Theme: Numbers & Data

Around twenty years ago, I received an intriguing query from a potato starch manufacturer (yes, consulting really was living in the fast lane…). The manufacturer’s competitors mainly used maize as their raw material and the manufacturer was curious about potential future price fluctuations of maize worldwide. I’m not sure why they asked me that. But it didn’t matter, the underlying question was a broad economic one that didn’t require deep industry-specific knowledge.

The first question was whether the maize price was fundamentally unpredictable, or in economic parlance what’s known as a random walk. Or – alternatively – if the long-term price trend had any predictive value and if so, in which direction. People tend to assume that if something goes up, it will continue to go up. That’s not the case with a stock price. The likelihood of a stock price rising is equal to the likelihood of it falling. So the fluctuation is as random as if you let a coin toss decide the stock price. Maurice Kendall demonstrated that in 1953.

That said, a trend may indicate an upcoming reversal of that pattern. Provided that a price sends out a signal, causing feedback—for example, if the price of steel is soaring, steel suppliers will try to increase their production—then a price rise is actually an indication that the price is likely to fall. This is also known as mean reversion. Price extremes correct themselves, balancing the price back towards its average. Benjamin Graham, the father of value investing, who was a great influence on Warren Buffett, cites mean reversion to illustrate that sellers or buyers will bring prices back in line with their average.

But the concept originally came from genetics. Towards the end of the nineteenth century, British researcher Francis Galton noted that children of tall parents were on average shorter than their parents. And children whose parents were short were on average taller than their parents. It’s not hard to imagine why that is. Let’s say you throw two dice at the same time, and the total is 12. Chances are you’ll throw something more in the region of 7 the next time. (7 is the average and also the most frequently occurring combination of two dice. In fact, the probability is so great that rather than somewhere between 7 and 12, you’ll likely come closer to 7). In genetics, this is simply the outcome of a random variable. To a certain extent, that’s also true in the economy, but other factors influence the economy’s underlying mechanism.

In the case of maize, it seemed logical to assume that there would be mean reversion in the long term. So the price was not a so-called random walk. Maize’s long-term price was defined by how fast the producers, farmers with fields, could respond to high prices. So that became the basis of our market model.

But you have to understand how something works if you want to make a sound prediction about it. The best tactic for finding this out is just to ask people who know. And that’s where I had not one, but two bits of luck. By coincidence, former INSEAD professor Philip Parker had built a whole stock and flow model of the worldwide maize market. He was willing to sell it to me for a reasonable price and even explained how it worked. The model was structured by country and showed maize reserves as well as its production, consumption, import, and export figures. The latter four aspects informed the flow between countries and continents.

And that immediately revealed an interesting detail. Almost all big markets were more or less autarkic, or self-sufficient. Some, such as South America and China, had surplus maize reserves while others, such as Europe and the US, had not quite enough, and that determined trade between continents. This also illuminated fast-growing demand in China—a steeper growth curve than in the rest of the world, but given their surplus, that wouldn’t be a problem.

My second bit of luck was that through a contact, who happened to be living in China at that time, I was put in touch with someone who had recorded statistics on maize for the Communist Party for many years. He told me (back when that was still possible) that data on maize stockpiles were regularly falsified and that the surplus probably didn’t exist. That meant that very soon China would have to import, rather than export, and it was safe to assume that the US would be the supplier of choice to compensate for the shortage. If that were true, the price of maize would go up, and stay high for quite a while. There was not enough land available in China to allow the expansion of maize cultivation, and other countries were equally unable to scale up production fast enough. So that’s what we predicted.

And our prediction was realized. With the exception of the 2008 financial crash, the price of maize was to rise for more than fifteen years, until the mean reversion began to take effect. That almost always happens in the end: what goes up, must come down. Timelines are problematic to predict, but market spikes always have a ceiling.