There is clearly a strong correlation between the percentage of a region's non-White population and the rate of violent crime in that region. Combining Blacks and mestizos for simplicity (hey, the FBI does it to Whites, so fair's fair), we get the picture shown in the graph entitled Crime Rate versus Percent of Population that is Black or Hispanic (it's in Part 9). The correlation coefficient for that data spread is +0.85, indicating that there may be some sort of relationship between the variables being graphed. (There is a class of 'spurious correlations' in which two variables both depend on a third, so a correlation of two graphed variables doesn't prove a direct cause-effect relationship between them.) Notice the point in the upper right corner of the graph. That point represents the District of Columbia, which has a higher percentage of Blacks (68%) than any US state does. Combined with mestizos, DC is 71% non-White. Let's remember that percentage, because I'm going to use it to test a theory about the non-linearity of the dependence of crime rate on the percentage of Blacks in a region's population.

Consider, now, the data in the following three graphs showing the crime rates for the most populous counties in Georgia, Alabama and Virginia.

Hey Liberals! Do you think that I put any of these points in the wrong place? Check my work!

Hey Liberals! Do you think that I put any of these points in the wrong place? Check my work!

Hey Liberals! Do you think that I put any of these points in the wrong place? Check my work!

Notice the upward-swinging curve of the least-squares fit to the data in each case. The equations of the curve fit are given on the charts, but I'll repeat them here.

Georgia:
Y = 11.1973 (1.04516)^X

Alabama:
Y = 13.2727 (1.04551)^X

Virginia:
Y = 4.98918 (1.06096)^X

where Y is the FOUR TIMES the murder rate in "murders per 100,000 population per year",

where X is the percentage of the population that is Black or Hispanic.

More precisely, Y is the combined murder rate per 100,000 population for the years 1990, 1991, 1992 and 1993 for the selected counties in these states.

Remember that Washington, DC, had a "Black or Hispanic percentage" of 71%. If we were to substitute 71 in the equations for those states, in order to predict the violent crime rate of Washington, DC, we would get

Y = 257.7 murders per 100,000 population per FOUR years (using the Georgia model)

Y = 312.8 murders per 100,000 population per FOUR years (using the Alabama model)

Y = 333.2 murders per 100,000 population per FOUR years (using the Virginia model)

Y = 299.5 murders per 100,000 population per FOUR years (using the overall model)

Dividing by four to get the annual murder rate predictions for the District of Columbia,

Y = 64.4 murders per 100,000 population per year (GA model)

Y = 78.2 murders per 100,000 population per year (AL model)

Y = 83.3 murders per 100,000 population per year (VA model)

Y = 74.9 murders per 100,000 population per year (overall)

In 1990, there were 472 murders in the District of Columbia. The population of DC that year was 606,900 persons. Thus the actual murder rate for DC in 1990 was: 77.8 murders per 100,000 population per year.

(In 1991, it was 80.6. In 1992, 75.2. In 1993, 78.5.)

Having given the data (and their extrapolation) more thought, I consider that it is unlikely that the exponential fit will continue to hold up for much past 70 percent. I think that it is more likely that the rise of Black dominance in an area shifts the rate of violent crimes from one (nearly) linear pattern to another, with the change-over occurring somewhere around a 40 percent Black infestation. Perhaps there would have to be a transition curve between the White dominant and Black dominant patterns. More data is needed to answer the question in a definite way.