Trends, Hot Hands, and Patterns

you may.jpg

Steph Curry (an American basketball player) is having one of those games. He's seven of nine, five of six on threes. He's matched against an average defender, and the team he's playing has a good defensive scheme. It's just a case of Steph being Steph. Everyone, from the casual Dubs fan to coach Steve Kerr wants the ball in his hand. As the half winds down, Klay Thompson gets him the ball and he takes the jumper. Is he more likely to make the shot or miss it?

Thirty-two years ago, three psychologists published a paper "On the Misperception of Random Sequences"--also known as the "hot-hand fallacy." They looked at several data sets and concluded that the previous performance of the player had no effect on the likelihood he'd make any given shot. Steph Curry had a 50-50 chance (his regular shooting percentage), no matter how hot he looked. Nevertheless, a large number of people believe in the hot hand theory--that a shooter on a streak is going to continue on that streak. As the paper puts it, "Our survey shows that basketball fans believe that a player's chance of hitting a basket are greater following a hit than following a miss."

There is a closely related phenomenon dictated by the same failure of logic: that current performance--say of a segment in the beer market--is predictive of the future. Data may exist that helps one predict the direction of a company or market segment, but recent performance is not it. I bring this up because I'm getting tired of reading think-piece articles talking about the state of craft beer that depend entirely on recent performance. Up until two years ago, articles about craft beer touted craft beer's amazing growth potential by citiing ... craft beer's growth. This was the data set:

The trend line had a magnetizing, persuasive effect. Except for a few morose characters warning of a bubble, most people took this to be a durable trend. Many people opened expensive new breweries or invested in expensive new equipment. The number of breweries more than doubled between 2010 and 2014. No trend is permanent, however; the craft segment continued to expand, but the growth percentages declined. The line in the chart above started bending earthward.

In 2015, articles were bundling together brewery and segment performance in statements like these: "The craft beer industry — led by Sam Adams producer Boston Beer, Sierra Nevada and New Belgium — has reported sharp volume growth in recent years as consumers favor smaller brewers willing to try new flavor concoctions and market their home-grown, small business stories." Now the story has reversed and you get ledes like this: " Some of the country’s biggest craft brewers are struggling with falling sales, hurt by a glut of competitors crowding retail shelves and moves by megabrewers to scoop up some of their rivals."

What's revealing is the way even seasoned economic reporters take this sole data point and begin inventing causal connections to wholly speculative theories. In that first quote, "consumers favor smaller brewers willing to try new flavor concoctions." Really? The second quote puts declining growth on clogged markets and rival actions. Do we have a source for this speculation? No, of course not. Next, take The Economist, which this July published "Craft Beer in America Goes Flat." After reporting on declining growth, the magazine (sans byline) offers these strange "reasons" for the change:

The dip is the result of two problems, one old and one new. First, the consumption of wine and spirits is growing more quickly than that of beer, and has been for nearly 20 years. Women are drinking more booze but often prefer wine and spirits. Men are turning to a wider range of drinks, including whisky and wine.

The second difficulty is that after years of effervescent growth, craft beer has gone flat. Volumes grew in 2016, but half as quickly as in 2015 (see chart). In the 13 weeks to June 17th craft-beer sales and volumes both dropped, by 0.7% and 1.5%, respectively.

To account for sales that began declining two years ago, The Economist offers two of the lamest reasons I've ever read. First, craft growth, white hot three years ago, is now declining because of a twenty year trend toward wine and spirits. What?  But hey, that's at least a theory. The second "reason" is pure tautology: growth has stalled because "craft beer has gone flat." Rule of thumb; it's bad analysis when the cause and effect are identical.

And yet, this is a perfect example of the way humans think and the way so much of the analysis of the craft beer market has been over the past twenty-five years. We are pattern-recognizers. Our brains are wired this way, and it has on balance been very good for our individual and collective survival. But it does lead to the occasional gross error when we see patterns that aren't actually there.

If you look at any historical trend line for anything, you see it bounce around. There's even a term for this--"noise." Focus in on too brief an interval and all you can see is noise. And even when you pull back to see a greater span of time, you see longer periods of rising and falling. Many go roughly in one direction, but bounce around along the way. Some go up (think of the price of movie tickets), and some go down (think sales of leisure suits). Some mostly go in one direction but have long-term mini trends. Consider the Dow Jones, which has a a fairly consistent century-long trend line.

Complex conditions account for long-term trends. There are some suggestive conditions that make one wonder if the current growth in the craft beer category isn't due to contract for awhile. But there are other conditions that make it plausible that current slowing growth is just a blip--statistical noise. There are reasons certain breweries may be doing well while others do poorly that have nothing to do with larger trends. The thing is, unless you have confidence you understand the relevant variables, looking at current performance tells you nothing. Your mind may want to see evidence in it, but a trend isn't itself evidence of anything.

I have no idea whether craft beer will stay on its current course; I know I don't have all the relevant variables. But the mere fact that craft beer has slowed over the past two years is evidence of nothing. And that I do try to remind my pattern-recognizing brain of that whenever I remember.