Lessons for investors from the 1800s
New data suggests that factors such as value, momentum and low beta have a long history of success
The hunt for ways to beat the market means that the investment industry has an enormous appetite for data on how different types of stocks have performed over time. The problem is that the data we have is more limited than you might expect. It’s quite decent for US stocks back to the 1920s, for example, because in the aftermath of the crash of 1929 and the Great Depression, American researchers began collating more financial and economic information. There are also long-term stock prices for many other countries, but there’s a shortage of long-term fundamental data.
This is an issue when you want to know whether a pattern of returns you have found only holds true for the limited amount of data you are working with (known as “in sample” in statistics), or whether it tends to occur across different markets and across time (“out of sample”). Findings that are only tested in sample will often be misleading – they may lead you to make wrong forecasts if you try to apply them more widely. Results that can be robustly replicated out of sample and in very different environments can be trusted much more as the basis for an investment strategy.
Reconstructing the past
So anybody interested in market history will be pleased by a new study from researchers at Dutch asset manager Robeco and Erasmus University at Rotterdam, who have put together data on US stocks from 1866 (when the country was just emerging from civil war) to 1926. This involved a great deal of work scanning old newspapers and other records to build up a database of 1,488 individual stocks (which explains why this kind of historical work is rare). The information available is limited – there are no earnings – but includes prices, market capitalisation and dividends.
This allowed the researchers to test whether factors (see below) such as yield, size and momentum worked in this very old, very different out-of-sample data in the same way that they have in more recent history. Their results suggest that yield and momentum effects were present, as was low beta (the tendency of less volatile stocks to outperform more volatile ones). The size effect (smaller stocks beating larger ones) was not significant (one could speculate that some issues driving small-stocks today, such as less public research and information, were true more broadly back then).
There is one huge caveat to historical studies like this. It is possible that any apparent anomalies could not have been profitably exploited given higher trading costs or lower liquidity. That aside, this data seems to strengthen the idea that many factor effects are long-term consequences of human behaviour in markets rather than misleading patterns that appear solely by chance.
I wish I knew what factor investing was, but I’m too embarrassed to ask
A factor is a characteristic that has been shown to contribute to a stock, bond or other security outperforming the market. Research into factors was originally driven by academics trying to figure out why certain stocks tended to generate higher returns than theories about efficient markets would have predicted. Widely accepted factors include size (the observation that small companies tend to beat large firms over time); value (cheap companies beat expensive ones); yield (high-yielding stocks do better than low-yielding ones); and momentum (stocks that go up just keep on going up). These factors will not always beat the market over any given time period, but they have generated superior returns in many different global markets over the long run.
Factor investing has become increasingly popular in recent years, as the investment industry tries to exploit both the rapid growth in computing power that makes it easy to construct indices based on different factors and the growing disillusionment with active fund managers. Factor-based strategies are often described as “smart beta” – a label that implies they fall between traditional index funds (which aim to deliver the market return, or “beta”) and active funds (which aim to deliver above-market returns, or “alpha”). Today, investors can easily buy a momentum ETF that constantly rebalances into stocks with strong momentum or a value ETF that holds stocks viewed as cheap on certain metrics.
The problem with the race to find new factors for smart-beta funds to exploit is the risk of data mining. If you look at enough historical data, you can always find patterns that turn out to be simply statistical flukes. However, the established factors (such as those listed earlier) are widely accepted as valid – although whenever they endure a long period of underperformance, there will be questions about whether they still work.