Data Normalization

Demographic groups are not evenly represented in the population. Canada’s population is >70% White, but White people only account for 35% of police involved deaths. Meanwhile, Canada’s population is only 5% Indigenous and 3% Black, but these groups account for 18% and 8% of police involved deaths respectively.

View Image in New Tab

In order to account for this, we need to Normalize our data. Normalization, is the process of scaling one variable by another. For example, to find the proportion of Canada’s population made up by each racial group, we can divide the population of each racial group by the total population. This tells the percentage of Canada’s population each group makes up. Normalizing by the sum of a dataset is the simplest example of data normalization. Doing the same operation to the police involved deaths data allows us to plot them side by side because they are on the same scale (Percentage).

The Police Involved Death Rate

Comparing the proportion of police involved deaths for each demographic group to their respective proportion of the population is informative. However, its more meaningful to combine the police involved deaths and the population into one statistic. The Death Rate is the number of police involved deaths per unit of population (ie. million) per unit of time (ie. year).

This does not describe the whole picture, because there are large disparities in the police involved death rate between demographic groups. The Death Rate for Indigenous and Black people are 3.8 and 2.3 per million per year. The White people, it is only 0.5 per million per year.

View Image in New Tab

Statistical Testing

What would we expect from a uniform proportional distribution? A Chi Square test can be used to check if the observed number of police involved deaths by racial group is significantly different than what would be expect if the PKR were equal between racial groups. Here is an example for the country as a whole.

• If Systemic Racism did not exist in Canada, as the RCMP commissioner Brenda Lucki claimed in a press conference … This same rate would apply to each demographic group. We can get the expected distribution by multiplying the population of each demographic group by the total rate and the record length (21 years).
  • The Chi Square test, will compare the expected and observed distribution to see if the deviations in the observed involved deaths are beyond what would be randomly expected.
  • The test is significant to p < 0.0001, meaning there is VERY STRONG evidence showing of systemic racial bias. This distribution is impossible by chance.

Power_divergenceResult(statistic=533.1042844134683, pvalue=6.0930082003129934e-111)

• Even assuming all Unknown are White, which is highly unrealistic, the test is still shows irrefutable evidence of systemic racial bias.

Power_divergenceResult(statistic=364.8967491570321, pvalue=7.959950923149648e-75)

---