Table 1 shows that the impact of the various constraints is even more damaging globally. Specifically, it shows our estimate of the TC of a fully constrained portfolio and compares it with the TC of a simulated portfolio with different individual constraints removed. When the long-only constraint is lifted, the resulting portfolio has a TC of 0.845, an increase of more than 200% in informational efficiency. This increase primarily is due to the concentrated cap-weighted nature of the Morgan Stanley Capital International World index. The benchmark contains about 1,900 securities, yet more than 25% of its weight is composed of the largest 40 securities and more than 50% consists of the largest 160, greatly restricting a managers ability to underweight most securities in the benchmark.
Benefits of shorting
With the long-only constraint removed, how can the gains from shorting be measured? Even a modest amount of short selling adds to portfolio informational efficiency. By way of illustration, we define a modestly constrained realistic baseline portfolio of $100 million that has no large style bets; the size, value and momentum exposures of the portfolio have moderate limits. In addition, all industries and countries are limited to allocations within plus or minus five percentage points of the benchmark and positions in individual stocks are within plus or minus three percentage points of the benchmark allocation.
Table 2 considers variations of the baseline portfolio with different levels of shorting. All portfolios are constructed to have a beta of one relative to the benchmark and have an expected tracking error of five percentage points vs. the MSCI World. They differ only in the amount of short selling allowed. For comparative purposes, the portfolio with practical maximum shorting abilities is considered to be the +200%/-100% example. This becomes the target portfolio, and the table shows what percentage of that portfolios TC is obtainable through various relaxations of the long-only constraint.
When limited to long-only positions, the portfolio has a TC of just 0.341. At the 130/30 level, this jumps almost 52% to 0.517, representing 69% of the efficiency of the 200/100 portfolio. It is important to remember that all of the portfolios have the same estimated risk and beta and that they all have the same style, industry and country constraints. The rise in the TC is thus solely attributable to the portfolios increased ability to use information.
What is the short availability?
A common concern related to long/short portfolio management is the availability of short inventory. For larger U.S. securities, such as those in the MSCI World index, there is typically ample inventory available for shorting. In our research, we draw on proprietary inventory data from a leading prime broker to examine the ability to borrow in order to short non-U.S. securities within the index. As all portfolios in Table 2 (where shorting is allowed) assume up to a 3% short position relative to the benchmark weight in any security; these gains in informational efficiency would be spurious if there were very limited ability to actually go short.
An examination of the inventory of non-U.S. MSCI World securities available for shorting shows a surprising breadth of coverage. For example, of the nearly 1,300 non-U.S. securities in the index, short inventory was recently available for all constituents at this prime broker. On average, each non-U.S. security has about 9% of its outstanding free-float shares available for shorting. There is no discernible pattern in the amount of short inventory available as a percent of each securitys free float.
This is not to suggest that the levels of inventory are random, only that there appears to be no relation to the size of the security. All else being equal, a small non-U.S. company appears equally likely to have 9% of its outstanding shares available for shorting as a large company. It is possible that the short inventory available for each security might be related to the streets consensus forecast for the security. That is, securities for which the forecast is highly negative are already lent, thereby reducing the available inventory.
In order to be confident that the $100 million portfolios in Table 2 can actually be realized in the marketplace, we must make sure the inventory available to be borrowed does not constrain active positions beyond the constraints already in place during the portfolio construction stage. We assume that we can access 1% of the prime brokers inventory and then examine the ability to meet the maximum 3% short position. We find that for most securities, a 3% short position is attainable. For some small benchmark holdings there is insufficient short inventory if we constrain ourselves to 1% of the prime brokers inventory, but this might be an overly conservative allocation of the inventory. In general, the depth of the short inventory data gives confidence that the gain in informational efficiency from a modest relaxation of the long-only constraint presented in Table 2 is valid. It should also be noted that there are many nuances associated with borrowing a security to sell short. Here we rely on inventory data from a single prime broker. In reality, there are more ways to borrow to sell short than are demonstrated by this example.
In conclusion, managers need not use an inordinate amount of leverage to gain a great deal in terms of efficiency within a global equity portfolio. And there is more depth in the global short inventory than is probably generally realized.
Brendan O. Bradley is a senior vice president at Acadian Asset Management Inc., Boston.