Hello, welcome back to Papers with Backtest podcast. Today we dive into another Algo trading research paper. Hi there. Yeah, today we're tackling a really interesting one from 2016. That's right. It's called Timing Smart Beta Strategies. Of course, buy low, sell high. It's by Rob Arnott, Noah Beck, and Vitaly Kelesnik from Research Affiliates. And it really digs into something I think a lot of us wonder about.
Can you actually improve your results by, you know, actively timing smart... beta or factor tilts. Exactly. Is it really feasible to like buy these things low and sell them high? That's the core question they're exploring. And they look at quite a few things, don't they? Like eight different smart beta strategies. Yeah. Things like fundamental index, equal weight, low vol, quality dividend weight. A pretty good mix. And also eight factors, value, momentum, size. Right.
And they look at factors as long, short. portfolios, whereas the smart beta stuff is long only compared to a cap weighted benchmark. That's an important distinction. Okay. And the main argument, I guess, is that you can potentially do better by leaning into factors or strategies that look cheap compared to their own history. That's pretty much the central idea. Yeah. Emphasize the cheaper ones, dial back the expensive ones. Okay. So let's get into the training rules and results.
The paper starts by warning about just chasing performance, right? It mentions Revaluation alpha versus structural alpha. Uh huh. That's a key concept here. Revaluation alpha is basically gains you get just because the market decides to pay a higher price, a higher multiple for the same thing. It's often not sticky. You know, it can reverse. So it's like the price goes up, but the underlying substance hasn't necessarily changed that much. Exactly.
Whereas structural alpha is the outperformance that's left over after you account for those valuation changes. It's hopefully more about the inherent. quality or characteristic of the strategy or factor itself. That makes sense. So just because something did well recently doesn't mean it'll keep going. It might just have gotten expensive. Precisely. The people warns this is a big trap in smart beta and factors, just like anywhere else in investing.
They give examples like value looking pricey back in 77 or size near its peak popularity in 81. And their point is a lot of people are timing these things already, but maybe badly. Buying after a good run, selling after a bad one. Yeah, effectively buying high and selling low. The classic mistake. So if chasing winners is dangerous, what's the alternative they propose? Well, they focus on relative valuation.
The idea is that how expensive or cheap a strategy is compared to its own history can give you clues about its future performance. Okay, so not just cheap in absolute terms, but cheap relative to its own normal range. Exactly. They find that this relative valuation seems to be a useful signal for timing. Strategies or factors that are historically cheap tend to do better going forward and vice versa. Interesting. So it's a contrarian signal fundamentally. It is. But there's a caveat.
They do warn that if you get too aggressive, like betting everything on just the single cheapest thing, you might hurt your diversification and your risk adjusted returns, your Sharpie ratio. So moderation is probably key then. Maybe tilting rather than making huge all or nothing bets. That seems to be the implication. Yeah. Moderate tilts. based on valuation. Okay, let's talk about how they tested this. They compared trend chasing versus a contrarian approach using some hypothetical portfolios.
Right. First, they set up a baseline, just simple, equally weighted portfolios of the smart beta strategies and the factors. Diversification helps here, lower risk, often better Sharper ratios. Simple but effective sometimes. Then they simulated a trend chaser, someone who invests in the, say, three best performers based on the last year or three years, five years, 10 years of returns. Following the momentum basically? And how did that strategy do? Not well.
Across all those look back periods, the trend chasing strategy actually underperformed the simple equal weight approach, both for smart beta and for factors. Really? Underperformed, even following the winners? Yep. And worse, because you end up concentrating your bets, the risk actually went up, lower Sharpe ratios. So that intuitive ride the winner's idea just didn't seem to work here. Wow. Okay. So what about the opposite? Investing in the losers. The contrarian approach.
Ah, now that's where it gets interesting. They simulated investing in the three worst performers over those same past periods. Yeah. And that contrarian strategy generally beat the trend chaser, often by quite a bit. It frequently even beat the simple equal weight portfolio in terms of raw return. So buying the laggards actually paid off better than buying the leaders. In their tests, yes.
Now, sometimes for factors, the Sharpe-Ray ratio wasn't necessarily better than equal weight because you are less diversified, but the overall return boost was often there. And did it matter which time frame they used for worse performers, like one year versus five years? Interestingly, no. The contrarian approach tended to outperform regardless of whether they looked at the past one, three, five, or ten years. Actually, the three-year laggards often gave the best results going forward.
That's quite a strong finding against performance chasing. It really is, and they even did a Fama-French factor analysis on the difference between the contrarian and trend strategies. What did that show? It suggested the difference had, you know, positive exposure to value, negative exposure to momentum, which kind of makes sense. But even after accounting for those known factors there was still some unexplained alpha left over. Okay, so let's circle back more directly to valuation.
They looked at the link between how expensive something is, its relative valuation, and its later performance. They did. Using a blend of metrics, PE, price to sales, price to dividend, price to book, all relative to the market, they found a pretty clear negative relationship. Negative meaning expensive is bad for future returns. Exactly.
High relative valuation tended to be followed by lower subsequent returns, and low relative valuation, being cheap compared to history, tended to precede higher future returns. And this connects back to performance chasing, right? Winners get expensive, losers get cheap. That's the link they make, yes. Good past performance often drives up valuations, making things expensive, while poor performance can make them cheap, potentially setting them up for a rebound.
So, to really nail this down, they did an out-of-sample test Using these relative valuations without looking ahead. Uh-huh. They essentially simulated making decisions in real time. At each point, they'd look at the available history, figure out which factors or strategies were cheap or expensive relative to their own past, and then form portfolios. How did they compare cheap versus expensive in that test?
They compared investing in the three cheapest versus the three most expensive based on that historical relative valuation measure. Standardized it so they could compare across different factors and strategies. And the results of that out-of-sample test? Pretty compelling. Over the almost 40-year period, the portfolio of the three cheapest smart beta strategies and the portfolio of the three cheapest factors both significantly outperformed the equal weight approach and the cap weighted market.
And the expensive ones? They tended to underperform. It really supports the idea that relative valuation has predictive power, even out of sample. But you mentioned earlier diversification still matters. How does that fit with focusing on just the three cheapest? Good point. They also looked at a strategy they called tilted diversification.
Instead of just the three cheapest, you hold all the strategies or factors, but overweight the cheaper ones and underweight the expensive ones on a sliding scale. And how did that compare? It often produced better risk-adjusted returns, a better sharp ratio than just holding the three cheapest, even if the absolute return was slightly lower. It suggests that while timing based on valuation is useful, you probably don't want to abandon diversification completely. Moderation again.
So finding that balance between the timing signal and keeping diversification benefits. Precisely. Now, the paper does acknowledge something important, data mining. How does that factor in? Yeah, they're upfront about it. Like a lot of investment research, finding factors or strategies often happens after they've already had a good run in the historical data. So they might look amazing in backtests, but... part of that could be because they were identified because they looked amazing. Exactly.
And that good run might have included a period where they became more expensive, that revaluation alpha we talked about. So by the time a factor is discovered or an index is launched, it might already be priced relatively high. Do they show any evidence of this, like performance fading after discovery? They do. They present evidence of what they call phantom alpha, comparing the excess returns of these strategies and factors before for their index launch or academic publication versus after.
And returns dropped off. Generally, yes. Returns tended to be lower post-launch or post-publication. Could be data mining bias, could be that the effect gets arbitraged away, or it could be that the earlier returns included unsustainable valuation gains. That's a really crucial point for anyone looking at back-tested strategies. Absolutely. Always question how much of the past performance is real structural alpha versus just luck. or revaluation. Did they look internationally at all?
Does this valuation timing work outside the U.S.? They did briefly look at developed ex-U.S. markets. The results were maybe a bit weaker. The difference between cheapest and most expensive wasn't as dramatic as in the U.S. Any thoughts on why? They suggest maybe the shorter data history internationally makes it harder to reliably judge historical valuation norms.
Also, post-financial crisis, there was a big flight to safety, which might have pushed up valuations for certain less risky factors globally. potentially distorting things for a while. But the general tendency was still there. Contrarian beating trend. Generally, yes. The cheapest still tended to outperform the most expensive, just maybe not by as much as in the U.S. data. Okay, so pulling it all together, what's the key takeaway trading rule from this paper? I think it boils down to this.
Be a contrarian based on valuation. Don't just chase past performance periodically, maybe annually. Look at your smart beta strategies, your factor exposures. See which ones are trading cheap. relative to their own history. And lean towards those. Yeah. Overweight the cheap ones, underweight or avoid the expensive ones. It suggests a systematic way to potentially add value beyond just holding a static mix.
So timing is possible, the paper argues, but it's about buying low based on historical norms, not buying high based on recent returns. That captures it perfectly. Consider the price you're paying relative to history. Don't just extrapolate recent alpha blindly. This has been a really insightful deep dive. Lots to think about for implementing factor and smart beta strategies. Definitely. Thinking about valuation as a whole other dimension beyond just the factor definitions themselves.
Thank you for tuning in to Papers with Backtest podcast. We hope today's episode gave you useful insights. Join us next time as we break down more research. And for more papers and backtests, find us at hgtps.paperswithbacktest.com. Happy trading.