Questioning the Results

Hi everyone,

I’m wondering if one of you can help explain this…

I set up a screen today, just seeing what would happen if I input zero rules. In other words, what if I just bought the stocks in index and kept rebalancing:

Settings–> As of today (July 13, 2016), Universe = S&P 500, Benchmark = S&P 500. No maximum number of stocks, no ranking system, no factors. Rebalance once a year and slippage at 0.25%

As expected, the portfolio held all 500 stocks in the S&P throughout the backtest. The weird thing is the alpha…annualized at 3.41% over the last ten years?? This would suggest that just holding the same stocks in the S&P 500, and rebalancing once a year, you would’ve outperformed over the last 10 years by 3.41%. If you use the equal weight S&P 500 as the benchmark (to account for the equal weight portfolio in the screen) the alpha is even higher??

I find these results hard to believe. Sure there are advantages to rebalancing over the long-run, but it can’t be that easy to outperform the index?

Any insight into this (or a confirmation that it is indeed very easy to beat the market through disciplined annual rebalancing) would be much appreciated.

I believe there are two differences there. The first is the weighting. Your screen is equal weighting the stocks, but the index isn’t. You can use the S&P 500 Eq Weight as the benchmark to eliminate that.

And then I believe the rest of your difference is attributable to dividends (paid and reinvested), which are included in your screen results but not the benchmark.

dmacdonal9,

Thanks for the reply.

I believe that the equal weighting in backtesting a screen has a lot to do with it. I was reading something today from Vanguard that shows that equal weight strategies can significantly outperform because you’re taking advantage of the small-cap, value risk premiums:

https://personal.vanguard.com/pdf/s287.pdf

Even if you use the S&P 500 equal weight as your benchmark though, the alpha is still unreasonably high.

At the end of the day, it just means that our backtest screen results have to be taken with a huge grain of salt. In the real world, you have to incur transaction costs/slippage in order to maintain an equal weight strategy. Current backtesting only takes into account these costs when stocks come in or out of the portfolio, but not costs incurred with just rebalancing current holdings back to equal weights. Would be interesting to try this in a simulation instead of a screen and see what the results are…

Not sure about the dividends comment, but perhaps you’re right. Maybe someone else can weight in on that…

The available S&P 500 equal weight index is a price return index. The equity curve you’re generating represents total return; price + dividend.

Walter

Current dividend yield on RSP is 1.55%. So between that and using the right benchmark index, alpha probably approaches 0%.