There was considerable evidence to suggest that the crew-calculated take-off weight (330,625 pounds) at Gander was less than the actual take-off weight. Determination of the actual weight was difficult due to inconsistent load documentation and, in some cases, an absence of adequate load documentation. Nevertheless, the Board estimates that the actual take-off weight exceeded that calculated by the crew by about 14,000 pounds. The most significant contributing factor to this underestimation was the use of an average passenger weight that was significantly less than the actual weight of a U.S. Army soldier with web gear, weapon, and the quantity of other carry- on baggage described by witnesses. Also contributing to this underestimation was the use of a basic operating weight and cargo weight that were each about 1,000 pounds in error. However, despite this underestimation, it is clear that the maximum authorized take-off weight was not exceeded for the accident flight, nor did the take-off weight exceed that allowable for the runway length available for take-off. Nor was the centre of gravity position altered significantly because of the relatively even distribution of the higher weight values.

This underestimation of weight would have, however, resulted in the use of take-off reference speeds below those appropriate for the actual take-off weight. The take-off reference speeds for the crew-calculated weight are between three and five knots lower than the reference speeds for the Board's estimate of the actual weight, that is, 344,500 pounds. According to information provided by Douglas Aircraft Co., the use of these lower speeds would have had little effect on the take-off performance of the aircraft. Early rotation would have resulted in a slight increase in take-off distance and time to take off. A slight decrease in initial climb rate would have also occurred. The stall margin would have been reduced by three knots if the 330,625-pound V₂ value was used as a reference speed by the crew.

Rotation results in a slight decrease in the rate of acceleration because of the normal increase in induced drag associated with lift production. When rotation is initiated too early, this decrease in acceleration rate occurs earlier in the take-off and results in slightly lower acceleration to lift-off speed, hence a slightly longer take-off roll, in both time and distance. With the exception of this slight lengthening of the take-off roll, there are no other adverse effects.

Other evidence suggests that the crew may have inadvertently used take-off reference speeds for a take-off weight about 35,000 pounds below the actual take-off weight. Examination of the wreckage suggested that the reference bugs on the co-pilot's airspeed indicator may have been set at the reference speeds appropriate for a take-off weight of 310,000 pounds. It is possible that the reference bugs moved during the breakup sequence and that their positions as found were not those set by the flight crew prior to take-off. Furthermore, parallax errors could account for small differences between the reference bug positions found on the face of the instrument and the positions observed and set by the first officer. Tests indicated that the possible parallax error was as much as three knots for the bug found at 144 knots and two knots for the bug found at 185 knots. There was no parallax error for the internal bug found at 158 knots. Nevertheless, with parallax errors considered, all three reference bugs were found at speed values less than those appropriate for the take-off weight calculated by the crew, and two of three were found at speed values appropriate for a take-off weight of 310,000 pounds. The positions of the three bugs at speed values less than those which corresponded to the take-off weight calculated by the crew may have been more than coincidental.

Although use of speeds applicable to a take-off weight of 310,000 pounds would result in an even longer take-off roll, slower time to lift-off, and slightly reduced climb rate, a successful take-off would follow. In certification testing, the aircraft manufacturer was required to demonstrate the aircraft's ability to perform a successful take-off when rotated 10 knots below normal rotation speed. The occurrence of a successful take-off under these conditions was further demonstrated in the computer simulations conducted by UDRI and the simulator testing conducted by the Board.

The post-accident position of the internal bug on the co-pilot's airpseed indicator was eight knots lower than the corresponding V₂ speed predicated by the actual take-off weight. If the lower V₂ speed is used as a reference, the 18-knot stall margin that would be available under normal conditions would be reduced by eight knots. If, for whatever reason, the stall speed was increased, the stall margin could be reduced to zero if lower than normal reference speeds were selected and flown.

The post-accident position of the inten al bug on the captain's airspeed indicator did not correspond with any published V₂ speed for the DC-8-63. It was suggested by a colleague of the captain that it was common practice for pilots to set this bug at a position that corresponded with V₂ plus 10 knots. If in fact this bug had been set to a position that corresponded to V₂ plus 10 knots, the corresponding V₂ speed is 162 knots, the V₂ value appropriate for the crew-calculated take-off weight. Representatives of Arrow Air could not confirm that setting the bug to V₂ plus 10 knots was common practice among their pilots. Nevertheless, it is possible that the internal bug on the captain's airspeed indicator had been set to V₂ plus 10 knots. If such was the case, it would indicate that the captain had set the bug with reference to the speeds appropriate to the crew-calculated weight.