examines questions about the efficiency of communal decision making (Bonin, Jones and Putterman 1993), the existence of a moral economy in antebellum New York (Rothenberg 1992, Brooks 1996) and the tendency over time of cooperatively organized firms to employ greater proportions of non-member labor (Ben-Ner 1984).

II. Markets for apprentices
The Shakers are remembered today in large measure for their commitment to celibacy, and in fact in Shaker history, celibacy preceded their communalism (Stein 1992). One consequence of this policy was a continual need for converts to perpetuate the group. Acceptance of children into the community from the outside was one such source of potential members - children such as apprentices who were bound to the community by indentures signed by their parents or guardians.
Shaker indentures exist from several communities. In order to link the indentures to membership records listing the date of departure of the apprentice, however, only the records from the New Lebanon, New York, Shaker community were used, as these could be linked to Shaker membership records that showed the outcome of the indenture1. The Shaker sample was similar to other contemporary apprentice samples in terms of trade to be learned and proportion of boys.
Influences on the probability that apprentices would complete their indentures were estimated via a logit regression in which the dependent variable was 1 for apprentices who fulfilled their contracts and 0 for those who left early. No Shaker apprentices died during their indenture. Age had no significant effect on probability of completion. Boys were 38% less likely to complete their indentures than were girls. Proximity to family mattered: those apprenticed to the New Lebanon Shakers from nearby were less likely by half to fulfill their contracts, but we cannot tell if this was due to more runaways or parental takeaways. Children who signed the documents (rather than marking them with an X) were a third more likely to complete them. Probability of completion increased over time, while elsewhere in contemporary America the tendency of apprentices to run away before fulfilling their contracts was leading to the demise of formal apprenticeships (Elbaum 1989).
The market for Shaker apprentices was subject to all the conflicting incentives that are typical of the master-apprentice relationship. All parties to the exchange were aware of these incentives:
Some Believers grew disgusted because so many children were snatched from their Shaker homes just as they were reaching an age when they might begin making a meaningful economic and spiritual contribution. In 1865, after a couple named Sherman came to leave a seven year old and take away several older progeny, a New Lebanon Brother complained about what he called a "real money making scheme for Shakers to raise them up give them a good education board and clothe them till they get to be old enough to be of some benefit and then have them up and kick the bucket." (Brewer 1986: 150)
Indeed, a prominent Elder noted in 1874 that the Society no longer admitted children, believing that adult converts made the most committed Shakers (Nordhoff 1966).

III. Markets for monthly contract laborers
It is useful to distinguish between workers hired by the day and by the month, as these seem to have been two different labor markets in the antebellum United States. This section examines records of hired hands on long-term contracts at the New Lebanon and Groveland, New York, Shaker communities2. Entries in Shaker account books provide date and value of wage payment and number of months and days of labor. The dates of these records fill an implicit gap in our understanding of Shaker economies. The use of hired laborers in the later 19th century has often been interpreted as a sign of Shaker decline. The present data illustrate that hired workers were employed in a systematic fashion from early in the histories of the two communities. Hiring non-member workers from outside the community coincided with communal prosperity, rather than foretold its eventual destruction.
In calculating average wages, the few cases in which the Shakers compensated their workers with in-kind payments such as a saddle, whiskey, shoes, and bricks were excluded. About a fifth of Groveland contract workers provided skilled labor, either at a carpentry or joinering or a mill related task. At New Lebanon a quarter of the monthly workers were associated with milling, while only 5% had other tasks listed, which were hatting and blacksmithing. The average length of contract at Groveland was just over three months, while the New Lebanon contracts were unusually long, lasting on average over seven months.
The Groveland Shakers paid their contract workers at rates somewhat above those paid by other employers, but the New Lebanon Shakers did not. The Groveland mean monthly pay of $15.86 was well above Lebergott's (1964) estimates of $10.00-13.00 for New York at this time3. McNall (1952: 88) found that by the mid 1840s the monthly rate for hands in the Genesee Valley of New York, that included Groveland, was also in the range of $10-13 a month. Bidwell and Falconer (1925: 275) extended the pay range to $10-15 a month by the end of the 1840s. The New Lebanon Shakers paid, on average, thirteen and a half dollars a month to its contract work force, well within this range.
To determine influences on the wages paid to contract labor, hedonic wage regressions were estimated. The dependent variable was the natural log of nominal monthly wages. Skill premia were noticeable at both locations. At Groveland, carpenters and mill workers received wages two-thirds and two-fifths again as much as common laborers, respectively. At New Lebanon, blacksmiths and hatters were paid half again as much as laborers. The skill premium estimated by Margo and Villaflor (1987) in the Northeast at this time was one-half to two-thirds, about the same as that paid by the Shakers. Although the New Lebanon Shakers hired their contract workers to perform mostly manual labor, a premium for literacy of about a fifth was detected, suggesting that literacy may have stood as a proxy for other work related habits, or was in fact useful in a rural setting (Petterson 1996).
The Groveland Shakers paid a premium of about 2% for each additional month that a worker was willing to extend his contract. If an employer fears being unable to find workers more than the worker fears being unable to find employment, longer term contracts will pay a premium to lock in worker availability. In the Groveland case, wage premia for longer term workers are consistent with claims in the secondary literature that the Genesee Valley was chronically short of hired laborers (McNall 1952).
Another perspective on Shaker contract labor can be seen in the duration of their contracts. While duration of contracts in Rothenberg's (1992: 187) sample of Massachusetts farm workers were roughly constant over time, New York Shaker laborers were hired on ever longer contracts. Groveland contracts in the 1850s were a month and a half longer, holding other characteristics constant, than earlier contracts. The increasing duration of contracts in the late antebellum period, combined with the higher wage rates that the Groveland Shakers had to pay longer term workers, indicates that the Groveland Shakers' strong demand for hired men predated the Census of 1860, which found 33 males aged 14-80 operating a farm of some 1800 acres. Thus, dependence on hired labor was not a new development at Groveland during its decline, but was typical during its relatively prosperous years as well. New Lebanon contracts were significantly shorter in the 1820s than later, and somewhat longer in the 1850s than they had been earlier, although the latter difference is not significant at standard levels. These results are consistent with increasing length of contracts, which may have indicated ever greater market dependence (Rothenberg 1992).

IV. Markets for day laborers
Account books from the Shaker Community at Groveland show that 2105 payments were made to individual workers for their daily labor between 1837 and 18584. These were condensed into observations that consisted of month, year, wage, task, number of working days in that month at that task and wage, and number of such payments. Thus, for example, in August 1837 payments for the following days of "harvesting" were made: 8 payments at 10/ (i.e., shillings) per day for a total of 12 half days of work, and 7 payments at 8/ per day for a total of 10 days of work. These combined to make two observations, one at each wage rate. This procedure yielded 797 observations. The number of days worked in each observation was used to weight observations in the regressions.
The daily workers were occupied with nearly every conceivable task that operation of a large, diversified commune could require. One hundred thirty three different tasks were specified. Unspecified work or labor composed 322 observations, or about 40% of the sample. Mill related tasks, such as "at the mill," "at the mill race," "fix mill," "in flouring mill," "millwright," and "tending mill" accounted for 8% of the sample. Mason work, which included "cutting stone," "drawing flagstone," "laying brick," "laying stone wall," "quarrying stone," and "tending mason," as well as "mason" and "mason work" described another 76 or 9.5% of the sample. Work with teams of horses or oxen made up 2%. Many occupations had too few mentions to justify a separate dummy variable, so the "assorted" category includes, among other tasks, blacksmithing, "grubbing," "loading boat," "shoveling," and the mysterious "work in storm."
In a typical year, demand for hired labor at Groveland was greatest in July through September. The peak number of worker days at unspecified labor was in September, while those with specific tasks were most commonly employed in July. This is not surprising for a largely agricultural organization, and in fact the monthly employment of hands at a nearby Genesee Valley farm (McNall 1952: 112) shows a similar pattern.
There was no visible trend upward in the number of worker days per year. Since Groveland's Shaker population was steadily falling over this period, from 124 in 1840 to 94 in 1860, this suggests that at Groveland hired hands were not replacing Shakers in the operation of the commune. If the Shakers had been substituting hired labor for that of Shaker Brethren we would expect an increase in the number of worker-days during the Shaker population decline. Nor do Groveland data support Ben-Ner's hypothesis (1984) that the proportion of work force that is not a member of a worker-owned firm should increase over time. These figures also correct a misstatement made in the standard history of the Society that also has implications for the viability of communalism. Stein (1992: 200) wrote that "in 1860...Groveland began to hire outside help," but clearly the Groveland Shakers had been hiring non-Shaker laborers from a much earlier date. The quantitative record thus aids in establishing just how common it was for hired labor to be present at Groveland. A much smaller sample of daily wages from New Lebanon is also available, from 1825-1860. The vast majority of these observations consisted of payments for unspecified work, although a few were for blacksmiths or teamsters.
Hedonic wage regressions reveal regularities in payments to Groveland and New Lebanon day laborers that show greater nuance than do the patterns in contract labor payments. Masons and millers at Groveland enjoyed substantial wage premia over unskilled labor. The premium paid to teamsters at Groveland was large as well, twice the payment for unspecified labor, but some of this was actually a rental payment to the teamster for the use of his capital (the team of horses or oxen). The Groveland premia were much larger than those found by Rothenberg (1992) in her study of roughly contemporary Massachusetts farmers. Part of the difference may have been due to the greater variety of tasks on the Shaker communes, as they operated not just farms but small manufacturing units as well.
The harvest premium is noteworthy due to its similarity between the two groups. Rothenberg (1992: 198) estimated that day laborers on Massachusetts farms around this time were paid a premium for harvest work in July and August, due to the increased demand for relatively inelastic hired labor around harvest time. Using a similar methodology, consideration of farm related occupations at Groveland (excluding, e.g., carpenters and millers) in July and August yielded 157 observations, 67 of which were specifically for harvest work (including haying related tasks). Harvest wages were then compared to wages for the other farm related tasks performed in July and August. The premium paid to the harvest workers, 27%, was very close to the 30% premium found in Massachusetts.

V. Regional labor market integration
A high degree of correlation has been found among commodity prices and among interest rates across regions in the antebellum period, indicating that the growth of nationwide markets for output and capital predates the Civil War (Rothenberg 1992, Bodenhorn and Rockoff 1992). For antebellum labor markets in the northeast, however, the state of integration is not as clear. Rosenbloom (1996) established that manufacturing labor markets in the northeast were integrated as early as 1879. Shaker wage records can be examined for correlation with contemporary wage series from nearby areas, to see if wages throughout the northeast moved together before the Civil War. Evidence of such co-movement would suggest an integrated, regional labor market.
Groveland Shaker wage data were divided into two series, one for unspecified work or labor and the other for skilled labor, which can be defined conveniently as all the other, specified tasks. Annual average wages were then estimated for both types of labor over the 1837-1858 period. Since a reasonable degree of product market integration in the Northeast over this period has been established (Rothenberg 1992), the difficulties of creating a price series with which to deflate the Shaker wage series can be circumvented by considering nominal wages for all available markets.
Four roughly comparable wage series are available, but the construction and context of each differs from the others. Rothenberg's data (1992) consisted of day laborer wages from farms in Massachusetts. Goldin and Margo's (1992) records came from civilian employees of Army forts throughout the Northeast and included forts in upstate New York. Adams's (1944) data on Vermont farm prices and wages of farm workers are well known. A geographically useful series is Smith's data on Erie Canal workers (1962).
Estimated correlation coefficients suggested a high degree of labor market integration from 18375. Among the unskilled wage series, each was significantly correlated with at least one of the others. The Vermont farm worker series was significantly correlated with all four other series, while the Erie Canal worker series was correlated with each of the others except the Army civilian worker series. The Shaker series was significantly correlated with the Erie Canal and Vermont series. Given the proximity of Groveland to the Canal, that is not surprising. In the period after 1843, the estimated correlation coefficient of the Shaker wage series with the Massachusetts series was 0.50, significant at the .09 level; for the Shakers and the Army series it was 0.57, significant at the .03 level. The graphs and statistics suggest that by the mid to late 1840s the Groveland Shakers were participating in what was becoming a unified Northeastern labor market for farmhands and unskilled labor. Similarly, high levels of correlation among the skilled wage series suggest that markets for skilled labor were largely integrated in New York State by an even earlier date, perhaps the late 1830s.

Endnotes
1 The indentures can be found in manuscript number 751 and the membership records can be found in manuscript number 1078, Edward Deming Andrews Memorial Shaker Collection, Winterthur Museum and Library, Winterthur, Delaware.
2 Wage records of hired workers can be found in the following sources. For Groveland, see Western Reserve Historical Society Shaker Collection manuscripts II:B-11, II:B-12, II:B-15, II:B-16, and II:B-27. For New Lebanon, see Edward Deming Andrews Memorial Shaker Collection, Winterthur Museum and Library, items numbered 1103, 1108, 1114, 1115, 1116, 1130, 1137, 1152.
3 Payments to Groveland workers appear to have consisted of cash and board, as was typical in the market for contract labor. One worker who lived nearby was paid 14/ extra for three weeks since he did not board with the Shakers (July 1847, WRHS II:B-15). The implied value of boarding, $0.58 per week, is a good bit less than Bidwell and Falconer's estimate (1925: 275) of $1.25 for New York in 1849. In the same volume, however, this man was reimbursed at a rate of 2/ per day, which yields a weekly equivalent of $1.50, much closer to Bidwell and Falconer.
4 By comparison, Rothenberg (1992) used 3018 observations covering a period of about a century, while Margo and Villaflor (1987) had 3555 observations in the Northeast alone for a period of 36 years. Thus, Rothenberg's sample yielded about 30 observations per year, Margo and Villaflor's about 100 per year, and the Groveland sample also about 100 per year.
5 Sophisticated statistical models exist for the testing of market integration with price data. A recent study (Baulch 1997) indicates that no one of the following is superior to the others: the Law of One Price, the Ravallion Model (useful for a market with a well defined center region), cointegration tests, and Granger causality tests. Given the mixture of means and modes, stratified means and regression coefficients, agricultural and non-agricultural tasks that characterize the five available samples, the conservative course seemed to be the presentation of simple correlation coefficients, and significance levels thereof.