Can we really learn from the past?

To form an investment strategy, it helps to have some theory of how financial markets work. We need some guiding principles to help establish which choices are likely to help us shelter or grow our wealth, and which are likely to diminish it. A good first step would be to look at what happened in the past and attempt to draw out insights that could help us navigate the future.

That’s why investors study past market events and try to reconstruct what happened. For example, keen students of financial history will examine the circumstances leading up to past market crashes and look for similar signs today. These red flags could then be used to estimate the likelihood of it happening again, and investors might be able to position their investments accordingly. But unfortunately, past performance is not a guide to the future. History doesn’t leave neat lessons lying around.

This article isn’t personal advice. Remember, investments and any income they produce can fall as well as rise in value, you could get back less than you invest. If you’re not sure if an investment is right for you, then you should ask for financial advice.

Learning from data

Computers and modern statistical techniques allow us to learn from history in new ways that were unavailable to previous generations. We can now take an unimaginably large number of measurements and have the processing power to meaningfully analyse them.

Sometimes this can look mundane. For example, did you know that over the last ten years the US stock market only rose on 53% of trading days? Despite only going up around half the time, the US market delivered a 269% return including dividends over that period. Source: Refinitiv Eikon 29 March 2021.

But other insights can be more surprising – and potentially more useful. With the advent of huge datasets, academics began noticing strange patterns in stock market performance. Patterns that challenged existing theories of how financial markets worked.

Patterns in the data

Researchers noticed that smaller companies tended to deliver superior returns to large companies, and companies trading on low valuation multiples, like price-to-earnings or price-to-book ratios, tended to outperform their pricier counterparts.

These findings were surprising because they appeared to offer a direct challenge to some academic views on market efficiency at the time. If it was possible to form a winning portfolio by following simple rules, why didn’t savvy investors simply buy up all the cheap, small stocks until the possibility of outperformance disappeared?

This mystery is still debated, but characteristics that have historically forecast superior stock market returns are called ‘factors’. A vast academic literature has grown up and lots of new factors have been discovered – or so their proponents claim. These include:

• Value – stocks trading on low valuation multiples have tended to outperform those on high multiples
• Size – while higher risk, smaller companies have tended to outperform big ones
• Momentum – stocks that have performed well have tended to continue to do well, and stocks that have done badly have tended to remain laggards
• Profitability – stocks that make high returns on investment have tended to outperform stocks which don’t

The list of proposed factors is constantly growing as data analysis improves.

Some providers, including Exchange Traded Fund (ETF) providers, now offer factor index funds – basically funds that use a rules-based approach to try and beat the market. The big hope is these funds can offer the same potential for outperformance as traditional stock pickers, while charging lower fees. It’s important to remember that these funds typically rely on wide diversification.

And the real-world results… well, we’ll come to those.

Beating the street

Data has since thrown up other, more insidious, challenges. Investors have always been drawn to the idea of beating the market, and those who do so successfully are often heralded as geniuses – and usually charge lucrative fees for their services. But data analytics and factor analysis have thrown the idea of investing ‘skill’ into question.

Logically, we can see that not everyone can beat the market. As the saying goes, you’re not stuck in traffic – you are traffic. And as a group, stock pickers can’t beat the market – they are the market. If you’re buying low and selling high some other poor soul must be selling low and buying high. Add in trading costs and charges and it’s clear that stock pickers must collectively underperform the market.

Nonetheless, some investors do beat the market – and stock pickers can point to very successful investors as apparent proof that skill and due diligence can pay off. But how can we distinguish luck from skill? Legendary stock picker Warren Buffett explained the problem this way:

Imagine we asked every adult in the US (then about 225m people) to wager one dollar on a coin flipping contest. Each person tosses a coin each morning and we eliminate anyone that flipped tails. After a day, half the group gets eliminated, and the remaining flippers have doubled their money. The odds of tossing ten heads in a row is less than 0.1%, and yet about 220,000 people will achieve this feat. After another ten days, the odds of getting heads 20 times in a row is below 0.0001%, but 215 will have done it.

Buffett picks up the story: “By then, this group will really lose their heads. They will probably write books on “How I turned a Dollar into a Million in Twenty Days Working Thirty Seconds a Morning”. Worse yet, they’ll probably start jetting around the country attending seminars on efficient coin-flipping and tackling skeptical professors with, “If it can’t be done, why are there 215 of us?”.

So, are stock markets really coin flipping contests? Are successful investors simply deluding themselves (and their clients) by asserting that their special insight and skill caused them to beat the market? Or do superstar money managers really have the Midas Touch?

This is a tricky problem to disentangle. Fortunately, some very clever cookies dreamt up an ingenious way to use data to offer a solution to the coin flipping problem posed above.

Nobel Prize winners Eugene Fama and Kenneth French (incidentally the same two that really got factor analysis going) published a paper called “Luck Versus Skill in the Cross Section of Mutual Fund Returns” in 2009. The authors tried to find evidence of skill in large datasets of fund performance.

They simulated lots of imaginary funds to show what the distribution of returns would be if there were no skill. Some will be very successful purely by chance, and some will do poorly – again purely by chance. Then they compared this skill-less distribution to actual fund managers investing in the US.

If stock picking skill really exists, then according to Fama and French “the worst performing funds should perform worse than we expect just by chance… and the best performing funds should perform better than we expect by chance.”

The answer is, well, complicated.

If you exclude charges, then Fama and French found some evidence for skill – some fund managers really do seem to add value. But once these American stock picking superstars have charged their fees, investors are left no better off than they would be by chance.

It’s important to remember that we’re just looking at the US here. The US market is one of the most widely covered in the world, making it harder to outperform. Other markets with fewer analysts covering them could offer more scope for outperformance.

And those factor models?

The graph below shows Fama and French’s value factor among US stocks. When the line is going up value is outperforming growth, and when the line is going down growth is outperforming value.

Fama & French value factor

Past performance isn't a guide to the future. Source: Fama & French datasets, HML US Equities, 28/02/2021.

Fama and French published their famous paper in the 1990s. If you look at the graph above you can see that value was doing really well up until that point, albeit with some hiccups along the way.

And then the Dot Com Bubble began to inflate in the late 1990s, and you can see that sharp drop in the graph as value badly underperformed growth. If you’d read the Fama & French paper and mechanically loaded up on value stocks you’d have felt pretty silly.

But eventually history reasserted itself, and when the bubble burst value investors felt vindicated and the graph leapt to new highs.

And then, following the Financial Crisis, and especially in the last few years, value did very poorly.

What should investors make of this?

Maybe value is about to make a spectacular comeback and investors should keep the faith. After all, value has had bad periods of performance before. Or maybe the world is just different now, and value stocks offer no potential outperformance. Or perhaps Fama and French were seeing patterns that didn’t exist, and value’s past decades of success were just more coin flipping.

What can we learn?

Bringing it all together, our findings point to a fundamental limit on our ability to learn from the past, especially through data. Sometimes this time really is different, and what worked before might not anymore. The future will always bring surprises, and no amount of data will change that.

We’re not saying that it’s impossible to get better returns than the wider market and that we should pay less attention to active stock picking funds. While it might be more tricky in the US, that’s just one market. There are lots of other markets, with funds investing in them that could outperform, even after charges.

All investors can do is stay humble and recognise that unpredictable things will happen. Whatever your investing style, the key is always stay properly diversified and invest for as long as you can.