The causes of the black-white education gap

It is common knowledge nowadays that there is a black-white education gap. What is not so well-known on the other hand is that this gap was steadily shrinking until the 1980s. This convergence was due to an increasing black achievement. But then this process abruptly stopped in the late 1980s.

The gap once again widened to some degree until the early 2000s. In essence, much of the progress was wiped out by this divergence. Black high school graduation rates in 2003 were back at their 1972 level.

The trend also seems to have been driven more by males than females. These facts are illustrated in the figure below. The left plot is the difference in black-white high school graduation rates between 1965 and 1997. The right one is the same broken down by gender.

The Black-White education gap

Evans, Garthwaite and Moore (2014) (ungated and more up-to-date version) have an amazing paper on this topic. Their hypothesis is that the emergence of crack was an important factor behind the widening of the education gap. The idea is to look at differences in the arrival of crack across various metro areas and see whether it explains the drop in educational achievement. So cities where crack arrived earlier should see a drop in their educational achievement earlier.

The arrival of crack. Fortunately, there is enough variation in the arrival dates of crack by city. The table below summarizes when the authors estimated crack arrived in large (population >800,000) metros.

Arrival of crack by city

What really affects the arrival date of crack? There are two main predictors: proximity to initial crack hubs and a higher population both lead to an earlier adoption of crack. The initial crack hubs were New York, Los Angeles and Miami. The graph below (based on figures reported in the paper) shows the minimum median driving distance from these hubs by year of crack arrival. We can clearly see that crack appeared earlier in cities closer to the hubs.

Distance to hubs vs. crack adoption date

An interesting corollary of this analysis is that the date of crack adoption is not related to economic characteristics (i.e. poverty, unemployment) of the cities.

The adverse effect of crack on young black males. The arrival of crack had a much more pronounced negative effect on the lives of young black males. This can be seen by looking at murder and incarceration rates. Murder rates on average did not vary much in the 1980-1995 period, but this overall trend hides significant changes in particular subgroups. Specifically, young people saw a huge increase in murder rates in the early 90s. Moreover, this was mostly driven by young black males as the figures below show.


But was this due to crack? Since crack was mostly affecting the youth, the increase in murder rates does seem to be connected to its appearance. This hypothesis is further corroborated by the authors when they show that after the arrival of crack the murder rates of young people increased substantially more than the murder rates of 40+ year-olds in the same racial/gender category.

After the arrival of crack, the murder rates of 15-24 year-old black males increased by 94% more than 40+ year-old black males’. For similarly aged white males, this figure is 51% when compared with 40+ year-old white males. The effects are strongest in black males, then white males. They’re the weakest in females (where little if any racial differences exist). The figures below show the differential increase in murder rates for black and white males.

Murder rates vs. the appearance of crack

In every category, young people saw their murder rates rise significantly more when crack was introduced than older people. The younger the group considered, the more murders increased. This is thus some indirect evidence that it was crack that was increasing murder rates, and much more among young black males than among any other group.

We can also look at incarceration rates to see that the arrival of crack had a more pronounced effect on young black males. This is well illustrated by the figure below.

Prison intake by race, 1980-2000

Such a huge increase in the prison population was partially brought about by a shift in federal policy. Mandatory minimum sentencing for drug offenders was introduced, and crack-related offenses were generally punished much more harshly than those related to other drugs. To get an idea of the extent of this, take a look at this quote from the paper:

California […] made crack cocaine count for between two and four times the multiple of its actual weight in sentencing. […] Michigan […] introduced one of the largest sentencing disparity laws in 1989 with a ratio of sentencing to actual weight of 75 to 1. In 1990, Maryland introduced a sentencing disparity between crack cocaine and powder cocaine at a ratio of 9 to 1 , while North Dakota and Alabama set a 10 to 1 ratio in the same year. New Hampshire enacted a sentencing disparity of 28 to 1 in 1994, and a year later a sentencing disparity of 10 to 1 was introduced in Ohio.

High school graduation rates vs. crack. If it is indeed the appearance of crack that’s causing the divergence in graduation rates, then we should see that divergence happened earlier in cities where crack appeared earlier. The authors fit two trends to their data: a linear trend for black graduation rates relative to white ones before and after the arrival of crack. One would expect both to be statistically significant with the former’s coefficient having a positive, the latter’s a negative sign. This would indicate that there was convergence in pre-crack, and divergence in post-crack years.

This hypothesis is tested with various specifications including metro and state-level data, using place of residence and place of birth to assign people in the sample to a geographical area, various controls for the environment (such as income or unemployment), and also controls for school environment such as expenditures per student or racial fractions in the schools.

Altogether, the above hypothesis is supported by the data. The main specification implies that black males were gaining an annual 0.23 percentage points in graduation rates on white males in pre-crack years. In post-crack years, however, they started diverging at an annual rate of 0.30 percentage points. Both figures are statistically significant and they are statistically different from each other. Convergence and divergence rates for females are similar in direction (though somewhat less significant) but smaller in magnitude, as expected.

What we can see is that there does seem to be a clear convergence trend in pre-crack years, and a clear divergence in post-crack years that is not driven by environmental changes such as unemployment, poverty (as measured by income) or school-related things.

The severity of the crack problem and educational attainment. Next, the authors look at by how much a more severe crack problem could decrease high school graduation rates in a given city. To proxy for the severity of the crack problem, the researchers construct a perceived mortality rate and perceived incarceration rate. These two are purported to measure the risks of getting killed or being sent to prison for a teenager; both of which are expected to reduce investment in education.

The perceived mortality rate is calculated for 16-18 year-olds as the murder rate of 20-24 year-olds in the same ethnic/gender group. The perceived incarceration rate is calculated similarly.

Data on perceived mortality is available on the metro-level, and it is shown that indeed perceived mortality significantly and robustly decreases high school graduation rates for black males. However, for black females it has no effect.

The perceived incarceration data is more limited. In a sample with 15 states, it is shown that the perceived mortality and incarceration rates together explain about a 7 percentage point drop in black high school graduation rates. In a larger (but more limited) sample of 36 states, this effect drops to around 4 percentage points. In both cases, perceived mortality and incarceration both explain about half of the drop.

An alternative measure of crack severity is also used as a robustness check. This one also shows no effects for black females. As for black males, this index indicates that in the 1983-1993 period about a 2 percentage point drop in high school graduation rates can be explained by the increasing severity of the crack problem.

Robustness. One concern may be that the strong negative relationship between the increased risk (of mortality or incarceration) and educational attainment is caused by some third variable. For instance, urban decay or some other economic processes could cause both of these things. This would mean that what we see is just spurious correlations.

To test this, the researchers check whether the risk-education relationship exists in other time periods as well. Sure enough, it is there and similar in magnitude in pre-crack years alone. Using different measures of risk (aggregate death rate and non-homicide death rates) also supports a negative relationship between risk and educational attainment.

There doesn’t seem to be any difference in the magnitude of this relationship between races either. This indicates that there was a huge shock to the (mortality/incarceration) risk rates of young black males (but not other racial/gender groups), which caused a drop in their educational attainment relative to other groups’.

Alternative mechanisms. The authors finally consider whether alternative mechanisms could have widened the gap. But they find (mostly based on data from other papers) that (i) kids are unlikely to drop out of school to deal crack, (ii) teenagers did not use crack, and (iii) changing family structure and/or school quality did not affect black males differently or more harshly than other groups.

Moreover, based again on other research, the authors conclude that the changes seen in murder and incarceration rates could have reasonably had a large enough effect on educational attainment to explain most of the divergence.

The present and the future. Violence in the crack market has subsided in the late 1990s and early 2000s. This is because the market matured – so to speak -, and boundaries/property rights had been established. Indeed the downward trend in black educational attainment stopped, but growth in attainment has not reached its pre-crack levels.

This may partially be explained by the fact that while murder rates declined, the risk of incarceration (prison intake) for young black males still remains high. A second reason could be that there was a huge negative shock to the human capital of the black population during the years crack appeared (mid-1980s to mid-1990s). Given how important intergenerational transmission is for things like educational attainment, this could have delivered a huge blow to the community as a whole.

There can then be two possibilities: either this large negative effect will slowly dissipate as the community recovers from it in the long-term (assuming that no other large negative shocks come); or this big shock pushed human capital investment onto another (lower) equilibrium. In this latter case recovery is not expected (again unless some other (positive) external changes/shocks happen).

Which of these two situations hold and what can be expected in the future is definitely an interesting and important direction for future research. For those interested, check out a related post on the educational investment of minorities here.

Finally, I don’t normally embed videos in my posts, but I feel this one complements this article quite well.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s