Reprinted from 99 Columbia L. Rev. (1999). Copyright Jack M. Balkin 1999. All rights reserved.
Free Speech and Hostile Environments
One major concern about sexual harassment law is that employers will restrict employee speech in order to avoid hostile environment liability, thus violating free speech principles. In this Essay, Professor Balkin argues that this “collateral censorship” is constitutionally permissible when there are good grounds for vicarious liability. Because employers actively control workplace culture, and because they are better able to prevent hostile environments than individual employees, vicarious liability for employee speech is more justified than in the case of distributors or common carriers.
Professor Balkin also argues that captive audience doctrine, generally thought to apply only to speech in the home, is actually better suited to workplace speech. Hostile environments are a method of sex discrimination that maintains gender segregation; a hostile environment does its work precisely through making the employee a captive audience.
The Essay concludes that First Amendment challenges to sexual harassment law should not become a defense of employer prerogatives presented in the guise of worker liberties. Without the incentives created by sexual harassment law, employees will not be freed from censorship; they will simply be remitted to the economic and social control of employers.
Does sexual harassment law conflict with the First Amendment? A number of commentators now argue that it does.2 Generally, these objections focus on employer liability for speech by employers and employees that creates hostile environments. Virtually no one finds fault with regulating quid pro quo sexual harassment: Employers who tell employees “sleep with me or you’re fired” make threats that are not protected by the First Amendment.3 In this Essay I offer an account of why hostile environment doctrine, properly interpreted, does not violate freedom of speech.4 I address two basic kinds of objections. The first concedes that the creation of a hostile environment is not protected speech, but argues that fear of Title VII liability will lead employers to censor employee speech and that this chilling effect violates the First Amendment.5 I hope to show why these concerns are overstated through a discussion of the idea of collateral censorship—an important but largely overlooked concept in the theory of free speech. Collateral censorship occurs when A censors B out of fear that the government will hold A liable for the effects of B’s speech. I believe that the question of when collateral censorship is a permissible effect of regulation and when it is unconstitutional will be increasingly important in the law of telecommunications and cyberspace. In this case, however, the concept of collateral censorship helps us understand the constitutionality of hostile environment law.
The second, and more radical attack on hostile environment liability argues that when employers create a hostile environment that materially alters working conditions to the disadvantage of women and minorities, they are engaged in protected speech.6 Rebutting this more radical attack helps us understand another important but undertheorized category in First Amendment law—captive audience doctrine. It also helps connect the captive audience doctrine more closely to what I regard to be the deeper purposes of antidiscrimination law.
I. Collateral Censorship
Although threats are not protected by the First Amendment, hostile environments do not always involve threats. A hostile environment is made up of individual acts of discriminatory speech and other conduct by all the persons who inhabit a workplace, including managers, employees, and even occasionally clients and customers. In hostile environments, “the workplace is permeated with ‘discriminatory intimidation, ridicule, and insult’ that is ‘sufficiently severe or pervasive to alter the conditions of the victim’s employment and create an abusive working environment . . . .’”7 Some of this behavior may be directed at particular employees; other elements may be directed at no one in particular but may help foster an abusive environment. Even if individual acts do not constitute a hostile environment separately, they can be actionable when taken together. The test is whether the conduct, taken as a whole, would lead to an environment that the employee reasonably perceives as abusive.8
Employers can be liable for maintaining a hostile work environment even if management did not personally engage in any of the predicate acts. In Burlington Industries, Inc. v. Ellerth9 and Faragher v. City of Boca Raton,10 the Supreme Court held that employers are liable for harassment by supervisory personnel, subject to a number of affirmative defenses where the harassment did not result in a tangible employment action like firing or demotion.11 The degree of vicarious liability for non-supervisory personnel (such as co-workers) is still contested, but currently most courts hold an employer liable if the employer knew or should have known of the harassment and did not take prompt corrective action.12
Employers who want to minimize hostile environment liability cannot merely prohibit individual instances of harassing conduct. They must also limit conduct that might, in combination with other conduct, contribute to a hostile environment. Hence employers are tempted to create prophylactic rules against all the potential components of a hostile environment. Some of these will be unwelcome physical advances, assaults, and forms of abuse. Others will be largely verbal: sexual jokes and innuendoes, taunts and threats, sexually oriented cartoons, pictures, and pornography. Some of this verbal abuse may even be couched in political or factual terms; for example, co-workers might oppose affirmative action programs for women or quote scientific studies arguing that women are less competent at certain jobs. Because employers have no general interest in preserving employee speech rights unrelated to efficiency, they will impose regulations as broad as they think necessary to insulate themselves from liability.13 The most important complaints about the constitutionality of sexual harassment law stem from these incentives to censor employee speech that might contribute to a hostile environment.14
The argument that sexual harassment law produces unconstitutional employee censorship is actually an instance of a more general problem in free speech law—a phenomenon I call collateral censorship.15 Collateral censorship occurs when one private party A has the power to control speech by another private party B, the government threatens to hold A liable based on what B says, and A then censors B’s speech to avoid liability. The offending speech may be defamatory, obscene, fraudulent, or a violation of copyright. In most situations A has greater incentives to censor B than B has to self-censor. That is because B has an additional interest in promoting his or her own speech that A usually lacks. Hence A can be expected to censor B collaterally with little regard for the value of B’s speech to B or to society at large.
Although to my knowledge no court has yet recognized collateral censorship as a distinct doctrinal category, it appears to be a fairly common phenomenon. For example, editors and publishers, driven by fear of defamation suits, may refuse to run stories by their reporters. Internet service providers, fearing that they may be held liable for contributory infringement of copyrighted materials, may attempt to ban messages from parties suspected of disseminating such materials.
Once one recognizes the ubiquity of the phenomenon, it should be obvious that not all government regulation that leads to collateral censorship is unconstitutional. The question is which varieties are permissible and which are not. It is tempting but incorrect to argue that collateral censorship is never unconstitutional because there is no state action. It is not the state that is censoring the employee or the customer but a private party—the newspaper or the Internet service provider. In fact, there is state action in every case of collateral censorship, because the government has created incentives for private parties to censor each other. Moreover, in virtually all cases of collateral censorship, the government not only knows but expects and desires that the private party will censor the unprotected offending material, even if it does not know exactly how much constitutionally protected material will also be censored in the process.16
Even so, collateral censorship is not necessarily unconstitutional. For example, the editors and the publishers of the New York Times are liable for publishing defamatory articles written by their reporters. Hence editors and publishers of large metropolitan dailies like the New York Times exercise considerable oversight over their reporters’ stories in order to avoid liability or the costs of defending a defamation lawsuit. Often acting on the advice of attorneys, editors and publishers severely edit or even discard much of what their reporters produce, even when a reporter insists that a story is accurate. The reporter’s statements may be matters of intense public concern and core political speech. They may be constitutionally protected so that the government could not impose liability for them directly. Nevertheless, reporters who insist on writing what the editor or publisher forbids out of fear of possible liability may be disciplined or even fired.
Do such limitations on employee speech violate the First Amendment? First Amendment law clearly does recognize constitutional limitations on liability for defamation, embodied in cases like New York Times Co. v. Sullivan,17 Gertz v. Robert Welch, Inc.,18 and their progeny. One reason for these limitations is the fear that valuable speech will be chilled by liability for defamation. But the doctrines do not appear to make any distinction between chilling the speech of editors and publishers and chilling the speech of their employees, the reporters. Both reporters and editors are entitled to the same constitutional privilege. Indeed, in Cantrell v. Forest City Publishing Co., the Court approved of a jury charge which permitted the imposition of vicarious liability upon a publisher for the knowing falsehoods written by its staff writer.19 Here, in effect, the Court allowed the jury to hold the publisher strictly liable for an employee’s defamation. Such a rule clearly gives a publisher strong incentives to censor employee speech. Yet the Court found no constitutional problem with applying the traditional doctrines of respondeat superior in this context.20
Federal securities laws require investment houses, brokerage firms, investment advisors, and even corporate officials to avoid making misleading statements about company profits, securities, and related investments. Companies are strictly regulated concerning what they may say about these matters, particularly in highly regulated procedures like proxy contests.21 Statements made by their employees, even politically motivated statements, may subject them to liability. Thus, rational companies will often severely limit the kinds of public statements their employees may make, and discipline or terminate employees who disobey. Do such rules violate the First Amendment because they chill the speech, not of the organizations themselves, but of their employees?
My sense is that neither of these situations presents a serious First Amendment problem, even though both involve collateral censorship. But this simply raises the question of when collateral censorship is constitutional, and when it is constitutionally troublesome.
The question we should ask is whether it makes sense, given the purposes of a regulatory regime, and the kind of harm that the legislature has a right to prevent, to treat the private censor and the private speaker as the “same speaker” for purposes of First Amendment law. Clearly this judgment is a legal fiction. The two speakers are not really identical, but the law is entitled to treat them as if they were one, and to hold the first liable for what the second does. The claim that the censor and speaker are the “same speaker” is just a shorthand way of saying that the private censor (the employer) has the right to control the content of the speaker’s (the employee’s) speech and that the private censor is properly responsible for the harmful effects of that speech.
Why might it be permissible to hold one speaker liable for the harms of another? One reason is that the private censor and speaker are part of the same enterprise that produces speech-related harm: They either collectively produce a single product that causes harm (a libelous publication), or their collective efforts create a harm or a risk of harm (misleading or fraudulent information about investments). A second reason is that the private censor is in the best position to avoid the harm. For example, we might think that the private censor is particularly good at distinguishing protected from unprotected harmful speech, that the private censor can avoid harms more easily and effectively than the speaker, or that the private censor has better information than the speaker.
Thus, we can identify three considerations that justify treating the private censor and the speaker as “the same speaker”: (1) the private censor’s right to control the speech of the private speaker, (2) the joint or collective production of a harm or danger of harm, and (3) the private censor’s superior ability to avoid the harm. Not surprisingly, these reasons for treating the private censor and the speaker as “the same speaker” for purposes of First Amendment law resemble traditional justifications for vicarious liability, in which courts treat employer and employee as the “same tortfeasor” for purposes of liability.22 (They also suggest why the Court found little difficulty with the application of respondeat superior in Cantrell, even though the use of that rule in the defamation context clearly chills the speech of reporters.)
It is easy to see why the justifications for vicarious liability are relevant to the constitutionality of collateral censorship. If we hold the private censor responsible for the private speaker’s harmful speech, it is reasonable to expect the private censor to censor. Conversely, if we don’t want to encourage the private censor to censor (because we value the free flow of ideas), we should ensure that the private censor is not held responsible for the private speaker’s harmful speech. Thus, collateral censorship is most acceptable from a First Amendment standpoint when vicarious liability is most acceptable, and it is least acceptable from a First Amendment standpoint when vicarious liability is least acceptable.
Moreover, these three considerations—the private censor’s right to edit and control content, the joint production of the harm by censor and speaker, and the private censor’s superior ability to avoid the harm—help explain why collateral censorship does not seem to violate the First Amendment even when the censor and speaker are not employer and employee, or part of the same business enterprise. Take the case of authors and publishing houses, for example. Book publishers employ legal staffs to inspect author manuscripts for possible liability for defamation, fraud, or copyright infringement. Publishers do so because they will suffer the consequences of their authors’ violations of the law. As a result, publishers often demand that authors rewrite or even omit troublesome passages as a condition of publication. No one doubts that these practices affect authors’ practical ability to speak. But this collateral censorship does not violate their First Amendment rights, even when they engage in explicitly political speech. If defamation laws are constitutional with respect to suits against the author directly, they are also constitutional with respect to suits against the publisher, even though the publisher clearly has different incentives from those of the author and therefore will exercise collateral censorship in situations where the author would not self-censor.
Thus, the case of the author and publisher is much like that of the reporter and the editor. Like the newspaper editor, the book publisher possesses and exercises the right to editorial control over how authors express themselves. This editorial control over content is the price that reporters or authors must pay if they want to publish in a particular newspaper or in a book produced by a particular publisher.
Conversely, collateral censorship seems to pose the greatest constitutional problems when it is most troublesome to treat the censor and the speaker as the “same speaker” for First Amendment purposes. The most obvious example occurs when courts and legislatures impose liability for harmful speech on a distributor, a common carrier, or some other conduit that is not part of the same business enterprise as the censored speaker, lacks the right to exercise editorial control, and lacks information about the nature of the content flowing through its channels.
In fact, the one Supreme Court case that comes closest to recognizing the problem of collateral censorship seems premised on this distinction. In Smith v. California,23 a California statute made it a crime for bookstore owners to stock books that were later judicially determined to be obscene, even if the owner did not know of the books’ contents. The Supreme Court struck down the statute, arguing that “if the bookseller is criminally liable without knowledge of the contents . . . he will tend to restrict the books he sells to those he has inspected; and thus the State will have imposed a restriction upon the distribution of constitutionally protected as well as obscene literature.”24 Hence, “[t]he bookseller’s self-censorship, compelled by the State, would be a censorship affecting the whole public, hardly less virulent for being privately administered.”25 What the Court calls “self-censorship” in Smith is actually collateral censorship that arises from the different incentives of the bookseller and the book author.26 In Smith the Court saw through (or, more correctly, did not even notice) the state action objection that a private party was doing the censoring.27
Note that the Supreme Court suggested that it was unfair to hold the bookstore owner liable because the bookstore owner lacked information about the content of each and every book. This injustice is not merely unfairness to the individual bookseller. Nor is it purely a concern about inefficient sorting. If our only concern were keeping harmful speech out of bookstores, we could accept a blunderbuss approach. But in the First Amendment area we should be as concerned with false positives (non-obscene books that don’t get stocked) as with false negatives (obscene books that wind up on the bookstore shelves). We should be concerned about closing off means of expression to the authors on the one hand, and closing off information to audiences on the other. That is why it is a bad idea, from a First Amendment perspective, to squeeze the distributor (such as a bookstore owner or a common carrier) in the middle. Concerns about mismatched incentives and inadequate information lie at the heart of the constitutional objection to collateral censorship.
The common law of defamation features similar policy concerns about collateral censorship, although the courts do not use the term and the relevant doctrines are not constitutionalized. The law of defamation recognizes the problem of collateral censorship through what is called the distributor’s privilege. Generally speaking, a person who repeats a defamatory statement is as liable for publication as the original speaker (assuming the person also acts with the requisite degree of fault).28 However, a distributor of information, such as a newsstand or a bookstore, is generally not held to this standard unless the distributor knows of the publication’s defamatory content.29 The fear is that if distributors were held to be publishers, distributors might restrict the kinds of books and magazines they sold, greatly reducing the public’s access to protected expression.
To receive the common law privilege, a distributor does not have to be a common carrier, which must take on all customers without oversight.30 Although distributors make some content-based judgments—for example, in choosing what books or magazines to stock—their editorial control is very different from and much more limited than that of the book publisher or magazine editor.
In the telecommunications industry, collateral censorship poses a genuine and recurrent constitutional problem: Cable companies and Internet service providers regularly act as conduits for the speech of unrelated parties. Treating them like publishers or editors would have the predictable effects noted above.31 Thus, in the Telecommunications Act of 1996, Congress extended a special privilege to Internet service providers whose customers post indecent, obscene, or “otherwise objectionable” matter in cyberspace, declaring that, as a matter of law, they should not be considered the publishers of such material.32
Hostile environment law surely produces collateral censorship. But the collateral censorship does not involve a distributor or conduit relationship. I believe that the employer who censors employees for fear of creating a hostile environment is more like the employer who censors employees out of fear of liability for defamation or securities fraud, and less like the Internet service provider who censors his or her customers, or the bookstore owner who refuses to stock certain books in order to avoid liability.
Unlike the case of the bookseller and the author, the employer and employee in a hostile environment case are part of the same business enterprise, and the employer has the contractual right to control the employee’s speech and conduct. Perhaps more important, the law has good reasons to hold the employer accountable for the acts of its employees. First, the employer is better able to see the larger picture about what conduct might contribute to sex discrimination. This is especially so given that hostile environments emerge from a combination of behaviors that not all employees may have knowledge of. Second, and for similar reasons, the employer is also better able than individual employees to prevent hostile environments from emerging, especially when they result from collective actions that no individual employee may be able or willing to prevent.
Third, precisely because the creation of a hostile environment does not stem from any single act, but from many acts taken together, it makes sense to treat the harm to equal opportunity in the workplace as a single harm. Indeed, because collective action problems may reduce the ability or the incentives of individual employees to prevent hostile environments, it may make even more sense to treat employer and employees as a “single speaker” engaged in a single harm than it does in the case of defamation or securities fraud.33
Fourth, the employer faces incentives to acquiesce in hostile environments that have no analogue in defamation or securities fraud situations. This makes the case for employer liability—and the need for incentives to police employees—even stronger. Sexual harassment is a form of sex discrimination that works to keep jobs and employment opportunities sex-segregated according to traditional gender roles—for example, by keeping women out of higher-paying construction positions and in lower-paying secretarial positions, or by imposing obstacles to advancement for women even in integrated workforces.34 Employers may accept (or ignore) sex discrimination by their male employees (including discrimination through sexual harassment) in order to avoid labor disruption and preserve esprit de corps and loyalty among a particular class of valuable (male) workers.