Every year between Christmas and New Year’s I look forward to catching up on some reading. This year I’m reading Nassim Nicholas Taleb’s Antifragile. Looking for some suggested reading yourself? Here are my personal nominations for Best Business Books 2012:
The Signal and the Noise: Why So Many Predictions Fail—But Some Don’t (Nate Silver)
I was unexpectedly impressed with this book. Because Silver writes the FiveThirtyEight political blog for the New York Times, I assumed The Signal and the Noise was also about politics, but in fact it is a much broader analysis of why some predictions are more reliable than others and what we can do to make our own forecasts more reliable.
According to Silver, some common forecasting pitfalls include:
- Extrapolating from past data, assuming a trend will continue indefinitely. Trends may continue, but history is full of examples where this assumption was incorrect. For example: certain forecasts regarding world population or food production, the spread of certain diseases, or the amount of oil left to be discovered and extracted.
- Failing to account for the uncertainty in a prediction. As Silver points out, if in 1997 you were a town like Grand Forks, North Dakota, with a 51-foot flood levee, you should have cared a lot about the range of uncertainty in the National Weather Service’s 49-foot flood level forecast. Similarly, if you are business with a high debt load, you should care at least as much about the uncertainty in your profit forecast as in the profit level you think is most likely.
- Using an overly simple model when the real world is more complicated. Despite the simplicity of Occam’s Razor or K.I.S.S., the real world is often nuanced and requires more complex explanations.
- Essentially the opposite problem is creating an overly-fitted model. A forecaster can develop a complex model that appears to explain a lot of past variation, but the model may be fitted to noise in the historical data and fail to accurately predict the future.
- Failing to consider the underlying plausibility of a hypothesis. It is often possible to find correlations between variables, but researchers need to have an underlying hypothesis about how one variable affects another. Otherwise, they may get statistically significant but “manifestly ridiculous” findings.
Some predictions fail because the phenomena they deal with are inherently difficult to predict:
- Dynamic systems. In dynamic systems the value of one variable affects the value of another variable (or other variables). Weather and the economy are dynamic systems, which is one reason they are difficult to predict.
- Non-linear relationships. In non-linear systems a very small change in an input can have a huge impact on the output. For example, suppose you are estimating the value z as yx. The true value of both x and y is five, so the result should be 55, or 3215. But due to measurement error, you think x is six. Your estimate of z will be 56, or 16,625, and your estimate of z will be off by 500%.
- Where direct measurements are difficult. For example, seismology is difficult because scientists are not able to take direct readings of pressures miles down in the Earth’s crust.
- When forecasts are self-fulfilling and self-cancelling. If economists suggest the economy will shrink, in theory the Federal Reserve or federal government will enact measures to prevent that. If those measures are successful, the original forecast will not be accurate.
Silver also points out that it may be rational for some people to create biased or overly dramatic forecasts. For example, your local weatherman probably gets more angry emails when he predicts no rain but it does than when he predicts rain but then it doesn’t. Similarly, managers in firms may have incentives to make aggressive forecasts that support their requests for resources.
Finally, Silver makes some observations about how much effort we should spend on prediction: “…It is not so much how good your predictions are in an absolute sense that matters but how good they are relative to the competition… When a field is highly competitive, it is only through this painstaking effort around the margin that you can make any money.” The corollary is that in fields where the quality of forecasting is limited, it may be possible to gain a large advantage with relatively little effort.
The Black Swan: The Impact of the Highly Improbable (Nassim Nicholas Taleb)
I was late getting to this book, which was published in 2007. But it is still well worth a read. Or, you may want to start with Antifragile, Taleb’s most recent work.
For those with a weak ornithological education, black swans are native to Australia. Prior to discovery of that continent, Europeans thought that all swans were white. Taleb uses the black swan as an analogy to point out the risks of generalizing from historical observations.
A Black Swan event, in his language, has three characteristics: First, it lies outside of the realm of normal expectations. Nothing in past events convincingly suggests a Black Swan should exist. Second, it has a major impact. Finally, human nature makes us come up with explanations for the event after the fact which seem to make it predictable.
According to Taleb, a small number of Black Swans account for the state of the world today. Consider: Pearl Harbor, the fall of the USSR, 9/11, or the iPhone. This makes “what you don’t know far more relevant than what you do know.” And, since we can’t predict Black Swans, we can’t predict the future.
Consider how much time you might have spent refining or debating your sales forecast for 2013. Is it five or ten percent too high or too low? According to Taleb, it doesn’t matter. What will really determine the fate of your business is some Black Swan you can’t anticipate!
So what should we do? First, be humble about your ability to predict the future. Second, set yourself (or your organization) up to benefit from serendipitous Black Swans. Finally, make your organization robust to negative Black Swans.
How to Measure Anything: Finding the Value of Intangibles in Business (Douglas W. Hubbard)
How to Measure Anything takes statistics out of the realm of the mathematician and makes it more applicable to the types of business problems we typically work on at Woodlawn Associates. It addresses questions such as how to measure things that at first glance appear to be too intangible to measure, whether small sample sizes are worthwhile, how to compute the value of additional information, and how to improve the accuracy of human estimates.
For example, one of the questions that we often get is “how big a sample size is necessary?” Many people often have a vague recollection of the idea of statistical significance, perhaps from an undergraduate mathematics course they took years ago. While Hubbard doesn’t provide the math to calculate the required sample size, he does a good job of articulating how even a very small sample size can be extremely insightful if (a) you don’t know much about the population beforehand or (b) the cost of having a wrong estimate is high. In either of those cases, even a very small sample can be valuable because it can quickly reduce uncertainty or potentially save a lot of money at low cost.
Blah, Blah, Blah: What To Do When Words Don’t Work (Dan Roam)
Blah, Blah, Blah is about getting ideas understood. Roam’s premise is that we have come to believe the best way to share our thinking is by talking or writing. Half right, he says.
Roam believes effective communication requires both verbal and visual communication. Visual clues help us get deeper into an author’s meaning. The combination of verbal and visual allows us to to simultaneously consider both the details and the overall context of the idea.
One of the other valuable lessons from Blah is that we must vigilantly edit our presentation of concepts if we want to be effective. But to get ideas down to their very essence requires a lot of work. The good news is that we are often as much or more changed by this as our audience: “As we evolve our idea— iterating, revising, reconsidering, trying options—we modify more than just the idea; we also modify our ability to think about it. Staying with an idea long enough to see it through to completion creates a bond between us and the idea that changes both: Our idea gets better, and we get better at thinking about it.”
Thinking, Fast and Slow (Daniel Kahneman)
I loved this book, which is a comprehensive summary of many of the major findings of psychological research from the past 50 years as applied to problems we commonly face in businesses and other organizations. There were four topics in the book I found particularly interesting:
- Biases in planning and forecasting, and how to avoid them
- Loss aversion, and why companies throw good money after bad
- When to trust intuition
- The superiority of algorithms in personnel selection
Click on the links above for blog posts I made earlier this year about each of these.
The Quest: Energy, Security, and the Remaking of the Modern World (Daniel Yergin)
Daniel Yergin’s 1991 book about the history of oil, The Prize, won the Pulitzer Prize that year for nonfiction. In The Quest he updates and picks up the story for the decades since the first Gulf War and expands his narrative to include other energy sources.
It’s an important story. Yergin’s thesis is that energy is power, and so the future of energy will determine the future not only of firms in the energy business, but also of the global economy, the position of nations, and the environment.
Through more than 700 pages, Yergin does a great job of harnessing this underlying drama. The Quest reads more like a fast-paced novel than a treatise on economics and geopolitics.
I’d love to hear about any business books you particularly enjoyed this year. May you have a healthy and prosperous 2013!