Author: rgearyiii

Subject: Russell 2000 Not Torpedoed

Date: 2/19/2014

Recommendations: 39

I`m sure Wall Street is breathing a deep sigh of relief as it wakes up to this thread title this morning, and I wouldn`t be surprised if the Superior Six comes a few days early as a result of this good news. (While I`m obviously kidding, it`s always a relief when the GTR1 backtester gives reasonable results for well-defined benchmarks because having its own torpedoes come back to sink it would be awfully embarrassing.) This post started out as a reply to BarryDTO`s most recent thread (whose main points I intend to address in that thread itself later), but I decided to put this in its own thread to make it easily located GTR1 reference material. In his thread, BarryDTO mentioned that he tried to create an ETF proxy for IWM (the iShares ETF tracking the Russell 2000) but that it didn`t appear to track the index or the ETF very well. I agree that the match between what he posted and IWM wasn`t great, but with some work, there`s room for improvement. Most of what you need to know about the Russell 2000 index can be found here: https://www.russell.com/indexes/americas/indexes/us-construc... The key features of the index I have attempted to replicate in my backtest are the following: Characteristics in Common with the Russell 2000 1. Preferred stock, structured products, exchange-traded notes, exchange-traded debt, warrants, rights, and OTC pink sheet and bulletin board stocks are all excluded by Russell, and are excluded by definition from the GTR1 universe. 2. Closed-End Funds, Exchange-Traded Funds and Unit Trusts are excluded with styp.a. 3. Russell includes REITs as long as they don`t distribute unrelated business income (a tax nightmare for otherwise tax-exempt investment entities); I simply include all REITs. 4. "Foreign" companies are excluded. Russell has a complicated procedure for determining whether a company is "foreign" or not (based on country of incorporation, country of headquarters and global distribution of assets), but I simply treat a company as "foreign" if it is incorporated outside the US, as indicated in styp.a, which is the only criteria I have back to 1974 (SI Pro gives country of headquarters, but only from 1997). Russell automatically excludes all ADRs, as do I using styp.a. 5. Existing stocks may only enter the Russell 2000 at its annual reconstitution toward the end of June, though IPOs are added to the index if they qualify when they begin trading. I have created a complicated formula called "OPEN", which produces a value of 1 if a stock is permitted to enter the portfolio and 0 otherwise. OPEN is only equal to 1 in three cases: a) In January, 1974 (the beginning of the backtest). b) On the 17th market day of June (approximating the "end of June" reconstitution date). c) Within 5 market days of a new security opening for purchase. Stocks that make it into the portfolio are held as long as OPEN equals zero, meaning they will be held at least until the next annual reconstitution. 6. Market capitalization and price for determining inclusion in the Russell 2000 are based on the last market day of May; I have therefore lagged these measurements by 18 market days (which turns into a lag of 17 after the lag adjustments are applied) so that they are from the end of May when OPEN is 1 at the approximate annual reconstitution date. 7. My complicated formulas for Market Capitalization attempt to calculate a meaningful value using whatever data might be available, with preference for SI Pro, then Compustat and then actual security-specific (as opposed to company) shares outstanding. 8. Like Russell, I require a (lagged) price of at least $1.00 per share and (lagged) market capitalization of at least $30 million for a stock to enter the portfolio and be retained at reconstitution. 9. Like Russell, I require rank by (lagged) market capitalization to be between 1001 and 3000 for a stock to enter the portfolio. 10. I have set both price lag and file lag to -1 to ensure a final lag of zero in market capitalization used for weighting. If the lag were anything other than zero, then daily rebalancing would result in constant whipsaw trading from daily price changes. Differences from the Russell 20000 My backtest differs from the Russell 2000 index in the following ways: 1. The Russell indexes keep stocks that delist to OTC bulletin board or pink sheet trading until annual reconstitution. The GTR1 backtester, by contrast, assumes that delisted stocks are liquidated at the earliest possible opportunity. 2. Russell defines "available shares" for calculating market capitalization, which can differ significantly from common shares outstanding and even float. I do not have the data necessary for replicating these figures, so I simply use common shares outstanding. 3. Russell keeps "available shares" fixed throughout the year, unless there is a significant change in shares outstanding. I, on the other hand, let shares outstanding update throughout the year. 4. Russell applies some "hysteresis" at the market capitalization boundaries of the Russell 2000 in order to cut down on turnover. I haven`t bothered with trying to replicate their complicated procedure, not only because it would be a lot more work, but I suspect that it wouldn`t improve the accuracy of my backtest by much. By trying to hold stocks that are retained by the index even though they wouldn`t qualify as new entries, I`d probably end up holding some that never belonged in the first place and displace others that do belong. Instead, I`ve focused on trying to capture all stocks that qualify for entry into the the index. 5. The criteria for allowing IPOs into the index mid-year are quite complicated. I have simply included all stocks that are new to the GTR1 universe as a whole if they meet the annual entry requirements (with updated, but lagged price and market capitalization) within five market days of their initial trading. Note that this includes many spin-offs or promotions from OTC trading, neither of which constitute "IPOs" for Russell`s purposes. gIWM vs IWM Here are the results of "gIWM" versus the real IWM, a Russell 2000-tracking ETF from iShares: 20000526 to 20140218 (1) (2) gIWM IWM Diff CAGR: 8.65 8.37 0.28 TR: 210.09 199.43 10.66 GSD(20): 25.86 25.29 0.57 DD(20): 18.11 17.82 0.29 MDD: -59.76 -58.86 -0.90 UI(20): 16.17 15.72 0.45 Sharpe(20): 0.40 0.39 0.01 Beta(20): 1.20 1.17 0.02 TI(20): 7.65 7.50 0.15 AT: 0.39 N/A N/A Total Return for year ending: 20001229 2.86 5.61 -2.75 20011231 3.62 1.80 1.83 20021231 -19.88 -20.37 0.49 20031231 48.62 47.89 0.73 20041231 17.43 18.06 -0.63 20051230 4.72 4.47 0.25 20061229 16.90 18.27 -1.37 20071231 -3.18 -1.79 -1.40 20081231 -34.37 -34.49 0.13 20091231 33.18 28.47 4.71 20101231 27.90 26.93 0.97 20111230 -5.77 -4.43 -1.35 20121231 18.05 16.70 1.36 20131231 41.20 38.70 2.50 20140218 -0.03 -0.10 0.07 (1) http://gtr1.backtest.org/2012/?s20000526lf-1lp-1h1::MktCapWt... (2) http://gtr1.backtest.org/2012/?IWM It appears that gIWM is not too bad of a proxy for IWM. gIWM vs ^RUT Keeping in mind that gIWM includes dividends while ^RUT does not, here is a comparison of the two back to the index`s inception: 19870910 to 20140218 (3) (4) gIWM ^RUT Diff CAGR: 9.66 7.59 2.07 TR: 1035.47 587.39 448.08 GSD(20): 23.14 23.25 -0.11 DD(20): 16.25 16.68 -0.43 MDD: -59.76 -59.89 0.13 UI(20): 14.85 15.42 -0.56 Sharpe(20): 0.39 0.30 0.09 Beta(20): 1.09 1.09 -0.01 TI(20): 7.32 5.54 1.78 AT: 0.42 N/A N/A Total Return for year ending: 19871231 -26.63 -28.73 2.10 19881230 24.04 22.38 1.66 19891229 17.30 14.21 3.09 19901231 -21.05 -21.45 0.40 19911231 47.20 43.68 3.53 19921231 18.80 16.36 2.44 19931231 18.54 17.00 1.54 19941230 -1.89 -3.18 1.29 19951229 28.45 26.21 2.25 19961231 16.47 14.76 1.71 19971231 22.71 20.52 2.19 19981231 -1.00 -3.45 2.45 19991231 29.79 19.62 10.17 20001229 -6.87 -4.20 -2.66 20011231 3.62 1.03 2.59 20021231 -19.88 -21.58 1.70 20031231 48.62 45.37 3.25 20041231 17.43 17.00 0.43 20051230 4.72 3.32 1.39 20061229 16.90 17.00 -0.10 20071231 -3.18 -2.75 -0.44 20081231 -34.37 -34.80 0.43 20091231 33.18 25.22 7.96 20101231 27.90 25.31 2.59 20111230 -5.77 -5.45 -0.32 20121231 18.05 14.63 3.42 20131231 41.20 37.00 4.20 20140218 -0.03 -0.19 0.16 (3) http://gtr1.backtest.org/2012/?s19870910lf-1lp-1h1::MktCapWt... (4) http://gtr1.backtest.org/2012/?!RUT Robbie Geary