Assignment Kim

Chapter X Randomness and Uncertainty

Reports that say that something hasn't happened are always interesting to me, because as we know, there are known knowns; there are things we know we know. We also know there are known unknowns; that is to say we know there are some things we do not know. But there are also unknown unknowns – the ones we don't know we don't know. Donald Rumsfeld, February 12, 2002

What do randomness and uncertainty have to do with clear thinking? Isn’t randomness the antithesis of thinking? It might be surprising that there is an element of randomness in most things we do. Without randomness, we would get exactly the same result each time we repeated the exact same action. The drive to work would be completely predictable, friends would always react the same way, and sports would be boring to watch. Even if something seems like it should be completely predictable, inherent variability comes into play. If the alarm is set for exactly the same time every day, there are still bound to be a few minutes of difference between the time it goes off and the time you are ready to leave each day. Traffic is affected by any number of variables; weather, the number of other drivers, problems caused by other drivers cutting in and out of lanes, and road construction. On any given day, these factors may or may not affect what is usually a fairly predictable trip. While clear thinking isn’t the result of randomness, it acknowledges and accounts for randomness and uncertainty when making choices and planning for the future.

Most people chronically underestimate the effects of randomness. Good luck rarely gets the credit it deserves, while bad luck receives too much blame. When we make plans, we often forget to factor in variability.

Randomness is intrinsic to the laws of probability, with which people also have trouble. However, it should be given its due. Many times one’s efforts seem to be highly effective when, in reality, external circumstances may be more responsible for success. The converse is also true – a great decision cannot always compensate for the effects of the economy, Mother Nature, or changing consumer tastes.

What are the benefits of understanding randomness and uncertainty? With the flood of information we are constantly subject to, we need to know what to believe and what to ignore and how to use information to make realistic decisions. Too few people understand the difference between correlation and causality, whether a new product or medical treatment will make any difference in our well-being, whether we should risk an investment, or what news is credible. A little skepticism about claims can go a long way toward developing a realistic view of the world. Understanding of the variability of the conditions that shape our decisions will foster improved choices and plans. The ability to recognize that something is a coincidence, and not inherently meaningful, keeps us from developing false beliefs. It’s important to know when information is reliable and when it’s not. Choices, both personal and professional, work out better when they are based on reality, not assumptions or misperceptions.

What is Randomness?

What does it mean for something to be “random”? People use the word random to describe events that are unexpected or seem to be unrelated to the topic at hand (“That was a random comment”). The typical definition is “without any discernable pattern.” An easy way to understand randomness is to look at examples from gambling.

The bouncing ping-pong balls that determine lottery winners are drawn at random. Every ball has an equal chance of being selected every time the lottery is played, despite beliefs about lucky numbers or relatives’ birthdays. Although the balls can’t remember which ones were drawn in the past, some people persist in trying to find patterns, thinking they will improve their chances of winning the jackpot.

Many people don’t know what randomness looks like. If someone were asked to pick a random number between 1 and 50, few would select 1 or 50, even though those numbers are as likely as something more “random-sounding” like 19 or 37. If we flipped a coin repeatedly and saw the patterns HTHTHTHT, HHHTTT, HHHHHT, TTTTTT, and HHTHTT, most people would say the last one is random, but the others aren’t. The truth is that they are all equally likely, because each coin toss is an independent event – the coin doesn’t remember what the outcome of the last flip was. Even though we eventually expect an equal number of heads and tails from repeated flips, it takes many, many flips to get this kind of result. This is due to the “law of large numbers.”

Simply put, the law of large numbers says that as the number of trials (flips of a coin, dice rolls, spin of a roulette wheel, pulls of a slot machine lever, etc.) increases, the more likely the average result will be the expected value (in this case, 50% heads and tails). While a long series of trials will converge on the expected value, short series seldom do. Most people know that there is supposed to be a 50:50 chance of heads or tails, but relatively few understand that this is the long-run outcome. When there is a streak of several heads or tails in a row, it seems surprising.

One phenomenon that sports fans wholeheartedly believe in is the “hot hand.” This is the idea that an athlete is on a winning streak (or conversely, a losing streak). The usual explanations point to momentum or the confidence from one success leading to another success. From a probability perspective, a hot hand implies that when a player scores, the probability that he or she will score on the next try should be higher than average. Psychologists Robert Vallone and Tom Gilovich wondered whether the hot hand could be documented, so they analyzed the shooting records of each player on the Philadelphia 76ers for 48 games. Much to the dismay of players, coaches, and fans, they found no evidence of a hot hand for any player. The reaction to this finding was, and continues to be, disbelief. However, think back to the coin-flipping example; remember that a series of flips doesn’t usually alternate between heads and tails, even though the average over the long run is 50:50. In a short series, a streak of heads or tails may not look random, but it is. It’s the same with the hot hand. Great players make more shots than average players, but the likelihood that he or she will make the next shot isn’t a function of the last shot. Since people are generally not very good at recognizing randomness, and the idea that momentum and confidence affect performance is very appealing, the myth of the hot hand rings true despite reality.

The “gambler’s fallacy” is another common belief. When someone is betting on a random outcome, like a particular number on a roulette wheel, a common misperception is that the longer he or she goes without winning, the more likely the desired number is to come up. The problem is that each spin is independent and the roulette wheel has no memory. Luck doesn’t self-correct. The same is true for slot machines, the lottery, and just about any other kind of gambling. Thinking that they are due to win on the next spin, or the one after that, or maybe the one after that, gamblers keep betting, often ending up with significant financial losses.

What do these examples have to do with everyday life? You don’t have to be a gambler to encounter problems caused by misunderstanding randomness or probability. Believing that success will continue based on prior success can lead to overconfidence and less careful decision making. Continuing to make risky decisions in an expectation that a win is due is wishful thinking. There are three main areas in decision making where understanding randomness will help you make better choices and plans:

· Understanding cause and effect

· Developing more accurate expectations about future outcomes

· Being a smart consumer of information

Understanding Cause and Effect

Many athletes swear by pre-game rituals to give them an edge, from lucky shirts to a specific way to tie shoes to special foods. Michael Jordan, famed Chicago Bull basketball player, always wore his University of North Carolina uniform shorts under his Chicago uniform. These rituals may give athletes a boost of confidence, but do they really cause better performance?

On a more serious note, a number of parents in the U.S. refuse to vaccinate their children against childhood diseases such as measles and whooping cough. The basis for this practice was a now widely discredited paper by Andrew Wakefield, a British doctor who claimed that childhood vaccination caused autism. He subsequently lost his medical license for falsifying data. Still, some Hollywood celebrities helped spread the idea that vaccines contain harmful ingredients that cause autism, giving legitimacy to the anti-vaccination trend in the eyes of some parents. Despite wide agreement in the medical community that there is no link between vaccines and autism, many parents persist in refusing vaccinations for their children.

Vaccination provides “herd immunity” – if the majority of a population is immune to a disease, it’s much less likely to spread widely. In populations where the anti-vaccination movement is strong, diseases such as measles, mumps, whooping cough and chicken pox are on the rise. For most healthy individuals, these illnesses cause minor discomfort for a few days. However, for those with a compromised immune system or infants too young to be vaccinated, these illnesses can be severe or even fatal. How can we determine whether vaccination causes autism?

If you have ever taken a statistics course, you will have heard “Correlation does not imply causation.” Correlation is a measure of the relationship between two variables, such as total revenue and the amount of money spent on advertising or time spent exercising and cardiovascular health. Correlation is necessary to demonstrate causal relationships, but it’s not enough. Two variables can be highly correlated such that an effect is present when a possible cause is present and an effect is absent when a possible cause is absent. That’s because other variables might be responsible. For example, deaths from drowning are highly correlated with ice cream consumption. When ice cream consumption is high, deaths by drowning are high. When ice cream consumption is low, deaths by drowning decrease. Would water safety be improved if the ice cream supply were restricted? Do people go back into the water too soon after eating ice cream? In this case, the answer is obvious. There is a correlation between deaths by drowning and ice cream consumption because both swimming (and, unfortunately, drowning) and eating ice cream occur more frequently in hot weather and less frequently in cold weather.

To assess whether a causal relationship exists between two variables, we need information about each variable. Let’s look at the relationship between vaccination and autism. The variables are whether or not a child is vaccinated and whether or not the child is diagnosed with autism. According to the Center for Disease Control, the current prevalence of autism in the U.S. is about 1.5% among children aged 3 to 10. With a sample of 100,00 children of whom 10% are not vaccinated, this is what we would expect to see.

Vaccinated Not Vaccinated

Autism 1,350 150

No Autism 88,650 9,850

Total 90,000 10,000

The number of autism cases is proportional to the number of children in each group. There are more autism cases in the vaccinated group because there are 9 times as many children, not because they were vaccinated.

If vaccinations did cause autism, our table should look more like this.

Vaccinated Not Vaccinated

Autism 90,000 0

No Autism 0 10,000

Total 90,000 10,000

Of course, there might be cases of autism unrelated to vaccination, and not every vaccinated child would end up with an autism diagnosis, so these numbers are an exaggeration. But the general pattern would look like this.

Here’s what you need to determine cause and effect:

Cause Present Cause Not Present

Effect Present Yes No

Effect Absent No Yes

If the possible cause is present, it should lead to the effect the majority of the time, and it should seldom lead to cases where there is no effect. If the possible cause is absent, there should not be an effect, and most of the time, absence of the possible cause should mean no effect.

Why do people falsely believe that one thing causes another, when in reality there is no relationship? Essentially, they only look at one cell of the table above – the cell for Cause Present and Effect Present. When two events happen close together, people sometimes think the first one caused the second one. They forget to check whether other causes account for the effect or whether the effect ever happens without the possible cause.

Interestingly, even pigeons can be conditioned to act “superstitious” by providing food at predictable intervals that have nothing to do with the bird’s behavior. (Pigeons are usually trained by receiving food after they perform a specific task.) The pigeons engage in behaviors like whirling around or flapping their wings in a certain way – whatever they were doing when the food first arrived. They look as though they believe their behavior caused the food to appear and continue to repeat the specific behavior so the food will keep coming.

When people hold strong beliefs, they are likely to see causality when there is only coincidence. In the case of superstitious sports stars, a good performance coincides with a lucky shirt (or meal, socks, etc.). When the athlete seeks a reason for the performance, attention falls on the shirt. Superstitions like this are harmless, but when mistaken beliefs about causality affect public health and policy decisions, we are worse off.

In business settings, there are numerous occasions when it’s important to know whether two variables have a causal connection. Do training programs improve employee performance? If more funds are allocated to the social media budget, will brand image improve in proportion to the extra spending? Does increased customer satisfaction really increase sales? Many online firms conduct A/B testing to determine whether one variable has a causal relationship with another. Too often, businesses don’t have the luxury to conduct those real world experiments and must work with the data that are available. In these cases, it’s important to look at all the information that bears on the question, not just that which supports the idea of a causal relationship.

Expectations about the future

Will the future be like the past?

It’s human nature to wonder what will happen in the future. Most of us end up basing our predictions on our prior experiences, or those of people we know. When thinking about how you will do on a final exam, it’s natural to think about how well you did on the midterm. If you have an exceptionally good meal at a restaurant, you look forward to sampling it again. How could randomness be part of predicting your performance on an exam or the quality of a restaurant meal? If you aced the midterm, shouldn’t you expect to ace the final?

You may well ace the final, but making that prediction just on the basis of your midterm score is a mistake. Performance on exams, quality of restaurant meals, stock prices, race times, heights of siblings, download speeds, and almost anything else that can be measured are a combination of an average performance plus some random variation. Performance varies from one time to the next, so a truly exceptional performance (either positive or negative) is unlikely to be followed by another that is equally exceptional. This is due to a phenomenon called regression to the mean. The basic principle is that over time, extreme values are followed by more moderate values. Simply put, scores typically return to their long-run average. That doesn’t mean extreme values can’t be followed by other extreme values, just that it’s unlikely. With no additional information, the average value is the best prediction.

If a student consistently aces all exams, his or her average performance is pretty high and the student may well ace the next one. For more typical students, an exceptionally high or low score will likely be followed by something closer to his or her usual score. If a restaurant meal is exceptional, it’s more likely that the next one won’t stand out as much unless the average quality is very high.

An easy way to understand this is to think about peoples’ heights. This is actually where the idea of regression to the mean originated, with British scientist Francis Galton in 1886. He noted that very tall people usually had tall children, but at least some of them were shorter than their parents. Very short people usually had short children, but at least some of them were taller than their parents. If the children of tall people were always taller than their parents, eventually their descendants would be extremely tall. The same holds for short people. Without regression to the mean, the range for adult human height eventually might go from 1 foot to 12 feet, or even more extreme sizes.

Regression to the mean should be taken into account when making plans and predictions. One of several factors contributing to the 2008 recession was an unrealistic belief that housing prices only went in one direction – up. Had that been the case, the risky loans made to homebuyers with bad credit and few resources would have been secured by continually appreciating assets. Instead, as was inevitable, home prices fell. Because so many risky loans had been made, a cascade of bad debt severely impacted the economy.

A similar phenomenon is the “Sports Illustrated effect,” where some people believe a team that appears on the cover of Sports Illustrated will be jinxed and perform worse following the cover feature. Similarly, the performance of CEOs who appear on the cover of Business Week often declines following the cover story. Does this publicity really affect performance? It’s much more likely that the events that prompted the athletes and executives to be featured on magazine covers were outliers and their performance returned to historic averages after the magazine covers appeared.

The problem with over-specified plans

When we think about the future, we often engage in daydreaming about what we think our lives will be like when we finish graduate school, have a new job, move to a different part of the country, or whatever other event we hope will actually happen. The more detail we add, the more real it seems. Daydreaming about the details of your future life is fun, but it shouldn’t be the basis of planning. While details make your daydreams seem more real, the more detail you add, the less likely it is that those details will be correct.

This may seem counterintuitive, but the reason lies with a simple rule of probability. The probability of two independent events co-occurring is always lower than the probability of either individual event. Probabilities are always between 0 and 1: a probability of 0 means the event will never happen and a probability of 1 means that it is certain to happen. To determine the joint probability of two events co-occurring (e.g., taking a specific job in a specific city) you multiply the individual probabilities. So if you have a 20% chance of being hired for a specific job and a 30% chance of finding a job in a specific city, the probability of both happening is 6%. Every time a detail is added, the joint probability is reduced. We will see more about the probability of multiple events in later chapters.

So, how should people think about the future? Do we need to be statisticians before we can start making good plans? Should uncertainty strike fear into our hearts? Absolutely not. The most important thing to remember is that there is variability around future events. Rather than making plans depend on a specific outcome, we need to try to figure out a likely range of outcomes. Remember that trends rarely continue in a single direction indefinitely. Investment firms always include the statement, “Past performance does not guarantee future results.” It’s true well beyond the domain of stock prices. Rather than evoking fear, accounting for uncertainty will lead to plans that are more realistic and flexible.

The best way to account for uncertainty is to first establish what is known and what is unknown, then develop estimates for the likelihood of different situations. With the combination of what is known and what is estimated, different contingency plans can be developed. This may seem a bit formal, but for important decisions it’s worth taking the time to be as accurate as possible.

Following some significant intelligence failures, such as the prediction that weapons of mass destruction would be found in Iraq prior to the Gulf War, the Intelligence Advanced Research Projects Activity funded research into how to improve predictions. In response, psychologists Philip Tetlock and Barbara Mellers developed the Good Judgment Project to understand the characteristics of people who were good at predictions and what might make them even better. The key factors turned out to be training in basic probability theory, education about cognitive biases, and working in a team that included both specialists and generalists. Keeping track of results and forming teams of “superforecasters” led to accuracy that was almost double that of people with no training.

Being a smart consumer of information

More than 60 years ago, Darrell Huff published a small book titled How to Lie with Statistics. The purpose of the book was to help people understand how statistics in the news and advertising could be technically correct, but misleading, depending on the purpose of the news report or the ad. This slim volume had dozens of printings and ultimately over half a million copies were purchased. The examples Huff used were tied to 1950s era concerns, but decades later the underlying message is still important.

We hear statistics about government, sports, political races, traffic accidents, crime and a myriad of other topics. Are we in a recession or a recovery? How can the unemployment rate go up when more new jobs are being created? The news is full of reports about purported causes of cancer, heart disease, and other health issues. Advertising makes promises that products will make us more attractive, energetic, and slimmer. Should we eat dark chocolate for its antioxidants or avoid it because it might contribute to obesity and diabetes? Should we run for cardiovascular health or walk to avoid joint damage? Do we need to buy a standing desk to avoid the effects of too much sitting? We often forget that news programs shape their programming to maximize ratings and advertisements are designed to influence our spending, not to help us make good decisions.

Many of us glaze over at the mention of statistics. But statistics is interesting because it enables us to summarize information in order to learn about the world. Statistics is a tool to understand whether a change has happened or not, whether variables are related; a way to detect a signal in the noise of randomness. Unfortunately, someone with an agenda can easily “lie with statistics” to mislead us. We don’t have to look too far for examples.

During the lead-up to the Brexit vote, in which Britain voted to leave the European Union, the Vote Leave group repeatedly claimed that the United Kingdom sent £350 million every week to the European Union. This was true – but something was missing. The European Union refunded about two-thirds of that amount, so the net figure was actually £100 to £125 million.

A recently published study reported in the Wall Street Journal (8-29-16) was titled “Eating Fruit While Pregnant May Boost Your Baby’s Intelligence,” with a subtitle of “Infants whose mothers ate more fruit were smarter one year after birth, a preliminary study shows.” Fruit is part of a healthy diet, so this news is not exactly earthshaking. However, the claim that the fruit eaten during pregnancy is the reason for a baby’s higher intelligence is stretching what the scientists found. Researchers looked at cognitive development scores for 688 infants and related the scores to data from a survey the mothers completed during pregnancy. The finding was that there was a statistically significant relationship between self-reported fruit consumption and a composite of the scores on the Bayley Scales of Infant and Toddler Development at age one. Test scores are not the same as intelligence, and the increase in scores was 2.38 points per serving of fruit, well within the standard deviation of the Bayley Scale, which has a mean of 100 and standard deviation of 15. The authors of the research study were careful to state that these results are preliminary and that cognitive development scores at one year don’t predict cognitive development scores at the age of three. The journalist made a claim in a catchy headline about intelligence, but the researchers were talking about test scores at age one, not intelligence, which is a much more complex concept.

Questions to Ask

There are a few things to keep in mind when someone is using statistics to support a point of view. In How to Lie With Statistics, Darrell Huff characterized these issues in a chapter titled “How to Talk Back to a Statistic.”

Who Benefits?

First, does the sponsor of the research have a reason to favor one side of the argument? Here are two examples from nutritional research where this question needed to be asked. The California Walnut Commission sponsored a study that found eating walnuts improved the health of people at risk for diabetes. Another study found that Concord grape juice improved driving performance and spatial memory among mothers of pre-teens included an author who was an employee of a major grape juice provider. It’s entirely possible that these findings are legitimate, but in many cases, studies that are funded by organizations with a vested interest in the results tend to show more positive findings than studies funded by neutral organizations.

How Do They Know?

What Sample?

A second issue to consider is the nature of the sample. Two factors matter here: the size of the sample and how the people in it were selected. When a sample is large, the data it provides is more likely to be true of the population the sample represents because of the law of large numbers. When the sample is small, you really can’t draw solid conclusions from the data.

Problems with sample selection occur for a number of different reasons. The ideal sample is one that accurately represents the population of interest. Finding a truly random sample to answer a pollster’s survey is difficult. If you select people from a telephone directory, you’ll miss the growing number of those who only use cell phones. With the prevalence of caller ID, many people won’t answer the phone unless they recognize the caller. If your survey is online, you miss the population that doesn’t use the Internet.

There are many reputable polling organizations that take pains to sample respondents and report statistics properly. Gallup, Pew Research, Harris and NORC (National Opinion Research Center) all apply sophisticated approaches to sampling and analyzing opinion data, so you can be confident in what organizations like these report.

Which Average?

There is a joke about Microsoft founder Bill Gates walking into a bar and everyone in the bar being happy because their average income just went up dramatically. Technically, a scenario like that would be true (about the average, not necessarily the happiness) – if the mean is the average that you use. Income distributions are almost always positively skewed, meaning that there are some individuals whose income is high enough to distort the mean in a positive direction. If the distribution weren’t skewed, the mean would be very close to two other average measures – the median and the mode. The median is the number that divides the distribution in two, so that half of the people make less than the median and half make more. Medians are usually used to report income, housing prices and other government statistics because they aren’t sensitive to extreme values like Bill Gates’s income. The mode is the most frequent value in a distribution and isn’t used as commonly as means and medians. You would use a mode if you wanted to figure out which item (or flavor or size) was the most popular. So, when you hear a news story that reports average income, prices, scores on educational tests, or any of a host of other topics, keep in mind which average is being reported.

What’s Missing?

When a new medical study comes out, we are often warned that the risk of contracting a disease is increased by 50% among people who fit a certain profile or promised that a new drug will reduce the time required to recover from an illness by 20%. What is left out is what is called the “base rate;” how many people are affected by the disease or how long people are typically sick. For example, Tamiflu is widely prescribed for the flu because it cuts the duration of the illness by 20% when taken within 36 hours of symptoms. The flu will make most people miserable, but the misery usually lasts about 5 to 7 days without medication. Tamiflu reduces the duration by 20% - to about 4 to 6 days (from 123 hours with a placebo to 98 hours with the drug, according to a 2015 study).

Since 1997, direct to consumer advertising for pharmaceuticals has become widespread in the U.S. Although ads must include disclosures about possible side effects, they rarely discuss the risks and benefits of drugs in a transparent way. Most ads mention benefits as a relative risk, such as a 50% reduction in developing a disease. What is missing is absolute risk, without which you can’t tell whether the 50% reduction is meaningful. Does the 50% reduction mean that only 100 of 1000 people would develop the disease compared to 200 of 1000 without the drug? Or does it mean that only 1 of 1000 people would develop the disease, compared to 2 of 1000 people without the drug? The 50% reduction in relative risk is correct in both cases, but the extent of the absolute risk is different by two orders of magnitude. You can’t really get an idea of the risk unless you know the base rate. That’s why (from a marketing perspective) many pharmaceutical ads mention benefits only in relative terms without including information about the absolute risk.

Does the picture tell the true story?

Many arguments are made using information presented in charts. Well-constructed charts convey information more quickly than tables and make it easy to understand relationships that otherwise might be difficult to discern. Unfortunately, charts are susceptible to the same kinds of manipulation as statistics. Can you tell what’s wrong with the following chart? It documents gun deaths over time in Florida, with a special emphasis on 2005, the year the “Stand Your Ground” law was passed.

The vertical axis starts at 1,000 rather than zero, so what you might normally interpret as a decline when the law was enacted in 2005 is actually a steep increase. This chart drew media attention because it was so misleading.

There are many ways charts can mislead. As in this example, axes can be misleading, especially when they start at a number other than zero. Pie charts are often used inappropriately (they should only be used to indicate proportions within a whole), and sometimes add to more than 100%. Some figures on infographics represent more of a difference between items than is warranted, because the area of the figures varies in two dimensions when the numbers they represent vary only in one. When someone has a point of view they are trying to sell you, be sure to look at how they are presenting the data.

Applications

The benefits of understanding the basics of randomness, uncertainty, and probability are similar in both personal and managerial settings. You will be at a significant advantage because the evidence is that far too few people understand these topics, even those who are educated. You will be less susceptible to questionable claims and better able to assess possibilities. Your plans will account for uncertainty and be more realistic. There are two major types of benefits associated with understanding randomness, probability and uncertainty. The first is greater clarity in your thinking. The second is that you will be able to make plans more successfully. Both benefits apply to personal and business life.

Clarity

The ability to discern when something is random or not is helpful when you are trying to understand why things happened and whether a causal relationship exists. When you see a true causal relationship, your actions will be more effective and you will be able to avoid problems. When you know something is random, you can stop wasting time trying to change it. You won’t be fooled into thinking something will succeed just because there’s been a long string of misses.

When you understand the principle of regression to the mean, you will have more realistic expectations about future events. Spectacularly good and spectacularly bad events can occur to anyone, but they are unlikely to be repeated and shouldn’t be taken as an indication of how future events will unfold. Investors who do the best tend to be the ones who don’t react on the basis of day-to-day swings in the market. Instead, they recognize that outliers occur on both the positive and negative side and focus on the long-term return. The less fortunate investors are those who check their portfolios daily, reacting to what is essentially random noise.

Understanding which events are meaningful and which are just noise requires a skeptical eye. Inclusion of base rates helps you understand whether a risk or benefit is significant or not. Statistics are so easily distorted that it’s worth your while to consider the source and ask the basic questions:

· Who says so?

· How do they know?

· Are they comparing apples to apples?

· Do they have an interest in a particular interpretation?

Planning

Planning involves making choices about what we will do in the future on the basis of what we expect the state of the world to be in the future. The problem is that the future is uncertain, except as Benjamin Franklin famously noted, “… in this world nothing can be said to be certain, except death and taxes.” What we want to be true in the future doesn’t necessarily have an impact on what will happen. If you don’t smoke, eat wisely, and stay fit, you will be more likely than not to enjoy a long and energetic life, but there’s no guarantee. You may want to win the lottery and quit your job, but the probability remains 1 in 292 million, so you’ll likely need to find an alternative for retirement. Rare events do happen, but they are by definition rare.

How can understanding randomness and probability help in planning? If your plans depend on economic conditions, competitors’ responses, and consumer demand, you are already well aware that the past doesn’t predict the future. Certainly the present and recent past provide a baseline to initiate planning, but how can you go beyond looking at the past and present to predict the most likely future?

As mentioned above, regression to the mean should be taken into account when trying to determine whether trends are likely to continue. Extreme results are most often outliers, so unless you can identify the specific causes and can expect those causal factors to continue to impact your business, you are better off with a more moderate forecast. If you are experiencing phenomenal success, how much of it can be attributed to you or your firm’s actions and how much can be attributed to external factors? Similarly, if you’ve had a disastrous year, can you identify the causes? Was it something over which you had control?

To make good predictions, you need to distinguish those aspects of your life or your business that you can’t control. For each of these, what is most likely to happen? How much variability exists? For example, if you are a manufacturer, what factors affect your supply chain and how likely are they to occur? The 2011 earthquake and subsequent tsunami in Japan led to massive shortages in the automotive supply chain. These shortages affected not only Japanese carmakers, but an estimated 350,000 – 400,000 fewer vehicles were produced in the US due to parts shortages. While it isn’t possible to predict specific earthquakes, Japan is part of the “Ring of Fire”, a seismically active area that stretches around the Pacific from New Zealand to Chile and is home to about 90% of the world’s earthquakes and most of the active volcanoes. Earthquakes are a fact of daily life in Japan, although most are quite minor. They are unpredictable as far as timing, but they are unsurprising due to Japan’s location. It is more surprising that automakers did not already have plans in place to deal with the aftermath of a severe earthquake. In the spring of 2016, two major earthquakes again struck Japan, but this time the impact on the supply chain was less severe – automakers had adopted a policy of multiple sources for parts. While they didn’t know when the next big earthquake would be, they knew it was coming eventually and developed a back-up plan.

Our plans are typically affected by factors that are much more predictable than earthquakes. Most guides to business planning recommend a standard list of items to consider. That’s a great starting point. How can we improve on that list?

A useful planning exercise, developed by psychologist Gary Klein, involves imagining the project you are planning has failed, then coming up with as many plausible reasons for failure as you can. The benefit of the exercise, which is usually done with other members of your workgroup, is that you have to think carefully about threats to your success. In the process, issues often surface about which no one has thought much, but many will recognize as potentially important. These are examples of “unknown unknowns,” to use Donald Rumsfeld’s phrase.

Like earthquakes in Japan, severe weather events can be hard to predict. A truly unusual event, like a blizzard in Georgia, is probably an outlier; but a blizzard in Chicago is a typical winter event. There are regions of the US where floods, blizzards and tornadoes occur often enough to be included as a risk in plans. We can’t really plan for an unexpected extreme event, but we should have contingencies in place for the unsurprising extreme event, the “known unknowns”.

Planning should include estimations of probabilities for events that can affect you or your business, along with what the consequences of those events are. For example, how likely is a significant increase in the price of gasoline? If you drive a hybrid car, it wouldn’t affect you significantly, but someone with a fleet of delivery vehicles could be severely impacted. The probability of the price increase is the same in both cases, but the consequences are very different. Thinking through issues in this way will help you distinguish the risks you should worry about from the ones you can let go.

Along with probability estimates, remember that the probability of independent events co-occurring is always a lot lower than the probability of each occurring separately.

When we don’t incorporate randomness and uncertainty into our thinking, our vision of the future tends to be flawed. We can mistake coincidence for causality and develop false beliefs. We make plans as though the present state of the world will continue into the future. That’s fine as a starting point, but it’s important to remember that the future comes with a range of outcomes, not just the ones we want.

Quick Tips to Deal with Randomness

Before accepting a claim that one thing causes another, ask yourself

· Does the outcome ever occur without the cause?

· Does the cause always lead to the outcome?

· Does the person making the claim have a strong belief about the topic?

When assessing risks, be sure to include the base rate for the risk occurring.

Planning and decisions should include a process to account for the following:

· What is the most likely outcome if you continue your current actions?

· Are you keeping track of what happened as a result of prior decisions?

· What are the uncontrollable factors in your situation?

· What is the range of outcomes that could result from uncontrollable factors?

· Are you paying attention to base rates?

When you hear news about polls, health, the economy, and potential risks, ask yourself

· Who says so?

· How do they know?

· Is the quantitative information communicated appropriately?

· Are comparisons being made on the same scale?

1