“The Signal and the Noise” (by Nate Silver) — Summaries: EP5

For thousands of other business book summaries, check out: https://summarized.biz/

1. Historical data shows that predictions in all industries, are very inaccurate and difficult. Despite this, experts are often very confident about their predictions. This book focuses on attempting to isolate the few key pieces of critical information in predicting economic development from the vast amounts of data available.
2. People make whole careers out of making predictions in the stock market, meteorology and sports. In some cases, companies have a system that collects up to four million economic indicators. Despite all this, their track records on future forecasts are bad.
3. As an example, say a prediction reads: “Next year, the GDP will grow by 2.7 percent”. This conclusion could’ve resulted from a system output which said that there’s a 90% chance that GDP will be between 1.3% and 4.2%. So they are just taking the average of that, which is misleading. Overall, the track record for economic predictions is 50% accurate.
4. A particularly bad area of predictions is: Economic Depressions. In the 1990s, only 2 out of the 60 depressions occurring worldwide were predicted successfully. Therefore we should really be careful, when considering economic predictions.
5. The economic system is extremely complicated and dynamic. It’s hard to figure out casual relationships between different factors.
6. Feedback loops might compound effects of factors, for example: Higher unemployment influences consumers’ buying power negatively, which reduces the quality of the overall economy, which loops back and further increases unemployment.
7. Rising prices could either be an indicator of positive health in a sector, but they could be artificially set by the government, in which case that’s not the case.
8. Predictions themselves can also have an effect on the economy directly, merely by influencing investors.
9. The global economy itself is constantly evolving, and therefore the basic tooling of predictions does not stay the same.
10. Data source accuracy is also a big issue, where for example, a last-quarter of 2008 government data, indicated a 3.8 percent decline in GDP, but the reality was a 9 percent decline (way off).
11. Some experts rely on massive amounts of juxtaposed statistical patterns, but this approach has the problem of confusion by coincidence. The system will think certain factors imply causality of the outcome, but they were merely a coincidence.
12. An example of this accidental causality is: between 1967 and 1997, 28 out of 30 times, if an NFL team won, the stock market would gain during the rest of the year, but if an AFL team won, the stock market lost. The statistical likelihood of this relationship being a coincidence is 1 in 4,700,000, however this was in-fact the case. Since 1998, this trend has been reversed.
13. When you are inspecting 4 million economic indicators, then you will, by definition, get accidental correlations that are just a coincidence. And they have occurred merely by chance, as opposed to causality.
14. A human still needs to analyze all the candidates for potential correlation, and decide if causality is plausible. A lot of people get this backwards, and instead try to introduce as much technical analysis information as possible. This, however, just hides the valuable signal with a higher percentage of noise.
15. In 2008, we knew that historically: a meteoric rise of housing prices combined with a record-low savings rate have, have always previously led to a crash. So how come so many people failed to realize this and predict the 2008 housing crash?
16. In 2008, people were making too much money in the booming housing market. This caused them to not question whether a recession is just around the corner.
17. The other issue in 2008 was: Collaterized Debt Obligations and the AAA ratings they received. CDOs were a new type of financial instrument which took on bundles of mortgage debts, in exchange for profits earned from debt-holders mortgage payments. But Standard & Poor’s made the mistake of only evaluating the possibility of individual mortgage defaults. They did not take into a possibility of a large-scale housing crash, which could bring down house prices across the board. So they were handing out estimations like a 0.12% chance of default, but in reality a 28% default rate resulted among CDOs.
18. The other problem in 2008 was that the entire industry collectively thought that a recession was impossible. Lehman Brothers leveraged themselves so heavily that only for every $1 they owned, they had $33 worth of financial positions. This allowed them to make excessive profits at the time, so an idea of recession was not interesting to them. In reality, if they had lost even 4% in their portfolio, they would have faced bankruptcy.
19. Then in 2009, when the first stimulus packages were being created, the government failed to recognize the nature of the correction which just happened. They thought it was a regular recession and unemployment figured would be back within one to two years. But history shows that in the case of a recession triggered by a financial crash, unemployment rates stay high for four to six years. Therefore, the stimulus package of 2009 was inadequate to address the situation.
20. The Bayesian approach (developed by Thomas Bayes) is a mathematical system which helps update predictions based on newly incoming information. We subconsciously prefer recent new information due to our inherent biases, forgetting more global, long standing information.
21. Take an example in which breast cancer occurs in 1.4% of women in their forties. This is a long-standing “prior probability”. A woman then gets a mammogram, and the mammogram shows a positive result. We then find out that mammograms discover breast cancer accurately only 75% of the time, and will actually produce a false-positive 10% of the time. Plugging all the numbers into Bayes theorem, how likely is it that this woman actually has breast cancer? The surprising result is that the likelihood she has breast cancer is still only 10%. This is confirmed by clinical data. In this case, the false positive of the erroneous mammogram test has tricked us into believing that new evidence outweighed the previous data, which was statistically correct and outlined the overall low chance of a woman developing breast cancer in her 40s. (1.4%).
22. In 1987, a psochologist and political scientist, Philip Tetlock, recorded predictions of politics and the economy by many experts. He then analyzed the accuracy of their predictions, their personalities, and styles, and derived a clear pattern. There are two types of prediction makers: “Hedgehogs” and “Foxes”. Hedgehogs are those that made predictions based on a few large pieces of data or only a few big facts. Foxes, however, were very cautious, using multiple perspectives of contemplation, and empirical data, to carefully weigh their pros and cons. They were quick to throw out their preconceptions and ideologies.
23. Foxes were much more successful than Hedgehogs at predicting future outcomes. In fact, hedgehogs’ predictions were only right about half the time.
24. Stocks on the stock market tend to go higher over the long run. This, however, is not as useful to most people, because they try to make shorter term gains in attempts to “beat the market”.
25. As a single person, it’s very hard to beat the market. A study was done which compared the predictions over a multi-year period between a group of 70 economists, against those of individuals. The predictions of the group have always been better, than those of the individuals’, when results were aggregated.
26. Stock markets are very efficient, because the bulk of the financial transactions are made by very experienced traders who are part of large financial companies with huge amounts of data, expertise and resources. Therefore, if a stock is underpriced or overpriced, the market corrects for that very quickly.
27. The only other way to beat the stock market is to know specific information, which nobody else knows, but such information can only be illegal insider info. One group of investors who outperform the overall market by 5 to 10 percent per year are members of Congress. They have insider information through lobbyists and can also affect businesses directly by using legislation.
28. The times when markets are not efficient, is during bubbles. One telltale sign of a bubble is a sharp increase in stock prices in general. When the S&P 500 stock market index increased at twice of it’s long-term five year average, this resulted in a crash 5/8 times.
29. The P/E ratio (Price / Earnings) is another indicator that signals a bubble. A normal P/E ratio of the market average itself is 15. So if the P/E ratio is way over that, such as 30, as it was during the dot com bubble in year 2000, that means a bubble is most likely forming.
30. Investors know that bubbles are forming. They invest anyway. This is because they are rewarded on performance of their clients’ funds. Bubbles for them are actually a really good thing, because if the market is so hot, they just get huge bonuses and make their clients money. If the bubble busts, they lose their clients’ money, not their own. When a bubble catches an investment firm, only 20% of the people lose their job. So investors still have an 80% chance to keep their job, even if they get caught by a crash. Therefore: investors are not incentivized to be responsible, but to simply make money.
31. The Climate system is also very complex, interrelated, and difficult to predict. The International Panel of Climate Change (IPCC) has used a very complicated model with lots of data in 1990 to predict the global temperature increase over the next 100 years of 2–5 degrees, settling on 3 degrees. But they were proven wrong with observational experiments conducted over the next 11 years, that the rate of global warming was only 1.5 degrees every 100 years. The climate scientists themselves are not confident in their own models and acknowledge the overwhelming complexity of this modeling problem.
32. The real signal in the Climate Change problem turned out to be the level of CO2 in the atmosphere. In the case of the atmosphere, a simple model such as this one, is more effective at predicting global warning. Understanding the cause and effect of the physical phenomenon of greenhouse gases, such as CO2, trapping heat in the atmosphere, allowed to identify this signal.
33. Regarding the 9/11 World Trade Center attacks: many people claim that these attacks should have been easy to predict, and the government should have known about them. There were warning signals in 2001, of heightened Al-Qaeda activity, and in August 2001, an Islamic fundamentalist was arrested for a suspicious request to practice flying on a Boeing 747 simulator. Previous plots have also been discovered, which involved flying commercial jets into buildings. But what we need to keep in mind is that these pieces of evidence were buried in a much larger amount of noise and other leads, most of which were not helpful.
34. Statistically, however, terrorist attacks follow a math pattern called Clauset’s curve. This pattern says that very devastating attacks can not occur frequently. The more devastating an attack is, the less frequently it can occur. An attack of the 9/11 magnitude, according to Clauset’s curve, happens roughly once every eighty years. Therefore, as that timeframe was coming up, the government could have predicted an attack of such a massive scale. Israel prioritizes almost all of its efforts on prevention of large scale attacks, and not small scale ones. Because of this, since 1979, no attacks in israel have claimed more than 200 people at any given time.
35. Summary: many experts are very sure about their predictions, which turn out to be incredible inaccurate. They all look for relationships in their data, but the chance of having accidental relationships is high, which will result in erroneous learnings and predictions from their efforts.

If you enjoy my business book summaries, please consider using my link to become a Medium member: https://franticrock.medium.com/membership, or giving me a follow.