What do you call a comprehensive stock market study that goes back 20 or 30 years? If you are one of the deep-diving quants at Erasmus University Rotterdam or Robeco Institutional Asset Management, a Dutch money management firm, you might call it “narrow and short-sighted.”

The Netherlands-based team of Guido Baltussen, Laurens Swinkels, and Pim van Vliet put together a giant study of market factors going all the way back to 1800 — an incredible 217 years of market data.

The study covered not just international stock indexes, but bonds, commodities, and currencies, too. The findings were published in a 63-page white paper titled “Global Factor Premiums.” (You can access the white paper here.)

The study’s output looks at six of the major factors driving modern-day quant investing, with each of the six factors applied to four separate asset classes (stocks, bonds, commodities, and currencies) for 24 “global factor premiums” in total.

The six factors are:

  • Momentum (the tendency of winning stocks to keep winning)
  • Trend (the tendency of strong trends to persist)
  • Value (the long-term outperformance of cheaply priced stocks)
  • Carry (the tendency of high-income securities to be favored)
  • Seasonality (buying at a historically favorable time of year)
  • BAB (‘Betting Against Beta,’ favoring low-volatility vs high)

Baltussen, Swinkels and van Vliet took great pains to avoid “p-hacking,” the researcher’s mortal sin of fudging or tweaking the data in pursuit of a desired statistical result. Their rigor was key because it is all too easy to corrupt a set of back-tested results by accident, even with the best of intentions.

Going all the way back to 1800 further ensured the study would cover all manner of market conditions, reducing the odds of a sample-biased finding. There were 43 years of bear markets and 74 years’ worth of recessions (not to mention multiple World Wars, panics, boom-bust manias and depressions) within the 217-year span.

When all was said and done, the team probably felt compelled to triple-check their results, because what they found was amazing.

Out of the 24 variables tested — six factors across four asset classes — an incredible 19 of the 24 maintained a statistically robust edge over the span of more than 200 years.

“Correlations between factors and between asset classes are remarkably constant over time,” Bloomberg columnist John Authers wrote, “even in periods that far predate cross-asset trading, algorithms or even the telephone.”

“Over the two centuries, trend-following worked better and more reliably than any other factor,” Authers added, “with a strong and consistent risk-adjusted return.”

Another surprising finding was the strength of seasonality factors. As it turns out, “Sell in May and Go Away” is not just Wall Street folk wisdom. It is now backed up by centuries of data.

The sheer persistence of these factors — again, for a period of 217 years across all manner of market conditions — raises a head-scratching set of questions:

  • Why haven’t factors weakened substantially after all this time?
  • Why do trend-following and momentum and value still work?
  • Why is, say, market seasonality as viable an input as ever?

If markets were truly rational and efficient, these factors would not exist.

Or if computers are truly crowding out human beings, as many now believe, the computers should have spotted the opportunity by now in exploiting these factors as anomalies — which would mean competing away all profits until the factor edge disappears.

A leading theory for why factors still work is varying levels of risk tolerance — the notion that factors are a way to get paid for taking on more risk.

Value stocks might be seen as more risky because cheaply priced companies tend to have damaged reputations, troubled business models, or flawed balance sheets, for example, which means value investors are compensated for taking on extra risk.

It can also be argued that, say, trend-following entails extra risk because prices tend to mean-revert more often than they trend on average, which suggests there could be an exploitable risk premium in choosing to stick with trends.

But this theory only goes so far, and fails to address other factors (like seasonality), and there are multiple holes in it regardless. As the quants write in their white paper:

“The global factor premiums are at best marginally explained by downside risk explanations. Furthermore, they are consistently present across various macroeconomic states, nor can be explained by macroeconomic risk models. Consequently, our results seem hard to reconcile with explanations based on risk…”

As to why factors still persist after all this time: We draw the conclusion that human nature is more powerful than technology. No matter how professional things look, emotion and instinct still rule.

The stock market is driven by human behavior, and the stewards of investment capital are still human beings. (If human investors don’t like the results that a robotic program is generating, for example, they may shut off the program or fiddle with it, or even take back their funds.)

If the six most powerful factors can span a 217-year period that saw multiple World Wars, countless recessions, and multiple full-blown manias and panics — along with the invention of the telegraph, the rise of electricity, the automobile, the airplane, the telephone, the microchip, and even the rising dominance of computerized stock trading — then those very same factors are probably robust enough to last another 217 years more. The longer a phenomenon has lasted, especially under harsh conditions, the longer it is likely to persist.

After all, the one thing that hasn’t changed a bit these past few centuries is human nature — and investing remains a human-centered activity.

At TradeSmith, we have built most of these factors directly into our investment software, with the goal of helping investors take direct advantage of factor-produced alpha by way of our proprietary algorithms and strategies.

We are also constantly on the lookout for new ways to apply these old factor concepts. And now we have a better sense of just how “old” — dating all the way back to 1800.

TradeSmith Research Team