This article takes a look at the population racial split of the largest U.S. states, and any differences in who is employed in those states. For simplicity and consistency with Department of Labor data, only self-identified White individuals were analyzed versus all others. The vast majority of the non-Whites are African Americans (or Blacks), and for this non-parametric analysis the results were so great in just some states that further refinement (say per industry) wasn't necessary to provide guidance. With the focus lately on the racial circumstances surrounding purportedly-prejudiced law enforcement, this is just one perspective to jointly learn more about disparity, and probability measures. Others may show analysis on the demographics, specifically on police activity, though that's not what's analyzed here.
What we do show, however, is that in a third of the most populated 15 states (with an evident bias towards the most populated of these 15 states) there is a modest tilt towards less favorable outcomes for non-Whites. There are a few of the middling of the most populated 15 states (e.g., Georgia and North Carolina), but which really move the needle by having a relative depression in jobs availability and a disproportionate number of non-Whites in the state. There is otherwise not much of a pattern, as we have states such as Michigan (with its gutted industrial base similar to Saint Louis area or to Baltimore) that appear slightly more balanced among the most populated states! And it should be striking that even as this state hosts a center city of Detroit, we haven't seen the type of civilian uprising that we have seen in other post-industrial areas.
For each of the 15 largest U.S. states, we looked at the population levels for Whites, and non-Whites. We also looked at the job holders in those states by race. Comparisons were made versus an ideal scenario where Whites and non-Whites were employed equally, in equal proportion to their underlying state population. In other words, race made no probabilistic difference when looking at the data. Differences in survey statistics can occur within many categories that don't matter as much when examining racial adversity (e.g., discriminatory bias), such as Whites who are unemployed, or from non-Whites who are employed. For more on the topic of probability and inequality, without regard to race, see one of our more popular articles here.
So the focus of this study was to customize a non-parametric statistic that would correctly do two difficult things jointly: (A) give a mathematical direction to the disparity statistics, and (B) provide normalcy in the overall statistic so that it could be further studied with advanced machine learning analytics.
The highest quality metric is shown below under "NONPARAM NORM", which also satisfied a number of normalcy statistics tests, such as Shapiro Wilks. One can see the three probabilistic clusters partitioned meaningfully, from the data.
We also see in the orange section, the inclusion of the overall "U.S.", where the White population is in the high 70% range, and their employment:population ratio is roughly 1 percentage point higher than the overall employment rate, regardless of race. When it comes to labor economics, a single percent difference in outcomes is a bit. Also for completeness, the U.S. is simply included in the table but it was not included in the numerical comparisons versus the other states.
Also just below Michigan, we see the most balanced state of Florida in the turquoise cluster. And further below that we see states such as Arizona and Texas, where there is much more enjoyment of economic positions by non-Whites. Or not as much amplification for non-Whites from lack of employment opportunities. Last, we see in the red cluster Georgia, which is the median of the 15 most populated states.
We also see one of the larger of these 15 states (New York), where the White employment rate is 1.5 percentage points more than the state average, regardless of race. And the White unemployment rate there is 6.6%, or higher than the U.S. average for Whites. Yet the overall averages, regardless of race, is 7.6% or higher and by a greater spread versus the U.S. average. So while this is not the problem everywhere, it is in this red cluster of the largest 15 states (where you can see the highest statistics scores shown), that we see the most acute adverse differences for non-Whites in their proportional attainment of jobs relative to Whites.
What we do show, however, is that in a third of the most populated 15 states (with an evident bias towards the most populated of these 15 states) there is a modest tilt towards less favorable outcomes for non-Whites. There are a few of the middling of the most populated 15 states (e.g., Georgia and North Carolina), but which really move the needle by having a relative depression in jobs availability and a disproportionate number of non-Whites in the state. There is otherwise not much of a pattern, as we have states such as Michigan (with its gutted industrial base similar to Saint Louis area or to Baltimore) that appear slightly more balanced among the most populated states! And it should be striking that even as this state hosts a center city of Detroit, we haven't seen the type of civilian uprising that we have seen in other post-industrial areas.
For each of the 15 largest U.S. states, we looked at the population levels for Whites, and non-Whites. We also looked at the job holders in those states by race. Comparisons were made versus an ideal scenario where Whites and non-Whites were employed equally, in equal proportion to their underlying state population. In other words, race made no probabilistic difference when looking at the data. Differences in survey statistics can occur within many categories that don't matter as much when examining racial adversity (e.g., discriminatory bias), such as Whites who are unemployed, or from non-Whites who are employed. For more on the topic of probability and inequality, without regard to race, see one of our more popular articles here.
So the focus of this study was to customize a non-parametric statistic that would correctly do two difficult things jointly: (A) give a mathematical direction to the disparity statistics, and (B) provide normalcy in the overall statistic so that it could be further studied with advanced machine learning analytics.
The highest quality metric is shown below under "NONPARAM NORM", which also satisfied a number of normalcy statistics tests, such as Shapiro Wilks. One can see the three probabilistic clusters partitioned meaningfully, from the data.
Also just below Michigan, we see the most balanced state of Florida in the turquoise cluster. And further below that we see states such as Arizona and Texas, where there is much more enjoyment of economic positions by non-Whites. Or not as much amplification for non-Whites from lack of employment opportunities. Last, we see in the red cluster Georgia, which is the median of the 15 most populated states.
We also see one of the larger of these 15 states (New York), where the White employment rate is 1.5 percentage points more than the state average, regardless of race. And the White unemployment rate there is 6.6%, or higher than the U.S. average for Whites. Yet the overall averages, regardless of race, is 7.6% or higher and by a greater spread versus the U.S. average. So while this is not the problem everywhere, it is in this red cluster of the largest 15 states (where you can see the highest statistics scores shown), that we see the most acute adverse differences for non-Whites in their proportional attainment of jobs relative to Whites.
No comments:
Post a Comment