The Empirics of Free Speech and Realistic Idealism: Part II

3. Harmful Speech: Sticks and Stones

This section aims to go through the most popular areas where people attempt to show the negative impacts of free speech. I will dismiss some concerns whilst accepting others. The general gist of my argument is that the negative effects are often overstated though not always.  I am not making any claims as to how to deal with the negative consequences and correlations laid out below. There are no policy prescriptions (at least as far as free speech is concerned) in this Part. To be honest, the main reason for going through this is because this is what I think a discussion about free speech should be about. Too often, people extol the virtues or vices of free speech as though this is a strictly normative question. It is not. And to the extent that it is, the discussion is boring.

I’m going to have to apologise for the length of this post. If there is anyone in the world who actually reads all of this, I will buy you a pint. Here is a contents so you can skip sections you don’t want to read:

1. News Media: Murdoch and the Purple Land
2. The Effects of Money and Lobbying in Politics  
3. Video Games: Crash Bandicoot Shouting Fire in a Crowded Theatre
4. Porn: Having an Orgasm in a Crowded Theatre
5. Sexist Speech: Crash Bandicoot Making Rape Jokes in a Crowded Theatre
6. Race Related Speech: Hollywood, Skokie and Umugandas in Rwanda
7. Incitement, Obedience and Speech Act Theory: Eichmann to Jihadi Twitter
8. Conclusion: Epistemic Humility

3.1 News Media: Murdoch and the Purple Land

If you spend enough time around Communist Twitter, you’ll find one of their main views is the idea that the media serves corporate or U.S. imperialist ends. For example, take Chomsky and Herman’s Manufacturing Consent wherein they argue that the following propaganda model has taken control:

[the] propaganda model focuses on this inequality of wealth and power and its multilevel effects on mass-media interests and choices. It traces the routes by which money and power are able to filter out the news fit to print, marginalize dissent, and allow the government and dominant private interests to get their messages across to the public…

the dominant media firms are quite large businesses; they are controlled by very wealthy people or by managers who are subject to sharp constraints by owners and other market-profit-oriented forces; and they are closely interlocked, and have important common interests, with other major corporations, banks, and government

[the] propaganda model suggests that the "societal purpose" of the media is to inculcate and defend the economic, social, and political agenda of privileged groups that dominate the domestic society and the state. The media serve this purpose in many ways: through selection of topics, distribution of concerns, framing of issues, filtering of information, emphasis and tone, and by keeping debate within the bounds of acceptable premises (p.2, 14, 298).

This is relevant to the discussion of free speech. If the media, who utilise free speech, do so to promote their own interests and are successful at doing so, it’s fair to say that this is a Bad Thing. But I don’t think Chomsky and Herman are right. There are two reasons I disagree with this view. The first relates to the views of the media themselves and the second relates to the effect that the media actually have. To take the first, it just does not seem obvious to me that concentrated corporate ownership skewers the perspective of a particular paper.

Take the example above from the last general election. You can take a very cynical view of this: Murdoch is backing the SNP in Scotland and the Tories in England to erode Labour as much as possible. I don’t take that view. On Radio 4, the political editor of the Scottish edition of The Sun was quizzed on this apparent discrepancy. The exchange that follows below was transcribed by me (unfortunately the Today Programme has taken the episode down) and is quite indicative of the true position. The interview began by asking why the Scottish edition is now backing the SNP given that they didn’t support independence.  

Andrew Nicoll: Point was at that time, as the results show, the nation and our readership was pretty divided. Looking at the polling evidence now, considering offer of the parties, it seems clear the nation is united behind the SNP... People are crying out for them so we're backing them [i.e., the SNP].

James Naughtie: Does that mean you just follow the tide rather than think what’s right?

AN:  In this case, the tide can run in the direction of what’s right... The offer that she’s [i.e. Nicola Sturgeon] making at this time which coincides with this the unique set of circumstances whereby its almost certainly going to be a hung parliament... will give Scotland a stronger voice at this particular time and its opportunity for her and we’ve chosen to back.

JN: ... So, The Sun in London is saying she's barmy, you're saying she's wonderful. What are people supposed to make of that?

AN: I’m sure readers of The Sun in London will have their view and our readers will have another view. I see this criticism often [whereby people] will contrast editorials of the two editions but they really are distinct... we have a great deal of self-determination to plough a different furrow.... People might find that surprising, but I’m sure Mr. Murdoch wont find it surprising.

I buy this. Here is what I am seeking to show in the next few paragraphs:

1. Corporate ownership of the media does not lead to corporate-friendly media output arising from a conflict of interest.
2. The main driver of media output is consumer demand (i.e., people read what they already agree with) as the above extract indicates.
3. This could create a new negative effect of a free media: people living in a bubble where their views are reinforced by an uninformative partisan press.
4. I do not believe this bubble exists: reputational effects and consumer demand for truth rather than reinforcement of existing beliefs means that the partisan media does not, uniformly or consistently, distort the truth.

Gentzkow and Shapiro (2010) look at what actually drives media slant and output. They look at whether the corporate ownership of the media drives the message of newspapers. What they find is that output is actually related to what readers already believe. As in, the reason the Guardian produces left wing commentary is because their readers are left wing. How do we account for reverse causality (i.e., the Guardian is making people left wing)? Gentkow has a pretty cool control: religiosity. It’s less likely, for obvious reasons, that religiosity would be affected by media output. And here’s what they find:

…consumer demand responds strongly to the fit between a newspaper’s slant and the ideology of potential readers, implying an economic incentive for newspapers to tailor their slant to the ideological predispositions of consumers.

We find little evidence that the identity of a newspaper’s owner affects its slant. After controlling for geographic clustering of newspaper ownership groups, the slant of co-owned papers is only weakly (and statistically insignificantly) related to a newspaper’s political alignment. Direct proxies for owner ideology, such as patterns of corporate or executive donations to political parties, are also unrelated to slant.

Another study that looks at the question of corporate ownership is Dellavigna and Hermle (2014). They look at the question of whether media concentration leads to more favourable reviews of films made by papers owned by the same media conglomerate. Turns out, it doesn’t:

Using a data set of over half a million movie reviews from 1985 to 2010, we find no statistical evidence of media bias due to conflict of interest in either the News Corp. conglomerate or the Time Warner conglomerate. The null finding is not due to imprecision… Overall, reputation-based incentives appear to be effective at limiting the occurrence of bias: we find no evidence of bias by commission, no evidence of editorial bias, no systematic evidence of bias by omission, and no evidence of bias among the aggregators.

What the figures above show is that Time Warner reviews of Fox films are about as negative/positive as their reviews of Time Warner films. The same finding applies to states (in the West): they don’t seem to drive media slant. [1] We can think a bit more deeply about the question: what specific incentives are at play here? Well, consumer demand drives slant but what if people wanted accurate news? We’d still expect slanted news in this situation. An earlier study by Gentzkow and Shapiro (2007) notes that ‘For free markets to produce accurate information requires three things: that consumers want to hear the truth, that markets provide incentives to give consumers what they want, and that firms respond to these incentives. None of these is a given.’

In this study they look at the viewing audience of a host of TV news stations in the Middle East. They first confirm that consumers seem to be driven toward viewpoints they already agree with (in line with the evidence above):

Of those who say they turn to CNN first for information about world affairs, the average favorability toward the US is .43. In contrast, the average favorability among those who turn to Al Jazeera first is .34. The difference in the mean favorability between these two groups is equal to about a third of a standard deviation and is strongly statistically significant.

Whats more interesting is a potential problem: free speech can’t actually allow good ideas to out if everyone just reads stuff they agree with, right? The interesting thing about this study is that it also looks at the quality assessments that people make of the channels looked out:

The coefficients show that consumers relatively predisposed to share the views expressed by Al Jazeera - those who are less favorable toward the U.S. or say religion is important in their lives - rate its quality significantly higher. These same consumers view CNN as lower quality. A one standard deviation decrease in favorability toward the U.S. increases the perceived quality of Al Jazeera by .1 standard deviations (p < :001) and decreases the perceived quality of CNN by 0.6 standard deviations (p < :001).

This provides a potential pathway for us to reconcile the fact that (i) people seem to read what they are predisposed to agree with and (ii) they think what they’re reading is accurate. Essentially there is confirmation bias: you think what accords with your pre-existing view is accurate. But this is not a good argument in favour of free speech. In fact it gives us reason to doubt the positive role that news media can play in advocating policies or holding parties to account.

But there’s something we haven’t considered: the suppliers rather than the demanders of media. As Gentzkow and Shapiro note, ‘the fact that the ultimate driver of the market is a demand for accurate information means this distortion [i.e., simply reinforcing existing views] can only go so far.’ How do we assure that our press is trying to be informative, rather than serve the interests of a corporate of government entity? Gentzkow et al (2004) look at two (partisan) scandals in American history: the Credit Moblier scandal in the 1870s (Republican Congressmen Ames bribing individuals with stocks) and the Teapot Dome scandal in the 1920s (the leasing of land for oil expropriation that similarly involved bribes).

The 1870s scandal was covered in a highly partisan fashion. The Republican press covered less and disparaged the sources for the story (see figure above). By the 1920s, the American press had changed. It became less partisan (see figure below). There is still a difference, but there’s a clear equalisation. The press, as a whole, was more informative than the 1870s. What caused this?

Declining costs and increased city populations caused a huge increase in scale. In 1870, a newspaperman might make more money pleasing a local politician than in selling papers and advertisements. By 1920 newspapers had become big business, and they increased readership and revenue by presenting factual and informative news. Following these financial incentives, newspapers changed from being political tools to at least trying to present a façade of impartial reporting.

During the decades from 1870 to 1920 when corruption appears to have declined significantly within the United States, the press became more informative, less partisan, and expanded its circulation considerably.34 It seems a reasonable hypothesis that the rise of the informative press was one of the reasons why the corruption of the Gilded Age was reduced during the subsequent Progressive Era.

So, from the available evidence, I’m just not convinced that corporate ownership of media is a problem. The reputational effects in a competitive market seem to do a reasonably good job of making sure the media doesn’t serve the interests of either the state or commercial interests. The study quoted in the first footnote below also indicates that its only when you have high market and political incentives (i.e., lack of competition) do you have a partisan press that is substantively uninformative.

The Effects of Media Consumption

Even accepting that media output isn’t driven by corporate interests and that competition generally leads to an informative press, you might still consider there is a case for regulating the press in some way because it has a negative impact. Even if the media isn’t pursuing some corporate interest, it could still misinform the public – and, therefore, you could argue that it should be regulated. Which brings us to John Zaller’s The Nature and Origins of Mass Opinion (1992). I think its a book that is commonly misinterpreted. In his dissertation, Zaller (1984) argued that:

...most new attitudes originate among the nation’s scientific and policy elite, spread outward to professional politicians and the press and (through the intermediation of these groups) diffuse gradually among the public

Not unreasonably, this is seen as a key takeaway from the Nature and Origins. If this is true, it provides a clear mechanism for how the media can have a real impact on the views of the electorate. But I don’t think that’s what Nature and Origins is really about. He clearly says that further external evidence is required for the idea that the elite are the ones doing the diffusing (p.272). I think Zaller’s methodology has some significant weaknesses. Zaller (1994) uses the Gulf War as an example of his model: as “mainstream” (read, elite) cues changed in the media, so too did the views of the public. I find this mechanism strange given that we saw above that media output is driven by consumer demand.

In any event, Zaller in that same article states that there ‘there is room for doubt’ as to the extent to which the elite are leading mass opinion (p.202). Zaller (1998) is more explicit: he looks at the effect of Monica Lewinsky revealing the affair with Clinton (and the initial Clinton administration denials). In this study, Zaller is clear that the model of elite cues cannot explain the fluctuation in support for Clinton:

… while media coverage of the Lewinsky matter explains part of the opinion change that occurred, it cannot explain all of it. In particular, the notion that the public responded mechanically to media coverage cannot explain how Clinton ended up with higher job approval ratings than he began with.

The tradition of studies on economic and retrospective voting, which maintains that the public responds to the substance of party performance, seems strengthened by the Lewinsky matter. On the other hand, the tradition of studies that focuses on the mass media, political psychology, and elite influence.. seems somewhat weaker… However poorly informed, psychologically driven, and "mass mediated" public opinion may be, it is capable of recognizing and focusing on its own conception of what matters.

Zaller’s A Theory of Media Politics (1999) was even more explicit: he found news media seemed to have no impact of presidential outcomes. Rather, as in the Lewinsky article, people’s perceptions about the ‘bottom line’ (peace, prosperity and ideological moderation) seemed to have predictive power. I think Zaller’s evolution over time is fascinating (see Bartels (2013)) and it warns against us arguing the media has an overly important impact on mass opinion.

This isn’t an easy view to sustain: Reeves et al (2015) look at the effect of The Sun’s endorsement on the 1997 and 2010 general election:

… we estimate The Sun’s decision to switch parties generated about 525,000 votes for the Labour party in 1997 and about 550,000 votes for the Conservative party in 2010.

However, from my reading of the literature, this is not representative. Most studies show that media influence is marginal (see for example, Druckman (2005), Curtis et al (1994)). Martin and Yurukoglu (December 2015) find that watching watching Fox News for 2.5 additional minutes per week increases likelihood of voting Republican by 0.3pp, watching MSNBC for 2.5 additional minutes is zero. (Ben tweeted this paper earlier in the year where the January version of the working paper found 0.9 for Fox News and 0.7 for MSNBC). A particularly interesting part of the literature looks at the effect of gaffes and their coverage and finds little effect (see graph below, taken from here)

The fact that media output is determined by consumer demand also makes me doubt the effect that the media has. Zaller’s (1992) theory is based on elites and the media, but even here, I find the theory unpersuasive: Saeki (2013) finds 'the legislators’ ideology in the House or Senate has no impact on the preferences of partisan voters.' In fact, there is a greater impact going from the publics’ ideology to the ideology of a Congressmen (see panels 2 and 3 in the graph below).. Elite and media cues seems relatively unimportant (see here).

If you aren’t convinced that (i) corporate/partisan ownership of media isn’t problematic and (ii) the media doesn’t have that much of an effect, there’s something else that assuage your concerns about the news media: our actual consumption of media. As John Sides notes on The Monkey Cage:

Most Americans do not get their news from some ideologically congenial set of outlets. First, most Americans watch very little partisan news at all. People report watching partisan news in surveys but data on what they actually watch reveals that these surveys exaggerate. For example, Princeton political scientist Markus Prior found that about 18% of Americans call themselves “regular” viewers of Fox News, but only 5% actually watch at least an hour of Fox News every week.

Second, most Americans get news from non-partisan sources or a variety of sources... Here is his graph of whether news consumption was skewed to the left- or right-wing.

A positive score means watching and listening to media that is conservative, and a negative score means watching or listening to media that is liberal. But most people are clustered near zero. They have a pretty balanced news diet.... [This is] consistent with the research of Matthew Gentzkow and Jesse Shapiro, who examined news consumption on-line and found that most consumers read ideologically diverse new outlets.

To the extent that media influence does matter, it’s only because there is an active engagement with ideas, our biases and an evaluative process in each individual. Fox News and The Sun might make you aware of something you didn’t know beforehand – and that’s okay. And no, Fox News doesn’t make you stupid. News media does not, therefore, appear to be an area where we need regulation. Chomsky’s ideas about an unthinking public and an avaricious media have little basis in fact.

3.2 The Effects of Money and Lobbying in Politics

Many people are concerned about the role of money in the political sphere. Piketty in Capital barely touches on the reasons he is concerned with inequality but when someone raised the issue with him, he stated his ‘main concern’ with inequality was ‘it creates capture of the political process.’ The concern isn’t difficult to understand: money is used to buy ads, pay lobbyists (after an election) or directly finance candidates (prior to election).

…widespread lobbying threatens the political culture and the principle of equal representation that undergirds democracy. Inasmuch as it is effective, the function of paid lobbyists is to make their clients more represented than the general public. They are hired as alchemists, to turn money into power through the production of information and the careful use of influence… One of the persistent concerns is that lobbying facilitates extractive behaviour (Teachout (2014)).

This, it might be argued, creates a corporate-centric political system that meets the interests of the wealthy which are distinct from the interests of the population (see table above from Baumgartner et al (2014)). Each one of the aforementioned acts, though, is inextricably linked with the freedom of expression: a lobbyist does not use force, he uses words. As for the link between speech and money, its worth exploring some U.S. Supreme Court jurisprudence.

In the U.S. Supreme Court decision of Buckley v Valeo 424 U.S. 1 (1976) the Federal Election Campaign Act (FECA) was considered. FECA imposed a number of limits on campaign finance and political financial support, namely it limited individual political contributions to any federal candidate in an election to $1000; limited individual contributions to political committees to $5000 per year and limited “independent expenditures"- expenditures not coordinated with the candidate or his campaign-by individuals and groups "relative to a clearly identified candidate" to $1000 per year. Volokh (2002) gives an example of what FECA meant:

Say you wanted to put a modest ad in a medium-sized newspaper saying “I’m outraged by Bush’s stand on abortion, and urge everyone to throw him out of the White House.” Under [FECA], it probably would have been a crime for you to express yourself in this way.

The Supreme Court held that some of these provisions violated the First Amendment. Broadly, the provisions that related to contributions to candidates were constitutional because of the compelling state interest in avoiding corruption. The expenditure (i.e., not going to candidates directly) limits were unconstitutional because there was little risk of corruption as compared with direct contributions and therefore it burdened speech with no compelling state interest. I disagree that a distinction between contributions and expenditures exists[2] nonetheless, the underpinning rationale in relation to expenditures is worth quoting:  

A restriction on the amount of money a person or group can spend on political communication during a campaign necessarily reduces the quantity of expression by restricting the number of issues discussed, the depth of their exploration, and the size of the audience reached. This is because virtually every means of communicating ideas in today's mass society requires the expenditure of money. The distribution of the humblest handbill or leaflet entails printing, paper, and circulation costs. Speeches and rallies generally necessitate hiring a hall and publicizing the event. The electorate's increasing dependence on television, radio, and other mass media for news and information has made these expensive modes of communication indispensable instruments of effective political speech.

In an endnote attached to this paragraph, the Supreme Court stated that ‘Being free to engage in unlimited political expression subject to a ceiling on expenditures is like being free to drive an automobile as far and as often as one desires on a single tank of gasoline.’[3] A restriction, then, on campaigning, lobbying and the like is a restriction on the speech of the contributors. More than any other precedent, it was Buckley that was relied on in the much-hated decision of Citizens United. Given that corporate and union lobbying has divergent interests from the electorate, is there a case for restricting their speech to avoid policy outcomes in their favour?

Campaign contributions and expenditures (prior to election)

This section looks at campaign contributions and expenditures and whether they actually have an effect on policy. The conventional view is that campaign contributions and expenditures ‘buy’ favours from politicians. Ansolabehere et al (2002) have three key findings which challenge this view:  

1. [Campaign spending as a form of consumption:] …almost all money in the existing campaign finance system comes ultimately from individuals and in relatively small sums. Individuals give because they are ideologically motivated, because they are excited byte politics of particular elections, because they are asked by their friends or colleagues, and because they have the resources necessary to engage in this particular form of participation, namely money.

2. [Previous literature:] Corporations and other investor contributors may still have substantial influence on policy. Evidence for this idea, however, is thin. We have surveyed an extensive literature [looking at nearly fourty articles in economics and political science that examine the relationship between PAC contributions and congressional voting behavior]. Legislators' votes depend almost entirely on their own beliefs and the preferences of their voters and their party.

PAC contributions show relatively few effects. In three out of four instances, campaign contributions had no statistically significant effects on legislation or had the wrong sign (suggesting that more contributions lead to less support). Also, given the difficulty of publishing\non-results" in academic journals, we suspect that the true incidence of papers written showing campaign contributions influence votes is even smaller

3. [Their statistical analysis:] An additional $60,000 in corporate PAC contributions (approximately one standard deviation) changes the voting score by at most 2 points; an additional $50,000 in labor PAC contributions changes the voting score by 6 point. By comparison, changing the party of a district's representative changes the voting score by more than 30 points...

Controlling for voters' preferences using district fixed effects almost completely eliminates the effects of contributions on legislative voting, in both the OLS and IV estimates...  Using legislator fixed effects eliminates the effects of contributions entirely, in both the OLS and IV. The estimated coefficients are tiny and statistically insignificant. Evidently, changes in donations to an individual legislator do not translate into changes in that legislator's roll call voting behavior.

The most interesting study in this area is Levitt (1996) who looks at the effects of campaign spending on the election results. He finds ‘campaign spending has an extremely small impact on election outcomes regardless of incumbency status’ and that ‘an extra $100,000 (in 1990 dollars) in campaign spending garners a candidate less than 0.33 percent of the vote.’ He goes on to estimate what would happen if there were limits on campaign spending. He finds that 'impact on spending caps on election outcomes is extremely small... less than 1percent of the elections during the time period examined [would be different].'

More recently, Vox found that, in the context of the Republcian presidential nominee race, the correlation between ad spending and polling outcomes was  ‘-0.2 — that is to say, ad spending is negatively correlated with polling averages in the Republican primary so far.’ The concern for money in politics seems to me to vastly overstated – and important given that Buckley is right that any restrictions on money are necessarily restrictions on speech.

Lobbying (post-election)

This ties in with another view I have: lobbying isn’t necessarily effective. Lobbying essentially encompasses speech by lobbyists directed toward legislators once they are elected; a discussion about lobbying is usually devoid of any consideration within a larger sphere of free speech because of the preoccupation with corruption. Baumgartner et al (2014) is a useful starting point. They looked at a random set of 98 policy changes, identifying the major actors, goals and different ‘sides’ each actors was on. They then followed the issues as they went through the legislative process. They found:

Our simple question was whether the policy change under consideration ever occurred. We found, overall, across the 98 issues that 58 cases showed no change; that 13 saw marginal changes; and that 27 cases saw significant policy change.

Baumgartner et al then put together thirteen separate measures of interest group resources of the actors involved. They then mapped these resources (essentially money spent on lobbying) and the outcomes. Their results are clear and counter-intuitive:

Table 3 shows the correlations between resources bT3and outcomes, but this time not for individual actors but for the aggregated ‘‘sides’’ which all seek the same outcomes. Now all of the individual-level measures of resources become statistically insignificant as businesses and citizen groups and other types of groups join together and aggregate their resources into heterogeneous sides. None of those relationships remain.

[Table 4] asks, for each type of material resources that might potentially matter to a lobbying group, whether the side with more of that resource won… The purely financial types of resources mattered little. Numbers ranged from 50percent success for those with greater levels of campaign contributions to 53 percent success for those with greater business resources, and none of those percentages is statistically significant. Those with greater levels of the financial resources and those with lower levels of the financial resources achieve their policy goals an equal amount of the time (my emphasis).

Walt and Mershimer in their book The Israel Lobby argue that ‘loose coalition of individuals and organizations that actively work to shape U.S. foreign policy’ (p.112). Absent lobbying, there is no interest in maintaining the current levels of support for Israel because, they claim, Israel is not a strategic asset. In fact, it is a liability because it attracts criticism if U.S, encourages terrorism and makes nuclear proliferation more likely. They further argue that there is no convergence of values because of the way Israel treats Palestinians. These are all incorrect statements which I will not be discussing, the part I want to focus on is the purported power of the lobby. If their view is correct, there are serious reasons in favour of lobbying being significant and therefore regulating that exercise of speech.

Koplow (2011) undertakes the only quantitative study I know assessing the claims of Walt and Mershimer. The first thing to note that that up until 1974, financial aid and grants to Israel were relatively low. In 1974, they rose to $2.6billion up from $474million the previous year. Koplow says that  'strategic considerations [are] a simple way of explaining this jump in support—Israel’s role in supporting Jordan during the PLO revolt and Syrian invasion in 1970' but 'the coalitional logrolling argument has no comparable explanation.' He goes on to say:

If level of aid is used as the benchmark of political effectiveness, there is no explanation as to why the American Jewish community, as embodied by the Israel lobby, was relatively ineffective before 1974 and has been so remarkably effective since. Support for Israel among American Jews was the same before and after the increase in American support, and thus the greater lobbying success does not reflect a change in American Jewish attitudes. Pointing to the activities of pro-Israel lobbying groups does not account for the enormous policy shift that took place in 1974, or the long-term patterns in military aid..

This strategic rationale is also consistent with military sales to Syria in 1973 increasing to $2.46 billion, up from $580 million in 1972. There was, obviously, no Syrian lobby controlling the stakes. But whilst this is a persuasive example, it doesn’t tell us about wider trends when it comes to lobbying. Koplow, however, notes that even here the argument is quite weak:

From 1998 through July 2010, pro-Israel groups have spent $25,248,057 on lobbying activities, and from 1990 through October 2010, pro-Israel contributions from groups and individuals have totaled $93,878,940. Yet, there is actually a reverse correlation over this period of −.47 between lobbying totals and yearly aid to Israel, and reverse correlation of −.68 between contributions and yearly aid to Israel… if the power of the pro-Israel lobby is driving policy, it must be doing so in a way that cannot be measured, since lobbying activities and donations have no correlative, let alone causal, effect on levels of U.S. support.

I’m a big fan of making the jingoistic claim that the American people are the most powerful Israel Lobby. And that is precisely the mistake that Walt and Mershimer make, and one of the reasons I doubt the Gilens and Page study: they ignore public opinion. They simply state their view about Israel as a strategic liability with differing values and then seek to explain policy on the basis that everyone should agree. Except they don’t. Koplow notes that:

Only in four out of twenty-one instances have less than 50 percent of the public indicated holding a very favorable or mostly favorable view of Israel, and those holding an unfavorable view of Israel have never outnumbered those with a favorable view during this time period, with an average gap between the two groups of thirty-one points.

This support has a causal effect on U.S policy. Koplow gives the example of the failure of the ‘Israel lobby’ in stopping the sale of 5 AWACS to Saudi Arabia in the 1990s and their failure in obtaining the release of Jonathan Pollard. Significantly ‘by looking at cases where the public’s preferences differ from those of pro-Israel groups [in opinion polling which he cites], we can see that the causal arrow runs between voters’ preferences and politicians’ policy views.’[4] I want to focus on a more recent example: AIPAC’s botched attempts at controlling everything in relation to the Iranian negotiations on their nuclear programme. Late in 2014, there was an attempt to table a sanctions bill. Here is how Eli Lake describes in The Daily Beast how AIPAC failed:

AIPAC was forced, in the wake of Democratic opposition, to retreat for the moment on the Iran sanctions bill the group had been pushing for months. Then, nearly every Republican in the Senate ignored AIPAC’s call for a retreat on the bill, and decided to keep on pushing for a vote on it, anyway.

Somehow, on the issue arguably of most importance to both the Israeli government and America’s pro-Israel community—Iran and its nuclear ambitions—AIPAC didn’t merely fail to deliver. It alienated its most ardent supporters, and helped turn what was a bipartisan effort to keep Iran in check into just another political squabble.

There are many other instances, whether its U.S rejection of redlines or meetings. The most important point from Koplow’s study is that lobbying doesn’t seem to matter much and public opinion does.

Alternative Argument

I want to be very honest and say that the literature cited above in terms of lobbying is one sided. There is evidence that lobbying does matter. Gilens and Page (2014) is a study that undermines both the claim that lobbying doesn’t matter and that public opinion matters. They find that

…a proposed policy change with low support among economically-elite Americans (one out of five in favor)is adopted only about 18 percent of the time, while a proposed change with high support (four out of five in favor) is adopted about 45 percent of the time. Similarly, when support for policy change is low among interest groups (with five groups strongly opposed and none in favor) the probability of that policy change occurring is only .16, but the probability rises to .47 when interest groups are strongly favourable..

When the alignments of business-oriented and mass-based interest groups are included separately in a multivariate model, average citizens’ preferences continue to have essentially zero estimated impact upon policy change, while economic elites are still estimated to have a very large, positive, independent impact.

I have serious reservations about this study: the views of the median voter and the ‘economic elites’ are highly correlated (0.78) which makes the conclusion that the median voter is not significant hard to sustain. Moreover, as John Cassidy notes, ‘the explanatory power of some of the equations that Gilens and Page use is weak.’ The three-variable probability model explains less than 10% of the variation in the data (R-squared = 0.074). To repeat: R-squared = 0.074; the standard deviation of its errors is around 5% less than the standard deviation of the dependent variable! And, yet, this is the study that many use to bash the U.S. as an oligarchy (see here for the fringe left journalist, Adam Johnson, basing most of his argument against Neil DeGrasse Tyson on the basis of this study). Given this weak statistical underpinning, I am inclined to believe the contradictory studies quoted above.

I would also add the following as reasons for downplaying Gilens and Page. Going back as far as Page and Shapiro (1983) we know that there is a huge congruence between public opinion and policy. They found in 66% of cases, policy change was congruent with public opinion change. For the remaining 34%, 'approximately a quarter vanish when lags longer than one year are allowed for.' The 34% also drops 24% where there are ideal policy measures. The number is reduced more still when they account for temporary opinion changes: the final number that Page and Shapiro give as ‘'better estimate of congruence’ is 87%.

Nonetheless, even whilst rejecting Gilens and Page’s study, you can make a reasonable argument that lobbying and campaign contributions are related to positive outcomes for corporate and union interests. Take Kalla and Brockman (2015) who run an experiment to see how far campaign contributions go in allowing for access:

In the experiment, a political organization attempted to schedule meetings between 191congressional offices and the organization’s members in their districts who were campaign donors. However, the organization randomly assigned whether it revealed to congressional offices that prospective attendees had contributed to campaigns [Constituent Condition v Revealed Donor Condition]. When informed prospective attendees were political donors, senior policy makers made themselves available between three and four times more often.

Note that their findings ‘leave open the question of whether increased access translates into tangible influence’ but access might be a reason in and of itself to try to limit contributions. More importantly, Denes and Duchin (2015) find 'connected firms are 10% more likely to win a contract. Connected firms receive larger contracts, with longer durations and weaker incentive structures.’ Canayaz et al (2015) find that the companies and organisations with former officials have abnormal returns:

The equally-weighted portfolio of revolver-hiring firms delivered average returns of 18.95% per year in the three years prior to the hiring. These returns compare favorably to average returns of 11.98% per year for all other firms in our sample during the same 1990-2012 period [i.e., those firms without officials joining]. The annualized difference between the two, 6.97%, is statistically significant, and it remains so once risk-adjusted using the one-, three-, or four-factor asset pricing models. For instance, using four-factor alphas the difference between these two portfolios is a highly statistically significant 7.43% per year.

Our results are also stronger in the years immediately before the hiring and significantly weaken as we move further away from that date, as we would expect if they were the consequence of revolvers helping their future employers before making their move.

I should emphasise that the literature is mixed. In addition to the studies above, for example, Chen et al (2014) find that whilst ‘lobbying is positively correlated with future excess returns’ that ‘that most lobbying expenditures are not associated with abnormal returns, and that simply spending the most on lobbying does not necessarily lead to better stock market returns.’ Smith’s (2001) notes that ‘serious studies of legislative behaviour have overwhelmingly concluded that campaign contributions play little role in floor voting.’

Rather than simply say that the literature is mixed and most support my view, I would say that even granting lobbying is effective, this isn’t necessarily a bad thing.  Take the following fun example from 2012 from the New York Times about the “United States Lobby” that exists in Israel:  

…behind the scenes, the Obama administration was conducting a quiet campaign that would strengthen the view, already circulating among Israeli security professionals, that prematurely attacking Iran would not advance Israel’s interests and would damage Israel’s relationship with America. Instead of holding Israel at bay or threatening punitive action, the administration was upgrading American security assistance to Israel…

….increased American assistance has been accompanied by closer institutional links between the two countries’ defense and intelligence communities, as well as more intimate personal ties between both communities’ top echelons. Through numerous meetings in Tel Aviv, Jerusalem and Washington, the Obama administration has used these connections to convey an unambiguous message: Do not attack before all nonmilitary efforts to roll back Iran’s nuclear program have been exhausted.

Ever deeper American-Israeli defense ties have created what might be labeled a “United States lobby” among Israeli security professionals, who now have a strong interest in continuing the close partnership. It is no accident that the security institutions have become among the most vocal opponents of attacking Iran. No one knows better than they what is at stake if they ignore Washington’s concerns.

The point being that another word for lobbying is persuasion or making the other person aware to agree with a certain position. It’s not some economic quid pro quo, its an exchange of ideas. Take another example from 2013 reported in the New York Times which is about banking legislation:

One bill that sailed through the House Financial Services Committee this month — over the objections of the Treasury Department — was essentially Citigroup’s, according to e-mails reviewed by The New York Times. The bill would exempt broad swathes of trades from new regulation…. Citigroup’s recommendations were reflected in more than 70 lines of the House committee’s 85-line bill. Two crucial paragraphs, prepared by Citigroup in conjunction with other Wall Street banks, were copied nearly word for word. (Lawmakers changed two words to make them plural.)

Icky, right? Maybe not, as the report goes on to say:

Industry officials acknowledged that they played a role in drafting the legislation, but argued that the practice was common in Washington. Some of the changes, they say, have gained wide support, including from Ben S. Bernanke, the Federal Reserve chairman… Lawmakers who supported the industry-backed bills said they did so because the effort was in the public interest… Citigroup executives said the change they advocated was good for the financial system, not just the bank.

What if Citigroup know the effect of the legislation better than others? What if the reason why there is ‘wide support’ including the Fed isn’t because of an attempt at political capture but a sincere attempt to have a workable policy? What if being close to the U.S. intelligence community reminds Israeli security officials of how important their relationship is? None of these questions have an easy or consistent answer. But I think Smith (2001) is right when he notes that:

Campaign money is of little value if it cannot be turned into an electoral victory and it makes little sense to betray one person convictions, lose the support of one’s party, and offend public opinion in order to obtain a contribution…to find corruption, we must assume that the representative is acting against his or her own best judgment and principles, against the wishes of a majority of his or her constituents, and against the intense preferences of a minority… How likely is any legislator to do such a thing for a mere contribution? (p.55, 59)

For clarity: my primary argument is that things like campaign contributions and lobbying don’t matter. But, in deference to how mixed the literature is, I would say that our aversion to interest groups is misguided. Whether it’s Save the Children campaigning for minimum levels of aid or Citigroup lobbying for certain legislation, we needn’t jump to accusations of corruption or cronyism. Democratic politics is about legislators listening, being persuaded in a marketplace of ideas – and it really doesn’t matter if the person putting forward that idea is Exxon Mobil or a constituent. The burden for suggesting that there is impropriety is necessarily high and I simply haven’t seen any convincing evidence that there is necessarily or mostly a link between money, lobbying, politics and impropriety.

3.3 Video Games: Crash Bandicoot Shouting Fire in a Crowded Theatre

I haven’t really played video games since I was 11. And even then, my favourite games were Simpsons Hit and Run and Crash Bandicoot. I also played GTA but mostly liked putting in cheats to fly cars around (I don’t think I played any of the missions). Fortunately for me, there is a vast literature that exists on the subject of the negative effects of video games. Ferguson (2014) provides a starting point for what the experimental literature tells us:

For both movies and videogame violence, some studies find evidence for effects on increased aggression (e.g., Ivory & Kaestle, 2013; Turner & Berkowitz, 1972), null effects (Ramos, Ferguson, Frailing, & Romero-Ramirez, 2013; Tear & Nielson, 2013) or even reduced aggression (Feshbach, 1961; Mueller, Donnerstein, & Hallam, 1983; Shibuya, Sakamoto, Ihori,& Yukawa, 2008; Valadez & Ferguson, 2012). Overall, making clear, declarative statements from this body of work is difficult.

Even the meta-analyses on this question are divided (see this admirable critique of a study I put a lot of weight on below). My favourite example of an experimental study is Barlett et al (2009) which finds that individuals who play violent video games are more likely to ‘give hot sauce to hypothetical individuals who do not like spicy food.’ But, this inconsistent data seems to be heavily impacted by a lack of ecological validity. This is shown by two recent studies looking at real world (i.e., not in the lab) effects of media violence. Ferguson (2014) looks at both movie violence and video game violence. To start with the former, Ferguson obtained the top five grossing films over five year periods from 1920 to 2005. These films were then rated by trained raters for their violent content. Ferguson then applied a bivariate analysis tacking the results of movie violence with homicide. He found:

When only the years from 1970 on are considered, the relationship reverses in trend with homicide rates [negatively] correlated r =−.28 (df = 8, p = .50) with frequency of movie violence and a strong r =−.61 (df = 17, p = .11) with movie graphicness. For the years prior to 1940, movie violence demonstrated an almost perfect inverse [positive] relationship with societal violence with the two variables correlated r =−.98.

This explains why his paper is called ‘Does Media Violence Predict Societal Violence? It depends on What Look at and When’. Recent movie violence (1970 to 2005) appears to have no role in violence but it seems to be highly correlated with homicide trends 1940 to 1970. Fergusson uses the same methodology for video games and finds:

As can be seen, videogame violence consumption in society is inversely related to societal youth violence. The bivariate correlation between these two phenomena is r =−.85 (df = 15, p = .001). [I.e., more video games is correlated with less violence]

Markey et al (2014) is another study that looks at real world effects. They look at homicides and aggravated assaults. Their study investigated the associations between violent crime and video game sales, Internet keyword searches for violent video game guides (the idea being that people would look for ‘cheats’), and the release dates of popular violent video games. They find:

Annual trends in video game sales for the past 33 years were unrelated to violent crime both concurrently and up to 4 years later. Unexpectedly, monthly sales of video games were related to concurrent decreases in aggravated assaults and were unrelated to homicides. Searches for violent video game walkthroughs and guides were also related to decreases in aggravated assaults and homicides 2 months later. Finally, homicides tended to decrease in the months following the release of popular M-rated violent video games

These results should not be that surprising. South Korea and the Netherlands spend more on video games than most countries and are far safer than the U.S (source). The literature above from experimental studies are inconsistent and, as Ferguson (2014) and Markey et al (2014) show us, they seem to lack ecological validity.

For the sake of transparency, Markey et al’s 2014 study had some criticism from Bushman et al (2015) because, apparently, ‘trends in gun violence in youth are actually consistent with gun violence trends in PG-13 movies.’ Except, they use CDC statistics on 0 – 19 year olds victims. This is wrong because the majority of the perpetrators would have been adults. When Markey et al (2015) re-ran the regressions, their 2014 results stand for youth as well. Interestingly, in that latter Markey et al (2015) riposte is the following graph:

I have given my reasons for being in the ‘unsure’ camp in respect of the first question (if we are talking about lab-condition measured aggression) and being in the ‘disagree’ camp for the second question (especially since the 1970s).

3.4 Porn: Having an Orgasm in a Crowded Theatre

The first question to consider about porn is whether it can really be part of ‘free speech.’ Free speech typically covers the conveyance and marketplace of ideas. John Finnis (1967) is probably the leading proponent of the view that porn is not speech. In his essay 'Reason and Passion: The Constitutional Dialectic of Free Speech and Obscenity', he argues that the rationale for free speech is rational criticism of government. This can only be done through ideas which engage reason rather than passions. Accordingly, ‘to the extent that expressions derive from the passion end of the reason-passion continuum, the rationale for that freedom disappears.’ In the following passage, Finnis seems to getting at the Yiddish saying ‘a stiff prick turns the mind to shit’:

The designed effect of these techniques [pursued by pornographers] is always the same-the replacement of aesthetic attention to the material with an attitude in which the practical concerns of the reader or viewer (in this case, a concern to achieve the emotionally aroused states which he desires for himself) [i.e., passions] intrude upon and suppress an understanding contemplation of the created symbol [i.e., any room for reason].

I don’t buy it. There is, of course, the argument that the government is inept to define what constitutes an idea and what constitutes an appeal to the passions and I will return to this in the next section. But at its core, I think Finnis is wrong. I think sex and projections of sex are vital for people developing their own sexual identity and provoking traditionalists to change (think Lady Chatterley). I could be wrong though, I am unfamiliar with the literature on this point (please do let me know if you know of any relevant literature). I would note though, that in a survey of men who have sex with men (MSM), pornography was the most common source for finding out about ‘enjoyable sex’ and respondents ‘were more likely to rate pornography as helpful or very helpful, than unhelpful or very unhelpful’.

In any event, I think that the debate about whether porn constitutes speech is largely made redundant by the issue of government incompetence and value pluralism. The far more important question is whether we pornography causes harm. There are two lines of research that are worth examining: the impact of pornography on sexual violence and the impact on sexism. In terms of making these claims, the big dog is Catherine MacKinnon (1996) in Only Words she states that

Sooner or later, in one way or another, the consumers want to live out the pornography further in three dimensions. Sooner or later, in one way or another they do. It makes them want to; they feel they can get away with it, they do... As pornography consumers, teachers may become epistemically incapable of seeing their women students as their potential equals... Doctors may molest anesthetized women, enjoy watching and inflicting pain during child birth (p.19).

The message of pornography, is ‘addressed directly to the penis, delivered through an erection, and taken out on women in the real world’ (p.21). I will return to the act-speech theory below. MacKinnon states that the empirical literature is on her side (p.37) – but is it? Ferguson and Heartley’s (2009) meta-analyses on the of experimental studies ‘conclude that results are generally mixed and the type of research methodology used in the study often greatly affects the outcome.’ The experimental literature ‘reveal that effects appear negligible, temporary and difficult to generalize to the real world.’

As with pornography, it is better to look at the real world effects literature. One line of studies looks to self-report questionnaires to assess this. Bonino et al (2006) carried out a cross-sectional study (n = 804) on the basis of self-report questionnaires investigating link between exposure to pornography and self-reported sexual harassment and rape. They found that

...reading pornographic comics and magazines significantly increased the likelihood of having sexually harassed a peer or having forced somebody to have sex, while viewing pornographic films or videos increased the likelihood of being a victim of sexual violence.

So it’s not like MacKinnon is making up the idea that there are some studies that support her argument. But again, is this study representative of the literature? Ferguson and Heartley look at the literature on pornography consumption and rape rates. My reading is that the literature, if anything, suggests that porn has a negative impact on rapes:

Scott further found that prevalence of adult entertainment venues such as strip clubs and book stores were not correlated with rape rates but circulation rates of neutral magazines (e.g., Field and Stream) did correlate with rape rates... [Looking into porn consumption and rape rates in West Germany, Denmark, Sweden and the U.S] revealed that increases in pornography consumption were associated with constant or decreased rape rates in each of the countries except the United States.

[But Ferguson’s own data shows even] rape rates [in the U.S] are negatively related to increases in pornography releases. Increasing availability of pornography in other words is associated with declining rape rates.

For those interested, Field and Stream is a magazine about fishing, hunting and other outdoor activities. This is representative of the literature. The question of whether porn contributes to sexist ideas of women, the perpetuation of rape myths etc., though is not directly addressed by these studies. It is suggestive that it does not have an effect but no more than that. Garos et al (2004) found 'a positive correlation between pornography use and benevolent sexism, such that participants who used pornography more frequently displayed more benevolent sexism'. It is very easy to find studies that conflict with this. Diamond (2009) notes the following:

[Padget et al. 1989] compared the attitudes toward women in a sample of patrons of an adult movie theater with a college sample of men and women. Their findings essentially and significantly showed the patrons of an adult movie theater had more favorable attitudes toward women than either male or female college students... [Reiss (1984)] reported on six different National Opinion Research Center annual General Social Surveys that found that those men and women who had seen an X-rated movie in the past year were more gender equal than those who had not seen any.

Interestingly, a recent study found that porn-consumers are more likely to have egalitarian views but less likely to identify as a feminist (see table above from Kohot et al (2015)). I am unsure what is representative of the research but it seems to be mixed with no definitive answer (see Layden (undated) for the literature show positive correlations, Diamond (2009) for negative correlations). Hald et al (2013) describe the literature as ‘equivocal’ and say it ‘does not fully allow for casual conclusions’.

3.5 Sexist Speech: Crash Bandicoot Making Rape Jokes in a Crowded Theatre

Sexist Humour

“It’s just a joke” so goes the common refrain. The point that many on the left make – and that mostly goes unanswered because free speech advocates are too focused on normative discussions – is that it’s not just a joke. It has real world effects. Take jokes about rape or sexist jokes – what does the literature tell us? Thomae and Viki (2013) who look at whether exposure to sexist jokes vs non-sexist jokes has an impact on rape proclivity. They find:

….the relationship between HS [hostile sexism, which was measured prior to exposure] and rape proclivity was significant in the sexist joke condition, β = .51, t = 4.04, p < .001. This relationship was also significant in the neutral joke condition, but was weaker than in the sexist joke condition, β = .31, t = 2.47, p=.017. These results are in line with our hypotheses: exposure to sexist jokes appears to strengthen the relationship between HS and rape proclivity.

Sure, it only affects those who are already sexist – but it’s not ridiculous to suggest we should try to reduce the amount of sexist results as a result, right? Again, I happen to disagree with this conclusion but the point is this is not a straight forward normative question. Empirically, this study isn’t strong: n = 96 men from the University of Kent. It might be true. It might not be. The things that cause men to rape seem to have a broader origin than their environment so its likely that it does not. Take Långström et al (2015) who find that 'genetic factors (40%) and non-shared environmental factors (58%) explained the liability to offend sexually more than shared environmental influences (2%).'

Ford et al (2001) suggest another way sexist humour may have a negative effect:

(1) for men high in hostile sexism, exposure to sexist humor creates a perceived social norm of tolerance of sexism relative to exposure to nonhumorous sexist communication or neutral humor, and (2) due to this ‘relaxed’ normative standard in the context of sexist humor, men high in hostile sexism anticipated feeling less self-directed negative affect upon imagining that they had behaved in a sexist manner.

Some of this is tautological: hearing something get approved makes you think its.. approved. But do we want all ideas approved? The argument from a free speech advocate (in fact, it’s an argument I made in Part 1!) is that if you give air to an idea, you can expose it. But what if that’s not the case? You can bleat on about ‘exposing ideas’ and free speech being great but you haven’t evidenced anything. And there’s a slightly persuasive reason why this may not be the case for sexist jokes. Here’s an extract from the Thomae and Viki study:

a humorous communication activates a conversational rule of levity, resulting in a non-serious mind-set on the part of the receiver, which prevents messages from being interpreted critically. By switching to a non-serious mind-set, the recipient accepts the local norm implied by the humor.

Video Games, Again

I don’t care about GamerGate. It doesn’t show anything about society, the ‘left’, men, feminists – nothing. It’s boring. My cursory glance at the literature leads me to conclude that the most popular video games are, on average, sexist. For example, Burgess et al (2007) looked at 250 video game covers and found that over two thirds of the female characters (compared to 10% of male characters) were represented in stereotyped gender roles or the subject of physical objectification.

I’m more interested in the literature of what the effect of video games on sexism is. Anita Sarkissian’s series on video games and sexism is probably the most well known exposition of the view that is has a negative impact. I find that people either give Sarkissian too much credit or not enough. One of her most prominent critics is the Youtuber Thunderfoot. Thunderfoot doesn’t really seem to accept the nuance in Sarkissian’s videos. Here is what she says:  

Likewise engaging with these games is not going to magically transform players into raging sexists. We typically don’t have a monkey-see monkey-do, direct cause and effect relationship with the media we consume. Cultural influence works in much more subtle and complicated ways, however media narratives do have a powerful cultivation effect helping to shape cultural attitudes and opinions.

Sarkissian is caricatured as saying that video games make people sexist, this is clearly not the case. But the reason I think people give her too much credit is because this nuance is frequently lost. Here is what she says elsewhere in her video series:

.. the negative impacts of sexual objectification have been studied extensively over the years and the effects on people of all genders are quite clear and very serious. Research has consistently found that exposure to these types of images negatively impacts perceptions and beliefs about real world women and reinforces harmful myths about sexual violence.

You can very easily read these statements as not contradicting each other but I think its incredibly slippery. Apologies for the cross-referencing, but this is precisely the problem with the Eustonite response to my post about Islamism. “I’m not saying x causes y, but x creates a culture in which y can thrive or it creates the “mood music” for y” – this is the refrain of the some feminists clamouring for censorship, Eustonites who have a deep desire to make Islamism relevant and, to my mind, people making meaningless claims. While we’re here, take Sarah Ditum’s recent New Statesmen article:

It is not a question of whether pornography “caused” Matthews and Hoare to commit their crime. What matters is this: in a world sodden with violence against women, pornography is one more form of it.

Oh right, cool, I get it. There’s no causal claims being made. It’s just that pornography is one example of sexism. I got it. That seems like a reasonable claim. And yet, in the same article that utters that this is not about causation (or even correlation), you will get this:

Through [pornography], men and women alike learn what women are supposed to be for: something to fuck, something to use, something to hurt if you’d like to, and something to dispose of when you’re finished. Matthews and Hoare dismembered Becky Watts with a circular saw. [List of rapists and murderers].. And this pattern does not apply only to confirmed criminals and obvious monsters… And this is how porn operates: first through the eyes, and then in the mind, and then back through the body, against other bodies.

What?! You just said you weren’t making a causal claim! Sarah Ditum is either making a meaningless argument (if we are saying there is no causal impact) or being inconsistent. Moving on to what the literature actually says, for someone who prides themselves in being a scientist, Thunderfoot doesn’t cite any research in support of his position that video games do not lead to violence/sexism. Anita Sarkissian does provide studies which support her claims. Yao et al (2010) took a group of individuals (n = 74) some of which played games with high sexual content and others with a control game. They then carried out a likelihood-to-sexually-harass scale and found:

A simple one-way ANOVA of participants’ LSH [likelihood to sexual harass] scores revealed a significant effect F (2. 74)=5.97, p<.05). Specifically, players of Leisure Suit Larry reported a significantly greater tendency to sexually harass (M=105.37, SD=20.25) than did players of the Sims (∆M=22.50,  p<.05) and PacMan II (∆M= 14.30, p<.05)... [This] provides strong empirical evidence that a sexually oriented video game with themes of female “objectification” may prime thoughts related to sex, encourage men to view women as sex objects, and increase the likelihood of self-reported tendencies to behave inappropriately toward women in social situations.

Another interesting study is Dill et al (2008). Their study has a similarly low n (n = 181, 120 of which were female). There were two groups: one (their Sterotype group) subjected to a Powerpoint of images of women from GTA: Vice City, GTA: San Andreas, Dead or Alive Xtreme Beach Volleyball 2, BMX XXX, Saint’s Row, Resident Evil and Gears of War etc. The other was subjected to a Powerpoint of both male and female senators (the control group). They find:

Tolerance for sexual harassment was greatest for males in the Stereotypical group (M = 41.0), followed by males in the control group (M = 47.64), and females in the control group (M = 48.47). The group with the least tolerant attitudes towards sexual harassment were the females in the Stereotypical group (M = 49.8)... [In particular] men showed a greater tolerance for sexual harassment of a young woman by an older man (her professor)

The authors suggest that the reason that women subjected to the stereotype condition have lower sexual harassment judgments is because they become galvanised after seeing sexist depictions. The idea, then, that Sarkissian is pulling her arguments out of thin air should be roundly rejected. That said, as I was reading through these studies issues of ecological validity entered my mind. We have already seen that in the context of porn and video games that the experimental literature doesn’t tell us a great deal about the way the world actually is because of the short term effects. I know of one (kind of alright) study that looks beyond this and finds video game use predicts benevolent sexism (Stermer and Burkey (2012)).

But the whole idea is grounded in cultivation theory (i.e., the idea that media shapes cultural attitudes). The literature above on video games and porn should make us doubt it. A meta-analysis on 5,799 separate findings from 97 studies of cultivation analysis ‘reports that the average correlation across all those findings is .10 and that the average partial correlation is .09... This means television exposure predicts only about 1% of the variation in cultivation indicators’ (quoted in Potter (2014)).

The paragraph above was drafted as the final paragraph to this section – and it bothered me because I couldn’t provide a real-world-effects study that gave us a better answer – until I came across Breuer et al (2015). Breur et al note that all of the previous literature (much of it cited above) does point to ‘second order effects’ (i.e., sexism) but that:

[…they] assessed only the short-term effects of adolescents or college students... the longitudinal studies on violence have been limited to periods of no more than 1 month, and all of the studies in that area have relied on self-selected or convenience samples composed mostly of adolescents or college students.

So, they undertook a 3 year longitudinal study to examine the effects of video game use on sexism. Their results, like the literature in relation to video games and porn, contradict the experimental findings above:

…no cross-sectional association between sexist attitudes and overall video game use for both men and women. On the longitudinal level, the only statistically significant finding was a negative association between video game use at time 1 and sexist attitudes at time 2 for males ( p = 0.027). However, the size of this effect (b = –0.08) can be considered negligible. All other longitudinal associations were both small and nonsignificant (b < 0.13).

Fortunately, my approach to free speech does not require me to make definitive statements about any of this literature (for reasons that will be apparent in Part 3). It appears the experimental literature mostly agrees with Sarkissian but the empirical literature otherwise is in its early stages with contrary results. I’m sure depending on which side of the issue you’re on, you’ll cherry pick. I would note the literature above should not allow you to be confident in any view. Breur et al is the first study of its kind and I’m sure there will be a lot more which goes in either direction. If I had to express a view, it would be one that took into account all the literature in this post which suggests little effect.

3.6 Race Related Speech: Hollywood, Skokie and Umugandas in Rwanda

I really enjoyed Spike Lee’s film ‘Bamboozled.’ The protagonist of the film is a black American who works as a producer in a U.S television network. In an attempt to shock his boss, he proposes a ‘New Millennium Minstrel Show’ where two individuals, Mantan and SleepN’Eat, live on a watermelon plantation. The two individuals are black but also in black face. The excerpt below is from the script of the film, and starts with an exclamation from Mantan in the pilot of the new show:

Mantan: Cousins, I want all of you to go to your windows.  Go to your windows and yell.  Yell, I'm tired of the drugs, the crack babies born out of wedlock to crack head aids infested parents.  I'm tired of the inflated welfare rolls while good wholesome Americans bring less and less of their paycheck home every two weeks... Aren't you tired of these basketball-dunking, football-running, hop-hip rapping ebonic-speaking sex offenders who got ten kids from ten different Ho's?... Go to your windows and yell out, scream with all the life you can muster up inside your assaulted, bruised and battered bodies: I’m sick and tired of niggers and I’m not going to take it anymore! [emphasis in original]

If the ‘New Millennium Minstrel’ show was on our TVs – would it have a negative societal or individual impact? This subsection is divided into four parts: the effect of racist media depictions on (i) the individual, (ii) on wider societal attitudes and (iii) violence.

Effect on the minority group

The first study looks at the impact that a racist statement might have on an individual. I should say at the outset that n = 50 black Americans and 46 white American, so over-reliance on this study is misjudged but it is, at least, the beginning of the conversation. Lyle (2015) looks at the effects of a racist statement on black peoples’ political efficiency (‘the self-assessment of one’s personal capacity for politics’) and race-specific collective self-esteem (used a proxy for ‘privately regarded their racial group and activate beliefs about public regard for their racial group). The racist statements that the participants of the study were given was either made by an ‘ordinary American’ or a ‘prominent political figure.’ Lyle found:

... when the anti-Black message was attributed to an ordinary American, it did not have statistically significant main or interaction effects on either self-esteem measure for African–Americans... however, when the anti-Black message was attributed to a political figure, it was significantly associated with lower private R-CSE among African–Americans

...the anti-Black political message also had a significant effect on internal political efficacy among African–Americans... African–Americans exposed to the anti-Black political message reported feeling significantly less competent to participate in politics than those in the control group.

But what if its just more than feelings of alienation? Mullen et al (2004) look at the impact of ‘ethnophaulisms’ (in essence, hate speech) on suicides in European immigrant populations. The dataset was taken from Allen (1984) which is a ‘compilation of more than 1000 terms used in hate speech.’ These were then rated according to their negativity: 1 being very negative and 7 being very positive. So, for example, ‘dumb Polack’ would score negatively whereas ‘taffy’ would rate relatively mildly. A low valence score indicates ' extreme negativity in cognitive representation’. Mullen et al regress these scores against suicides in immigrant populations in the U.S. and find that:

…after partialing out the variability due to origin suicide rates and ethnophaulism complexity, ethnophaulism valence was a significant independent predictor of immigrant suicide rates, β = -0.431, t (6) = 2.464, p = .0244…the suicide rates for ethnic immigrant groups in the United States were significantly predicted by the negativity of the ethnophaulisms used to refer to those ethnic immigrant groups.

[Moreover,] even after taking into account ethnic immigrant group size and the suicide rates for those ethnic immigrant groups in their countries of origin, the suicide rates for ethnic immigrant groups in the United States were significantly predicted by the negativity of the ethnophaulisms used to refer to those ethnic immigrant groups in hate speech.

As stated, this study is looking at the negativity / valence of the language toward immigrant groups. It does not tell us about how much this language is used which is unfortunate but if we take valence as a proxy for use (admittedly, a moderate leap), perhaps its plausible. Here’s another interesting study: Gubler and Kalmoe (2015) who find that ‘that mild “fighting” words (e.g., “battle,” “fight”) combined with a reference to the outgroup provoke significantly greater support for policies that harm the outgroup among some citizens, even when the intended use of violent words is clearly rhetorical.’ The graph below shows the effect applies even more to those who are not pre-diposed to racism.

Effect on the majority group

Tukachinsky et al (2015) is another interesting study that looks at the 345 most viewed U.S. television shows spanning the years 1987 to 2009. They look at how ethnic minorities are presents in these shows and then look at how good those qualities are at predicting white attitudes toward Black people and Latinos.

Overall, all ethnic minority characters tended to be presented as highly positive, but Blacks were slightly more positive than Latinos... The majority of Black (76.7%) and Latino (74.1%) but less than a half (47.1%) of Asian characters were likable. Comparisons of means of characters’ likability suggest that on average, Asian characters were presented as significantly less likable (M = 2.47, SD = .51) than Blacks (M = 2.75, SD = .47) and Latinos (M = 2.74, SD = .44)

Model 1 reveals that the number of highly professional and social Latinos characters had a significant positive effect on Whites’ attitudes toward Latinos. The number of Latino characters and the number of hyper-sexual Latino characters were associated with more negative attitudes toward Latinos (however this result did not reach conventional levels of significance [p = .08], possibly owing to the small N at the media representation level). The number of good and liked Latinos characters was not found to have a significant effect on attitudes toward Latinos, when controlling for all other variables.

Model 2 examined the media’s contribution to public attitudes toward Blacks. The model revealed that both the prevalence of Black characters (i.e., the overall number of Black characters), and Black characters’ professional and social status had a positive and significant effect on attitudes toward Blacks. The number of good and liked Black characters and the number of hypersexual Black characters significantly reduced the support toward Blacks

One clear finding is that hypersexual depictions of Latinos and Black people have a negative impact on how favourable people felt toward said group. But another element of the results is quite strange: why would the overall number of Black characters have a positive impact but the numbers of good/liked black characters reduce support toward black people? The results seem counter-intuitive, especially for those who complain about the lack of ethnic minority representation: what if likeable black people on TV make things worse? The authors provide one possibility:

...to promote overall positive attitudes toward the outgroup, a likable media persona need also be perceived as typical of the group... It is conceivable that many of ¨ the positive and likable characters in this study were deemed nonrepresentative of Latinos and Blacks as a whole... For example, in the case of The Cosby Show, Jhally and Lewis (1992) suggest that many White viewers felt that the Huxtables were characteristically “White.” If this is the case, it appears that in some instances, the race/ethnicity of minority characters... is almost incidental

Other studies look at the representation of Blacks in (real) crime segments in news broadcasts. Essentially exposure to representations of Blacks committing crime leads to the view that black people are violent – and this seems to be a pretty consistent finding (Dixon (2015)). This is a free speech issue because if we want to avoid the consequence of people holding the belief that black people are violent – these studies can easily be used to argue we should censor coverage of Black crime.

Violence

In Rwanda, there was (and is) a kind of community service in the form of ‘Umugandas.’ The etymology of the word can be traced back to the Nguni proverb ‘umuntu ngumuntu ngabantu’ which means ‘a person is a person through other persons’ – isn’t that lovely? What it means in practical terms is that on every Saturday, every able citizen is required to do some form of community work: sweep the streets, paint a church, plant a tree etc.

By way of background, the Rwandan genocide entailed the slaughter of over 800,000 Tutsis by the Hutu majority. Amusingly, many on the fringes of the left like to deny that the Tutsis were the victims of a genocide. John Pilger, for example, endorsed the denialist account by Herman and Peterson’s The Politics of Genocide (and Chomsky wrote the foreword to it, though is reluctant to endorse it). But anyway, for our purposes, its worth exploring the use of umugandas in the period prior to the Rwandan genocide. Here is how Uwimbabazi and Lawrence (2013) summarise it:

Umuganda in the post-colonial period is best understood in the context of the mythical peasant, with the ideology that only the Hutus were the real peasants of Rwanda’... This ideology also explained who was a true munya-Rwandan, which was in turn used against the Tutsi, who were not known as cultivators but aliens pastoralist... Hatred that led to divisions was increasingly planted under the stream of development during umuganda

Umuganda then turned into a means of promoting oppression and exclusion among Rwandans. This was done through colonial legacy of ethnic construction. For instance, in 1994 the idea of umuganda was used and served as a means of mass mobilisation during the genocide… Those in power argued then that only one particular group of people, the Hutus, had the right to exist, and other groups, Tutsis, were targeted for extermination in the name of umuganda

In essence, ‘learning from the peasant ideology… and the everyday propaganda during umuganda had also motivated people to see their fellow ba-Tutsi as enemies’ in the run up the genocide. When the genocide finally hit, umugandas were used more directly in the genocide:

During the genocide, umuganda did not involve planting trees but ‘clearing out the weeds’ – a phrase used by the genocidaires to mean the killing of Tutsis. Chopping up men was referred to as ‘bush clearing’ and slaughtering women and children as ‘pulling out the roots of the bad weeds’... The slogan, ‘clearing bushes and removing bad weeds’, were familiar terms used in the course of ordinary agricultural labour undertaken in umuganda.

Cowen’s Second Law really holds up because we have a study on this! Bonnier et al (2015), in a study that can be seen as semi-fuck you to Putnam, they investigate whether attendance at umugandas is related to violence being carried out. They do this by measuring attendance by the impact of rainfall: they ‘expect the meetings to be less enjoyable when it rains and furthermore to be cancelled altogether in the case of heavy rain.’ They look at the period 1990-1994 for umuganda attendance and then the civilian participation rate in genocide by reference to court records. They find that:

…results indicate a negative relationship between Umuganda intensity and civilian participation in genocide: one additional rainy Saturday is associated with a 5 percent decrease in the civilian participation rate…  If we assume a one-to-one relationship between the number of rainy Saturdays and the number of canceled Umuganda meetings, then an additional canceled meeting reduces the average civilian participation rate by 5.4 percent (interpreted at the mean number of civilian perpetrators per Hutu, which is 7.7 percent).

One more Saturday with rainfall above 10mm corresponds to a 0.41 percentage point reduction in the civilian participation rate. Those who wish to stop curtail certain forms of hate speech might very easily rely on studies like this. But there is an even better study which they can rely on in doing so: RTLM was the radio station in Rwanda and much like the umugandas: referring to Tutsis as cockroaches and dirty. The International Criminal Tribunal for Rwanda contains many extracts of what RTLM was putting out:

Prior to 6 April 1994: I have nothing against Tutsis, or Twas, or Hutus. I am a Hutu but I have nothing against Tutsis. But in this political situation I have to explain: “Beware, Tutsis want to take things from Hutus by force or tricks.”

After 6 April 1996: The Good Lord is really just, these evildoers, these terrorists, these people with suicidal tendencies will end up being exterminated. When I remember the number of corpses that I saw lying around in Nyamirambo yesterday alone; they had come to defend their Major who had just been killed.

Yanagizawa-Drott (2014) attempts to measure the impact of RTLM radio coverage on actual deaths in the genocide:

The estimates are significant at the five and one percent levels and imply that a one standard deviation increase in radio coverage increased participation in militia violence by 13–14 percent. RTLM broadcasts were also shown to have increased individual violence. The estimates (significant at the ten percent level) imply that a one standard deviation increase in the share of the village with radio reception increased individual violence by 10–11 percent.

Figure III [above] graphically illustrates results using a specification with dummy variables for various levels of RTLM reception. For militia violence… there is evidence of scale effects. For increases in radio coverage at low levels, the overall pattern indicates that there is no increase in participation, but once a critical level of coverage is reached there is a sharp increase in violence. By contrast, for individual violence there is no similar discernible pattern

These results are bad enough but Yanagizawa-Drott goes on to measure potential ‘spillover’ effects from this violence. The idea is that RTLM would have catalysed some who would then go on to engage more people so that more violence would be committed. In measuring this, he finds that ‘a one standard deviation increase in the share of the population in nearby villages with radio reception (0.18) increases participation in militia violence by 47.6 percent.’ The total effects are significant:

Focusing on the mean, the estimates imply that 9.9 percent (approximately 51,000 persons) of the total participation in genocidal violence was caused by the propaganda… The evidence also shows that spillovers were important, as only 7.7 percent of the militia violence is estimated to be due to direct effects. 22.3 percent of the militia violence can, therefore, be attributed to spillover effects.

The distinction between spillover effects and ‘direct’ effects might be key for some but I believe that would be a red herring: whether the speech comes through the radio (“direct”) or through an individual (“spillover”), in both cases, a link between speech and violence is being made. Other studies find much the same: Straus (2007), on the basis of interviews with 200 perpetrators, concludes:

….radio alone cannot account for either the onset of most genocidal violence or the participation of most perpetrators. That said, [he] find[s] some evidence of conditional media effects. Radio catalyzed a small number of individuals and incidents of violence [and these] hard-liners achieved dominance and were able to persuade individuals to join attacks against Tutsi civilians.

To repeat again, I am not advocating any particular policy prescription in this Part. But one particular prescription worth considering is raised by Samantha Power’s brilliant A Problem From Hell. In it she states

The county best equipped to prevent genocide broadcasting... was the United States. The United States could destroy the antenna... Pentagon planners understood that stopping the genocide required a military solution [though] it was clear that radio jamming would have been no panacea... In early May, the State Department Legal Adviser's Office issued a finding against radio hamming citing international broadcasting agreements and the American commitment to free speech. When Bushnell raised radio jamming yet again at a meeting, one Pentagon official chided her for naiveté: "Pru, radios don't kill people. People kill people!" (p.371-2).

This is a question that all free speech opponents and proponents must engage with: would you bomb or jam RTLM knowing what you know now about Yanagizawa-Drott (2014) and Straus (2007)? This is a far more interesting question rather than the replaying of the A Level question of “should offensive speech or hate speech be banned?”

3.7 Incitement, Obedience and Speech Act Theory: Eichmann to Jihadi Twitter

The studies above about Rawanda are not necessarily studies about incitement. Rather, umugandas can be said to be a cultural priming mechanism. By contrast, Austin’s (1957) How To Do Things With Words gives a good basis for discussing ‘incitement’ laws. His compartmentalisation of speech has been done over and over, so I will simply quote a recent exposition of it by Dawson (2015):

Austin proposed a tripartite classification of speech acts. To perform a locutionary act is simply to say something with meaning. To say it with a particular force is to perform an illocutionary act: to make a statement, make a promise or ask a question and so on. The perlocutionary effects are the further effects of the illocutionary act on an audience. For example, saying ‘I am walking the dog’ has the illocutionary force of making a statement and may have the perlocutionary effect, if uttered as part of a mobile phone conversation, of reassuring someone of your safety. Saying ‘I promise to walk the dog’ has the illocutionary force of committing you to walk the dog and may have the perlocutionary effect of causing someone else to make a cup of coffee and relax.

Hence, when we make contracts, we are not simply making locutionary statements of fact, we are doing something. This ties in with what is known as an ‘Austinian excertive’ which is ‘acts enact[s] permissibility facts and thereby determine what is permissible in a certain realm’ (McGowan (2004)). McGowan gives the example of a university chancellor stating that ‘no music will be played in dorms after 11pm.’ This ‘utterance has exercitive force because it takes away certain privileges’ and it can do this because the chancellor has authority.

There have been attempts to apply this line of argument to pornography. I believe this fails on empirical ground for reasons given above but also because, as McGowan has argued elsewhere, it simply does not meet the conditions of an excertive directive. This has been summarised wonderfully by John Danaher. I should note, Langton is making the arguments in premises 1 to 9 and McGowan is making the critical arguments in premises 10 to 12:

Maybe some producers of pornography do intend to silence and subordinate women, but many may not. Even if they did have that intention, the consumers of pornography would probably not recognise it. This will most often be caused by the fact that the semantic content of pornography will be exceptionally opaque — not an efficient means to convey the intended exercitive. Finally, and perhaps most fatally, producers and distributors of pornography do not have the authority to enact permissibility conditions concerning heterosexual behaviour or women’s speech acts.

Of course, McGowan believes porn to be a conversation excertive but I will not be assessing that argument. But, what if there can be excertive directives given in the context of other acts? Incitement laws seek to address precisely this issue. Again, the full extent of my objection to these kind of laws will not be well justified and apparent until Part 3. That said, it’s worth looking at some of the literature on the issue. Milgram’s studies are so engrained into popular culture that I wont even need to provide a citation: a figure of authority gives individuals an instruction to keep electrocuting someone and 65% go all the way to 450 volts. This might be great evidence for a harmful example of an excertive directive but Haslam and Reicher (2012) provide a significant and caveat:

…close analysis of the experimental sessions shows that participants are attentive to the demands madeon them by the Learner as well as the Experimenter [38]. They are torn between two voices confronting them with irreconcilable moral imperatives, and the fact that they have to choose between them is a source of considerable anguish… Ultimately, they tend to go along with the Experimenter if he justifies their actions in terms of the scientific benefits of the study

More importantly, ‘it was only when they had internalized roles and rules as aspects of a system with which they identified that participants used them as a guide to action.’ This not simply a blind acceptance, brainwashing, one-way control that authority has. It requires you to accept authority in the first place. Hence, their conclusion:

…the fundamental point is that tyranny does not flourish because perpetrators are helpless and ignorant of their actions. It flourishes because they actively identify with those who promote vicious acts as virtuous.

This position, which I share, inevitably comes up against Hannah Arendt’s view of Eichmann (i.e., about the ‘banality of evil’). However, as a New York Times review of a series of books notes, her line of argument does not seem withstand recent historiography:

Ms. Stangneth’s book cites that document and a mountain of others to offer what some scholars say is the most definitive case yet that Eichmann, who was hanged in 1962, wasn’t the order-following functionary he claimed to be at his trial, but a fanatically dedicated National Socialist.

If previous researchers have seriously dented Arendt’s case, Ms. Stangneth “shatters” it, said Deborah E. Lipstadt, a historian at Emory University and the author of a 2011 book about the Eichmann trial.

The facts about Eichmann in Argentina have been dribbling out, “but she really puts flesh on the bones,” Dr. Lipstadt said. “This was not a guy who just happened to do a dirty job, but someone who played a crucial role and did it with wholehearted commitment.”

This should also all tie in to recent discussions of people going to Syria and Iraq to join ISIS. These people are not brainwashed, it is not speech nor directive that leads them to joining these groups. It’s accepting, after active engagement with ideas and social associations, the false ideas of ISIS. The converse view has rightly be derided as the ‘Zoolander Theory of Terrorism’:

…what scholarly research on terrorism overwhelmingly shows is that terrorists, in the main, are not only not crazy, but also not stupid. Furthermore, it shows that people who join terrorist organizations tend to do so because they believe they are defending what they see as a just cause. They join because they want to, and because they think that what they are doing is right and necessary.

I myself have written elsewhere about the idea that terrorists are not prone to mental illness but have some acceptance of the specific ideology. Many examples of the internet seemingly having a role in radicalisation seem to collapse once looked at. [5] The first study to note is Benson (2014). Benson looks comprehensively at the effect of the internet on transnational terrorism. He finds that ‘the areas where there are increases in terrorism are exactly the opposite locations expected if the Internet were actually driving the increase; rather being in Internet-wealthy areas, the increases occur in areas where the Internet is almost non-existent.’ He concludes ‘enough data has been collected to provide evidence that transnational terrorists have not reaped a windfall in capabilities because of the Internet.’

Another study is von Behr et al (2013) which undertakes a literature overview and then applies it to through an idiosyncratic approach to 15 extremist individuals from the UK. Their headline findings include: ‘evidence does not necessarily support the suggestion that the internet accelerates radicalisation’; ‘evidence does not support the claim that the internet is replacing the need for individuals to meet in person during their radicalisation process’. So when people want to cut down on extremist material online, I think they’re missing the point.

3.8 Epistemic Humility

Why have taken so much time writing out potential negatives of free speech? There are two reasons. First, people need to have a sense of humility when talking about this issue. Yes, free speech is good (see Part 1) but not its not a clean, cut good that should be spoken about without regard to the evidence. Even when talking about things like pornography, video games etc., there is a lot of research – and it doesn’t always support what you would’ve necessarily thought. I have given my reasons above for siding with positions which are more amenable to free speech but I have been unable to do this consistently in this Part.

Scarily, Ben wrote a post on the day I finally got to this section which had the point I’m trying to make. He noted a new IMF study about inequality which contradicted his own view. This is what he said:

I can say that the methodology of the IMF work is shoddy and that you should look at credible meta-analyses—but why should Elliott et al. [the author of the Guardian write up of the study] trust me when I’m going against far more credible people. And surely we cannot expect everyone to dissect the details of every paper when they make a judgement on the area… So while I might criticise the rather silly framing of this new ‘revelation’ I can hardly give full-throated criticism of the Guardian economics section line in general—interpreting evidence is just too hard.

I disagree with only one part of this paragraph: we should expect everyone to dissect details of every paper. But aside from that, Ben is right: these issues are hard. Yes, free speech is important but its beyond arrogance to think that it’s a simple case of talking about ‘Enlightenment values’ and talking about free speech per se. I say this despite, in all likelihood, probably supporting less curtailment of free speech than even the most ardent of free speech advocates as will be seen in Part 3.

Talking about Charlie Hebdo, ‘offense’ and ‘blasphemy laws’ is too easy. In a debate with terrorists or fringe left people, that might be valuable. I can understand the desire to respond to the constant attacks on free speech with people like Will Self. But in a debate where people broadly agree that free speech is good, it means you’re not really testing your position if you ignore these complex issues. There is so much room for legitimate nuance. Talk about RTLM, the statistical differences in the studies on video game usage, umugandas and Becker’s model of discrimination.  

The second reason is to draw attention to a theme should have arisen from the above sections: in relation to race and TV in arose in relation to the idea that likeable ethnic characters only had positive impacts where that individual thought the character was representative; in relation to incitement it arises because extremists fish out views on the internet they already agree with – and so on. What does this tell us? Here are a few excerpts from the studies above which point to the right direction. Ferguson (2014):

Current “hypodermic needle” theories of mass media effects on behavior ultimately may imply simplistic modeling of behavior, focused too heavily on the development of automatic cognitive scripts (Ferguson & Dyck, 2012). Such theoretical models may, effectively, remove the user from the media experience except as a passive “victim” of a powerful, influential media…

[Different models, however posit] posit media as fulfilling pre-existing motivational structures. Thus, a particular form of media may have very different influences depending more on what individual consumers seek to achieve rather than on content specifically… User motivations determine what users watch and what influences they hope to experience from media.

Hald et al looks at the mediating role of sexual arousal in the effect of pornography. The idea is that sexual arousal --> engagement of existing sexist cognitions, attitudes and beliefs. Once you take into account sexual arousal, the effect of pornography on sexist attitudes becomes insignificant:

.... among low in agreeableness participants (i.e., the lower 33% in the distribution of agreeableness), the initial correlation between our initial variable (experimental exposure) and the outcome variable (hostile sexism) was .29 (p = .014; 95% CI: 1.59; 10.72), indicating a significant direct effect of experimental exposure to pornography on hostile sexism among low in agreeableness participants. When including sexual arousal as a mediator, the correlation between experimental exposure and hostile sexism, that is, the direct effect, dropped to –.10 and turned nonsignificant (p = .56).

…affective activation (herein sexual arousal) may serve as an important mediator of significant exposure-attitudinal relationships and be central to the priming of ‘‘associative networks’’ of emotions, cognitions, and attitudes which in content or feeling tone correspond to the attitudes investigated (i.e., herein sexist attitudes)

...among some individuals, sexist cognitions and attitudes may partly have been learned through specific environments, scenarios, and role models on the basis of reinforcement and vicarious learning. These sexist cognitions and attitudes may, in sexual situations, be activated by sexual scripts if these are attuned to content or feeling tones sexist in nature.

Pornography is of course significantly correlated with sexual arousal (.85) which is then significantly correlated with hostile sexism (.46). Therefore, the extent that the literature does find a link between porn and sexism, it may be finding something about us rather than porn. I think the view that us being sexist leads to uniformly sexist pornography is somewhat overstated, see a study that looks at the content of porn at [6] but the broader point that appears supported by the above literature, is that its not content, but the content viewer where the problem resides.

I apologise for the constant references to Part 3 in this Part but I hope how my views all tie in, even with a model of affective engagement, will be apparent there. To re-emphasise: there really are no policy prescriptions here. To give policy prescriptions, we must undertake an assessment of the state – and the market – along with other normative positions and that brings us to the final part of this series.

Endnotes

[1] Gentzkow (2014) uses data from 1869 to 1928 to look at the effect of a party being in power over media output. What they find is that increases in Democratic representation actually decreases the Democrat share of circulation (though this is insignificant). The point is that the state does not drive media slant. Our newspapers aren’t serving the elites in the state.

Gentzkow, however, does find that the state does have an effect on media output in certain situations. Gentzkow uses ‘an index of commercial and political incentives to measure the effect of control of the state government in places with both relatively strong political incentives and relatively weak commercial incentives, and find no evidence of an effect of control of the state government in such places.’ Put simply, Gentzkow tries to measure situations where there is government capture. Even then, Gentzjow fails to find an effect on media output.

Gentzkow tests for when market incentives (i.e., losing buyers to other newspapers because you are seen to be too partistan ) and finds this doesn’t really change things. Market incentives work best where there is high demand for newspapers and a competitive market with many newspapers. In addition when political incentives are high, there is still no effect. However, there are situations where low market incentives and high political incentives, working together, can make a difference. This is precisely what happened in the Reconstruction period:

The chaotic environment of the post–Civil War era provided unusually strong political incentives for Republican governments to support the entry and growth of Republican newspapers... Local advertising markets were limited. Roughly two-thirds of incumbent papers existing before the war had ceased publication by its end.

Market forces that might otherwise have restrained the desire to control the press for political aims were especially weak in the postbellum South, where newspapers faced low demand, high costs, and limited market competition… the evidence from the Reconstruction and post-Reconstruction South shows that state governments did sometimes exert meaningful influence on the press, and suggests that such influence was at its height when political stakes were greatest and market forces weakest

Before Chomskyites get too happy, Gentzkow applies this test to the 1932-2004 period and find ‘find no correlation between the party of the incumbent governor or congressional representative and the political slant of newspapers in 2005’ – i.e., the idea that governments, in the West, can control media output seems wrong.

[2] Justice Burger’s Opinion makes clear the reasons for thinking there is no distinction between political contributions and expenditure. He notes that

The Court dismisses the effect of the limitations on the second aspect of contributions: "[T]he transformation of contributions into political debate involves speech by someone other than the contributor." On this premise -- that contribution limitations restrict only the speech of "someone other than the contributor" -- rests the Court's justification for treating contributions differently from expenditures. The premise is demonstrably flawed; the contribution limitations will, in specific instances, limit exactly the same political activity that the expenditure ceilings limit, [n7] and at least one of the "expenditure" limitations the Court finds objectionable operates precisely like the "contribution" limitations [n8].

The Court's attempt to distinguish the communication inherent in political contributions from the speech aspects of political expenditures simply "will not wash." We do little but engage in word games unless we recognize that people -- candidates and contributors -- spend money on political activity because they wish to communicate ideas, and their constitutional interest in doing so is precisely the same whether they or someone else utters the words.

[n7] Suppose, for example, that a candidate's committee authorizes a celebrity or elder statesman to make a radio or television address on the candidate's behalf, for which the speaker himself plans to pay. As the Court recognizes, ante at 24 n. 25, the Act defines this activity as a contribution and subjects it to the $1,000 limit on individual contributions and the $5,000 limit on contributions by political committees -- effectively preventing the speech over any substantial radio or television station. Whether the speech is considered an impermissible "contribution" or an allowable "expenditure" turns not on whether speech by "someone other than the contributor" is involved, but on whether the speech is "authorized" or not…

[n8] The Court treats the Act's provisions limiting a candidate's spending from his personal resources as expenditure limits, as indeed the Act characterizes them, and holds them unconstitutional. As MR. JUSTICE MARSHALL points out, post at 287, by the Court’s logic, these provisions could as easily be treated as limits on contributions, since they limit what the candidate can give to his own campaign.

Professor Powe makes the point that the distinction between contributions and expenditures on the basis that the former restrict someone else’s speech is flawed because even expenditures, practically, do not involve the speech of the donator. As he notes, ‘In some circumstances an individual gives to a committee which in turn gives to professional or to a campaign treasury.. An individual choice to have a message with which he agrees prepared by professionals is no less speech. Proxy speech is simply a pejorative name for a political commercial. It is still speech’ (quoted in BeVier (1985)). BeVier also makes the following point which is quite apt:

Contributions by individuals to groups or to political committees, however, permit the pooling of resources. This amplifies the contributors' individual voices... "[e]ffective advocacy of both public and private points of view, particularly controversial ones, is undeniably enhanced by group association." [Accordingly,] Limits on political contributions obviously constrain the efforts of groups of individuals to increase the effectiveness of their advocacy.

I agree with Justice Burger that ‘freedom of association and freedom of expression were two peas from the same pod’ and so I think BeVier understates the power of her argument.

[3] It is put in different terms in the Brief of the Appellant:  

… those who contribute money to a candidate hope to further their political, social and economic views-just as those who engage in demonstrations and rallies, write books on political issues, organize like-minded citizens, draft and circulate petitions, buy advertisements, publish periodicals or newspapers, make speeches, sponsor seminars or teach-ins, or engage in litigation hope to further their political, social and economic views. Such activities are the hallmark-as well as the indispensable precondition-of a free society.

[4] Interestingly, Koplow goes through five cases studies to assess whether public opinion support is primarily based on ideological affinity or strategic considerations. For example, after the 1973 when there was an oil embargo, support for Israel was still high. Conversely, despite having little strategic impact, public opinion fell by 8-points during Cast Lead. Koplow’s conclusion from this is that the American public support Israel primarily on ideological grounds.

[5] As Gwern notes in his wonderful “Terrorism is not about Terror”:

Earlier ‘lone wolves’ like bombers Timothy McVeigh or Eric Robert Rudolph turn out on closer inspection to have ties, social & otherwise, to like-minded people; McVeigh lived with several other extremists and was taught his bomb-making skills by the Nichols, who also built the final bomb with him, while Rudolph remained on the run for several years in a community that wrote songs and sold t-shirts to praise him and was ultimately caught clean-shaven & wearing new sneakers. Lone wolves who genuinely had no contact with their confreres, such as Ted Kaczynski, are vanishingly rare exceptions among the dozens of thousands of terrorist attacks in the 20th century, and as rare exceptions, otherwise implausible explanations like mental disease account for them without trouble.

As with porn and video games, it’s not surprising that terrorists use the internet: these things are all ubiquitous. For this reason, studies that look into the video game/porn use of criminals are largely irrelevant without a comparison of how much the general population uses it; and more specifically, what proportion of those who use them end up committing violence.

[6] As mentioned, it is a fair argument to say that if the problem is with us then it should also come about in what we produce. Accordingly, it’s worth noting the literature on the prevalence of sexism in porn. This can be measured by looking at objectification, power relations and violence. Objectification, for our purposes can be subdivided into instrumentality and dehumanisation. The former entails the idea that a person’s use is defined by how much they gratify someone else. The latter entails removing agency from an individual. Klassen and Peter (2014) try to measure for these. I think the table is easy enough to understand so only include their conclusion alongside it:

...regarding objectification in mainstream pornographic Internet videos, women were more likely to be instrumentalized than men, as indicated by a strong focus on women’s sexual body parts as well as on sex acts and orgasms in which men rather than women gained sexual pleasure. However, there was no evidence of a general dehumanization of women. Notably, men were more likely to be dehumanized than women in that men’s faces were rarely shown.

In terms of power, Klaasen and Peter find that ‘the distribution of power between men and women was about equal in the hierarchal positions, referring to social and professional roles, in which they were depicted’ but ‘power differences became prevalent in the context of sexual activities as men were more likely depicted as dominant and women as submissive.’ The following results are also worth noting:

... men and women were equally likely depicted as not initially wanting to engage in sexual activity. Similarly, both men and women were almost never depicted as intoxicated while engaging in sex. Manipulation into sex was rare as well, but when it happened women were more likely to be manipulated into having sex than men

The mechanisms of direct social learning (i.e., normalising not obtaining consent) other than by manipulation and indirect social learning (objectifying women so as to remove agency from them) will need to be rethought for those who wish to argue that there is link between pornography and sexism.