I was wondering if anyone here looks to avoid certain sectors all together. I know some folks have mentioned avoiding financials (as some metrics look very different in this sector). What about sectors like energy, materials, or comm services. When I backtest on the stocks in these sectors individually, I see no good reason to consider them at all. Any thoughts?
One option is to create a custom formula that assigns different numbers to different sectors and then use that in a ranking system. You can also do this with subsectors or industries using GICS codes. You do this with a series of EVAL commands. For example, the following formula will rank sectors in the following order: Health Care, Industrials, Staples, Discretionary, Technology, Energy, Financials, Communications, Utilities, Materials, Real Estate. Note that I’m NOT suggesting anyone rank sectors in this order!
Eval(GICS(35)=1,10,0)+Eval(GICS(20)=1,9,0)+Eval(GICS(30)=1,8,0)+Eval(GICS(25)=1,7,0)+Eval(GICS(45)=1,6,0)+Eval(GICS(10)=1,5,0)+Eval(GICS(40)=1,4,0)+Eval(GICS(50)=1,3,0)+Eval(GICS(55)=1,2,0)+Eval(GICS(15)=1,1,0)
Here’s what I do. I test my ranking system on each sector or industry group and see how much alpha it adds over a random selection. I then favor those industry groups that my ranking system really works on and not those that seem relatively unaffected by it.
But I would advise you not to do a lot of work on this now. Once we switch to FactSet in Q2 2020, industries and subindustries will look very different on a component basis than they do now. I will be posting something on the forum about this shortly, but we’ll be replacing GICS with RBICS, which is a very different (and, in my opinion, better) classification system.
Thanks Yuval. I guess I was trying to open up a more philosophical discussion around the possibility that some sectors should just be avoided. They are either too volatile or too linked to the randomness of commodity prices, etc.
Yes, I exclude financials, bc. I trade small caps, there are so many regional banks in that universe, that I often found myself having 50-70%
Financials in my port. Also if you test the financials they do not give as much alpha as tech etc.
I also exlude china stocks, bc. they are usually rip offs and they do not backtest as well too!
Regards
Andras
Charles - You are asking a good question. I think there is an argument for not mixing sectors/industries that are immature versus those that are basically commodities. Energy, materials, semiconductors (to a certain extent) are commodities and basically cyclical industries. Software and healthcare are not. So you generally consider each sector or industry in isolation and devise your system accordingly. This is essentially what I do. For industries like cloud computing, it makes sense to have factors such as 3-Yr or 5-Yr composite sales growth. It would not make sense to apply such a factor to the energy industry. By the time a company reaches the top of the 5-Yr growth ranking, the cycle is over making the company the worst pick.
Also, within sectors and industries, there are certain characteristics that should be considered separately. For example, you may want to treat the health care services industry separately from health care equipment and supplies.
SteveA
I exclude banks, airlines and reits
I exclude financials and reits from my base universe for reasons you mention. I do have exposure to financials via vehicles like BRKB and do hold other financial stocks - but it’s outside the universe I use p123 for. I tried to model financials and just haven’t been happy with results. I have a background in the insurance industry and when I invest in insurance I still utilize traditional fundamental approaches and domain specific knowledge. I haven’t found financial quant models that I like very well.
I do have a separate model for utilities that I allocate to. It backtests well, but out-of-sample over the past 15 months of live usage I can’t tell if it’s any better than just using a broad sector etf like XLU.
I guess philosophically I like the idea of allocating more to defensive sectors like Utilities, REITS, Staples and maybe Healthcare when the macro is pointing that direction, but have not approached it systematically. I have a separate macro model I track in excel to help inform me on this, but I’m still learning. Nothing is in stone. I’ve used preferreds, utilities, and bond and reit etfs at various points to express a degree of defensiveness as the data changes, but it’s not part of a cohesive system. In practice the defensives have helped buffer equity volatility in slowing growth, falling rate environments.
Energy (mostly oil) is a tough sector for me to understand, so I tend to stay away from that. When I read analyses of companies in the sector it seems investors rely on considerable knowledge and data not found in historical fundamentals. Of the models I use, none work very well in energy stocks, so I feel at a distinct disadvantage. My history in oil stocks is also pretty bad, so I know I’m not seeing the right things when I look. It’s probably best I just use an ETF if I want exposure there.