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

Data Variables

The variables include information that, for convenience, have been grouped into eight categories: (1) vessel characteristics (name, tonnage, rig, guns, place and year of construction, owners); (2) the outcome of the voyage (3) the itinerary of the voyage; (4) the dates at which the vessel left or arrived; (5) the captain and crew of the vessel; (6) the numbers of captives; (7) the characteristics of the captives and their experience of mortality; and (8) the sources for the record. The Variable List in the “Understanding the Database” section presents a complete listing of the data variables as well as the imputed variables in the data set. Imputed variables are always marked with an asterisk. No voyage, however, includes information for all data or imputed variables. Table 1 provides a summary of the coverage for some of the more important data variables.

Generally, we attempted to preserve the written documentary record in adding to the data variables. Numeric variables, such as vessel tonnage, numbers of crew, and numbers of slaves, demanded a ranking of sources, particularly for the well-documented British trade. (5)

Sources often report different numbers of slaves embarked on or "taken on board" the coast of Africa or landed in the Americas. Furthermore, for some years there are inconsistencies in slave age or gender totals per voyage. Regarding slave exports, we were careful to distinguish between the number of slaves purchased and the number who in fact were shipped from the coast. We used slave departure totals, whether reported by slave traders, African merchants, or European captains, agents, or merchants. We included the slave departures reported in sources such as logs kept by the Dutch and English castles on the Gold Coast, even though the totals often were rounded numbers, such as 400 or 500 slaves, and even though the totals occasionally were significantly less than the numbers of slaves who were disembarked in the Americas. Users should keep these biases in mind, not least for any calculations of mortality they may wish to try. For slave arrivals recorded in customs documents or shipping gazettes, we decided to use maximum totals where there was conflicting information, under the assumption that these differences might indicate deaths of slaves before slaves disembarked.

Age categories

Age categories must also be used with care.(6) The Voyages Database includes variables for adult males ("men"), adult females ("women"), male children ("boys"), female children ("girls"), adults, children, and infants (reported often as "infants at the breast"). Unfortunately, age and sex definitions changed over time and among carriers. Arrivals in the early Iberian Americas were assigned a ratio of what a prime male slave would cost—the latter being termed a pieça da Indes. A child would receive a rating of half a pieça, a woman 0.8, and so on. It has not proved possible to infer age and gender breakdowns from aggregated pieças. In the 1660–1730 period, the London-based Royal African Company (RAC) defined children as about ten years of age or younger. For most of the British and French slave trades, a height (about four feet four inches) and/or age (about puberty) criterion distinguished adults from children. In the nineteenth century, captured slave ships of all nations, but mainly Spanish and Brazilian, had their human cargoes recorded by a variety of courts, some British, some international. There is little doubt that the criterion used to separate out adults was sexual maturity as assessed by physical appearance, which for most Africans at this time would probably occur in the mid-teens, but could vary according to the diet prevalent in the areas from which Africans were drawn as well as according to the eye of the purchasers. Yet another categorization emerges from Cuban slave trade data (1790–1820) taken from the Seville archives, which adds "men-boys" and "women-girls" to the previous categories. These we included among men and women, respectively. All these measurements are of course imprecise, with even a clear age definition of "ten years and younger" hinging on casual inspection by Europeans, because many African cultures did not attach importance to knowledge of precise ages. In nineteenth-century court records, different officials often recorded slightly different distributions of the same group of slaves. However, the physiological correlation of height (specifically the teenage growth spurt) and sexual maturity means that there is broad similarity among most of these concepts, except for the RAC’s definition, which excluded individuals that other definitions would have included as children. As the RAC records form the bulk of the age and gender information for 1660–1710, the share of children for this period is biased downward.