How To Predict

Most people when they make predictions, predict change. I’m the opposite in that I predict that things will tend to remain unchanged, which has worked well for me since 2011 (and in retrospect best agrees with the subsequent empirical evidence). The S&P 500 keeps making new highs every week. Same for home prices in the Bay Area. Bitcoin, which in late 2015 I predicted would go much much higher, has gone up 120% since then and is now at $1,130/coin. There is no guarantee these trends will continue, but I’m batting nearly 1000, whereas most people, who predict change, fare much worse.

Then the obvious questions is, how do you know which trends will persist. Why is Facebook so successful and Fit Bit and Go Pro are not?

There are three factors:

1st: Understanding markets. Facebook, Amazon, and Google are irreplaceable, in addition to having huge growth in users and advertising dollars. There are no substitutes for these three companies. Meanwhile, there are hundreds of wearable fitness device copying the same technology as Fit Bit. A Go Pro is just an expensive phone camera. Also, Facebook, Amazon, and Google cannot be ‘fads’, because they are such an important, indispensable part of the global economy and society. It’s like trying to create a substitute for indoor lighting. Bitcoin is another irreplaceable technology..although there are other blockchain currencies, Bitcoin by far has the most ubiquity. This is where the ‘Matthew Effect’ kicks in, because existing big and successful companies, such as Facebook, can use their market dominance to keep growing. Facebook becomes more valuable by having more users, because the total number of interactions grows quadratically (Metcalfe’s Law).

2nd: The role of HBD. This blog is the only site to incorporate an HBD into its predictions, which historically have proven to be unfailingly accurate:

High-IQ countries outperform low-IQ ones, on an inflation-adjusted and per-capita basis (Singapore vs. Brazil)

High-IQ industries and sectors (tech) outperform low-IQ ones (energy), and have less volatility. With the exception of the grand finale of the tech bubble (1998-2000) which negativity skewed the 2000-2006 returns for tech, this trend has held for many decades, and especially since 2009.

High-IQ companies outperform low-IQ ones (specifically, companies that have a high valuation relative to the total number of employees, indicating a very selective hiring process and a lot of value created per employee. An example is Whats’s App, which was valued at $18 billion but only had 150 employees. GM, on the other hand, has thousands of medium and low-IQ employees. The latter went bankrupt in 2008. )

The real estate markets and economies of High-IQ neighborhoods and cities outperform low-IQ ones (Detroit vs. Palo Alto)

So with the help of this sorting process, predicting becomes much easier.

3rd: Having an understanding of underlying macroeconomics economics and trends, but more importantly, an ability to put data in context and filter out noise. 90% of my economics prediction were accurate, from predictions about GDP growth, employment, the strength of US dollar, and low inflation. Understanding the fallacy of composition is especially important here. Weakness in one part of the economy (labor market) does not portend to weakness for the entire economy, as well as other misconceptions about the economy. Between 2009-2005, a lot of pundits wrongly predicted another bear market and repeat of the 2008 recession, by erroneously extrapolating a single negative data point and generalizing it for the entire economy and ignoring the healthier data, or a failure to put the negative data in its proper context. This is related to the confirmation bias. In 2013-2014, others said that the end of QE would cause the stock market to crash, because the economy was supposedly dependent on it. Nope.

Here’s an example: in June 2016, the ADP report, which is released every month, showed only 50,000 non-farm jobs created in the prior month, well-short of the estimate of 150,000-200,000 jobs. The usual pundits proclaimed that this was the start of another recession, and so on. I ran a statistical analysis on the numbers and determined that such a large miss was statistically due to happen and that the miss, although large, wasn’t a big deal. In retrospect, a year later, the miss turned out to be just an outlier, not the start of a recession.