A Reinterpretation of The Slavery/Profitability Question


Leonard A. Carlson* and Robert S. Chirinko

Department of Economics

Emory University

Atlanta, Georgia USA 30322-2240

*PH: (404) 727-6375

FX: (404) 727-4639



In a recent article (as measured in historical time), Conrad and Meyer (1958) revolutionized the study of economic history by examining slavery with the traditional tools of neoclassical economics. Specifically, slavery was analyzed as a potentially profitable economic activity, and chattel labor as a financial asset. Using traditional economic tools to study historical topics has gained wide currency, and underlies the cliometric approach that now dominates the field of economic history. However, recent developments in financial economics question the appropriateness of some traditional asset market valuation formulas, thus reopening the question of slavery's profitability.

This paper begins by showing, as in prior studies, that slave assets generated normal market returns. The K-period holding return to slave assets is the sum of the capital gains (adjusted for depreciation) and the marginal product of labor, the latter measured in terms of the value of the annual output of a slave less the costs of maintenance. The primary data source is Peterson (1929), supplemented by information from Kahn (1992), Ransom and Sutch (1988), Ransom (1989), and Watkins (1895). The normal return is defined by the return on financial assets, and we use the databases developed recently by Homer and Sylla (1996) and Schwert (1990), as well as a series developed by David and Solar (1977).

Valuing chattel labor with the traditional present value formula is based on the assumption that the asset market was efficient. This assumption has come under increasingly formidable attacks in recent years, and can be tested using recently developed methods in financial economics (see Campbell, Lo, and MacKinlay (1997, Chapter 7):

r(t,t+K) = A + B(K) [d(t)/p(t)] + E(t) K=1,2,3,4,5,10

where r(t,t+K) is the K-period return on slave assets, [d(t)/p(t)] is the dividend-price ratio for slave assets, A and B(K) are estimated parameters, and E(t) is an error term. The equation is estimated by OLS. If the asset market is efficient, then no variables dated prior to and including period t should have predictive power for returns. Prior work with Twentieth Century data has established that the dividend-price ratio has remarkable predictive power.

We present the R2(K)'s from the equation (1) estimated at different horizons for the U.S. equity market from 1927-1994 (Campbell, Lo, and MacKinlay (1997, Table 7.1)) and the slave market for 1800-1859 (authors' preliminary calculations). The Twentieth Century results are presented first; the Nineteenth Century results follow in parenthesis: R2(1): 0.073 (0.072); R2(2): 0.143 (0.162); R2(3): 0.207 (0.194); R2(4): 0.261 (0.231).

There are three key results. First, the dividend/price ratio has predictive power in the slave market, and hence the assumption of asset market efficiency fails dramatically. Second, the pattern of rejections are amazingly similar for both the slave market and post-1927 U.S. equity market. At a one- period horizon, the dividend-price ratio has only modest impact, but predictive power rises as the horizon is extended. Third, at a horizon of ten years, the predictive power of the dividend- price ratio, hence the failure of the efficiency assumption in the slave market, is a very sizeable 0.531.

Preliminary work indicates that this inefficiency and the widely-held view that slavery was profitable are traceable to a financial market "bubble" that increased ex-post returns. We identify the sources of the inefficiency by estimating asset prices based on a projection technique and alternative assumptions on the ex-ante information set used by slaveholders. The technique was introduced by Abel and Blanchard (1986), and involves a three-step procedure: 1) define the fundamental present value of an asset in terms of future marginal products and discount rates; 2) estimate the autoregressive process defining the evolution of these variables; and 3) based on variables known at or before period t, solve the autoregression forward and calculate the fundamental value of the asset at period t. The bubble is calculated as the difference between actual and fundamental values. In the absence of a bubble, the differences will be tightly distributed about zero. Alternatively, a systematic and prolonged deviation of actual prices from fundamentals is interpreted as a bubble.

These projections are used to reevaluate two issues surrounding this "peculiar institution." A bubble in the market for chattel labor has implications for the willingness of Southern slaveholders to abandon slavery. As discussed by Wright (1976, 1978), Southern slaveholders were deeply concerned about the value of slave assets after the election of 1860. If slave prices were artificially high due to a bubble in the late 1850's, this asset market inefficiency may have contributed to the hard- line taken by Southern secessionists (Ransom, 1989).

Furthermore, an asset market bubble would create false price signals that may be partly responsible for the (mis)development of the Southern economy. Starting with Phillips (1918) and discussed extensively by Bateman and Weiss (1981), there is a debate among economic historians as to whether the existence of slavery distorted investment in the South in the first half of the nineteenth century. Our series quantifying an asset market anomaly provides a new way to consider this unresolved and enduring issue.