Published in socio.philica.com
Mass incarceration of the scale present in the United States presents a significant bias to social surveys, which rarely ever reach the prison population. By failing to survey prisoner, general surveys of the populaton are biased in meaningful ways, especially within particular demographics.
Since 1980, the incarcerated population in the United States has risen more than 400%, so that it is now roughly 0.7% of the entire population. U.S. prisons contain a subpopulation the size of Houston. This group is by no means a random sample; it is disproportionately male, non-white, and adult. For certain demographics, such as black men in their twenties, the incarcerated subpopulation represents a very large proportion-upwards of 10%—of the total population.
This situation is extreme; of the countries for which data is available, the only one that approaches the U.S. level of incarceration is Russia. It is likely, given the spreads, that no other country, with the possible exception of China, incarcerates such a large proportion of its population. (Mauer 2003)
A great deal has been written about the social outcomes of this project, but very little has been written about its effects on social science per se. Prisoners are not canvassed in any normal surveys of the U.S. population. Their opinions, stories, and vital statistics are excluded from poll and survey results. When prisoners are surveyed, as by the Bureau of Justice Statistics, non-prisoner control groups are not surveyed using the same set of questions. Under any circumstances, this would create a systematic bias. But the fact that prisoners are, by definition, a highly deviant population makes this bias quite extraordinary.
Prisoners differ systematically from the overall population not only in terms of their background, but also in terms of their ongoing life experience. For example, the incidence of sexual assault substantiated by the Bureau of Justice Statistics (BJS) is twice as high as the same figure for the overall population, as measured by the Criminal Victimization Survey (CVS). The rate of reported sexual assault in prison is 14 times higher than the CVS figure.
This creates an odd feedback loop. Public policy, including policies regarding incarceration, often look to social surveys for advice. But those same policies can place people out of the reach of surveys, in statistically significant quantities. It means very little today to poll young black men in the United States, because the population has already been "sorted" and a vast number of them removed from sight.
Conceptually, it also suggests the possibility of an endlessly progressing change in objective conditions which is never noticed by our sampling techniques. Like Niemöller's poem, we can envision a society in which political minorities are quickly arrested. Even if millions of them are incarcerated, polls will consistently demonstrate that there is no opposition to the regime.
There is little that most social scientists can do to address this. Large surveys conducted using stratified samples might be able to survey on both sides of the bars, but only with considerable cooperation from the prisons. Moreover, the extraordinary power dynamics of a prison would tend to badly bias any results. More usefully, we can bear in mind that accurate social statistics are one of the many costs of mass imprisonment.
Beck, Allen J. and Paige M Harrison (2006) Sexual violence reported by Correctional Authorities, 2005. BJS, Washington.
James, Doris (2004) Profile of Jail Inmates, 2002. BJS, Washington.
Mauer, Marc (2003) Comparative International Rates of Incarceration: An Examination of Causes and Trends. The Sentencing Project. Washington.
Information about this Article
Peer-review ratings (from 3 reviews, where a score of 100 represents the ‘average’ level):
Originality = 163.73, importance = 154.92, overall quality = 162.00
This Article was published on 19th November, 2006 at 22:16:47 and has been viewed 7136 times.
This work is licensed under a Creative Commons Attribution 2.5 License.
The full citation for this Article is:|
Mitchell, E. (2006). Mass incarceration and social surveys. PHILICA.COM Article number 51.
1 Peer review [reviewer #82402] added 19th November, 2006 at 23:59:15
The author’s brief article (really a comment) is an accurate observation. However, it is a little trite out of context — in some situations the sampling bias will be relevant, and in others it won’t (e.g., voting behavior, as felons in many states are not allowed to vote). Similar sampling biases of traditional household surveys (involving probability sampling or other sampling designs) are present (and fairly well known among survey researchers) with respect to homeless persons, illegal immigrants, persons working on weekends and evenings (when survey staff contact households), and other categories of people. So a discussion of this particular bias is meaningful in relation to interpreting the results of a particular survey or class of surveys, but doesn’t amount to a contribution outside of this context.
Originality: 2, Importance: 2, Overall quality: 2
2 Peer review [reviewer #2144] added 20th November, 2006 at 04:07:28
I believe that the author has made a valid point.
From a statistical perspective, a sample must fulfill these 2 criterias:
1) It must be random
2) It must be representative of the population.
By excluding prisoners, the above criteria are compromised.
However, the article is rather lacking in concrete data, the only data being percentages. Perhaps the data could be found in the references though.
Originality: 3, Importance: 4, Overall quality: 4
3 Peer review [reviewer #187] added 20th November, 2006 at 10:34:57
I think this is an very important and interesting point. I teach social science sampling and this is an observation that has the pleasant effect of making me think “Of course!”. Had the author confined himself to pointing out that surveys typically do not include the prison population, this would have been a germane point but would have been easily dealt with by mentally substituting the words “non-prison population” for “population” in any survey. However, he goes further than this by reminding us that the prison population is itself a highly skewed one in both socio-economic and ethnic terms, which means the influence of its ommission on population statistics is potentially rather more powerful than we might otherwise think. Well done.
Originality: 6, Importance: 6, Overall quality: 6
4 Additional peer comment [reviewer #187] added 20th November, 2006 at 13:47:46
Having looked again at one of the above reviews, I agree that it would have been interesting to have more data. Perhaps the author could provide a follow-up study with some estimates — it shouldn’t be too hard to use data on the nature of the US prison population to show how a couple of surveys (or elections) might have had their results changed had the prison population been included.
5 Additional peer comment [reviewer #82402] added 20th November, 2006 at 17:00:07
Among survey researchers, this observation is not news. In fact, research articles based on large national surveys make the scope of the underlying samples quite clear. Most national surveys of the type on which the author comments are _household_ surveys, and termed as such. By definition, this excludes persons not living in households, that is, persons living in group quarters (jail/prison, military barracks, college residence halls, motels/hotels, rooming houses, homeless shelters, etc.).