A First Glimpse into the Short Side of Hedge Funds
After filtering the ANcerno data, for each ANcerno manager we consider each 13F manager, and count the number of firm-quarters for which the holdings changes computed from the ANcerno data match the holdings changes computed from the 13F data. Dividing this count by the number of available firm-quarters for the ANcerno manager, for each ANcerno manager the result is a ranking of how closely the holdings changes of each of the 13F managers match the holdings changes of the ANcerno manager. We also construct an alternative ranking in which we count the number of firm-quarters for which the absolute difference between holdings changes computed from the ANcerno data and the holdings changes computed from the 13F data is less than ten percent of the absolute value of the change in holdings computed from the ANcerno data. Using these two rankings, for each ANcerno manager we select the ten most closely matching 13F managers. Finally, for each ANcerno manager we examine the names of the ten mostly closely matching 13F managers to find a match that is exact up to abbreviations and minor variations in the manager names.7 Using this procedure we are able to match 194 ANcerno managers to the managers reported in the 13F filings.8
We identify management companies as hedge funds based on their SEC Form ADV filings, following the approach in Jame (2015). We classify a management company (asset manager, or fund company) as a hedge fund if first, item 5D of Form ADV indicates that more than half of the management company’s investors are high net worth individuals or pooled investment vehicles and second, item 5E of Form ADV indicates that the management company charges a performance fee. Of the 194 ANcerno managers that we match to 13F managers, we identify 53 as hedge funds and 141 as non-hedge fund investment managers. 2.3. Classifying Transactions as Short Sells, Short Buys, Long Buys, and Long Sells After matching the ANcerno investment managers to the managers in the13F filings we combine the manager positions reported in the 13Fs with the transactions (changes in positions) from the ANcerno data to construct estimates of the holdings of each manager in each stock on each date. One we know the managers’ positions in the various stocks on each date we can easily identify the transactions as short sales, short buys, long buys, and long sells.
For example, we identify a decrease in a non-positive position as a short sale, and an increase in a negative position as short covering, that is as a short buy. Similarly, an increase in a non-negative position is a long buy and a decrease in a positive position is a long sell. It is straightforward to estimate manager A’s holdings in stock X for each date as long as the manager A reports holdings in stock X for at least one quarter-end date. However, it is possible that ANcerno reports transactions by a manager in stock X but the manager never files a 13F reporting holdings of stock X because the manager’s position never meets or exceeds the Form 13F reporting threshold of 10,000 shares or $200,000. In this case we can construct upper bounds on the manager’s positions in stock X on each date by using the fact that manager A’s maximum position in stock X across all dates must have been less than the reporting threshold. Let I(t, A, X) represent the estimate of the inventory of ANcerno manager A in stock X at the end of trading day t. Negative values of I(t, A, X) mean A has an open short position in stock X.
For each date t for which a manager A Form 13F reports a position in stock A we set I(t, A, X) equal to the number of shares listed in the 13F. We then use the ANcerno data to assign values to I(t, A, X) on trade dates between 13F filings, before and after the first and last 13F filing reporting holdings in stock X, and if manager A never reports holdings in stock X. For each combination of manager A and stock X there are four cases: (1) dates t that fall between two end of quarter dates for which manager A reports holdings of stock X; (2) dates t that are before the date of manager A’s first 13F that reports holdings of stock X; (3) dates t that are after the date of
Full study below: