15 December 2018

Education, Substance Abuse and Prison

Executive Summary

Earning a high school diploma in the ordinary course, taking some college classes after earning a high school diploma in the ordinary course, and earning either a two year or a four year college degree after earning a high school diploma in the ordinary course, all lower the likelihood that someone will be incarcerated in prison incredibly (almost completely). If a person with those educational achievements also does not have a moderate to severe substance abuse problem, they are extremely unlikely to be incarcerated in prison.

People who end up in prison, overwhelmingly, have cognitive and behavioral health issues. Less than a quarter of Colorado prison inmates graduated from high school in the ordinary course (as opposed to getting a GED) (compared to about 85%+ of the general population), and just 1% have earned any college degree (compared to about 43% of the general population). Substance abuse is a problem for 79% of inmates. Just 6% of Colorado inmates both have a high school diploma (as opposed to a GED) and don't have a moderate to severe substance abuse problem, and some of the inmates in that 6% have moderate to severe mental health problems. The percentages in Colorado are typically nationally. (Although it is worth noting that mental health issues are much more of a risk factor for people without educational attainment and people with substance abuse problems than for people wh have educational attainment and don't have moderate to severe substance abuse problems).

While pursuing educational opportunities while in incarcerated reduces a person's risk of reoffending by about 15% (regardless of the exact nature of the educational program pursued), this is not nearly so protective as educational attainment prior to any incarceration.

As a result, someone who has earned a GED (which often happens while incarcerated), often is at more of an elevated risk of incarceration than someone who has never graduated from high school and does not have a GED, since many people in this category are simply immigrants from places where secondary education was less universal or had difficulty obtaining a high school diploma in the United States because they were not fluent in English. But, immigrants commit crimes at lower rates than comparable native born citizens (something that illustrates the "fit immigrant" hypothesis).

In contrast, a native born American who speaks English as his or her primary language, frequently doesn't graduate from high school due to discipline problems or other behavioral issues, rather than because they aren't intelligent enough to master the academic material. So, a high school diploma is more of an indicator of socialization than it is of academic ability in the United States for native born individuals.

The extreme protective effect of educational attainment, while notable in and of itself, also illustrates a point that the practice of sealing juvenile crime records otherwise obscures. This is that the lion's share of adult felons were engaged in disruptive behavior and frequently in criminal activity, before becoming adults. The earlier someone has run ins with the juvenile justice system, the more likely it is that they will have run ins with the adult criminal justice system.

More generally, "original sin" is a more accurate metaphor for reality than "innocence at birth". Almost everyone becomes less likely to engage in disruptive anti-social behavior as they get older. Young children are more violent than older children, and children are more violent than adults, and the age of an offender when released from prison is one of the strongest predictors of recidivism, because typically offenders "age out" of blue collar criminal activity as they get older. An older child or adult may have a greater ability to successfully attempt to commit a serious crime than a younger person, however, which is why the most serious crimes aren't committed by young children.

Education is so protective because someone who is incapable of refraining from anti-social behavior is unlikely to earn high school diplomas, to have steady enough attendance to make taking college classes make sense, or to earn college degrees. If you are sufficiently well socialized to graduate from high school and college before you are first incarcerated, you are unlikely to engage in the more disruptive behavior needed to get you incarcerated for a felony as an adult after you graduate because people get tamer with age, nor more criminal.

While the primary protective effect of age before a first offense and education comes from being socialized not to commit serious crimes, a secondary part of the protective effect is that people with educational attainment don't have extreme economic need that can pressure someone to commit a felony because the vast majority of felonies for which people are convicted have some economic motive, and a tertiary part of the protective effect is that people without a prior criminal record who have some kind of educational attainment are much more likely to be treated leniently in the criminal justice system when they do commit a felony, which is also, often a less serious felony.

High school dropouts are more often unemployed than anyone else, are more often fired for conduct or poor performance, get the lowest paying and least desirable jobs when they do find work, so they are often in poverty. And, judges imposing sentences and district attorneys negotiating plea deals usual see incarceration of high school dropouts as less harmful to society than the incarceration of people who have educational attainment and have demonstrated an ability to function in society in the past.

There is more analysis below the break.

Analysis

This data is from a post in 2011 describing 2009 data from the Colorado Department of Corrections, but the numbers haven't changed much in the meantime and Colorado is very typical of the nation in most similar kinds of statistics nationally:
Educationally, just 1% of those admitted to prison had an associates degree or more education although about 11% have some college, while 37% lacked a high school diploma with 36% being at least functionally illiterates who needed adult basic education instruction, rather than high school level GED instruction which would be too advanced for them. About two-thirds of those with either a high school diploma or GED had a GED rather than a high school diploma. So, less than a quarter of Colorado prison inmates graduated from high school in the ordinary course. In Colorado as a whole, 11% lack a high school diploma or GED, 89% of the age 25+ population has at least a high school diploma or GED, 65% have at least some college, 43% have an associates degree or higher degree, and 33% have a bachelor's degree.

About 8% had an IQ of under 81. A moderate to severe mental health problem is an issue for 30%. A moderate to severe substance abuse problem is an issue for 79%. A moderate to severe medical problem is present in 15%. Sex offenders make up 11% with another 5% suspected of having sex offense histories who are not convicted. An absence of adequate skills to get a job is a factor for 42%. Mental health needs differed considerably based on gender. A moderate to severe mental health problem was an issue for 22% of men and 55% of women. . . .

Only 30% of inmates without a substance abuse problem have a high school diploma and 24% have neither that nor a GED.
Another factor not mentioned is that about half of prison inmates in the Colorado are gang members. 

The Colorado Department of Corrections doesn't track the percentage of inmates who have suffered traumatic brain injury, but the percentage is much higher for inmates than it is in the genera population.

RELATIVE RISK

Highest Education Level..........General Population..........Prison Population.....Prison/General Ratio

2 or 4 yr college degree................43%....................................1%.......................0.023
some college, no degree...............22%..................................10%.......................0.455
high school diploma or GED........24%..................................52%......................2.167
no high school diploma or GED...11%..................................37%......................3.364

If you want to compare the relative risk ratio of two different education levels you have to divide one by the other. So, for example, the relative risk of someone with no high school diploma or GED being in prison relative to the risk that someone with either a 2 year or 4 year college degree being in prison is 3.364/0.023=146.3 to 1.

Relative Risks:

no high school diploma or GED v. 2 or 4 yr college degree: 146 times higher risk
no high school diploma or GED v. some college, no degree: 7.39 times higher risk
no high school diploma or GED v. high school diploma or GED: 1.55 times higher risk
high school diploma or GED v. 2 or 4 yr college degree: 94.2 times higher risk
high school diploma or GED v. some college: 4.76 times higher risk
some college, no degree v. 2 or 4 yr college degree: 19.8 times higher risk

Distinguishing High School Diplomas from GEDs.

This really understates the extent to which a high school diploma as opposed to a GED is protective, however.

What percentage of people with high school diplomas or GEDs actually have only a GED?

I estimated above based on a source examined in the previous post that it was about two-thirds.

About 3.8% of the U.S. population over the age of 25 has a GED (8.5 million people over age 25 with GEDs v. 221.42 million people over age 25 from here and here). Colorado is probably typical of the nation, so assuming that 4% of the population in Colorado has a GED but not a high school diploma is plausible. About 40.8% of people with a GED go on to get some college, although only about 6% of people who get a GED get either a certificate or a college degree of some kind. So, only about 2.2% of people have a GED but no high school diploma and no college.

If these figures are right, overall about 123,400 people in Colorado have a GED but no other educational credential (if the percentage is 4%), while about 7,967 people in Colorado who have a GED but no other educational credential are in prison. Thus, about 6.5% of people who have GEDs but no other educational credential are in prison and a significant share of people who have a GED but not other educational credential who are not in prison earned their GED while they were in prison.

The census bureau, however thinks that the GED only percentage is more like 4% in the general population (suspiciously close to the percentage who have earned a GED, whether or not they have gone to college) v. 20% in the prison population. If these figures are right, about 224,300 people in Colorado have a GED, and 4,592 people in prison do, so about 2% of people with GEDs are currently in prison (although a larger percentage earned them there).

The U.S. Justice Department has determined in an April 2003 report that in state prisons that 39.7% of inmates lack a high school diploma or a GED, 28.5% have a GED, 20.5% have a high school diploma only, and 11.4% have some post-secondary education. About 26% of State prison inmates said they had completed the GED while serving time in a correctional facility (just 9.1% of state prison inmates and 9.4% of federal prison inmates with a GED earned it while not incarcerated). About 2.4% of those with some post-secondary education in these statistics were college graduates or more. The U.S. Justice Department also says that 4% of the general population has a GED only.

The U.S. Justice Department percentage of inmates with post-secondary education is the same as the percentage in Colorado, and the percentage of inmates with no high school diploma or GED nationally in state prisons of 39.7% is only slightly higher than the 37% of Colorado, which makes sense given that high school graduation rates in Colorado are modestly above average. Thus, 58% of students with a high school diploma or GED only in state prisons have a GED only nationally.

The U.S. Justice Department figures are probably the most accurate for the percentage of prisoners with a high school diploma or GED only who have a GED.

This impacts the relative risk ratios a lot for high school graduates v. GED earners.

Highest Education Level..........General Population..........Prison Population.....Prison/General Ratio

2 or 4 yr college degree................43%....................................1%......................0.023
some college, no degree...............22%..................................10%......................0.455
high school diploma only.............20%................................21.8%.....................1.090
GED only......................................4%................................ 30.2%....................13.73
no high school diploma or GED...11%..................................37%......................3.364

The inversion of risk between non high school diploma or GED v. a GED only, to some extent, probably reflects the fact that a significant share of people in the Colorado's general population with no high school diploma or GED are probably older, fit immigrants who grew up in countries where dropping out of high school didn't have the same social significance, or at a minimum, where immigrants who arrived as fairly old children who didn't speak English fluently and didn't graduate from high school because they were not fluent in English, rather than because they were either academically incompetent or socially disruptive. 

This dovetails with the fact that 61% of non-citizens in state prison but only 38% of citizens in state prison, had neither a GED nor a high school diploma (according to the U.S. Justice Department). Similarly, the percentage of state inmates with neither a GED or high school diploma was 27.2% for white inmates, 44.1% for black inmates, and 53% for Hispanic inmates. Of 10.9% of white state inmates, 11.7% of black state inmates, and 27.9% of Hispanic state inmates never even made it to 9th grade. Meanwhile 14.9% of white state inmates, 10.0% of black state inmates, and 7.4% of Hispanic state inmates had at least some college.

The percentage of white state inmates age 20-39 without a high school diploma was 27.7% v. 13.9% in the general population, for black state inmates age 20-39 it was 43.9% v. 15.6% in the general population, for Hispanic state inmates age 20-39 it was 51.8% v. 41.3% in the general population. So, Hispanic state inmates are much more similar to the general population in education than white or black state inmates.

Both estimates of risk are somewhat misleading, however, because earning a GED while in prison reduces your chance of reoffending by about 15% relative to not earning a GED while in prison, if you enter prison without a high school diploma. The benefits of participating in any kind of education program in prison in terms of recidivism rates are similar. As further explored below, a high school diploma, some college attendance, and college degrees are all significantly more likely to reduce the odds of incarceration when obtained before ever being incarcerated than they are when obtained while incarcerated.

It is probably more illuminating to lump people without a GED or no high school diploma, with people who have a GED only. This gives you:

Highest Education Level..........General Population..........Prison Population.....Prison/General Ratio

2 or 4 yr college degree........................43%....................................1%....................0.023
some college, no degree.......................22%..................................10%....................0.455
high school diploma only.....................20%................................21.8%...................1.090
no high school diploma (incld GED)...15%................................67.2%...................4.480

Subcategories Of Those With Some College or Degrees

It is also worth noting that a significant share of people with a 2 year college degree in prison earned that degree while in prison, either after having a high school diploma out of prison or earning a GED in prison. Similarly, a significant share of people with some college in prison took some college classes in prison, either after having a high school diploma out of prison or earning a GED in prison.

According to the U.S. Justice Department, 54% of state inmates with a high school diploma or GED, 60% of state inmates with a GED, 42% of state inmates with a high school diploma, and 43% of those with post-secondary education had participated in education programs since they were most recently incarcerated in prison. About 9.9% of state prison inmates and 13.9% of federal prison inmates take some college courses while in prison. Realistically, at least 90% of state prison inmates and 80% of federal prison inmates with some college have only taken college classes while in prison.

Nationally, about 2% of inmates earn an associate's degree while in prison and 4-11% have taken college classes but not earned an associates degree while in prison (the percentage is definitionally unclear, the percentage earning a bachelor's degree while in prison is negligible). At any given time, probably on the order of 1% of inmates have an associates degree earned in prison and 2-8% of inmates have taken college courses for the first time while in prison.

This implies that the risk of incarceration reducing status of some college education earned outside of prison prior to incarceration in prison, and of a 2 year or 4 year college degree earned outside of prison prior to incarceration in prison, is probably underestimated.

Other Kinds of Correctional Supervision

Federal prison, which disproportionately has white collar prisoners (and makes up a quite small share of all prison inmates), has more educated inmates than state prisons: 26.5% lack of GED or high school diploma, 22.7% have a GED only, 27.0% have a high school diploma, and 23.9% have some post-secondary education. People with higher educations are, relatively speaking, much more likely to end up in federal prison than in state prison.

In local jails, according to the same data, 46.5% lack of GED or high school diploma, 14.1% have a GED only, 25.9% have a high school diploma, and 13.5% have some post-secondary education. One way to interpret these numbers is that more "regular" people get jail time than prison time, but that the least educated often end up in jail for minor offenses but don't have the wherewithal to commit a felony.

Of people serving probation sentences, according to the same data: 30.6% lack of GED or high school diploma, 11.0% have a GED only, 34.8% have a high school diploma, and 23.6% have some post-secondary education. People with higher educations are, relatively speaking, more likely to be sentenced to probation (even if they commit felonies) than they are to be sentenced to prison, in addition to committing serious crimes that demand prison sentences much less often.

4 comments:

Guy said...

Hi Andrew,

Off topic (at least somewhat), I read of IQ having a normal distribution when it's clearly bounded at the bottom. If someone was 3 sigma above normal they are prime candidates for grad school. Someone with the same size negative delta would be institutionalized. But if it's a Poisson distribution are the sigmas the same above and below the norm?

Cheers,
Guy

andrew said...

I understand what you are trying to get at. There is a significant subset of individuals at the low end of the IQ range who due to environmental causes (e.g. lead poisoning, or improper prenatal and early childhood nutrition or de novo single gene mutations causing autism) have cognitive impairments that are not due to the same ordinary variation in alleles of a massively polygenetic trait within a population genetic community. Often, these individuals have IQs so low that would be vanishing rare in a population but for this anomalous negative strike. There is basically nothing, in contrast, that gives that kind of boost on the high end.

As a construct, I think it is useful to think about heredity apart from de novo mutations with a large effect that lead to cognitive impairment, as establishing a maximum potential IQ in the absence of environmental factors or de novo mutations of large effect that prevent that potential from being reached. And, that potential IQ curve is pretty much a Gaussian normal distribution, but one that is shifted higher than the distribution of actual IQ. It isn't really bounded at the bottom per se.

There are some good statistical reasons to think that this is the case. First, we know that discrete causes of about 50% of cases of retardation that are environmental or due to specific rare individual mutations or genetic defects of known large effect. Second, the amount of variation in IQ explained by heredity is much higher in high SES individuals than it is in low SES individuals (basically poverty or near poverty) for whom the amount of variation in IQ explained by heredity is much lower. Third, you see large changes in the average IQs of entire populations (such as Flynn effects when they increase rapidly) that seem to correspond close to large changes in stature and big improvements in nutrition and public health and universal education (potential IQ is to some extent a use it or lose it thing).

andrew said...

Instead, the frequency of very low IQs in the upward shifted potential IQ curve is so low that the very small number of people who have very low IQ due to mere random variation get lost in the sea of people with very low IQs for which there are specific environmental or de novo mutation causes. Most people are severely retarded for a reason that has a very particular cause specific profile, while a minority of severely retarded people have a very general indistinct lack of cognitive capacity across the board for no really apparent cause except that most of their close relatives also have fairly low IQs within the normal range of variation. Also, while it has only a quite mild selective fitness effect, the rare person who by statistical fluke ends up with virtually all of the low end IQ alleles that existing within normal variation in the population often aren't fit enough to survive and reproduce. But, since the vast majority of people with some low IQ alleles (basically everyone) doesn't have almost all of them, the rare person who has all of them doesn't have much of an effect on an particular low IQ allele within the range of normal variation that is sufficient to purge it from the gene pool very strongly in any given generation.

This gives you a bimodal distribution, with one corresponding to a more or less Gaussian distribution but slightly skewed to the lower tail and away from the upper tail as a result of potential IQ minus mild environmental effects, and then a second mode far below the higher mode which is more irregular and only vaguely Gaussian to reflect Gaussian Bell Curve variation around several main degrees of large discrete impairment in characteristic amounts from particular causes whose magnitude also varies as a very narrow and steep Bell Curve.

This isn't readily apparent in IQ statistics, because there is no natural metric by which to measure the degree of IQ difference between to levels of performance on an IQ test other than the historic but largely abandoned practice of benchmarking to the cognitive abilities of a child of a given age relative to the age of the person being tested for children. Instead, what is done is that raw scores on IQ tests are force fit to a normal Bell Curve based upon percentile ranking. So, the fact that there is actually a bigger gap between IQ 100 and IQ 70 than there is between IQ 100 and IQ 130, by any reasonable consistent unit by which to measure the amount of cognitive ability differences, is obscured by the way that raw IQ test results are calibrated to reflect IQ test scores.

You can get a pretty decent approximation, however, by adding to Gaussian Bell Curves of IQ scores together to reflect the two modes and properly weighting the percentage of people who are in the ordinary Bell Curve and the impaired Bell Curve, even though the reality is more complex.

"The Poisson distribution . . . is a discrete probability distribution that expresses the probability of a given number of events occurring in a fixed interval of time or space if these events occur with a known constant rate and independently of the time since the last event. The Poisson distribution can also be used for the number of events in other specified intervals such as distance, area or volume."

This distribution is not appropriate for measuring IQ. First of all, the Poisson distribution is a probability distribution that applies for natural number inputs like numbers of events with a clear zero and very clearly defined events to tally. Secondly, the Poisson distribution is unimodal rather than bimodal, and the ordinary potential IQ curve from which the bulk of the apparent IQ distribution is derived with minor environmental adjustments is closer to a Bell Curve and less lopsided between the upper tail and the lower tail than a Poisson distribution.

andrew said...

CORRECTION: "You can get a pretty decent approximation, however, by adding to Gaussian Bell Curves of IQ scores together to reflect the two modes and properly weighting the percentage of people who are in the ordinary Bell Curve and the impaired Bell Curve, even though the reality is more complex."

Should have said:

"You can get a pretty decent approximation, however, by adding the two Gaussian Bell Curves of IQ scores together to reflect the two modes and properly weighting the percentage of people who are in the ordinary Bell Curve and the impaired Bell Curve, even though the reality is more complex."