BLACK LIVES MATTER Analyzed – A Conflict of Narrative Within the Left

Black Lives do Matter.

Black Lives Matter. Police reform is a critical issue that urgently requires attention and action. What happened to George Floyd was immoral, unlawful and a national disgrace. My wife is black. My children will be black. I want to live in a world where opportunity is not predicated on the color of a person’s skin. I want to live in a world where the law is enforced fairly and consistently across all demographic groups. 

Ultimately, I want Martin Luther King’s dream to come true:

“I have a dream that my four little children will one day live in a nation where they will not be judged by the color of their skin but by the content of their character.”

I agree with the self-evident expression “Black Lives Matter.” However, I’m concerned that the organization's narrative is grounded in ideology instead of empirical evidence.

Race. And only race.

The following passages from the About Us section of BLM’s website paint an unambiguous and bleak picture of what it is like to be a Black American in 2020. ¹

“Black Lives Matter Foundation, Inc is a global organization in the US, UK, and Canada, whose mission is to eradicate white supremacy and build local power to intervene in violence inflicted on Black communities by the state and vigilantes. By combating and countering acts of violence, creating space for Black imagination and innovation, and centering Black joy, we are winning immediate improvements in our lives.”

“We are working for a world where Black lives are no longer systematically targeted for demise.”

The common theme of the BLM narrative, which is voiced by activists and largely reinforced without reservation by the media, is that racism is either the sole cause of disparities between Black Americans and White Americans, or that racism is by far the most significant contributor to the variances between the two demographic groups. 

In this worldview, racism is generally characterized as “systemic.” Thus, the BLM notion of racism is very different than the MLK notion of racism, which referred to bigoted beliefs that individuals held about individuals belonging to other demographic groups.

BLM’s use of the term “systemic racism” implies that American institutions, laws, and cultural norms are structured in a manner that perpetuates racial inequalities in virtually all imaginable categories, from wealth to education to COVID-19 mortality. According to this view, omnipresent structural racism is the main cause of the perceived dire state of many Black American communities. 

This structural machine is said to nurture and sustain white supremacy, which is another term that has been redefined. White supremacy is no longer the ideology of Nazis and the KKK. Instead, it refers to the alleged outcomes of systemic racism, which harm minority communities. Thus, any disparity between Blacks and Whites that is in Whites’ favor is an outcome of white supremacy.

This causes individuals who hold this view to focus almost entirely on race when determining causality, instead of considering other potential contributing variables (in addition to race) to racial inequality. 

The Left's schism

For some time, there has been a conflict within the political Left. Both sides agree on the harmful and expansive consequences that result from large racial disparities. The conflict is centered around the variables that need to be addressed to rectify the issue.

On the one side are those who hold the BLM view, which is the intellectual offspring of academic Critical Race Theory. Critical Race Theory, and other intellectual descendants of the original Frankfurt School’s version of Critical Theory, was undoubtedly influenced by French postmodern philosophy. However, since the 1990s, scholars like Kimberlé Williams Crenshaw, Patricia Williams and others have significantly differentiated Critical Race Theory perspectives from orthodox postmodern perspectives (most notably via the insistence of an overarching meta-narrative of racial determinism). Critical Race Theory designates race as the cardinal factor driving racial inequalities due to systemic/structural racism (particularly in America). Additionally, adherents to the BLM/Critical Race Theory view tend to be hostile to the consideration of additional potential causes of racial disparities.

On the other side are those whose worldview is grounded in Marxist ideology more historically characteristic of the far-Left. Unsurprisingly, in addition to race, this faction believes that class is a critical driver of racial inequality.

A recent incident with the Democratic Socialists of America illustrated this schism. 

Adolph Reed, Professor of political science at the University of Pennsylvania, is one of the most prominent Marxist academics of the past several decades. A prolific Socialist scholar and critic of Obama-style Neoliberalism, Reed was raised in the segregated South and became an anti-war activist in the 1960s before venturing into academia.² Having “walked the walk” for his entire life, to most it seemed that he would be a perfect fit for a DSA lecture. 

However, Reed’s longstanding belief that the Critical Race Theory portion of the Left erroneously overstates the impact of race and ignores the impact of class has rubbed some Leftists the wrong way for quite some time. Unconcerned with offending his audience, Reed planned to talk about how the Left’s heavy emphasis on how COVID has disproportionately harmed Black Americans has lowered the likelihood of multi-racial organizing.² 

His history coupled with his planned topic was too much for the DSA’s New York City chapter to handle, so DSA leaders cancelled the talk.

Professor Adolph Reed, Jr.

The DSA’s Afrosocialists and Socialists of Color Caucus claimed that Reed’s planned lecture was “reactionary, class reductionist and at best, tone deaf.” It does not appear that the caucus was concerned about racial reductionism. They continued:²

“We cannot be afraid to discuss race and racism because it could get mishandled by racists. That’s cowardly and cedes power to the racial capitalists.”

Many found it shocking that the Democratic Socialists of America cancelled a Black Marxist scholar. Reflecting on the incident, Reed stated:²

“An obsession with disparities of race has colonized the thinking of left and liberal types. There’s this insistence that race and racism are fundamental determinants of all Black people’s existence.”

Obviously, Reed’s concern is not that race isn’t an important variable to consider when attempting to understand the drivers of racial disparities. Rather, Reed is concerned that the tunnel vision focus on race makes multiracial organizing, which he believes is necessary for substantive social change, less likely to occur. He believes that myopic narratives (which could arguably be construed as racially divisive) don’t generate diverse coalitions.  

Reed is far from the only thinker on the Left that exhibits a concern about “racial reductionism,” which refers to the univariate designation of race (via systemic racism) as the sole causal factor of racial inequality. For example, Touré Reed, a Professor of History at Illinois State University, recently penned a fascinating book, Towards Freedom, which echoes and elaborates on Reed’s concern regarding the BLM/Critical Race Theory Left’s tendency towards univariate oversimplification when ascribing causality to social phenomena. 

Professor Toure F. Reed

Which view does the data support?

Police Brutality
Figure 1: Total Police Shootings
Figure 2: Total Police Shootings of UNARMED Persons
Figure 3: Total Police Shootings (Chart)
Figure 4: Total Police Shootings of UNARMED Persons (Chart)

Terms like “systemic racism,” “white supremacy,” and other similar terms are not falsifiable because they’re defined so loosely. Yet, when activists use them, they act as if these terms are concrete, empirical terms like unemployment, illiteracy, or labor participation rates. These terms can be measured. “Systemic racism” is a vague, miasmatic moniker that cannot be quantified; it’s an abstract concept. However, just because a statement is not empirically falsifiable does not mean that it cannot be true; it just means that we cannot employ the scientific method to understand or test the claim. 

Some scholars have argued that the nebulous definition of systemic racism is an example of the logical fallacy of reification. Reification means that an abstraction is treated like a concrete thing. It is very easy to measure the education levels of two different groups to falsify the null hypothesis of equality in education. It is not easy to measure the level of systemic racism in an institution, state, or country. The claim that America is a systemically racist nation is not empirically falsifiable.  

The fallacy of reification should not be used as an excuse to dismiss concerns of systemic racism. We need to “steel man” the argument of systemic racism. Perhaps systemic racism is an umbrella term that refers to a collection of racial inequalities, such as wealth levels, incarceration rates, and proportion of Fortune 500 CEOs. So, in this view, systemic racism is a composite term. The individual phenomena within the domain of the composite term are measurable and falsifiable, so we’ll proceed with this collective understanding of systemic racism and tackle each issue separately to avoid the reification fallacy. We’ll analyze police brutality to see if race is the only or dominant causal factor.

Source of Data

The Washington Post administers a database that tracks all police shootings in America since 2015.³ Figures 1 through 4 above are produced from that database (updated 8/10/20). Figures 1 and 3 show the same information and Figures 2 and 4 do as well.

Not much can be deduced from the information above without proper context. While not entirely, much of the anger, and much of the conversation around police brutality, is directed at police killings of unarmed individuals. Due to availability bias and the media’s anecdotal coverage of shootings, most Americans feel that police shooting an unarmed person is an event with a high probability. 

Figure 5: Bureau of Justice Statistics Report (2018), "Contact Between Police and the Public, 2015"
Figure 6: U.S. Census Bureau Population Estimates (7/1/19)

According to the Bureau of Justice Statistics (shown in Fig. 5), in 2015, 21.1% of Americans 16 or older had some contact with police. We’ll use this as the average number of police interactions per year. This does not take population growth into account, which means that it’s a conservative estimate. The 21.1% translates to about 53,506,941 individuals with some degree of police interaction annually. Between 2015 and 2020, the highest number of unarmed police shootings in one year was 90 (in 2017). We’ll use this maximum as our annual number of unarmed police killings. Let’s assume that the 53.5 million individuals with some degree of police contact only had ONE interaction with police. Thus, the number of police interactions per year equals the number of people with a police interaction: 53.5 million. Obviously, this is a highly inappropriate assumption considering that many individuals have multiple interactions with law enforcement per year. However, this ensures that our estimate errs on the side of conservatism. 

If these unrealistic assumptions (zero population growth and a maximum of one police interaction annually) held and the average number of unarmed killings per year was the maximum level of the recorded five-year time span, it would mean that .0000168% of police interactions end up in the shooting of an unarmed person. Put another way, for every 10 million police interactions, there will be between 1 and 2 shootings of an unarmed person. Due to our unrealistic assumptions, the actual probabilities are much lower. While this shows that the marginal probability of being killed by the police while unarmed is extremely low, it does not help us evaluate whether law enforcement is racist against Black Americans (a critical component of claims of systemic racism). 

Without doing any analysis, the data shows that blacks are shot, both armed and unarmed by police, at a disproportionate rate when compared to other races. According to the U.S. Census Bureau, Blacks make up 13.4% of the population but account for 24% of all shootings and 30% of unarmed shootings. However, Native Americans are shot unarmed even more disproportionately. They account for 1.3% of the population but make up 17% of all unarmed police shootings. Whites comprise 76.3% of the population but only comprise 45% of total police shootings and 35% of unarmed police shootings. 

However, disparate outcomes alone do not provide conclusive evidence that American police are systemically racist against Black citizens. To provide us insight, we’re going to conduct a series of analyses on the Washington Post database using linear and non-linear regression models to ascertain whether race is the main driver of the probability of being shot unarmed by police officers.   

The variables listed below comprise the matrix X shown in Figures 7 and 10 (where UNARMED is a scalar value and ε is a column vector):

  • Year of individual’s shooting
  • Whether individual was armed (UNARMED)
  • Age of individual (age)
  • Race of individual
  • Gender of individual
  • State where shooting occurred
  • Whether individual showed signs of mental illness (MENTAL_ILLNESS)
  • Whether individual attempted to evade arrest or flee (NOTFLEEING)
  • Whether the police officer was wearing a body camera at the time of shooting (BODYCAMERA)

Linear Probability Model (LPM)

Figure 7: LPM

The linear probability model employs basic ordinary least squares (OLS) methodology to regress the binary independent variable UNARMED, denoting whether an individual was unarmed when he or she was shot by a police officer, on the independent covariates. The coefficients of the regression are added together to estimate the probability of being shot unarmed by police. Figure 9 includes fixed effects for the year and the state while Figure 8 does not control for these elements. Thus, Figure 9 is likely a more accurate representation of the population model.


In Figures 8 and 9, we see both expected and counter-intuitive results. Regarding the latter, the Figure 8 results indicate that a police officer with a body camera, other things equal, is 2.24 percentage points (pp) MORE likely to shoot an unarmed individual, and that a male is 4.14 pp LESS likely to be shot unarmed by a police officer than a female. When state and time variables are controlled for, the BODYCAMERA variable becomes more statistically significant and the coefficient increases to indicate a marginal effect of 2.56 pp on the probability of an individual being shot unarmed. The gender variable remains significant and the coefficient is negligibly affected by the additional covariates. 

While it’s difficult to come up with any rationale for why females would be more likely to be shot unarmed than males, it’s possible that the body camera result is a consequence of the moral hazard problem. The moral hazard problem is illustrated when people with health insurance feel empowered to engage in behavior that’s more detrimental to their health than if they were not insured. In this case, if an officer is confident that the body camera footage will vindicate him from accusations of wrongdoing, he may be more likely to shoot an unarmed suspect. 

Another potentially counter-intuitive result to the general public is the statistical insignificance of the MENTAL_ILLNESS variable: both models do not provide evidence to conclude that individuals showing signs of mental illness are more likely to be shot unarmed by police. However, when the fact that most people who suffer from mental illness are not violent is considered, this finding actually becomes intuitive.


Unsurprisingly, whether or not the individual is fleeing police or resisting arrest has a significant impact on the probability of being shot unarmed. Both models show this. When state and year fixed effects are controlled for, the coefficient for the NOTFLEEING variables increases in both magnitude and statistical significance. Figure 9 indicates that, other things equal, if an individual flee or resists arrest, the probability of being shot unarmed increases by 1.9 pp.

Also as expected, both models indicate that age is a significant driver of the probability of being shot unarmed by police. Other things equal and according to Figure 9, every one year increase in age is associated with a .14 pp increase in the dependent variable. While this marginal effect might seem trivial, we need to think about how sizable differences in age affect the dependent variable: other things equal, a twenty-year old is 2.7 pp more likely to be shot unarmed by police than a forty-year old.

Regarding the effects of race on the probability of being shot unarmed by police, there are some interesting changes in the predictions that occur when state and year fixed effects are controlled for (look at the differences between Figure 8 and Figure 9).

Figure 8: LPM #1 - No Fixed Time or State Effects
Figure 9: LPM #2 - Fixed Time & State Effects Included

Notably, when these additional covariates are included in the regression, the Hispanic racial coefficient changes signs and loses statistical significance. Additionally, both models provide little to no evidence that Native Americans and Asians are more likely to be shot unarmed by police than other races. 

The racial coefficients for Black individuals and White individuals are relatively unaffected by the inclusion of fixed effects in the regression. According to Figure 9, other things equal, a Black individual is 1.96 pp more likely to be shot unarmed by police than a White individual.


So, we DO have some non-trivial evidence to suggest that Black Americans are more likely to be shot unarmed by police than White Americans. However, we need to be cautious about this conclusion for several reasons.

First, the Washington Post’s database does not include a number of pertinent variables. For example, Adolph Reed and other academics would likely conclude that the results are flawed because there is no measure of class, income or wealth. Education levels, home environments, and whether or not the shooting occurred in an urban or rural setting are just a few other variables that could influence the probability of being shot unarmed by police. The R-squared values are .0157 and .0247 for Figures 8 and 9, respectively. This means that the independent covariates in Figure 8 only explain 1.57% of the variance in the dependent variable. When further controls are added in Figure 9, this measure only increases to 2.47%. 

While R-squared is a crude measure of a regression model’s fit, the low values do indicate that there are many other variables that need to be considered for us to make a firm conclusion. While 1.96 percentage points is not a trivial difference between Blacks and Whites, it does not match up with the BLM’s description of Black Americans being “systematically targeted for demise.” At the very least, the LPM suggests that there are many more important variables to consider than race alone, AND that race is not the most impactful covariate on the dependent variable.

It should be noted that the addition of more controls could INCREASE the significance and magnitude of the Black variable, meaning that our predicted probabilities described above could actually UNDERSTATE how much more likely Black individuals are to be shot unarmed by police than White individuals. 

Probit Model

Figure 10: Probit Model

Unlike the LPM, whose predicted probabilities are not restricted to the values between 0 and 1 (as all valid probability density functions are), the non-linear Probit’s fitted values are limited to the appropriate domain. In this model, the fitted probabilities are a function of the normal cumulative density function and the Xβ index is preserved. Figures 11 and 12 show the marginal effects of each covariate on the dependent variable. As with the LPM, the latter model includes fixed time and state effects. 


The predicted marginal effects of a body camera on the likelihood of being shot unarmed by police align very closely with the LPM’s predictions; the differences are trivial. The same is true for the gender variable, the Age variable, the variable denoting signs of mental illness, and the NONFLEEING variable. 

Like the LPM, the Probit does not provide evidence to support the claim that being Hispanic, Asian, or Native American has a significant impact on the probability of being shot unarmed by police. 


The Probit predicts a larger variance between the White coefficent and Black coefficient. According to Figure 12, which controls for fixed state and year effects, other things equal, a Black individual is 3.26 pp more likely to be shot unarmed by a police officer than a White person. While this is notably higher than the LPM’s prediction of 1.96 pp, it does not vindicate a view of racial reductionism that someone like Adolph Reed would criticize.

Figure 11: Probit #1 - No Fixed Time or State Effects
Figure 12: Probit #2 - Fixed Time & State Effects Included


This exercise is (obviously) not meant to replace the work of eminent scholars like economists Roland Fryer and Glenn Loury. Rather, I hope that it shows the danger of univariate reasoning, of which racial reductionism is JUST one example. 

When, due to political or ideological loyalties, we ascribe all causal power to one or two phenomena, we ensure that our understanding of the world is dangerously simplified. Solutions that are crafted based on such an understanding will inevitably fail. 

There’s no doubt that race plays an important role in racial inequality. However, as this exercise has shown, it’s not the only driving variable. Univariate reductionism is narratively satisfying but statistically dubious. I agree with Adolph Reed in that class must also be considered when evaluating the drivers of racial inequality. 

Sources Referenced

  1. Black Lives Matter. (2020) “About Us.”
  2. Powell, Michael. (2020). “A Black Marxist Scholar Wanted to Talk About Race. It Ignited a Fury.” The New York Times.
  3. The Washington Post. (2020). “Fatal Force: Police Shootings Database.” 
†. Bureau of Justice Statistics. (2018). “Contacts Between the Police and the Public, 2015.” U.S. Department of Justice.
‡. U.S. Census Bureau. (2020). “Quick Facts.”

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.