Evaluating public versus member short-sale breadth

Split the published total

Editorial reading: evaluate short-sale breadth by who is shorting, not by the headline total. The published total is only the starting figure.

Published exchange short-sale totals can be split into specialist, floor-trader, and other-member components. Summing those three isolates member short sales. Subtracting that sum from the published total isolates public short sales.

The public-to-member ratio

A ratio of public short sales to member short sales is constructed so that a high reading means the public is more pessimistic than exchange members. That high reading is treated as a constructive breadth signal.

The public-to-member ratio is smoothed with a 22 percent exponential moving average. Each new average combines 22 percent of the newest observation with 78 percent of the prior average.

The three-part short-sales composite

A three-part short-sales composite is defined as (A+B)/C, where A is an 18 percent exponential average of specialist shorts as a share of total shorts, B is a 29 percent exponential average of member shorts as a share of total shorts, and C is an 18 percent exponential average of public shorts as a share of total volume.

In that composite, a high reading means heavier professional shorting relative to public shorting and is treated as a negative regime signal.

Historical readings as described

On the New York public-to-specialist series, the first half of 1983 was described as a negative reading associated with the 1984 industrial-average decline, the last half of 1986 as a constructive reading ahead of the 1987 advance, and the first half of 1987 as a downturn to a four-year low before the August peak.

The American public-to-specialist series was described as offering no useful lead before the 1985-86 advance or before the 1987 crash, while remaining strongly constructive through 1988.

The three-part composite was described as negative in the middle of 1983 ahead of the 1984 decline, constructive in the last half of 1986 ahead of the 1987 industrial-average rise, unhelpful before the 1985 advance and the 1987 crash, and constructive through 1988.

Editorial reading: those descriptions are mixed. The participant split and the named smoothers are the objects under evaluation, not a settled lead time.