He's actually right. Just as an example: (Links are outdated but you can still find the studies if you want to)

Kovandzic et al. 1998](http://onlinelibrary.wiley.com.sci-hub.tv/doi/10.1111/j.1745-9125.1998.tb01259.x/full) took data from 190 cities looking for homicide variables. He produced six regression models, three for different measures of economic inequality and three for stranger, acquaintance, and family homicide. Across all six, % black remained statistically significant with all of the other variables being held constant. Not only that, but it also yielded the highest standardized regression coefficient (Beta) value across all models. Thus it was a better predictor of homicide than income inequality, poverty, unemployment, divorce, etc. This demonstrates that both family structure and economic inequality among other things don't adequately explain racial disparities in crime.

Kposowa et al. 1993 did the same thing across 3083, 409, and 1469 counties respectively. Once again, % black was a better predictor of homicide than education, unemployment, poverty, divorce etc. The amount of variance explained in rural and main was much lower than urban, but that's because rural crime is much harder to predict.

Kposowa et al. 1995 replicated these findings with better and much more robust models.

There's also some data from The Color of Crime 2005 which looked at government data for all 50 states + D.C. and produced a few neat bivariate correlations:

Violent Crime Variable (per 100,000) with Correlation

% black and hispanic |0.81|

% in poverty |0.36|

% unemployed |0.35|

% not completed high school |0.37|

% black and hispanic (all other variables controlled for) |0.78|