Ice Cream and Murder
Aaron Shaffer, Psy.D.
Consultant, coach, change leader, neuroscience geek, AI savvy, data scientist, cyclist, capitalist building high-performance people systems
Headline - ice cream sales are connected to murder rates. As ice cream sales rise and fall, so do the number of homicides. Is this the universe trying to tell me to ease off on the rocky road? That stuff is dangerous! What is it with ice cream that causes people to kill each other? Or is it that violence causes people to get a sweet tooth?
The key insight for this piece is that Correlation Does Not Mean Causation!
Although the relationship between ice cream sales and murder is absolutely true, it illustrates a flaw in the way most companies, researchers, the media, etc. often consume data. In this age of ‘Machine Learning’ and Artificial Intelligence, we can develop a blind confidence in ‘an answer’ being ‘the answer’. We often mistakenly assume that just because variables are related, that one causes the other.
Ice cream sales and murder rates really are correlated. When one variable changes, so does the other. Shoe size and income operates in the same way. The higher the shoe size, the more likely you are to earn more money. Does that mean bigger feet cause a higher income? Sweet! I wear a US size 13, and I think Bill Gates only wears a 10 ½. Or is the causal relationship more likely due to gender differences…. Hmmm???
Although variables must be correlated to be causally related, as in the case with ice cream and murder, there are often other variables involved that are actually causing the main effect. In the case of ice cream and homicide, the cause is likely seasonality. Ice creams sales rise in the summer, and more murders occur in the summer. Windows are open, people are out, etc.
Being a smart consumer of data is a critical business (and life) skill, and we’re finding more and more organizations are asking us how to get the right insights out of their data analytics investment. Although their business strategies, industries, and geographies often differ, we’re seeing common gaps in research design and statistical knowledge.
So how do we know when data is causally related? How do make sure we’re working with the right variables? What’s the key to deriving meaning from our data? Sorry for the cliff hanger, but we’ll tackle that in future posts.
Human Resource Executive
8yGreat article Aaron.
CEO and Dir of Application R&D, introducing executives to humaneering - 21st-century OS for the human side of business (humaneeringtech.com). Resolve people problems, unleash achievement, delight employees and customers.
8yIf we look closely, we see that the perilous "assumption of causation" pervades virtually every application of human science in the practice of organization management (e.g., recruiting, interviewing, compensation, incentives, performance, leadership, etc.). This is just one of the many reasons why many if not most organizations need a wholesale updating of today's still widely used yet fundamentally obsolete Industrial Era organization management principles and practices. We delight in reading about the organization-management innovators, yet may not realize that they are very few, and even those trying to copy these innovators are doing so without questioning their obsolete assumptions about work, workers, and workplaces. One of the major threats that organizations face from the fast evolution of technologies for managing is the potential to unwittingly automate outdated management practices, thus doing the wrong thing better, faster, and easier . . . and with fewer people around to workaround the resulting issues. Perhaps even a greater problem than assumptions about causation is the fact that organizations and the people working and managing within them are each living systems (vs. isolatable, cause and effect physical/machine systems) and so, no matter how well you try to isolate a single variable or jiggle the dials of statistics, the only way to study and develop each unique organization is to roll up ones sleeves and try higher probability improvements suggested, not by experts who claim to just happen to have the perfect solution on special this week, but by the people responsible for and performing the work. If the organization climate is toxic, then you might need some facilitators to first help you bridge the communications gap between workers and management, and second to help conduct the experiments so they get a reasonable opportunity to work.
cXo Headhunter talent:acquhired …ever serving, learning…
8yTimely, timeless.
Consultor, Auditor e Instrutor - Quality Lead Auditor, Lean Six Sigma Master Black Belt
8yProud, Aaron was my teacher at UCI .. Great reflection