On his Substack, scientist Noah Carl looks at the same Financial Times article on “gun violence“ and social trust in America that Steve posted on, above.
John Burn-Murdoch is one of the UK’s best data journalists, having produced dozens of great visualisations in his role as chief data reporter at the FT. Which makes his latest article on gun violence in the US something of an outlier.
Burn-Murdoch claims that “trust plays a significant role in helping to drive gun violence”. To support this claim, he presents a chart showing a strong positive association between interpersonal trust and the gun homicide rate across US states. It’s the third from the left:
When he combines his measure of interpersonal trust with a measure of gun ownership, the resulting “score” is strongly associated with the gun homicide rate (final chart above). This leads him to conclude: “it’s the interplay between guns and fear that sends homicide rates climbing”.
Incidentally, interpersonal trust seems to be doing most of the explanatory work here. As Twitter user AnechoicMedia points out, the bivariate relationship between gun ownership and gun homicide rate is non-significant. Though to be fair to Burn-Murdoch, he states in the article that the relationship is “weak”.
But does trust really play “a significant role” in gun violence?
It’s long been known that black Americans reportsubstantially lower trust than white Americans. We also knowthat black Americans are disproportionately involved in gun violence. I therefore checked whether the association between trust and gun violence is confounded by the black share of the population. [Emphases added]
The somewhat overdone emphasis added, by me, is because Steve Sailer did a CTRL+F search of John Burn-Murdoch’s FT article and did not find the words “black” or “African.”
Noah Carl’s conclusion:
Trust’s strong negative association with the gun murder rate becomes weak and non-significant when % black [i.e the percentage of the local population that is black]is included in the model. What’s more, adding trust to a model that already includes % black increases the variance explained by only 1 percentage point. % black is the most important predictor, single-handedly explaining 60–66% of the variance.
