Active Share

Would it be possible to get “active share” data - i.e., the extent to which our portfolio deviates from the benchmark? This would be helpful for most Books/Sims and even for Portfolios for some users.

Thx,
Ed

Active Return is in the Quick Stats to the right of the chart on the Summary page. Is that what you’re looking for?

Not Active Return but Active Share, the extent (in %) to which a portfolio’s holdings deviates from the benchmark/universe. It’s a metric increasingly disclosed by active managers, funds, etc.

Ah. So if the S&P 500 holds 2% IBM and your portfolio holds 1.5% then .5%?

As far as I know, we don’t get that data (percentage weightings, that is). It’s a huge reason why ETFs are cordoned off: If we had the holdings and the percentages, then you’d be able to do a lot more fundamental stuff with ETFs. As it is, the results are just kind of weird if you don’t rope them off from stocks.

The data is useful for analytics. Give us some perspective how much of active return and risk is due to active share, among other reasons.

<< If we had the holdings and the percentages, then you’d be able to do a lot more fundamental stuff with ETFs>>
I would love to screen/backtest ETFs based on their holdings-weighted stats and stuff.

Aren’t the holding for every ETF disclosed every day?

Maybe a great data partnership for P123: https://activeshare.info/

…may be value added for both the providers and distributor.

Unfortunately that website only has data for mutual funds. Actually I think this feature is an easy add-on: doesn’t P123 already have the components for each of the benchmarks available? If so the Active Share data is just a simple comparison search and calculation.

It might incur an extra cost for P123 to access index holdings and weights.

And I’m not sure if active share would be insightful for P123 strategies, as I’d be very surprised if any ports are in the “closet indexing” range, e.g., active share less than 50%.

I suspect that all P123 portfolios deviate significantly from any benchmark, given that most users seem to focus on micro caps with highly concentrated portfolios. What’s the difference between active share of 90% and 95%? Would the 95% active share model imply higher expected returns?

We get the point-in-time components of the S&P as part of our contract, as far as I know as a simple list. To get a usable database of the point-in-time holdings of indices and ETFs is an entirely separate data contract.

Sorry, we just don’t have the data at this time. The interest is noted for future discussions with S&P, though.

Sounds good, would be nice to see that data at some point. And no worries about ETF’s, “active share” is compared to indices/benchmarks, so if you have it for the S&P or Russell indices, etc., that would be helpful.

Accessing additional index data (e.g., Russell) typically incurs a cost, while most ETF data is publicly available.

I still think active share may not be as useful for P123 ports or books. How would one evaluate strategies that employ market timing or hedging, where active share could be either 100% or 0% at any given time?

Instead, I’d suggest using tracking error to evaluate risk / deviation versus a benchmark. Tracking error only requires performance data, as opposed to holdings, so it should be far easier to calculate.

Active share is literally what we do. Active management is building investment strategies that differ from the benchmark. Otherwise, we would be passive investors tracking some index.

Active share varies depending on the chosen benchmark holdings and your portfolio holdings. And while it’s true that you can’t beat the benchmark without active share, you can’t trail the benchmark without it either.

And so active share is really not predictive of outperformance. It’s simply predictive of different performance (than the benchmark).

Alpha is a good measure of outperformance.

Information ratio is the best measure of outperformance. See Grinold & Khan’s treatise on Quantitative Active Investment Management.

Active share is not binary. Share of, say, 30%, 60% or 90% may be very instructive in both model construction and out of sample performance Furthermore, as for cutting edge research, the ongoing debate between Cliff Asness of AQR and Rob Arnott of Research Affiliates about the distinction between factor-based investing and smart beta shed some important light on the subject of Active Share.

As an aside, I like Information Ratio but one needs to work within its limitations. With Expected Return as a key variable in Information Ratio, it necessarily involves assumptions, something that doesn’t afflict Active Share.

Most here build ports with 5-25 holdings with benchmarks holding 500 to 3000 stocks or more. Active share is going to be huge on all of those ports, and not very instructive.

Literally, if you are picking 10 stocks out of the S&P 500, you are taking the position (based on your information) that those 10 will outperform the benchmark as a group, and the remaining 490 will underperform the benchmark as a group.

The best way to judge that is the information ratio, which is not based on assumptions but rather historical excess return divided by the standard deviation of historical excess return.

Here is one definition of active share which is available from P123 data holdings:

In this definition, every portfolio objectively defines its own benchmark. Under this construct, active mangement value is added when holdings’ weights differ from the relative weighting’s within a standard float adjusted index.

It’s not perfect, because there is active management value added by deciding what to hold in addition to how much. Also, an equal weighted portfolio or inverse weighted portfolio would be assigned a high amount of active share, although there is clearly not much value added in either of these scenarios. In spite of the drawbacks, this information is tractable and — I would argue — very illuminating, especially when used against diversified holdings.

Also, I would argue that P123 can easily reverse engineer the weights of major indices. The following Port Sim replicates the S&P 500 weights and returns with very little tracking error versus the SPY: https://www.portfolio123.com/port_summary.jsp?portid=1517907

Hi all. Hi P123 and team.

I would like to revisit this topic. Even though probably most portfolios have very high active shares, I would like to also have the precise Active Share number for portfolios and books.

Is there any way you could implement it?