Foreword added 7/2/17

This post has been cited so often, and achieved such notoriety, that I feel it would not be right to edit it overmuch, despite the fact that I have updated the analysis since. A pdf of the talk I gave at UCL is here and a better laid out version of the original analysis as well as my updated analysis can be found here. I would point the reader in particular to the updated analysis, presented second in the latter link and based upon indictable offences only. I now regard this as more robust.

The original analysis is presented unchanged below but I have extensively revised the section summarising evidence from the USA (which was previously too thin, and actually misleading). Finally, I have added a note at the end regarding criticisms of my approach and my responses. Much of this has been addressed in detail in 5 in 6 Men? A Response to Ally Fogg but I include a brief summary here because I note that people tend to link only to this post and not the follow-up.

Original post…

The above graphic shows the prison population of England & Wales from 1900 to 2014. The persistent rise in the number of prisoners since the end of WW2 is due primarily to the rise in the number of male prisoners. One hundred years ago there were about five male prisoners for every female prisoner, now there are about twenty. To be precise, at the end of January 2014 there were 81,045 men and 3,932 women in prison in England and Wales: 20.6 times as many men as women, Ref.[1].

Why?

Most people would say that it is because men commit more crimes. And so they do. But twenty times more? Really?

The number of crimes committed by each sex can be gauged from the number of sentences passed each year – that is, the number of cases coming to court in which the verdict is guilty and a sentence is required. And the answer? In England and Wales there have recently been just over three times more men sentenced per year than women, and this ratio has been falling – see Table 1 for the actual data (taken from Ref.[2]).

Table 1: Numbers of Men and Women Sentenced per Year (England & Wales)

year	Men sentenced	Women sentenced	Ratio Men:Women
2006	1,135,228	277,712	4.1
2009	1,046,281	314,627	3.3
2012	929,720	297,034	3.1

But if only three times more men are sentenced, why are there twenty times more men in prison?

Well, not all sentences are prison sentences. In fact, only a small percentage are. The rest are fines, community service, suspended sentences or enforced remedial treatments. There are three contributing causes to there being a disproportionate number of men in prison. These are,

• A larger percentage of male offenders than female offenders are sentenced to prison;
• Men receive on average longer prison sentences than women; and,
• Men serve on average a greater proportion of their sentences.

We shall see that these three effects explain the factor of 6 by which 3.3 more men being sentenced becomes 20 times more men in prison (i.e., 6 x 3.3 = 20). But is this factor of 6 fair? Does it arise because men’s crimes are, on average, 6 times more heinous than women’s? Or is it simply gender bias?

Consider firstly the proportion of all sentences which are prison sentences. The data for a couple of example years are given in Table 2 (from Refs.[2,3]). In 2009, 8.8% of male offenders were sent to prison, but only 2.6% of women offenders. So, 3.4 times as many men are sent to prison than women (per 1000 offenders of each sex, say).

Does this disparity factor of 3.4 represent justice at work because men’s crimes are worse to a degree which justifies this disparity? This can be examined by looking at the disparity factor separately for specific categories of crime – see Figure 1 (data from Ref.[4]). This reveals that men are substantially more likely to be sent to prison than women for the same category of crime – whatever the category of crime (the only exception being drug offences for which there is negligible disparity). Thus, it is not the case that men are being sent to prison more often because they are committing more serious crimes (for example violent crimes as opposed to, say, minor shop-lifting). It does not matter what crimes a man commits – whatever it is, he will be far more likely to go to prison than a women committing the same crime. I conclude that the average disparity factor of 3.4 is predominantly due to gender bias.

Note also that this average disparity factor has been increasing (Figure 1) whilst men’s overall criminality, as judged by the number of men being sentenced, has been falling (Table 1).

Table 2: Numbers of Men and Women Sentenced to Prison as a Percentage of the Total Sentenced (England & Wales)

year	Men sent to prison	As percentage of total men sentenced	Women sent to prison	As percentage of total women sentenced	Disparity Factor
2006	88,173	7.8%	7,844	2.8%	2.8
2009	91,762	8.8%	8,098	2.6%	3.4

Figure 1

For people who are sent to prison, how do the lengths of their sentences compare between the sexes? In 2009/10, over all crimes, and averaging only over cases where a prison sentence is awarded, men received on average a sentence which was 64% longer than for women (Refs.[2-4]). Again we can ask, is this justice or is it gender bias? Are women receiving lesser sentences because they are committing less serious offences? Again we can examine this by looking at the sentences awarded to men and women for the same category of crime. For 2009, Figure 2 shows a histogram of the ratio of men’s sentence length to women’s for the same category of crime over 42 different crime categories (data from Refs.[2-4], see Table S5.8). This shows that men receive substantially longer sentences than women for the same category of crime – for virtually all categories of crime, with only a few minor exceptions. Why should men receive longer sentences for, say, “theft from a vehicle”, or “false accounting” or “fraud” or “causing death by reckless driving”, etc., etc.? The disparity in sentence lengths is again clearly gender bias.

Figure 2

Finally I looked at the effect of parole: what proportion of their sentence do men and women actually serve? Women serve rather less of their sentence than men, 48% versus 53% respectively on average (Ref.[5]), a disparity factor of 1.1. Again we can ask if this is fair, or is it gender bias? The key issue is the prisoners’ behaviour whilst in prison. You might think that women behave better in prison. But not a bit of it. Quite the opposite. Women prisoners are subject to between 20% and 50% more prison disciplinary actions (per 100 prisoners) than are men (Ref.[6]). This includes women prisoners being disciplined more frequently for acts of violence. So, the preferential treatment of women by parole boards would appear to be another case of gender bias.

We can now see why there are 20 times more men in prison than women. This factor of 20 is composed of four effects which numerically are: 3.3 x 3.4 x 1.64 x 1.1 = 20. Only the first of these (3.3) is fair, representing the greater number of crimes committed by men. The other three factors are all the result of gender bias. Men are 3.4 times more likely than women to be sent to prison for the same crime, men receive sentences which are 1.64 times longer than women’s for the same crime, and men actually serve 10% more of their sentence on average than women, for no obvious reason. So, all told, the discrimination against men is about a factor of 6. This leads me to the following staggering conclusion,

Men are subject to massive gender discrimination in the criminal justice system. If male offenders were treated in the same way as female offenders there would be only one-sixth of the number of men in prison. About 68,000 men would not be in prison if they were female, leaving a male prison population of only 13,000.

Analyses of USA Data

This Section has been considerably extended since the original version of the post because, (a) I had not accurately represented Mustard’s results (unwittingly), and, (b) I only alluded briefly to Starr’s results which I give here in slightly greater detail.

2001 Analysis by David B. Mustard, Ref.[7]. 

Mustard initially looked at sentencing disparity, i.e., a comparison of sentence lengths for people sentenced to prison, considering both race and sex. The first three rows of Table 3, below, are taken from Mustard’s Table 5. (Like my analysis, Mustard’s neglects life sentences on the basis that the sentence length is unknown). Men on average are sentenced to 2.78 times the average length of a woman’s sentence. However, the largest part of this can be accounted for by the greater offence level and worse criminal history. Mustard proceeds to control for these variables and concludes that the regression variable ‘female’ is associated with a sentence 5.51 months shorter across the whole dataset. He then divides this by the population average sentence (46 months) and concludes that there is a sentencing disparity in favour of women of 12%. However, this is misleading because the percentage disparity should be with respect to the average sentence for women, which is only 18.51 months, giving a sentencing disparity in favour of women of 30%.

I have performed a simple check of this as follows. Mustard’s Table 1 presents the USA sentencing guidelines against offence level and criminal history. Using the average values of these variables given in Table 3, below, one can evaluate by linear interpolation the corresponding expected sentence length, given as the 4th row in Table 3. Taking the ratio of the actual average sentence to that expected provides a simple measure for men and women in which offence level and criminal history have been normalised out. The ratio of these ratios is then an estimate of gender disparity, and Table 3 shows that it is 1.31, i.e., 31%, in good agreement.

In conclusion, where I have estimated a gender sentence length disparity of 1.64 based on UK data, Mustard’s analysis is (I claim) consistent with a sentencing disparity of 1.30.

Table 3: US Data on Sentences and Offense Level from Mustard’s 2001 Paper

	Men	Women
Sentence (months)	51.52	18.51
Offence Level	18.30	13.11
Criminal history	2.10	1.37
Expected sentence	34.93	16.44
Sentence/Expected	1.47	1.13
Disparity	1.31

blue text in above Table are mine, not Mustard’s – see text above

Mustard observes that 73% of the gender disparity originates from judges departing from the sentencing guidelines, which appears to be indicative of gender bias. In fact departures from the guidelines are more pronounced on the basis of sex than race. His Table 10 suggests (after translating logit to probability) that women have a 63% chance of a departure from the guidelines which reduces their sentence, but only a 27% chance of a departure which increases their sentence. He also makes this observation,

“One important result from Table 6 is that females receive even shorter sentences
relative to men than whites relative to blacks. The discrimination literature generally
argues that females are objects of discrimination and receive worse outcomes.
In sentencing, however, women receive better outcomes, consistent with
women’s being treated paternalistically in court. Although some contend that the
sentencing guidelines harm women, studies have usually concluded that females
are sentenced more leniently than males.”

Mustard goes on to consider the disparity in being sentenced to prison at all “when that option is available”. Here there may be a significant difference with UK practice in that judges’ discretion in the USA is more tightly constrained, having no option but to award a prison sentence after a guilty verdict for some offences. In the UK, judges have greater discretion, e.g., to award suspended sentences, community orders, etc. Mustard finds that “the results of these regressions are striking…..females are more likely than males to be assigned no prison
term”. The logit associated with this finding is 0.53 (significant at 99%CL) but I have not been able to translate that into a disparity factor since no overall logit for males was given.

Overall, Mustard’s analysis of US data supports the contention that there is a significant disparity in favour of women but suggests smaller disparity factors than derived above for the UK. This difference in quantitative estimate of gender disparity may be explained by Starr’s more recent analysis.

2012 Analysis by Sonja Starr, Ref.8

The paper by Starr, Ref.8, is a particularly thorough study of the US data. The headline conclusions are,

1. This study finds dramatic unexplained gender gaps in federal criminal cases.
   Conditional on arrest offense, criminal history, and other pre-charge observables, men
   receive 63% longer sentences on average than women do.
2. Women are also significantly likelier to avoid charges and convictions, and twice as likely to avoid incarceration if convicted.

I my terminology, conclusion (1) gives a sentencing disparity of 1.63, almost exactly the same as I derive above for the UK (1.64), whilst conclusion (2) implies a disparity factor for being sentenced to prison of 2. The latter is rather smaller than the imprisonment disparity factor of 3.4 derived above, but is very close to my more recent estimate of 1.9 based on indictable offences only (see here).

Some quotes from the excellent Starr paper are worth noting,

“Prior studies have reported much smaller sentence gaps because they have ignored the role of charging, plea-bargaining, and sentencing fact-finding in producing sentences. Most studies control for endogenous severity measures that result from these earlier discretionary processes and use samples that have been winnowed by them. I avoid these problems by using a linked dataset tracing cases from arrest through sentencing.

I ask: do otherwise-similar men and women who are arrested for the same crimes end up with the same punishments, and if not, at what points do their fates diverge?

In the United States, men are fifteen times as likely to be incarcerated as women are. But can this gap be explained by differences in criminal behavior or circumstances, or are courts or prosecutors treating genuinely equivalent cases differently on the basis of gender? The latter would violate the Constitution, undercut the criminal justice system’s punishment objectives, and contribute to the social consequences of demographically concentrated mass incarceration. So the reasons for the gender gap are of considerable legal and policy interest. This study explores them using a dataset that traces federal criminal cases from arrest through sentencing. I find that gender gaps widen at every stage of the justice process and that men and women ultimately receive dramatically different sentences.

Policymakers might simply be untroubled by leniency toward women…. But the gender disparity issue need not be framed in terms of how women are treated. One could ask: why are men treated so harshly, if women are (apparently) treated otherwise? It is hard to dismiss this question as trivial: over two million American men are behind bars. While males generally are not a disadvantaged group, men in the criminal justice system generally are; they are mostly poor and disproportionately nonwhite. The especially high rate of incarceration of men of color is a serious social concern, and gender disparity is one of its key dimensions…. Most defendants of both genders have suffered serious hardship, have mental health or addiction issues, have minor children, and/or have “followed” others onto a criminal path.”

Quite.

The Laughably Named “Equal Treatment Bench Book”

As if this staggering degree of anti-male bias were not enough, I make some alarming observations regarding the UK government’s Judicial College “Equal Treatment Bench Book”. This is the official guidance on how to treat people within the UK criminal justice system. The section on Gender Equality is headed by a summary of its Key Points, which are as follows,

• Women remain disadvantaged in many public and private areas of their life; they are underrepresented in the judiciary, in Parliament and in senior positions across a range of jobs; and there is still a substantial pay gap between men and women. (Aside: one might ask what a remark about the so-called “pay gap” is doing in guidance on criminal justice. This document couldn’t have been written by feminists, by any chance?).
• Stereotypes and assumptions about women’s lives can lead to unlawful discrimination.
• Factors such as ethnicity, social class, sexual orientation, disability status and age affect women’s experience and the types of disadvantage to which they might be subject; assumptions should not be made that all women’s experiences are the same.
• Discrimination is often unconscious and based on a person’s own experience and perceptions; it is important to be aware of the wide diversity of women’s experiences.
• Women may have particular difficulties participating in the justice system, for example, because of child care issues, and courts may need to consider adjustments to enable women to participate fully.
• Women’s experiences as victims, witnesses and offenders are in many respects different to those of men.
• As judges, we can go some way to ensuring that women have confidence in the justice process and that their interests are properly and appropriately protected.

If you are surprised that an official government guidance document with “Equality” in the title is concerned only that women should be treated nicely – but shows no concern for men at all – then I can only assume that you have not read one previously.

In the 2013 edition of the Bench Book the following final bullet point has been added,

Of course, men can suffer from gender discrimination too; this section reflects the reality that this is rarer.

What? When the very criminal justice system of which this guidance forms a part is guilty of massive discrimination against men, as proved conclusively by the data presented above? That the authors of the Bench Book have the audacity to include such a statement can mean only one of three things. Either they have not noticed the issues discussed above, in which case they are incompetent (this is the most benign possibility). Or they are completely unable to make a balanced and fair judgment due to being blinded by an anti-male sexist mindset. Or, finally, the authors are well aware of the injustice and are happy to see it persist, or even worsen.

Worsen? Yes, it seems that the degree of discrimination against men (or preferment of women, if you prefer) is not yet sufficient. The Equal Treatment Bench Book includes the following infamous quote from Lady Justice Brenda Hale:-

It is now well recognised that a misplaced conception of equality has resulted in some very unequal treatment for the women and girls who appear before the criminal justice system. Simply put, a male-ordered world has applied to them its perceptions of the appropriate treatment for male offenders…. The criminal justice system could … ask itself whether it is indeed unjust to women.

Words fail me. A factor of 6 preference in favour of women is not enough for them, clearly.

This is “women’s equality”. Women’s equality means, not merely preferential treatment for women, but ever increasing preferential treatment for women. One really shouldn’t treat women like we treat men. That would be insupportable. That is a “ misplaced conception of equality “.

Addendum added 8/2/17 – Answers to criticisms and additional observations

Four criticisms have been raised and addressed by me in 5 in 6 Men? A Response to Ally Fogg. I provide a brief precis here (see that post for the data sources),

(1) My analysis does not allow for the fact that men and women tend to be prosecuted for different types of crime: Actually the presumption here is not a fact. The pattern of offending by men and women is broadly similar. This is true both as measured at the end of the process (sentencing) and also at the beginning (arrests), supporting evidence is presented in 5 in 6 Men? A Response to Ally Fogg and later years’ data confirm this finding. [Of course, far more men commit crimes in virtually every category, but the pattern of offending relates to percentages of each sex]. Moreover, a key observation is that men are treated more harshly in virtually all the crime categories, both severe and more minor crimes.

(2) My analysis assumes that, within each crime category, men and women commit crimes of equal severity on average: That is correct, and it is valid to question the assumption. What it amounts to is questioning the granularity of the crime categories used, i.e., are there enough sub-divisions of categories to capture differences in severity? I used 42 categories. They are listed here (slides 18 and 19). For example, ‘theft’ is not one but five different categories. I note that Starr’s paper, Ref.8, refers to the use of 430 offence codes, so the closely comparable disparity results of that paper (especially when compared with my re-analysis based on indictable offences only) suggests that greater granularity does not alter the picture. Moreover, since the pattern of men’s and women’s offending across the major crime categories is similar, why should we favour an explanation of the disparities based on an unsupported assumption that the severity of their offending is significantly different within every crime category – and to a similar degree?

(3) My analysis does not account for the effect of offending history on sentencing, and it tends to be repeat offenders who are sentenced to prison. This invalidates my analysis because men are disproportionately the repeat offenders: The first sentence is correct, the second is contrary to the data. Men are twice as likely as women to be sentenced to immediate custody for a first indictable offence, whilst women are twice as likely to receive a conditional discharge. However, the large majority of both sexes who were sentenced for an indictable offence in 2013 were repeat offenders, namely 86% of women and 91% of men. 37% of male offenders sentenced for an indictable offence had 15 or more previous sanctions, compared with 30% for women. So the difference in recidivism for gaoled men and women is not very marked. In respect of the sentences received by such recidivists: “In 2013, the most common disposal for offenders convicted of an indictable offence with 15 or more previous sanctions was immediate custody for both males and females, reflecting that repeat offenders are more likely to get an immediate custodial sentence. A higher proportion of males (40%) with 15 or more previous sanctions received an immediate custodial sentence compared with females (31%)“. So, as anticipated, repeat offenders of either sex are more likely to be imprisoned, but there is again a disparity against males even for a comparable degree of recidivism.

(4) My analysis clearly is inapplicable to some categories of crime, such as sexual offences and crimes awarded life or indeterminate sentences: True. All analyses of gender bias in sentencing exclude life and indeterminate sentencing, simply because the sentence is indeterminate. The more numerically significant issue is sexual offences. There are approaching 12,000 men in prison for sexual offences, but fewer than 100 women, a gender ratio of perhaps around 140 to 1. Moreover, sexual offences attract particularly long sentences. Clearly sexual offences are in a class of their own. But they are not in a class which is treated with unusual leniency towards men, are they? The disparity here transcends the numerical. Our society regards essentially the same acts completely differently according to whether they are committed by a man or a woman (see the post Rolf Harris v Women Sex Offenders for example). Women tend to be awarded suspended sentences for offences which, if committed by a man, would attract a long prison term – and utter social and financial destruction. How many of the 12,000 men would be in prison if they were treated like women? Who knows, but the disparity factor of 6 would imply just 2,000, hardly unreasonable and probably an over-estimate.

I take the opportunity to make some further observations on my original analysis,

(5) I am now of the view that basing the analysis on indictable offences is probably more secure. The reason is that roughly 95% of prisoners have committed indictable offences, it being far less common to go to prison for summary offences. This re-analysis may be found here and here. Instead of an overall gender disparity of 6 an overall disparity of a factor of 3.6 is found based on indictable offences alone. This may be interpreted as a further reinforcement against the criticism that women’s offending tends to be less serious, because the decision to classify as indictable filters out the more minor crimes. This has the impact that, whilst the male:female conviction ratio across all crimes is roughly 3:1, confining attention to indictable offences alone raises this to a ratio of roughly 6:1. Of course, a preference for the overall disparity of 3.6, rather than 6, assumes that the judicial process of classifying an offence as indictable is not itself contaminated with gender bias. Given that gender bias exists in all other stages of the criminal justice process, this is obviously questionable.

(6) The estimated disparity factor (of x6 or x3.6, according to taste) refers to sentencing alone. It does not account for gender disparity in, (i) arrest, (ii) prosecution, (iii) conviction, and, as we allude to above, (iv) classification as indictable. The Starr paper, Ref.8, makes the observation that, in fact (and in the US process) disparities are encountered at all stages. Does it require data to convince you which of these scenarios is more likely to end in an arrest: a man punching a woman in the face in the street, or, a woman punching a man in the face in the street? Or, for that matter, two women fighting compared with two men fighting. Starr concludes, based on US data, that women are significantly more likely to avoid both charges and convictions. In other words, the UK conviction ratio, across all crimes, of 3:1 might already contain gender bias. What might it be if women were treated equally as regards arrest, prosecution and conviction? The gender bias is very deep.

References

1. Prison population data for England & Wales are published weekly and can be found, for example, at https://www.gov.uk/government/publications/prison-population-figures-2014 and similar web sites.
2. Criminal Justice Statistics, Quarterly Update to March 2012, Ministry of Justice Statistics Bulletin (and similarly for the other years).
3. http://www.statistics.gov.uk/hub/crime-justice/offenders/sentencing.
4. As [3], specific file “sentencing-stats-09-supp-tables”
5. http://brightonmanplan.wordpress.com/2012/10/20/five-myths-about-the-sentencing-of-female-offenders/
6. Ministry of Justice publication “Statistics on Women and the Criminal Justice System”, November 2010.
7. David B. Mustard, “Racial, Ethnic and Gender Disparities in Sentencing: Evidence from the US Federal Courts” by, University of Georgia, in Journal of Law and Economics, vol. XLIV (April 2001).

8. Sonja B. Starr, “Estimating Gender Disparities in Federal Criminal Cases”, University of Michigan Law School, August 29, 2012