Lies, damned lies, statistics and criminology

Longer texts

Differences in sentencing in similar cases?

This Long TEXT was the second non-peer review of the (now defunct) Dutch website ‘Emperors & Clothes’ (Keizers & Kleren). I discussed the article Differences in sentencing in similar cases that appeared in episode 2012-11 of the Dutch Legal Journal (‘Nederlands Juristenblad’ (NJB)).

In publishing this English translation, I realised how difficult the material it contains is. For a first impression of what is covered in it, one can start by looking only at the text in black letters on a yellow background.

Translated text starts here:

The underlying research is based on observations by students of the same criminology department to which the three authors are affiliated. The students were given observation lists that they filled in at a total of 541 sessions of the police judge. Of the collected data, data from 333 were included in a statistical analysis. The Results section cumulates in a table 2, in which the last row states: “Pseudo R2 (Nagelkerke): 72.1%”. 

‘Nagelkerke’ stands for a calculated value that gives an indication of the extent to which an explanatory model matches (fit) with found data. This size is designed to indicate values between 0 and 1 and is therefore normally not recorded as a percentage …

Before it was published in the NJB, the article was reviewed by ‘peers’. 

After studying the article and the design of the observation list, I not only have critical comments on the article itself, but also question how the NJB  peers have fulfilled their task. Did they feel partly responsible for safeguarding the good name of science? 

The editors were asked directly whether “among the reviewers there was also someone considered to have expertise in the field of statistics”. This was followed by an evasive answer.

I see my plea for non-peer reviewing painfully supported by the uncertainty that the article Differences in sentencing evokes about this.

When you look through the jargon, the piece comes across as an opinion piece for an average Dutch newspaper. However, the authors did not write a ‘letter to the newspaper’:  with great success they sought publicity in their capacity as scientists, via a journal that has its status to lose. Wermink, De Keijser and Schuyt however, explicitly present themselves as working at Leiden University in an area to which the article in question also relates.

Leiden University’s website reported on the publication in a way that one would expect from a sensationalist newspaper. Headline: Suspects with ‘foreign appearance’ receive heavier penalties ¹. The quotation marks turn out to have a very special meaning, as will become clear later.
Further in this article I will first go into the language that is too suggestive for a scientific article. 

Followed by six paragraphs on how the data was obtained and -statistically- processed, and on the nature of the study. 

Then I briefly consider an alternative statement for the findings presented and conclude with some comments on the theory designated as ‘underlying’: the Focal Concerns (FC) theory. 

It turned out that there are two theories with this name

The agreements and differences between the two shed a special light on the prevailing paradigm of Leiden criminology.

Suggestive language in the article

The lead mentions “the odds to disappear behind bars” for “perpetrators” with a foreign appearance who also do not speak the Dutch language”. These odds are “no less than twenty times higher” (My emphasis. All emphasis in quotes is always mine. FG). The immediate association is that this is the case -people who are victims of something– while it is really just about people who have been convicted of a crime or offense: perpetrators, therefore².

In the introduction we also find: 

In previous sentencing investigations (…) In previous research on sentencing (…) a realistic picture of what happens during the court hearing is missing. 

This creates high expectations. After all, the suggestion is irrevocable that this research has paid attention to that reality. In discussing the observation list it will become clear that there is quite a bit of ‘light’ between that which is or could have been observed there and what has been reported.

The Results section starts with this paragraph:

Table 1 shows that a typical offender who is sentenced to prison [comma is missing in original] has been in pre-trial detention, has a criminal record, is in the age group 31 to 40 years, has been convicted of qualified theft and has had a hearing which took place in Amsterdam.

This claim is false.

Table 1 does not show this: the data is insufficiently disaggregated. But also in another sense this paragraph is, to put it mildly, rather striking.

Why does one start talking about a ‘typical perpetrator’ and more specifically a typical perpetrator who received prison sentence? Was the research aimed at bringing them into the picture? How does this relate to the hypothesis which one through research wanted to test? 

What was the hypothesis? 
This is not explicitly formulated. 

We do find several references to the FC theory: “an authoritative social science approach”. According to the authors, that theory presupposes that: 

… the judge, when determining the sentence, takes into account, among other things, the estimated dangerousness of the offender, which also makes characteristics of the person of the suspect play a role in sentencing.

However, that concept of ‘estimated dangerousness‘ as such is no longer reflected in the article and a testable hypothesis do not distract the authors from it, at least not directly. 

Instead, the hypothesis is more or less deducible by combining the paragraphs ‘Punishment’ and ‘The suspect’s person‘.

‘Sentencing’:

The type of punishment that the police judge imposes on the suspect is central to this investigation. More specifically, it was investigated whether an unconditional prison sentence or another penalty has been imposed (Yes = unconditional prison sentence; No = fine, community service, suspended prison sentence, guilty verdict without punishment).

Yes, only the type of punishment.

The other paragraph zooms in on characteristics of the suspects who were mapped through the questionnaires, that is, on a selection from them. The phrase “more specific” is at the same time vague, crucial and incorrect: one has simply all penalties other than unconditional imprisonment lumped together. This has created a dichotomy. A dichotomy that is also required to apply the statistical tool used: logistic regression analysis.

Implicitly and scattered throughout the article, the authors hint why they have (re)divided the various types of punishment into these two categories.

At the end of the Results section we find this:

However, the group that is punished even more severely is the group …

It seems that the typical perpetrator was presented for the benefit of a more marked contrast with another group. On superficial reading it does not stand out, but the following passage from the introduction reinforces that suspicion:

 If judges want  to protect society from recidivism, they would sentence these perpetrators earlier to an unconditional prison sentence, so that in any case temporarily they cannot commit crimes in society.

With ‘these perpetrators’ the authors refer to:

Immigrants or (sic) suspects with a foreign appearance [who] because of stereotypes can be judged more dangerous by judges, for example by their disproportionate participation in criminal behaviour and overrepresentation in the prison population compared to Dutch offenders.

It seems that the authors believe that the intention to protect society from more crimes committed by the same criminal, by imprisoning him, should not be a proper motive for a prison sentence³.

With this sentence from the introduction, that aversion to imprisonment is even more pronounced on top: 

The prison sentence is the heaviest punishment that can be imposed in the Netherlands and has major consequences, both intended as unintended.

That the authors with these ‘unintended consequences’ not only think of the phenomenon that imprisonment can have the effect of a ‘crime school’, is evident from the following quote:

(…) (unintended) effects (…) such as a greater chance of separation and reduced opportunities on the labour market. These relationships (sic) can then lead to a long-term criminal career.

It seems worth taking another look at how in the research, referred to in this context, with the difference between correlation and causation is dealt with, but here I want to draw attention to something else. 

The researchers present both the unintended and the intended consequences to illustrate their view on imprisonment. It defies any kind of logical thinking lumping together these completely different kinds of consequences.

The type of offence

It is also questionable whether this dichotomy of heavy and not or less heavy, is experienced by everyone. By everyone I mean both the convicts as the judges. That doubt arises when we take a better look at the most striking of the presented findings.

The authors themselves do not pay attention to it, but many people will notice something else in be in the presented data than the apparent coherence between types of punishment and external characteristics of the offenders: the special, unexpected relationship between type of offence and type of punishment

The researchers also use in this context only the dichotomy of ‘prison sentences’ versus ‘otherwise punished’.  Table 1 of the article is not sorted on percentages of prison sentences⁴. When we do, we get the following table:

Type of offence (number convictions)  %prison
qualified theft (18)                                     72.2
theft (70)                                                      42.9
other property crime (23)                         34.8
threat (15)                                                   26.7
opium (23)                                                   21.7
serious violent crime (19)                         21.1
minor violent crime (49)                           12.2
other offences (148)                                    7.4

Even ordinary, unqualified, theft became significantly more commonly punished with a prison stay than minor violent crimes, and even more often than serious ones. Also ‘threat’ corresponds to a higher percentage of prison sentences than violence! 

Remarkable. Very remarkable.

Too sharp conclusions cannot be drawn on the basis of this data but it is clear that the claim about ‘the heaviest punishment’ is not as obvious as the authors suggest. Apparently (also) other very different factors play a role in the choice of judges for one type of punishment or the other. One of those statements was suggested by the authors themselves but does not seem to be otherwise studied; I will return to this in a future paragraph. 

Another possible explanation could be related to another gap: in the article nowhere the length of the prison sentence imposed or the amount of the fine etc. is mentioned. The observation list also provided before the recording of this type of data the mere question 7:7 

What punishmen does the judge impose (conditionally/unconditionally)? (literally write along). 

The effect is that, for example, suspended imprisonment, however long, is considered a lighter sentence than unconditional, no matter how short.

The opacity of odds ratios

The findings are essentially cast in the form of odds ratios⁵. Odds ratios are ratios between the number of times occurrences of different outcomes: ‘odds’.
This way of presenting results cries out, as it were, to be misunderstood or misinterpreted⁶. Scientific literature warns against this.

The Wikiwand-lemma on Odds-ratios summarises the problem as follows:

This may reflect the simple process of uncomprehending authors choosing the most impressive-looking and publishable figure. But its use may in some cases be deliberately deceptive. It has been suggested that the odds ratio should only be presented as a measure of effect size when the risk ratio can not be estimated directly.

The wiki authors derive this last warning from, among others, Huw Talfryn Oakley Davies and Iain Kinloch Crombie (link: British Medical Journal, http://www.bmj.com/content/316/7136/989 ), who warn:

If the odds ratio is interpreted as a relative risk it will always overstate any effect size:

the odds ratio is smaller than the relative risk for odds ratios of less than one, and bigger than the relative risk for odds ratios of greater than one.

It is very important to note that:

Odds ratios are a common measure of the size of an effect and may be reported in case-control studies, cohort studies, or clinical trials. Increasingly, they are also used to report the findings from systematic reviews and meta-analyses. (op. cit.)

Especially in that second context, in which no predictions are made or no experiments are done -when it comes to more than two possible values of an independent variable- it is of importance which odds you take as a basis to other odds to compare.

With one variable, that choice is not so strange. The convicts are divided into three groups: 1) people who have a ‘Dutch appearance’ and speak Dutch, 2) people with a ‘foreign appearance’ who speak Dutch and 3) the group that ‘looks foreign’ and does not speak Dutch. The first group has been designated here as the reference group.

For the variable on which the article does not report in the conclusions, the choice of the reference value is not so obvious. According to Table 2 with odds for imprisonment per type of offence, the choice was made to use theft as reference type.

A motivation for this choice is not in the article

The ‘typical offender’ who received imprisonment was guilty of ‘qualified theft’. Court cases related to ‘other offences’ became the most attended. So those were certainly also candidates to be used as a reference value.

The biggest difference for the Exp(B) values for this variable ‘type of offence’ is now found between ‘theft’ and ‘qualified theft’: that factor is over 16, and this means almost as high as that between Dutch speaking and Dutch looking convicts compared to non-Dutch speakers who also did not look Dutch (over 20). 

However, when one had used ‘other offences’ or ‘light violent crime’ as reference value, the value for ‘qualified theft’ would have been significantly higher than the 20 headlines that raised questions from (Social Democrat) Members of Parliament to the minister⁷. (Link: http://www.volkskrant.nl/vk/nl/2686/Binnenland/article/detail/3225546/2012/0 3/14/Research-with-foreign-appearance-rather-the-cell-in.dhtml)

The nature of the study

This choice, and especially the failure to explain it, fueled my distrust and urged me to try to understand the nature of this study and more in particular to look at the way in which statistics have been used here. 

This present investigation did not attempt to fathom anything. Who overlooks the use of concealing language and avoids self-concealing language, concludes that an attempt is being made here to ‘demonstrate’ that police judges ‘discriminate against foreigners’⁸.

The conclusions begin as follows:

The main findings of the study showed that perpetrators with a Dutch appearance who also spoke the Dutch language during the hearing had least chance of (sic) imprisonment.

On the next page but still in the same paragraph we find this:

That men and perpetrators with a foreign appearance who do not speak the Dutch language during the criminal court hearing have a substantially increased chance of an unconditional imprisonment,  could be at odds with the aim of consistency in sentencing in the Netherlands.

The second passage is more correct – after all, that is where the difference between men and women is mentioned – but the meaning of that did not make the publicity. Not even on the university’s website, while you would hope for more accuracy there.

There’s a particular kind of symmetry in the data related to women in comparison with persons with ‘foreign appearance’. 

The groups ‘women’ and ‘non-Dutch speakers with a foreign appearance’ are even exactly the same in number -45 out of 365- but that coincidence only draws attention to the intended symmetry in an amusing way.
Roughly speaking, among the convicts women are underrepresented by a factor of 7, and the people ‘with foreign appearance’ with about the same factor overrepresented compared to the ‘native’ men. When the odds for unconditional imprisonment for both of these groups are compared with ‘native dutch men’ there is about a factor 5 difference: again mirrored.

The headlines about this study could therefore also have screamed: ‘Female criminals punished less severely than male criminals’.

The authors’ comment that the differences found “can stand at odds with the goal of consistency in sentencing”, completely ignores that the groups ‘women’, ‘people with foreign appearance’ and ‘native Dutch men’ also do not behave consistently when it comes to crime. 

The groups ‘men’, ‘women’ and ‘foreigners’ who were convicted were not random samples from all men, women and foreigners: they are different in size and therefore different in nature. There are differences between those complete groups expressed in the sub- and overrepresentation of suspects and convicts. When the same or related differences also lead to a different sentencing you cannot, purely on the basis of the numbers, label that as inconsistent.

The authors themselves suggest the possibility that people: “… can be considered more dangerous than perpetrators with a foreign appearance who speak Dutch, due to an expected lesser bond with society, for example.” What’s wrong with that? Do they really not realise that crime is an expression of less connection with society?

A – not completely- hypothetical example  helps to correct the inaccuracy of the way in which the concept of ‘consistency in sentencing’ is used here, to be penetrated well. Suppose two other subgroups were distinguished among the suspects: nuns over the age of 65 and assassins flown in from abroad ¹⁰. The members of one group never commit crimes of any significance and the other does virtually nothing else. Would it be even remotely distressing or inconsistent when the odds for acquittal versus conviction for one group was 1000 times as large as for the other?

What actually happened at the hearings and the quality of the collected data 

Notwithstanding their claims of “realistic picture of court hearings” the authors base themselves, apart from a single detail, solely on numbers, on numerical values, which could have been determined even without direct observations

They mention both in the introduction and at the end of the article, however, the special contribution they have made by delving into “what happens at the hearing”. 

So this is one of the most shameful aspects of the article.

In the observation list, nine questions were included under the heading ‘personal characteristics of the defendant’, eleven under ‘judge’s motivation’ and 32 under ‘attitude’ suspect’. 

Of those 32, there are only two that come back in the article. 

Well considered, there is only one item in the article that is directly related to the presence of observers at the hearing: the expression of regret. And this turns out to be precisely the aspect that has no influence on whether or not to receive punishment in the ‘heavy category’ unconditional imprisonment. And not only that; while 15 to 20 questions from the watchlist relate directly or indirectly to this aspect of regret, footnote 29 reports on this item: “Observers have on the basis of their own interpretation indicated to what extent the suspect has expressed regret.” Oops.

541 hearings were attended. In the Data section 4 groups sessions are mentioned that have been omitted from the analysis for various reasons. These groups comprise 90, 7, 3 and 76 cases respectively.

The largest of these groups includes several subgroups. Indeed, it seems that much is to be said for disregarding arrested cases, but why cases in which the accused was acquitted have no meaning for the research question is not explained, while it is certainly not obvious.

The group of 76 concerns the hearings at which the convicted person did not appear. Simply omitting this group from the study comes across as a missed chance: after all, this is about the group where there has been no interaction at all between suspect and judge, particularly interesting as a reference group for assessing (the effect of) that interaction.

The omission is motivated by the finding that in those cases there was no question of expressing regret or not. A strange motivation against the background of the observed non-significant correlation with the dependent variable studied. Why are after this finding, these 76 not still added?

The 365 cases that remained are said to be 68% of 541, while normal rounding here would really yield 67%…

The mentioned number of 365 cases also appears in Table 1. The logistic regression analysis, according to the caption of Table 2, is based on 333 of these convictions. Why those 32 have fallen off has not been explained separately.  From Table 1 can be observed that on different variables in respectively  4, 13, 2, 16 and 6 cases the values have not been scored, so together it is 41. There were apparently convicts of whom more than one variable was not recorded. Those 13 include the 10 convicts mentioned in footnote 25 whose observers could not have determined or forgot to fill in whether these are a Dutch or had a foreign appearance, in addition to 3 of whom one could not even determine whether they spoke Dutch or not. Footnote 26 then states: “Of the other defendants who do not speak Dutch [of the 43 who did not rely on the right to remain silent], it is not known whether an interpreter was present or which language they spoke”.

Foreign appearance versus immigrant

For the way in which the item that did make it to the article was handled, “oops” is really too weak an expression. This is an unacceptable distortion.

The whole article and all the publicity which was generated with it, starting with the website of their own university, is full of references to the ‘foreign appearance’ of the convicts, although on that website they did use quotation marks. However, the observation list contained, at least in the version which was sent to me by one of the authors, no questions at all about the appearance.

However, question 5 did ask about the “country of birth” of the convicted person and question 6 reads as follows:

Origin:
o Native
o Ethnicity, namely: European / Non-European / unknown
o Don’t know

At the very least, a lot of uncertainty is imposed on the observing students here.  Someone can be lily-white and have four Dutch-born grandparents but is formally ‘immigrant’ because his or her father or mother was born when the grandparents were abroad. Conversely, people are called ‘native’ when all four grandparents were born abroad but both parents here, no matter how fanatically the convicts are dismissive towards the society in which they behave criminally.

The question is how students have interacted with perfectly integrated people of whom one of the parents was Dutch and the other an equally perfectly integrated immigrant from Vietnam or Chile, or with people whose great-grandparents already lived here but who still have a very dark skin tone and speak with an accent.

Knowingly, the authors did not mention this shortcoming in the article. In fact, in the Conclusion and discussion we find this:

In follow-up research, a distinction could therefore also be made according to which language the accused speaks during the court hearing. The latter also applies to the appearance of the perpetrator. For follow-up research, it would be advisable to make more distinctions between different ethnicities and also make use of characteristics such as the country of birth of the offender and the country of birth of the offender’s parents.

Why in follow-up research? 

What language the defendant spoke was already a question in the observation list and what one writes here about the country of birth is an indirect description of the dichotomy immigrant-native; both ‘country of birth’ and native/immigrant’ also appeared in the observation list. That which one already tried to uncover in a focused way, is now presented here as recommendation for follow-up research.

It is impertinent that one starts here about ethnicity and illusory that there is no question of, for example, the religion of the convicted or any factor that could say something about the ‘distance’ of the suspects from the non-criminal part of Dutch society.

For future research it is also important to examine the influences of the characteristics of judges on sentencing.

Here too, the observation list already included three questions about the demographic characteristics of the judge. These questions were under the heading ‘motivation judge’. What information has been collected under that heading about “what happened during the hearing”, is in no way reflected in the article.

All in all, it is difficult to get rid of the impression that the data collection through “structured and standardised observation lists” has been a complete fiasco. 

The quality of statistical processing: correlation and causation

Above I expressed all my reservations about the use of logistic regression, because the independent variable – the sentence imposed –has artificially been made into a dichotomy. This is especially questionable because of the combination with the determination regarding the nature of the study.

Why has regression analysis been applied here at all?

In short, regression (analysis) involves estimating the parameters of a model. In other words, it is not about testing a hypothesis regarding  a relationship between two variables, which was well considered was done here. 

There are several reasons to want to build a model. Either you look for a ‘formula’ that can provide the most accurate prediction possible for a certain result – in the medical context, for example, whether the patient does survive – or  you explore: try to fathom what is really happening, you search for the factors that, in combination, lead to a certain value of the dependent variable. 

In doing so, you look for the fullest possible explanation of the outcome based on the causative factors: not only correlation but also causality, in other words. That is why it is also customary to start with these analyses with a large number of possible explanatory factors and then via ‘backward elimination’ eliminate factors in order to arrive at a better model.

There is nothing to suggest that the authors were engaged in searching for a model. And even more serious: the obligatory mention of the Nagelkerke score and the not very enlightening remarks in notes 27 and 31, cannot remove the impression that little work has been done to check the reliability of the model.

Strangely enough, the word “model” in the article only appears in those two footnotes!
In number 31, in passing, reference is made to the highest value in the results table: the odds for imprisonment for people who have been in pre-trial detention. These are as much as 65 times as high as for people who lacked experience. The authors note: 

The results of Table 2 do not change direction and (sic) significance when pre-trial detention is not included in the model.

The surprising, even disturbing, thing is that the authors are not surprised by this: nor about that remarkably high value, nor about the small effect of omitting or including this variable in the model. No attempt is made to explain this in terms that refer to the reality of sentencing. 

[][y]]As a result, the authors charge themselves with the suspicion of not being curious, and yes, that is the most serious suspicion that can be expressed about scientists.

Because of the time lapse  the sentencing comes after scoring all data – there is no danger of reversing the direction of a causal relationship. However, the danger of other forms of confusion of correlation with causation is abundantly present. First of all, one should consider the possibility that factors play a role that have a (causal) relationship with one or more of the ‘independent’ variables, as well as with the dependent variable.

For this role, interesting candidates can be found in the observation list. For example: 3:7 “Does the suspect have fixed day time activities”, 3: “.. permanent residence or stay?”, or 3:9 about the living situation, in addition to numerous questions under the heading Attitude suspect.

When examining the overrepresentation of certain groups in crime, one would do well to take into account the possibility that not the most eye-catching difference, skin colour for example, is the explanatory factor but for example the socio-economic position, the (social emotional) intelligence, the culture of education or the relatively frequent occurrence of certain mental illnesses within the group in question.

The lack of this in this study is all the more striking because in general there is quite a lot of attention for it in the social discussion and in social science. In footnote 13, the authors themselves refer to the article ‘An integration of theories to explain judicial discretion’ by Mrs Albonetti ¹¹. Even in the summary of that article, reference is made to three other articles that “… found a direct effect of socio-economic status on the severity of the punishment“, and to yet other articles which contradict that.

Another explanation for the results found

Of a completely different order is the problematic aspect of the alternative explanation for the result found, with regard to the convicted who did not speak Dutch. The striking thing about this is that the authors themselves put forward in the part on theory as well as in the concluding part. 

In the first of these two, they are most clear about this:

In addition, practical objections may play a role in the determination of a punishment for offenders who do not master the Dutch language. If suspects do not speak Dutch, a judge may not want to impose community service because of the problems which may arise in its implementation as a result of not mastering Dutch. Namely, in the implementation of community service, there is no provision for interpreters. In the event of the largely disappearance of this penalty option, the judge would rather be inclined to impose a prison sentence.

A sensible, down-to-earth side note. 
The researchers perhaps wanted to get clarity about this by means of the questions in the observation list under the heading Justification by the judge. In the part Conclusion and discussion, however, we read: 

However, if the community service that would otherwise be imposed is relatively short, it would also be possible to rather impose a fine in such a case. It is therefore unclear how exactly this affects the likelihood of an unconditional prison sentence.

Compared to the introduction, only one additional recital has been added here, not any research result. Again the question arises: what went wrong with the data collection and why do we find virtually nothing about that failure in the article? Why was this not done by talking to judges or using the motivation in the verdicts?

TWO focal concerns theories

A superficial orientation to the Focal Concerns theory already provides an instructive result. The indirect references in the footnotes are to the modern variants. Like the one mentioned earlier by Albonetti. A look at the titles of the publications of her hand lead to a suspicion about her approach: Mothering from the Inside: Parenting in a Woman’s (sic) Prison, The Symbolic Punishment of White-Collar Offenders, A Utopian Prison: Contradiction in Terms?, Pretrial detention and guilty pleas: if they cannot afford bail they must be guilty.

Browsing through some articles in which this theory has an important place gives a disturbing picture of the worldview of these researchers. Some sentences from the article Attributions and Institutional Processing: How Focal Concerns Guide Decision-Making in the Juvenile Court by Alexes Harris for illustration. The subject is the application of adult law for criminal juveniles. Under the heading Re-Criminalization and Juvenile Justice you find this:

That is, centered in discourse and actions that prioritize community safety and punishment of dangerous offenders, a portion of juvenile justice work now focuses on the identification and management of such youth rather than treatment and rehabilitation. (…) The legislation assumes that ‘‘violent’’(sic) youth are no longer amenable to rehabilitation and are the target of transfers.

Further on in the section Focal Concerns and Causal Attributions we read:

The author [Harris refers to Albonetti] suggests that those individuals perceived as having a stable and consistent predisposition for criminal activity –labeled as dangerous- will receive more severe sentences than those not viewed as such. 

Harris’s quotes unmistakably reflect the indignation about the priority given to the safety of (potential) victims of crime: they are young criminals who should be subject to treatment and rehabilitation.

It is clear: this FC theory relates to what Wermink c.s. refer to as estimated danger. A central point of attention seems to be that offenders, in determining the penalty, are victims of the fact that they are considered –by the judges- to belong to a group that has a higher percentage of criminals: they are associated, also inappropriately, with the behaviour of others.

The central point of Walter B. Miller’s original FC theory was very different. The thrust of that was actually that remarkably high delinquency within certain groups should not be understood as the behaviour of individuals but as an expression of another (sub)culture.

In his story Lower-class Culture as a Generating Milieu of Gang Delinquency (Link to pdf-file: http://courses.missouristate.edu/karlkunkel/soc540/miller.pdf) we already read this in the first footnote:

The complex issues involved in deriving a definition of “delinquency” cannot be discussed here. The term “delinquent” is used in this paper to characterize behavior or acts committed by individuals within specified age limits which if known to official authorities could result in legal action. The concept of a “delinquent” individual has little or no utility in the approach used here; rather, specified types of acts which may be committed rarely or frequently by few or many individuals are characterized as “delinquent.”

Pay particular attention to the quotation marks. It’s like reading something by a French philosopher

Both FC theories are in a way each other’s mirror image. In the new theory, the approach is that the individual perpetrators are wrongly regarded as representative of a more criminal subgroup and in the old theory the approach was that the individual perpetrators are not sufficiently regarded as the product of a subculture that promotes behaviour that ‘civil society’ only regards as criminal.

However, the approach of both theories has a great similarity at the same time: do not pay too much attention to individual responsibility of perpetrators, consider perpetrators as victims. I am glad that Wermink c.s. with their article, via a detour, draw attention to the fact that this paradigm was already being worked on at the end of the fifties of the last century; and this in the US.
For that, my sincerest thanks.

NOTES:

¹ Link: http://law.leidenuniv.nl/studenten/nieuws-2012/verdachten-metbuitenlands-uiterlijk-krijgen-zwaardere-straffen.html.
It is also striking that the aericle speaks of ‘the Leiden criminologists’, apparently on the basis of their appointment. Wermink is a sociologist, De Keijser a political scientist and Schuyt studied Dutch law and Dutch Language and Literature.
² The document does not distinguish between crimes and offences.
³ This curious view is, by the way, in line with Nieuwbeerta c.s. of the first article that was peer reviewed here. About recidivism and type of punishment.
⁴ It is unclear whether these data are sorted here at all.
⁵ In the context of logistic regression analysis, the calculations are made using its natural logarithm, but at the presentation the values found were inverted via e-power, and the findings are also described as proportions of proportions.
⁶ The article by Wermink et al. was also subject of discussion by the NRC-ombudsman.
(Link: http://www.nrc.nl/ombudsman/2012/03/31/statistiek-in-dekrant-kansen-zijn-iets-anders-dan-kansverhoudingen/)
It focuses exclusively on this aspect of Sentencing. The discussion ends cheerfully: “Should the newspaper have difficult scientific terms? to use? No, as long as it is explained clearly and consistently. When in doubt: consult the science editors. Or statisticians, of course. Then the probability that things will go well increases.”⁷ Link leads to Volkskrant article entitled ‘Onderzoek: met buitenlands uiterlijk eerder de cel in’ (‘Research: with a foreign appearance more likely in jail’): “The Social Democrat faction in Parliament wants clarification from Minister Ivo Opstelten (Security and Justice). The party wants to know, among other things, what the minister will do about it if it is indeed the case that judges take into account the appearance of suspects, and the language they speak, in their judgement.”
⁸ A number of judges seemed to understand what was going on. In the NJB J.J. Bade and M.M. van der Nat responded: “We do see you: students at the back of the courtroom watching attentively, listening and completing questionnaires. But what you do exactly remains often unknown. Except when your research results make the news. And that happened on March 14, 2012 in, amongst others, Algemeen Dagblad and NRC Handelsblad: ‘Research: with a foreign appearance more likely in jail’ and  ‘Suspect with foreign appearance more likely to get a prison sentence’. That was a bit of a shock. Do we discriminate?”
¹⁰ To avoid irrelevant criticism: cases against flown-in assassins are not brought before the police judge. This is an illustration of the line of thought.
¹¹  Social Problems 1991-38, p. 247-266.

1 comment

Leave a comment

Your email address will not be published.

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