I. Hate speech laws and protected characteristics

I want to begin by clarifying the nature of the “Who?” question. There are numerous characteristics, social identities or statuses that either currently are or conceivably could be brought under the scope of hate speech laws.^{Footnote 5} These include:

• • age (e.g., Canada,^{Footnote 6} South Africa,^{Footnote 7} Tasmania (Australia),^{Footnote 8} YouTube^{Footnote 9});

• • age performance or ways of performing age such as acting young or old;

• • citizenship status, if distinguished from nationality (e.g., England and Wales (UK),^{Footnote 10} Northern Ireland (UK)^{Footnote 11});

• • criminal record, when not already included under social status;

• • disability including both mental and physical disability (e.g., Canada,^{Footnote 12} Finland,^{Footnote 13} France,^{Footnote 14} Hong Kong,^{Footnote 15} the Netherlands,^{Footnote 16} Northern Ireland (UK),^{Footnote 17} South Africa,^{Footnote 18} Tasmania (Australia)^{Footnote 19});

• • education, when not already included under social status;

• • employment status, such as employed or unemployed;

• • ethnicity including cultural heritage, ancestry or descent, physical appearance, homeland, origin (e.g., Australia,^{Footnote 20} Australian Capital Territory (Australia),^{Footnote 21} Canada,^{Footnote 22} Chile,^{Footnote 23} Connecticut (US),^{Footnote 24} Croatia,^{Footnote 25} Denmark,^{Footnote 26} England and Wales (UK),^{Footnote 27} Finland,^{Footnote 28} France,^{Footnote 29} the Netherlands,^{Footnote 30} Massachusetts (US)^{Footnote 31}, New South Wales (Australia),^{Footnote 32} New Zealand,^{Footnote 33} Northern Ireland (UK),^{Footnote 34} Queensland (Australia),^{Footnote 35} Republic of Ireland,^{Footnote 36} Russia,^{Footnote 37} South Africa,^{Footnote 38} Tasmania (Australia),^{Footnote 39} Western Australia (Australia)^{Footnote 40});

• • gender such as man, woman, male, female, or sex such as cisgender, transgender, cissexual, transsexual, third gender, bigender, pangender, agender, intersex, third sex (e.g., Australian Capital Territory (Australia),^{Footnote 41} Canada,^{Footnote 42} Chile,^{Footnote 43} France,^{Footnote 44} the Netherlands,^{Footnote 45} New South Wales (Australia),^{Footnote 46} Queensland (Australia),^{Footnote 47} South Africa,^{Footnote 48} Tasmania (Australia)^{Footnote 49});

• • gender performance or ways of performing gender identity such as being masculine, effeminate, metrosexual, when not already included under gender;

• • HIV/AIDS status (e.g., Australian Capital Territory (Australia),^{Footnote 50} New South Wales (Australia)^{Footnote 51});

• • immigration status, when not already included under citizenship status and nationality;

• • language including language status, mother-tongue and language accent identity, if not included under ethnicity (e.g., South Africa^{Footnote 52});

• • marital status including relationship status, when not already included under social status (e.g., South Africa,^{Footnote 53} Tasmania (Australia)^{Footnote 54});

• • medical status including serious disease, when not already included under HIV/AIDS status (e.g., Facebook^{Footnote 55});

• • nationality or legal relationship to a state, when not already included under ethnicity (e.g., Australia,^{Footnote 56} Australian Capital Territory (Australia),^{Footnote 57} Canada,^{Footnote 58} Chile,^{Footnote 59} Connecticut (US),^{Footnote 60} Croatia,^{Footnote 61} Denmark,^{Footnote 62} England and Wales (UK),^{Footnote 63} Finland,^{Footnote 64} France,^{Footnote 65} the Netherlands,^{Footnote 66} Massachusetts (US)^{Footnote 67}, New South Wales (Australia),^{Footnote 68} New Zealand,^{Footnote 69} Northern Ireland (UK),^{Footnote 70} Queensland (Australia),^{Footnote 71} Republic of Ireland,^{Footnote 72} Russia,^{Footnote 73} South Africa,^{Footnote 74} Tasmania (Australia),^{Footnote 75} Western Australia (Australia)^{Footnote 76});

• • parental status (e.g., Tasmania (Australia)^{Footnote 77});

• • personality traits or types;

• • physical appearance such as body weight, skin tone, hair colour, facial configuration and other bodily features, when not already included under race or ethnicity;

• • political beliefs, activities, or affiliations (e.g., Australia,^{Footnote 78} the Council of the European Union,^{Footnote 79} Penn State University^{Footnote 80});

• • pregnancy, when not already included under parental status (e.g., Penn State University,^{Footnote 81} South Africa,^{Footnote 82} Tasmania (Australia)^{Footnote 83}) or, even more specifically, breastfeeding activity (e.g., Tasmania (Australia)^{Footnote 84});

• • profession such as banker, politician, lawyer, or tax collector, when not already included under social status or veteran status;

• • race including colour and other aspects of physical appearance (e.g., Australia,^{Footnote 85} Australian Capital Territory (Australia),^{Footnote 86} Canada,^{Footnote 87} Chile,^{Footnote 88} Connecticut (US),^{Footnote 89} Croatia,^{Footnote 90} Denmark,^{Footnote 91} England and Wales (UK),^{Footnote 92} Finland,^{Footnote 93} France,^{Footnote 94} the Netherlands,^{Footnote 95} Massachusetts (US)^{Footnote 96}, New South Wales (Australia),^{Footnote 97} New Zealand,^{Footnote 98} Northern Ireland (UK),^{Footnote 99} Queensland (Australia),^{Footnote 100} Republic of Ireland,^{Footnote 101} Russia,^{Footnote 102} South Africa,^{Footnote 103} Tasmania (Australia),^{Footnote 104} Western Australia (Australia)^{Footnote 105});

• • regional identity including sub-national regional identity and trans-national regional identity, when not already included under nationality;

• • religion including religious beliefs, practices, or affiliations as well as lack thereof (e.g., Canada,^{Footnote 106} Chile,^{Footnote 107} Connecticut (US),^{Footnote 108} Croatia,^{Footnote 109} Denmark,^{Footnote 110} gland and Wales (UK),^{Footnote 111} Finland,^{Footnote 112} France,^{Footnote 113} the Netherlands,^{Footnote 114} Massachusetts (US),^{Footnote 115} Northern Ireland (UK),^{Footnote 116} Queensland (Australia),^{Footnote 117} Republic of Ireland,^{Footnote 118} Russia,^{Footnote 119} South Africa,^{Footnote 120} Tasmania (Australia)^{Footnote 121});

• • sexual orientation (e.g., Australian Capital Territory (Australia),^{Footnote 122} Canada,^{Footnote 123} Croatia,^{Footnote 124} Denmark,^{Footnote 125} England and Wales (UK),^{Footnote 126} Finland,^{Footnote 127} France,^{Footnote 128} the Netherlands,^{Footnote 129} New South Wales (Australia),^{Footnote 130} Northern Ireland (UK),^{Footnote 131} Queensland (Australia),^{Footnote 132} South Africa,^{Footnote 133} Tasmania (Australia)^{Footnote 134});

• • sexual preference such as preference for particular sex acts or practices, when not already included under sexual orientation;

• • social status including social origin and class (e.g., the Council of the European Union,^{Footnote 135} South Africa^{Footnote 136});

• • traveller community, when not already included under race or ethnicity (e.g., Republic of Ireland^{Footnote 137});

• • war record including veteran status (e.g., Northern Arizona University (US),^{Footnote 138} Penn State University (US),^{Footnote 139} University of Oregon (US)^{Footnote 140}) or pacifist status.

This is, of course, a very diverse list—diverse both in the types of characteristics and in the types of hate speech law. As far as the characteristics are concerned, we have:

• • affective states or patterns thereof (e.g., sexual orientation, sexual preference);

• • affiliations relating to communities, cultures, social groups or families (e.g., citizenship, ethnicity, language, marital status, nationality, parental status, regional identity, religion);

• • attitudinal dispositions, beliefs or ways of thinking (e.g., political, religious);

• • biological, genotypic, physiological, or physical(-phenotypic) attributes (e.g., medical status, race, sex);

• • conduct, (phenotypic-)behaviour, performance or ways of living (e.g., age performance, education, employment status, gender performance, marital status, profession, religion, traveller community, war record).

This diversity will be particularly relevant when we come to consider the third approach to specifying the proper scope of incitement to hatred laws, what I call the formal approach, because it appeals to intuitions about the formal qualities of characteristics and about the form of people’s relationships to their characteristics.

In terms of the diversity of hate speech laws, we have:

• • laws that proscribe group defamation based on protected characteristics;

• • laws that regulate negative stereotyping and stigmatization based on protected characteristics;

• • laws that disallow using insults, slurs, or derogatory epithets against, disseminating ideas based on the inferiority of, or using any words, signs, or symbols that are deeply insulting or offensive to, members of groups or classes of persons based on protected characteristics;

• • laws that ban stirring up, inciting, or promoting feelings of hatred or hostility toward or among members of groups or classes of persons based on protected characteristics;

• • laws that prohibit speech or other expressive conduct concerning members of groups or classes of persons identified by protected characteristics when it is a threat to public order;

• • laws that penalise denying, grossly trivialising, approving, justifying, condoning, or glorifying acts of mass cruelty, violence, or genocide perpetrated against members of groups or classes of persons based on protected characteristics;

• • laws that constrain speech or other expressive conduct directed at members of groups or classes of persons identified by protected characteristics when it constitutes the enactment of a dignitary crime or tort;

• • laws that forbid speech or other expressive conduct when it amounts to conduct that violates or interferes with people’s exercise of civil or human rights based on protected characteristics;

• • laws that interdict speech or other expressive conduct that constitutes a hate crime based on protected characteristics;

• • laws that restrict speech or other expressive conduct aimed at members of groups or classes of persons identified by protected characteristics via time, place, and manner restrictions.^{Footnote 141}

I do not have space here to discuss each of the different types of hate speech law nor particular instances thereof. The focus of the study will be stirring up hatred offences in the UK. It may be that different conclusions follow if the focus is shifted to other types of hate speech law. To put the same point a little more formally, it may be that when thinking about a hate speech law of type X it would be correct to consider inclusion of characteristics c1, c2 and c3 but not c4, c5 and c6, whereas when reflecting on hate speech law of type Y it might be fitting to include characteristics c4, c5 and c6 but not c1, c2 and c3, because of morally relevant differences between X and Y. I shall return to, and try to defend, this generalisation in the conclusion at the end of Part 2.

Before discussing the first of five approaches to specification, I first need to distinguish between two motives for addressing the “Who?” question and to clarify which of the two informs this investigation. One motive is deep scepticism about the moral justification for, and practical usefulness of, hate speech law including incitement to hatred law. Here the “Who?” question is intended or designed to create a slippery slope or adverse consequences argument against any such legislation by adding more and more protected characteristics to the list so as to make all hate speech laws seem unpalatable.^{Footnote 142} A second motive stems from a belief that current incitement to hatred legislation in the UK has developed over time in a piecemeal, reactive, politicised, and in many ways illogical and incoherent manner, and a desire to (re-)theorise the proper scope of such laws in a way that is far more comprehensive, systematic and analytical. This motive is open-minded over whether or not hate speech laws can be warranted all things considered. It is about seeking correct answers to both the “Who?” question and the more general warrant question without prejudging either. This article proceeds under the second motive.

II. Consistency specification

The first approach to specifying the proper scope of incitement to hatred laws emphasises norms of consistency. Consistency is a familiar and essential feature of the rule of law, of course. In the area of sentencing, if two people have broken the law in similar ways, then consistency requires that they should expect to receive similar punishments from judges. This is the principle of treating like cases alike (and unalike cases unalike). But consistency is also important in the area of the enactment of criminal laws, where similar conduct should have similar status as criminal or not criminal. So, if two examples of conduct are similar in that they are both incitement to hatred albeit one is incitement to hatred on grounds of characteristic c1 and the other is incitement to hatred on grounds of c2, then prima facie consistency requires that both examples of conduct should be dealt with in similar ways by the criminal law. This is the principle of treating like conduct alike (and unalike conduct unalike).

Then again, perhaps what matters is not only similarity in the treatment of similar cases and conduct but also similarity in the treatment of similar groups; which is to say, similar groups should receive similar protection in law. According to Alon Harel, for example, “treating the victims of racist speech more favorably than victims of sexist, homophobic, or other forms of abhorrent speech is itself a form of discrimination”.^{Footnote 143} So, if two similar groups of people are subject to similar forms of incitement to hatred, then consistency requires that they should enjoy similar legal protections or lack thereof. This is the principle of treating like groups alike (and unalike groups unalike) or the principle of parity for short.

However, consistency in the treatment of similar cases, conduct and groups are not the only relevant forms of consistency. Consider as well consistency in the way that lawmakers, for example, handle reasons or rationales for including some characteristics and not others under the scope of given hate speech laws. Some of these reasons are practical, some formal, some functional, and some relate to democratic values. But what really matters is that when lawmakers invoke and apply these reasons they do so in a consistent manner, rather than in highly politicised or even haphazard ways. The principle of parity demands equal treatment of similar groups, whereas what I am talking about now relates more to equity in the treatment of groups, meaning that groups have a right to expect that reasons or rationales will be applied in consistent ways. Thus, if two types of conduct are similar in that they are both forms of incitement to hatred but one type of conduct is incitement to hatred on grounds of characteristic c1 and the other is incitement to hatred on grounds of c2, and if the principled reasons for banning incitement to hatred on grounds of c1 would also apply mutatis mutandis to incitement to hatred on grounds of c2, then consistency requires banning both types of conduct qua forms of incitement to hatred, absent other relevant and equally compelling reasons for banning one and not the other. I shall call this the principle of treating like reasons alike (and unalike reasons unalike) or the principle of higher-order consistency for short.^{Footnote 144}

Keeping in mind these basic forms of consistency, let us now consider some concrete arguments about the proper scope of incitement to hatred laws in the UK. They have to do with anomalies or inconsistencies within anti-discrimination law, criminal law, incitement to hatred laws, and constitutional law. Starting with anti-discrimination law, the Equality Act 2010 ^{Footnote 145} makes it unlawful in England and Wales to discriminate against persons based on certain “protected characteristics”, namely, age, disability, gender reassignment, race, religion or belief, sex, sexual orientation. At the same time, however, it is currently unlawful to stir up hatred only on the basis of three of these characteristics: race, religion, sexual orientation. This raises a question of consistency between incitement to hatred laws and extant anti-discrimination law.^{Footnote 146} As Ivan Hare puts it,

In fact, such inconsistency is neither necessary nor unavoidable. In South Africa, for example, hate speech laws have been developed in concert with, and even written into, anti-discrimination legislation, thus ensuring that the wide scope of prohibitions of discrimination (in terms of the range and number of protected characteristics) is matched exactly by the wide scope of prohibitions of hate speech.^{Footnote 148}

Now it might be objected at this stage that there is a relevant difference between discrimination and hate speech: namely, whereas discrimination is an act, hate speech is speech. But the difference disappears as soon as one recognises that using words or behaviour to stir up hatred is itself a type of act and that some forms of discrimination are enacted to a large extent through speech or other expressive behaviour. So, for example, ss 26 and 27 of the Equality Act 2010 ^{Footnote 149} in England and Wales make it unlawful to harass or victimise persons based on protected characteristics including when this harassment or victimisation takes the form of speech or other expressive conduct.^{Footnote 150}

Turning to inconsistencies within criminal law, ss 145 and 146 of the Criminal Justice Act 2003 ^{Footnote 151} are hate crime provisions giving magistrates and judges in England and Wales powers to determine if criminal acts were made more “serious” or “aggravated” by hostility toward victims based on the characteristics of race, religion, disability, sexual orientation or transgender identity. Now s 146 relates specifically to disability, sexual orientation and transgender identity, and could apply inter alia to various public order and harassment offences that are typically enacted through speech or other expressive conduct.^{Footnote 152} Consider the offences of causing fear or provocation of violence (s 4), intentional harassment, alarm or distress (s 4A), and harassment, alarm or distress (s 5) under the Public Order Act 1986,^{Footnote 153} as well as the offence of harassment (ss 1 and 2) under the Protection from Harassment Act 1997.^{Footnote 154} But if hostility toward victims based on their disability or transgender identity can be aggravating factors in the case of someone using, say, threatening, abusive or insulting words with intent to cause harassment, alarm or distress (s 4A), then why should stirring up hatred offences not include stirring up hatred on grounds of disability or transgender identity? It scarcely seems sufficient merely to point out that these are separate regimes or schemes of law since that only invites the following question: why should they be considered separately or regarded differently when it comes to specifying the proper scope of such laws?^{Footnote 155}

Next, consider inconsistencies within incitement to hatred laws themselves. When the 1965 Race Relations Act ^{Footnote 156} introduced for the first time the stirring up racial hatred offence it was well recognised that there are groups in society whose categorisation as a race could be open to doubt. Yet the concern was to ensure parity of treatment for different groups. What if the law protected newly arrived immigrants from the Caribbean, for example, but not Jews? And so, speaking in the House of Commons in 1965, the then Home Secretary, Frank Soskice MP, opined, “I would have thought a person of Jewish faith, if not regarded as caught by the word ‘racial’ would undoubtedly be caught by the word ‘ethnic’, but if not caught by the word ‘ethnic’ would certainly be caught by the scope of the word ‘national’, as certainly having a national origin.”^{Footnote 157} The controversies have persisted however. For one thing, if immigrants arriving from the Caribbean are protected by the stirring up racial hatred offence (as defined by s 17 of the Public Order Act 1986 ^{Footnote 158}) on grounds of their colour or race, then what about people against whom hatred might also be stirred up not ostensibly because of their colour or race but on grounds of their immigration status as being economic migrants, illegal immigrants, so-called bogus asylum seekers, failed asylum seekers, genuine asylum seekers, or even refugees? Are they to be included under the technical term “race” on grounds of their nationality or citizenship?^{Footnote 159} For another thing, because Jews have been covered under the legislation as a racial or ethnic group, courts in England and Wales have on occasions convicted Muslim activists and clerics for inciting racial hatred against Jews—for example, R. v. Iftikhar Ali ^{Footnote 160} and R. v. El-Faisal.^{Footnote 161} However, ironically, the courts have not regarded Muslims as belonging to racial, ethnic or national groups for the purposes of interpreting the stirring up racial hatred offence—a view consistent with a leading case in the field of anti-discrimination law, Nyazi v. Rymans Ltd.^{Footnote 162} And so, a rabbi could not be convicted for stirring up racial hatred against Muslims. Perhaps not surprisingly, then, various Muslim groups and politicians argued on grounds of parity (treating like groups alike) for extending existing incitement to hatred laws so that they covered Muslims. In 2005, for example, the office of the then Home Secretary, Charles Clarke MP, allegedly wrote to several mosques to explain “[w]e cannot see why it is right to have protection in law for Jews and Sikhs, but wrong to extend it to other communities like the Muslim community.”^{Footnote 163}

Following on the heals of the Racial and Religious Hatred Act 2006,^{Footnote 164} which introduced the stirring up religious hatred offence and effectively extended protection to Muslims, yet further anomalies were identified in the treatment of other groups. Thus, in 2007 Chris Bryant MP declared that it was high time to introduce an offence of stirring up hatred on grounds of sexual orientation in order to “overcome anomalies” in the relevant laws.^{Footnote 165} Likewise, in 2011 the Equality and Human Rights Commission defended the creation of a new offence of stirring up hatred on grounds of disability for the sake of “parity”.^{Footnote 166} More recently, a significant number of the individuals and stakeholder organisations who took part in the Law Commission’s consultation exercise allied to its report Hate Crime: Should the Current Offences be Extended? ^{Footnote 167} argued for the creation of new stirring up hatred offences for both disability and transgender identity also on the basis of parity.^{Footnote 168} The Commission itself ultimately did not recommend this extension for mainly practical reasons, which I intend to discuss in the next section. However, in 2015 the House of Commons Women and Equalities Committee also heard evidence from a number of expert witnesses on the need for parity of protection for people with transgender identities.^{Footnote 169} Pace the Law Commission, the Women and Equalities Committee recommended that “[t]he Government should introduce new hate crime legislation which extends the existing provisions on … stirring up hatred so that they apply to all protected characteristics, as defined for the purposes of the Equality Act 2010.”^{Footnote 170}

Although much more could be said about these particular arguments, here I am interested in critically evaluating consistency as a general approach to specifying the proper scope of incitement to hatred laws. One potential weakness in the current approach is that the principles of consistency may underdetermine single best solutions to inconsistency. Consider parity in the treatment of groups and a situation in which one group enjoys protection via an existing stirring up hatred offence, whereas another, similar group does not enjoy such protection. There are two main ways to achieve parity in this situation. The first is to expand the existing stirring up hatred offence or else create a new stirring up hatred offence so as to protect the group that is as yet unprotected. This involves a kind of levelling up of protection: members of a group are said to have a prima facie right to the same high level of protection that other groups already enjoy. A second way is to revise or repeal the existing stirring up hatred offence without introducing any new offence. This involves a kind of levelling down of protection: groups who currently enjoy protections are said to have no right to a level of protection that other groups do not enjoy. The problem is that the principle of parity does not in itself dictate which of these two strategies is best. In terms of parity alone, either is acceptable. This means that the argument for the levelling up strategy over the levelling down strategy is dependent not merely on the principle of parity but also on treating the existing offence as given or as having a sound or generally accepted rationale. In other words, in the absence of a generally accepted rationale for the existing offence the argument would probably not be made for expanding it or creating a new offence like it. Conversely, the argument for the levelling down strategy over the levelling up strategy is dependent not merely on the principle of parity but also on questioning the rationale for the existing offence. Hence, it is partly because the rationale for the existing offence no longer commands widespread acceptance that the case for retrenchment is being made. The upshot is that consistency is only one of a number of principled considerations that are likely to be needed in order to determine the proper scope of incitement to hatred laws.

By way of illustration of this problem, consider once again the case of incitement to hatred against Muslims. Up until 2006 it could have been an offence to use threatening, abusive or insulting words with the intention or likelihood of stirring up hatred against Jews defined as an ethnic group—including words identifying or picking out Jews as an ethnic group partly on the basis of their religious beliefs—but not an offence to use threatening, abusive or insulting words with the intention or likelihood of stirring up hatred against Muslims defined as an ethnic group—including words identifying or picking out Muslims as an ethnic group partly on the basis of their religious beliefs.^{Footnote 171} Now it would be incorrect to say that prior to 2006 Muslims enjoyed no legal protections against hate speech whatsoever. In England and Wales the Crown Prosecution Service (CPS) already had the power to prosecute someone who used Islamophobic hate speech in the process of committing a religiously aggravated public order or harassment offence under ss 31 and 32 of the Crime and Disorder Act 1998.^{Footnote 172} Nevertheless, before 2006 Jews but not Muslims were protected under the stirring up racial hatred offence. This inconsistency mattered even more because the maximum custodial sentence for the stirring up racial hatred offence—an offence that was inapplicable to Islamophobic hate speech—was higher than for comparable religiously aggravated public order offences—offences that were applicable to Islamophobic hate speech. So, for example, at that time the maximum custodial sentence for the offence of displaying writing, say, which is threatening, abusive or insulting and is intended or likely under the circumstances to stir up racial hatred—such as against Jews—was seven years under s 18(1) of the Public Order Act 1986.^{Footnote 173} Yet the maximum custodial sentence for the discreet religiously aggravated offence of displaying writing, say, which is threatening, abusive or insulting and with the intention of causing harassment, alarm or distress—such as to Muslims—was just two years under s 31(1)(b) of the Crime and Disorder Act 1998.^{Footnote 174} The anomaly was further exacerbated by the fact that s 9 of the Public Order (Northern Ireland) Order 1987 ^{Footnote 175} made it an offence in Northern Ireland (under direct rule from Great Britain) to stir up hatred on grounds of religious belief. So Muslims in Northern Ireland enjoyed protections not afforded to Muslims in other parts of the UK. Consequently, some scholars insisted that England and Wales needed the same stirring up religious hatred offence that existed in Northern Ireland.^{Footnote 176} But herein lies the problem with the current approach. Put simply, the argument for a new stirring up religious hatred offence is underdetermined by the principle of parity. For, it would have been feasible to remove the inconsistency by adopting any one of the following courses of action: (i) legislate a new stirring up religious hatred offence; (ii) repeal the existing stirring up racial hatred offence; (iii) revise the existing stirring up racial hatred offence so as to explicitly name Muslims along with Jews as racial or ethnic groups, or else direct judges to change their working definitions of race and ethnicity so as to include Muslims along with Jews as racial and/or ethnic groups for the purposes of interpreting the offence; (iv) revise the existing stirring up racial hatred offence so as to explicitly exclude both Jews and Muslims as racial or ethnic groups, or else direct judges to revise their working definitions of race and ethnicity so as to exclude both Jews and Muslims as racial or ethnic groups for the purposes of interpreting the offence.

At this juncture one could, of course, try to make an appeal to the principle of higher-order consistency in the enactment of criminal law, which speaks to consistency in the application of legislative rationales. One possible rationale for banning incitement to racial hatred is that certain kinds of racist hate speech can contribute to a climate of hatred and fear. Applying this same rationale to the case of Muslims could justify course of action (i) based on a parallel concern that stirring up hatred against Muslims can contribute to a climate of hatred characterised in part by an increased likelihood of acts of discrimination and violence against Muslims, as well as an increased fear among Muslims of acts of discrimination and violence.^{Footnote 177} However, this argument for creating a new stirring up religious hatred offence is also underdetermined by the principle of higher-order consistency in the enactment of criminal law. This is because appealing to the principle of higher-order consistency and the aforementioned rationale would also justify course of action (iii).^{Footnote 178} Appealing to the principle of higher-order consistency does not by itself determine one course of action as being better than another if either represents the consistent application of legislative rationales. Indeed, the equivalent suitability of (iii) is all the more apparent given the development of Modood’s own thinking on Islamophobic hate speech in the wake of the Danish cartoons controversy.^{Footnote 179} Modood presented the Danish cartoons—or two of the cartoons^{Footnote 180}—not as pure expressions of religious Islamophobia but as instances of quasi-racist Islamophobia, a type of racism that comes close to ethnophobia.^{Footnote 181} As he explains, “[i]t is true that ‘Muslim’ is not a (putative) biological category in the same way as ‘black’ or ‘south Asian’, aka ‘Paki’, or Chinese. But nor was ‘Jew’ once: a long, non-linear history of racialization turned a faith group into a ‘race’.”^{Footnote 182} This explanation prompts the following question. If public expressions of Islamophobia amount to quasi-racism, and if the operative rationale for banning incitement to hatred is to tackle speech that contributes to a climate of hatred, then why not argue for the assimilation of Muslims into incitement to racial hatred provisions as opposed to creating a new stirring up religious hatred offence?^{Footnote 183} The key point here is that additional rationales must be adduced as determinative reasons for choosing course of action (i) over (iii).^{Footnote 184}

The inconsistencies did not end once the Racial and Religious Hatred Act 2006 ^{Footnote 185} came into effect, adding the stirring up religious hatred offence. The current body of legislation contains seemingly arbitrary differences in the thresholds for prosecution for the different offences. Part 3 of the Public Order Act 1986 ^{Footnote 186} covers incitement to racial hatred and the test remains a person who uses “threatening, abusive or insulting words or behaviour” and “(a) he intends thereby to stir up racial hatred, or (b) having regard to all the circumstances racial hatred is likely to be stirred up thereby”. By contrast, Part 3A of the Public Order Act 1986 ^{Footnote 187} covers incitement to religious hatred and the test is “a person who uses threatening words or behaviour” and “he intends thereby to stir up religious hatred”. Therefore, so long as courts continue to deal with cases of incitement to hatred against Jews under Part 3 (incitement to racial hatred) and cases of incitement to hatred against Muslims under Part 3A (incitement to religious hatred), then in theory it is easier for prosecutors to secure convictions in cases of incitement to hatred against Jews than it is to secure convictions in cases of incitement to hatred against Muslims. The anomaly did not go unnoticed by members of parliament. Paul Goggins MP, for example, argued that if we cannot have the anomaly of it being an offence to stir up hatred against Jews and Sikhs (qua racial groups) but not an offence to stir up hatred against Muslims and Christians (qua religious groups), then, by the same token, “we cannot have different rules [or thresholds] for Jews and Sikhs than for Muslims and Christians.”^{Footnote 188} Once again, however, appealing to the principle of parity does not lead to a single correct way of resolving inconsistency. For, it would have been feasible to remove the inconsistency by adopting either of the following two courses of action: (v) define the new stirring up religious hatred offence with the same lower threshold for conviction as the existing stirring up racial hatred offence; (vi) revise the existing stirring up racial hatred offence so that it has the same higher threshold for conviction as the new stirring up religious hatred offence.^{Footnote 189}

Would it help to make an appeal to the principle of higher-order consistency? What would the consistent application of reasons look like? Suppose part of the underlying function or purpose of incitement to hatred law is to combat the creation of climates of hatred and fear, and this applies equally to race and religion. And suppose this principled reason suggests that generally speaking we should prefer lower prosecution thresholds to higher prosecution thresholds. Based on this we now do appear to have a consistency-based reason to favour (v) over (vi). However, there remains incompleteness in the consistency specification precisely at the point at which the principle of higher-order consistency is introduced. The incompleteness consists in the fact that this principle tells us to treat like reasons alike; it does not tell us what those reasons should be. And so we must inevitably, I think, appeal to other approaches, such as the functional approach, in order to obtain the sorts of reasons that can be utilised by or fed into the principle of higher-order consistency. Without the other approaches, there would be nothing to go on.

There is one final area in which inconsistencies may emerge that will serve to highlight this incompleteness. The area is constitutional law and, in particular, the constitutional principle that governments must secure the basic rights and freedoms of all citizens such that it must not deny the protection of law to any citizens. It may be possible to interpret this principle as flowing from or serving the more abstract principle of parity (treating like groups alike).^{Footnote 190} What is more, it has seemed to some writers axiomatic to say that the principle of equal protection requires governments to extend any protections afforded to some groups who are the subject of harmful hate speech to other similar groups who are the subject of similarly harmful hate speech.^{Footnote 191} How do these ideas relate to the situation of hate speech law in the UK? The Human Rights Act 1998 ^{Footnote 192} sets out the basic rights and freedoms of all citizens in the UK and it does so by giving further effect to the European Convention of Human Rights (ECHR).^{Footnote 193} Now it might be argued that appealing to art 14 of the ECHR could justify an extension of the UK’s existing incitement to hatred laws to other groups when combined with art 10(2) of the ECHR. To explain, art 14 makes it clear that “[t]he enjoyment of the rights and freedoms set forth in this Convention shall be secured without discrimination on any ground such as sex, race, colour, language, religion, political or other opinion, national or social origin, association with a national minority, property, birth or other status”.^{Footnote 194} And art 10(2) states that the exercise of the right to freedom of expression protected under art 10(1) may be restricted by laws that “are necessary in a democratic society”.^{Footnote 195} Therefore, if art 10(2) can be interpreted as implying that hate speech laws are necessary in a democratic society to secure the rights and freedoms of those groups who are subject to it, then art 14 seems to imply that this securing should be done for all citizens or “without discrimination on any ground”.^{Footnote 196} How are these arguments incomplete? Put simply, these are not freestanding arguments but instead piggy-back on arguments about the conditions under which it would or would not amount to discrimination to draw distinctions between the different characteristics listed in art 14. Some such arguments may have to do with the underlying function or purpose of incitement to hatred laws. After all, if the function of incitement to hatred laws strongly implies that characteristics c1, c2, and c3 should be protected but not other characteristics c4, c5 and c6, then it would not be unfair discrimination for incitement to hatred laws to protect groups of people with characteristics c1, c2, and c3 only. In other words, it is hard to know what the relevant discriminatory grounds are in relation to the scope of incitement to hatred laws unless we know something of the function of such laws.

III. Practical specification

A second approach to specifying the proper scope of incitement to hatred laws focuses on the ostensible goals or apparent aims of such laws and then attempts to solve the problem of specification in a practical way based on the relevant goals or aims. It usually asks questions of the following form, “If the ostensible goal or apparent aim of hate speech law is X, then what would have to be the case in order for that goal or aim to apply not merely to characteristics c1, c2 and c3 but also to c4, c5 and c6?” So, for example, one ostensible goal of incitement to hatred law might be to deter acts of incitement to hatred. Another could be more simply to punish people who engage in incitement to hatred. Notice, however, that, in contrast to functional specification, an approach that concentrates solely on ostensible goals may ignore the underlying or real functions, purposes or objectives of incitement to hatred laws and as such may provide a limited or incomplete justification for the very existence of such laws.

If one ostensible goal of incitement to hatred laws is to deter words or behaviour that amount to stirring up hatred, then a basic practical requirement of any extension of such law to cover more groups must be that members of these other groups are in fact (or are likely to be) the object of words or behaviour that stirs up hatred against them. It will not suffice merely to point to the existence of at least some instances of hate speech against members of these groups; this is because not all hate speech is incitement to hatred. Thus, if one is very specifically arguing for an extension of incitement to hatred laws to cover sexual orientation, for example, it is not enough to show that instances of homophobic hate speech can be found in the media and other areas of public life.^{Footnote 197} And if one wanted to push for an extension to cover people with disabilities, it is not enough to flag up evidence of cyber-bullying of people with disabilities^{Footnote 198} or the existence of websites proclaiming hatred of “retards”, “spastics” and “cripples”^{Footnote 199} or the fact that negative media portrayals of people with disabilities may have increased off the back of the austerity policies pursued by the UK coalition government after 2010 or even that many people with disabilities in the UK have reported an increase in their own personal experience of direct, face-to-face verbal harassment and hostility potentially as a result of negative media portrayals of people with disabilities.^{Footnote 200} Likewise, if one wanted to justify an extension of existing incitement to hatred laws to also cover body weight and age, it would not be enough to refer to studies showing that “almost three quarters of overweight women [in the UK] have received derogatory remarks regarding their weight”^{Footnote 201} or to point to surveys revealing that 41% of respondents in the UK say they have experienced ageism in the form of subtle prejudice or lack of respect.^{Footnote 202} Instead, it would be necessary to show that actual instances of the various forms of hate speech that currently surround sexual orientation, disability, body weight and age have in fact reached the level of, or qualify as, stirring up hatred on a par with other forms of stirring up hatred that are already criminalised.^{Footnote 203}

In 2007 the parliamentary committee examining the Criminal Justice and Immigration Bill ^{Footnote 204} received expert witness evidence from the Chief Executive of Stonewall, Ben Summerskill, on the question of introducing a new offence of stirring up hatred on grounds of sexual orientation. He addressed the aforementioned practical issue head-on.

In the end the government concluded that introducing a new offence of stirring up hatred on grounds of sexual orientation was an appropriate response to an actual problem. Writing to the Joint Committee On Human Rights at the end of 2007, the then Minster of State for Justice, David Hanson MP, put the position thusly.

By contrast, when a year later in 2008 the same Minister was asked by the same Joint Committee to set forth the government’s view on the merits of creating a new offence of stirring up hatred on grounds of transgender identity he stated the following.

What might “compelling evidence” look like? Presumably it would have to be not only methodologically robust but also comprehensive in nature. Specifically, perhaps it might include data from large-scale quantitative discourse analyses of media and Internet content, as well as reliable statistics based on large-scale gathering and recording of reported incidents of incitement to hatred. At any rate, in 2010 the government reaffirmed its position that there was insufficient evidence that hatred was being stirred up against people with transgender identities to justify creating a new offence, and also insufficient evidence in relation to the stirring up of hatred against people with disabilities.^{Footnote 208}

Nevertheless, I believe that it is more difficult to draw policy conclusions from this putative evidence gap than one might at first assume. On the one hand, let us just imagine for the sake of argument (and almost certainly contrary to fact) that compelling evidence does exist and what it shows is that the extent of incitement to hatred relating to disability and transgender identity in the media and on the Internet is very small both in absolute terms and relative to race, religion and sexual orientation, as well as that there are few reported incidents of incitement to hatred relating to disability and transgender identity and once again fewer than for race, religion and sexual orientation. It would not necessarily follow from this evidence (if it existed) that the creation of new stirring up hatred offences for disability and transgender identity are unwarranted. In terms of incitement to hatred in the media and on the Internet, it may be that the “pressing social need” requirement is different for people with disabilities and people with transgender identities (people who may feel particularly vulnerable or socially excluded), and so a lower extent may be sufficient to warrant intervention. In terms of reporting, it may be that people with disabilities and people with transgender identities (again people who may feel particularly vulnerable or socially excluded), as well as the wider population (people who may not be used to looking out for or even recognising incitement to hatred against people with disabilities and people with transgender identities), are simply not yet reporting incidents at the same rate as for other forms of incitement to hatred, and this could be for various reasons other than that there are fewer incidents to report.^{Footnote 209}

What seems far more likely, on the other hand, is that there is currently a lack of methodologically robust and comprehensive evidence one way or the other. Now it is certainly true that non-governmental organisations (NGOs) like Disability Rights UK do from time to time commission research into negative media portrayals of people with disabilities and that NGOs like True Vision do gather and record self-reports of transphobic hate incidents, to give just two examples. But because these evidence-gathering practices are not coordinated by governmental authorities, and because each NGO focuses (or is seen to focus) on particular groups at particular times, and does so with limited resources, and does not seek to pinpoint incitement to hatred specifically, the evidence generated is patchy and unsystematic. At present governmental agencies do not regularly commission large-scale quantitative discourse analyses looking into the true extent of incitement to hatred on grounds of disability or transgender identity in the media and on the Internet. Nor do they engage in large-scale gathering and recording of reported incidents of incitement to hatred on grounds of disability or transgender identity, albeit the Home Office does capture reported incidents relating to existing stirring up hatred offences covering race, religion and sexual orientation within its recorded crime figures under the public order offences category.^{Footnote 210} But does this mean, therefore, that creating new stirring up hatred offences for disability and transgender identity cannot be warranted? Again not necessarily. For one thing, it would be hard to understand the logic of a decision to refrain from creating new stirring up hatred offences whilst leaving the existing offences in place rather than removing the existing offences if there is a paucity of evidence to call upon for any of the characteristics in question. For another thing, the paucity of evidence might reflect a lack of institutional impetus and political will on the part of governmental agencies. And there may be various reasons for this. One malign reason could be that lack of research is symptomatic of precisely the sort of attitudes that can sow the seeds of hate speech itself, namely, lack of concern for, empathy toward or solidarity with people with disabilities and people with transgender identities. A more benign reason is simply that the issues around incitement to hatred towards such people are relatively new to the agenda of civil servants, politicians, policymakers, and media professionals, if not to stakeholders, campaigners and academics.^{Footnote 211} Moreover, in a time of departmental budget cuts in the UK the resources available to the Home Office and Ministry of Justice to investigate possible extensions of the law may be limited, not to mention the fact that parliamentary time to discuss and push through necessary provisions is at a premium. As an illustration of these practical issues, consider the fact that as of June 2016 the UK government has still yet to respond officially to the Law Commission’s 2014 report on extending hate crime legislation including the stirring up hatred offences to cover additional groups.

Let us take it as read, therefore, that governmental authorities should take on the responsibility for creating or obtaining methodologically robust and comprehensive evidence. Then again, what should legislators do in the meantime whilst they wait for compelling evidence to come in? One strategy is to hold off creating any new stirring up hatred offences until the results are in. This is precisely the view of the Joint Committee on Human Rights.^{Footnote 212} The problem with this approach, however, is that some groups may continue to be the objects of incitement to hatred whilst they wait for governmental authorities to create or obtain compelling evidence, weigh it, draft an action plan, consult with stakeholders, and finally attempt to get legislation passed through both Houses of Parliament. It goes without saying that there is unlikely to be a similar hiatus among hate speakers during this potentially lengthy period. Indeed, members of groups who are subject to incitement to hatred and who want the government to create new offences may find themselves in the perverse position of welcoming a spike in hate speech against them, so that it can be picked up by researchers.

A second strategy is for legislators to go ahead and create new stirring up hatred offences based on suspicions or anecdotal reports in lieu of methodologically robust and comprehensive evidence. What, if anything, can be said on behalf of this strategy? It seems to me that some rationales are better than others. One is that creating new offences could give people the confidence they need to report incidents and this in turn will enable government authorities to build up a more reliable picture of the phenomena in question. Yet an obvious objection here is that the rationale proposes to create new offences in order to gain some certainty on the existence of phenomena whose existence is in fact a precondition for introducing the new offences in the first place.^{Footnote 213} Nonetheless, a second rationale is based on the old adage that “there is no smoke without fire”. This could mean two things. First, that if people are raising suspicions about or offering anecdotal reports of certain phenomena, the chances are that the phenomena do exist, even if nobody yet knows for certain. Of course, it might be countered that the people raising suspicions or offering anecdotal reports are biased or have a vested interest or are simply unreliable witnesses because of their own traumatic experiences. But what if their concerns are supported by a second application of the adage? Suppose we know for certain that members of groups who are the objects of incitement to hatred are also typically subject to discrimination or violence and that there is some reciprocal connection between stirring up hatred and the discrimination or violence. In which case, if we also know for certain that members of a given group are subject to discrimination or violence, this may lend additional credence to the suspicions or anecdotal reports of incitement to hatred against them. A third, related rationale is a conscious adoption of a certain form of the precautionary principle. In the absence of methodologically robust and comprehensive evidence of the existence of incitement to hatred against particular groups in society but mindful of the fact that this sort of hate speech can have significant harmful effects, especially for members of those groups but also for society as a whole, a perfectly rational course of action for government authorities to take is to not merely commence gathering the evidence but also to take substantive precautions against the risk of significant harmful effects. This could be done through an extension of the relevant legislation whilst they wait for that evidence to come in and during which time the burden of proof is shifted toward those people who advocate non-extension on the grounds that there is no problem.^{Footnote 214}

Let us now turn to consider another ostensible goal of incitement to hatred law: namely, to punish wrongdoers, that is, people who stir up hatred against vulnerable groups in society. Now it might seem fair to suppose that a practical requirement of law with this sort of ostensible goal is that it can be applied to prosecutable cases and that the CPS has a reasonable prospect of securing successful convictions. If the elements that make up existing offences together create a high threshold for prosecution, then creating new offences to cover yet more characteristics but based on the elements of the existing offences, could become a pointless exercise (it might be argued) in the event that few, if any, actual cases involving those newly protected characteristics are prosecutable. In 2014, for example, the Law Commission argued that it would be futile to create new offences of stirring up hatred on grounds of disability and transgender identity because, based on the elements of the existing stirring up hatred offences relating to religion and sexual orientation (i.e., intent, use of threatening words or behaviour), the threshold required for successful prosecution would be so high that there would be small numbers of prosecutable cases and vanishingly small numbers of actual successful prosecutions.^{Footnote 215}

However, I believe that there is a significant weakness in this line of argument and that this weakness is a good illustration of why it would be wrong to rely exclusively upon a practical specification of the proper scope of incitement to hatred laws. If the reason not to create new offences for additional characteristics is the lack of prosecutable cases, which itself reflects the high threshold for prosecution, then surely this reason could also equally support some very different conclusions about how we should proceed.^{Footnote 216} For instance, it might be argued that the best solution is not to refrain from creating new offences but instead to adjust the basic elements of the new offences to create a lower threshold. Perhaps this could be done by matching the basic elements of any new stirring up hatred offences relating to disability and transgender identity not to the existing stirring up hatred offences relating to religion and sexual orientation, which have relatively high thresholds for conviction (i.e., intent, threatening words or behaviour), but instead to the stirring up racial hatred offence, which has a relatively low threshold for conviction (i.e., intent or likelihood, threatening, abusive or insulting words or behaviour). Of course, it might be countered at this stage that even the lower threshold for incitement to racial hatred has not led to a significantly higher number of successful prosecutions. But this response only invites the following question. If the Law Commission concluded that low prosecution rates are a valid reason to refrain from creating new stirring up hatred offences covering additional characteristics, why did it not also conclude that low prosecution rates are a valid reason to repeal existing stirring up hatred offences? The answer must surely be that there may be powerful rationales for the existing offences that are neither undermined nor trumped by the issue of thresholds for prosecution and the extent of prosecutable cases. One such rationale might be the symbolic function of incitement to hatred law, to which I shall return in Part 2. And if this is true, it seems reasonable to consider whether or not the same or similar powerful rationale(s) might also apply to the new offences.

My point here is that arguments for and against the creation of new stirring up hatred offences cannot rely solely on practical considerations relating to thresholds for prosecution. After all, if authorities declared that they were going to lower the threshold for the new offences to ensure that legal professionals have plenty of prosecutable offences to work with, so that introducing the new offences is not a pointless exercise, members of the public might reasonably respond that it is not enough for new offences to create prosecutable offences; they must serve some underlying function or purpose. For example, if one wanted to justify the erection of a sign in a field that reads “People who throw stones at this sign will be prosecuted” it would not be enough to comment on how many prosecutable cases and successful prosecutions would be likely to occur. One would need to supply a good reason for creating the offence in the first place. Likewise, in order to justify enacting new stirring up hatred offences it is not enough to make arguments about what threshold would be needed in order to sustain a certain number of prosecutable cases and successful prosecutions. Instead, one would need to make more fundamental arguments about the underlying or real function or purpose of incitement to hatred laws, and one would also need to recognise that these functional arguments might be relevant to determining the thresholds.

I plan to explore the functional approach in detail in Part 2, but for now I can offer one brief illustration. One possible functional argument might support the creation of new stirring up hatred offences for disability and transgender identity but with lower prosecution thresholds because of the expressive or symbolic value of having these offences on the books with lower thresholds. A lower threshold for prosecution sends out a message that the government is genuinely interested in combating this speech because, for example, it has bona fide concern that people with disabilities and transgender identities should not face a climate of hatred and fear (to which the stirring up of hatred contributes). Of course, if this argument is accepted, then there may also be reasons of parity to adopt lower thresholds for all the stirring up hatred offences. What is more, there could be another functional argument for this lowering of thresholds for all the stirring up hatred offence once again couched in terms of expressive or symbolic value: namely, it sends out a message that the government has no greater or lesser concern for people with disabilities or transgender identities than for other groups in society; that there is no suggestion of a pecking order of sociolegal status among different groups based upon a hierarchy of prosecution thresholds. At any rate, it seems clear to me that these or other functional arguments, as well as consistency and democratic arguments, have just as important a role to play as purely practical considerations. That being said, some people might try to argue that characteristics such as religion, say, are less deserving of protection because of the formal qualities of these characteristics and because of people’s relationship with them. So, it is to this other type of argument that I shall turn next.

IV. Formal specification

A third approach to specifying the proper scope of incitement to hatred laws appeals to intuitions about the formal qualities of characteristics and about the forms of people’s relationships with their characteristics. These intuitions may lead us to suppose that some characteristics are deserving of, or appropriate objects of, legal protections whilst others are undeserving of, or inappropriate objects of, legal protections. Various formal distinctions have been drawn in the context of both public and academic debate on the “Who?” question. In what follows I shall submit five such distinctions to critical scrutiny.