Why is Etf (BOSS) so bad at replicating academic findings?

Research
1

I’m trying to research the returns of Founder led companies: https://anchorcapital.com/founder-led-companies/

I have then looked at the return in BOSS, as well as FIFNX. Which - but with a short history - does not do very well.

But then I have gathered all the shares in a Watchlist: "ADBE, ARE, HES, AFG, ADI, BRK-B, SAM, CPT, COF, GIB, CTAS, CPRT, COMMQ ^ 05, CSGP, CCI, DHI, DHR, DLTR, EQT, EL, EEFT, GRMN, MNST, HOLX, INTU, LEN, LPSN, MAR, MCRI, NBIX, NVDA, PCAR, PENN, PXD, PLUG, REGN, RMD, RH, SGEN, SPG, LUV, STLD, TOL, UNH, OLED, X, WCN, NFLX, BRKR, TGTX, CRM, MKTX, MORN, TDG, BLK, BX, MASI, MELI, BKI, FRC, Z, MN, CPRI, EPAM, FB, TCEHY, NOW, WDAY, RNG, VEEV, WIX, PCTY, MC, ZEN, JD, BABA, W, HUBS, FFWM, ASND, SEDG, GDDY, BPMC, PEN, SQ, TEAM, TWLO, FTV, TTD, BL, SNAP, ARGX, MDB, SE, ZG, SPOT, INSP, TENB, PDD, FTCH, ARCE, NFE, LIFT, ZM, BYND, UBER, RVLV, CRWD, REAL, PHR, TXG, NET, DDOG, BRP, BDTX, DKNG, FOUR, RPRX, PSTX, GOCO, NCNO, XPEV, SNOW, LSF, PRLD, VNT, SEER, DASH, ABNB "

It has done surprisingly well in the shorter and longer period. This also applies if I keep IT out, low volume, and set transaction costs, or have a small or large portfolio.

Has anyone done any research on this phenomenon, and is it possible to achieve excess returns? Or is this another strategy that works on paper but does not work well in the real world?

My idea was to use cloning strategies on either BOSS and FIFNX or both together:

https://quantpedia.com/strategies/alpha-cloning-following-13f-fillings/
https://mebfaber.com/2014/12/31/cloning-the-l...bridgewaters-all-weather/
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2411910
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=1566794
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2741129
https://citeseerx.ist.psu.edu/viewdoc/downloa...amp;rep=rep1&type=pdf
https://www.researchgate.net/publication/3209...ormation_to_all_investors

2

Your list represents current holdings. You can’t use those for a backtest because these holdings are tainted by survivorship bias. The fund manager typically buys stocks that have performed well in the past.

What you need to have are the historical holdings at 3 months intervals, as a minimum. Then you can enter the stock symbols and their dates as Imported Stock Factors into P123 and run a backtest on the Stock Factor data. That is the only way to get a realistic backtest result.

3

Thank you Georg, I have tried to look around a bit for the historical portfolios, but I can not find any. Where is it possible to find such information? I see that even on their own side it is not possible to find the older share position: https://www.globalxetfs.com/funds/boss/

4

What Georg says is absolutely true, but I have had some pretty good success with backtesting on current holdings. You just have to remember that the further back the simulation goes, the more suspect the results are. You want to keep any backtest simple (few rules), and just try to see what ranking factors perform the best. I did this for Inspector Sector Cloud Computing which uses the SKYY constituents as the base universe to choose from.
I optimized using current constituents with a 5-year backtest.

Using the BOSS constituents that you listed above, three-month rebalancing, and my favorite growth ranking system, I got some pretty good results on a 10-stock 1-year backtest.

Don’t let a lack of historical data stop you from creating portfolios! At least you can set them up and watch them for a while to see how they perform out-of-sample.

BOSS.gif
BOSS.gif662×674 46.2 KB

5

Thank you for your feedback. Yes, the strategy seems to give good results, and then I always get a little extra skeptical. I am therefore doing some extra research.

Anyway, does anyone have any idea where I can find the historical positions of funds or ETF?

Otherwise thank you for taking the time to test the strategy :slight_smile:

6

Whycliffes - thank you for bringing the founder’s ETF to my attention. I wasn’t aware of it. There are services that provide the historical constituents of ETFs and I think P123 was looking into it at one point in time. Just be aware that backtesting is only a tool and some people put too much reliance on it for predicting future results.

7

I see that in this thread, I forgot to ask if it’s possible to screen for founder-led companies?

I see that there are several insider “nodes”, but none specifically indicate “who” is making the insider purchase or has ownership.

From what I see in this study, it seem that they got the Founder Led data from Factset: https://clearingcustody.fidelity.com/app/proxy/content?literatureURL=/9893955.PDF

List of insider function that I found:

image
image753×644 15.8 KB

8

There might be some info in the insider transaction details data that we have yet to process. But I doubt I saw a “founder” label. Just your typical titles, CEO, etc.

For the most part founders are around when company is still small. Maybe “founder led” is synonymous with “small size” ? Perhaps that’s where the alpha is coming from?

In any event, buying ETF constituents data is on our to-do list so we would get this data “for free” using BOSS ETF . That would be the first step at least.

9

Thank you. Yes, I hope it will be possible to screen for Founder-led companies. It seems to be a very interesting indicator based on the studies I have looked at. Some of them are referenced in the article above.

Regarding your second question, yes, I have tried to look into whether the size effect has been significant, but if you look at the BOSS ETF, which is mostly equally weighted Large Cap, it seems that the effect applies more broadly: Solactive | Indices

When I look at the holdings in (allcap) FNDRX - FRC Founders Index Fund - Portfolio Holdings, AUM (13F, 13G), the numbers from this fund also suggest that the effect applies to more than just small size: https://www.businesswire.com/news/home/20191007005812/en/First-Republic-Launches-First-Republic-Founders-Index℠ (Look at the creator of the index: First Republic Bank (NYSE:FRC) :))

So if it is possible to retrieve the Founder-led indicator (numbers) from Factset, that would be great.