5.5: Spurious Correlation

When we find a significant correlation between one condition and another, it is tempting to assume this indicates that one condition causes the other. Indeed, often it does. The high correlation between flipping the light switch and the room lighting up is explained by the former action causing the later condition. But a correlation between one condition and another doesn’t aways work this way. It can also, for instance, be that both conditions have a common cause. For example, night routinely follows day, but day does not cause night. The correlation we find in night following day is caused by the rotation of the planet as it orbits the sun.

Official crime rates among Black Americans are higher than they are among white Americans. The statistics here need to be understood in the context of an assortment racial biases in the criminal justice system. This is not just a matter of individual police officers being racially biased, though some are. There are also a number of systemic factors involved. Poor neighborhoods are more heavily policed and these tend to be more racially diverse. The crack cocaine epidemic that plagued Black communities was aggressively prosecuted an sent many Black people to prison. The current wave of opioid addiction that more often afflicts white communities is treated with more a more compassionate approach. So, there is a good deal of institutional racial bias built into the official crime statistics. But even if we bracket these injustices, the correlation between crime and race in official statistics is spurious.

The gap in official crime rates between Black Americans and white Americans leads a significant number of people to the conclusion that Black people are just more criminally prone, as if race alone could explain this. But the racist conclusion is not well supported. The gap in crime rates closely mirrors the gap in unemployment rates between Black Americans and whites. Both official crime rates and official unemployment rates are higher among Blacks by similar factors. This suggests a causal explanation for higher crime rates among Black Americans that makes a good deal of sense and doesn’t attribute innate criminality to Black Americans as racists would have us believe. Typically, people turn to crime only when they are deprived of decent opportunities in life. Regardless of race, people with good jobs and some prospects for a decent life have lots to lose and won’t be very tempted to risk it all on criminal activity.

Correlations call for explanations. There is an inference to the best explanation involved in this. So, let’s recall how the surprise principle discussed in the last chapter works. The explanation that makes the correlation we want to explain least surprising is the one that explains best and is thereby inductively confirmed. The idea that race somehow explains criminality is rather mysterious. There have been many racially motivated attempts to substantiate this idea and none have panned out. But higher crime rates among people who have been denied opportunities in life is not at all surprising. Inference to the best explanation strongly favors the idea that unemployment is a significant causal factor in crime over the idea that race somehow explains crime. So, inference to the best explanation indicates that the correlation between crime and race is spurious, not causal.