Construction of the Trans-Atlantic Slave Trade Database: Sources and Methods

David Eltis (Emory University), 2010

Imputing Numbers of Slaves

A second set of inferences is suggested by the data on numbers of slaves leaving Africa and arriving in the New (and in some cases, the Old) World. Although 26,216 voyages in the data set arrived with slaves, and a further 7,192 could have done so, the sources provide the actual number on board at arrival for only 18,269 voyages. On the African side, the data are much weaker, with only 8,272 yielding information on the number of slaves leaving Africa, out of 29,095 voyages that left with slaves, and a further 3,335 that could have done so. Because most of those studying the slave trade are interested in the captives rather than the ships, some inference would seem appropriate for those ships that traded without leaving anything in the historical record about the slaves they carried. The first step in making reasonable inferences is to draw on the number of captives who might have been reported for the same voyage at an earlier or later stage of its itinerary.

Only 5,803 voyages in the revised database contain numbers of both captives embarked and captives disembarked, but for a further 2,404 we have the figure for departures alone and for 12,279 numbers arrived alone. Imputed totals for the missing information may be made from the large subset of voyages that provide information on deaths during the passage. The Voyages Database contains 6,438 voyages for which a ratio of deaths to slaves embarked may be calculated. Deaths as a proportion of those embarked differed markedly by African region of embarkation. Table 2 shows breakdowns of shipboard mortality as a percentage of those slaves taken on board.

Table 2. Slaves died on board ships reaching the Americas as a percentage of those embarked, by African region of embarkation, 1527-1866

Deaths/Embarked(%) Standard Deviation Number of Voyages
Senegambia 10.9% 13.7% 421
Sierra Leone 9.8% 16.5% 231
Windward Coast 9.6% 11.9% 111
Gold Coast 12.0% 13.5% 654
Bight of Benin 11.7% 14.3% 1,197
Bight of Biafra 19.1% 18.8% 646
West-central Africa 9.1% 11.7% 2,464
South-east Africa 19.3% 16.5% 356
Region cannot be identified 17.4% 17.7% 358
All Africa 11.9% 14.2% 6,438

The breakdown of mortality ratios by African region is used here as the basis for imputing numbers arrived in the Americas where totals leaving Africa exist, and for numbers leaving Africa where the numbers on board at arrival in the Americas are known.

There remain 12,819 voyages with no information about the numbers of slaves on board ship. Indeed, we do not even know if some of these ships carried slaves. The CD-ROM database divided such unknowns into just two groups, one sailing from North American ports and the other departing from all other ports in the Atlantic. Behind such categorization was the recognition that the slave vessels from North America were substantially smaller than the average. Separate means were calculated for the two groups, adjusted in the case of the larger group for region of embarkation in Africa where this information was available. In the Voyages Database a more refined strategy is adopted. A more complete data set now allows us to focus on the type of vessel as well as the route the voyage followed in forming estimates of captives transported where such information is missing. The number of captives on board was very much a function of the type of ship as well as place of trade in Africa, and to a lesser extent in the Americas. Moreover, the size of the type of vessel as reflected in its rig changed over time. The Appendix table attempts to take into account these factors by showing average number of captives both embarked and disembarked for 155 separate combinations of first, rig of vessel and time period; and second, where these were not available, place of trade in Africa; and third, a separate grouping of 18 types of vessels – smaller than those from the rest of the Atlantic World before 1800 - built in North America. A small group of vessels have no information on either rig or place of trade and estimates of captives for these are classed as “No rig” in the Appendix table. The means for these 155 categories were added to voyage records, as appropriate, whenever data on slaves could not be extracted from the sources. Users should note that in all cases these averages might be termed “running” in the sense that that as we added data to the database we recalculated the statistics reported in the Appendix table. And as augmentation of the dataset is set to continue then the imputed values will vary as data are added in the future. Users should thus not expect to find that the imputed values assigned to any given ship type are always the same or, indeed, will remain the same.

Inferring Places of Trade Regions of Embarkation and Disembarkation
Copyright 2008, 2009 Emory University. Software licensed under GNU General Public License 3.0 or later version. Some content licensed under Creative Commons Attribution-Non-Commercial 3.0.