concurring in part and dissenting in part.
I join parts I and II of Judge Easter-brook’s opinion. I welcome the acknowledgment that the parties to a collective bargaining agreement may not, whether deliberately or inadvertently, contract their way out of Title VII’s provisions. I *538welcome as well the court’s recognition that the assessment of an employer’s liability for sexual harassment must focus on whether the company took such measures as were reasonably necessary to stop the harassment rather than upon what steps an arbitrator might or might not have endorsed.
I also agree, at least in the abstract, that evidence of an employer’s concerns regarding the collective bargaining agreement or the likely results of arbitration pursuant to that agreement may be admissible to show that the employer’s failure to halt the sexual harassment occurring in the workplace does not merit the imposition of punitive damages. Whereas the liability inquiry focuses on the objective reasonableness of what the employer did or did not do in the face of harassment, the punitive damages inquiry calls upon the factfinder to assess the mindset of the employer. Thus, although neither the provision of the CBA nor the prospect of unfavorable rulings in arbitration can excuse an employer’s failure to take the measures reasonably necessary to stop harassment from recurring, an employer’s concern for the requirements of a CBA and the prospect of an unfavorable outcome in arbitration may help to establish that the employer’s actions and omissions, even if misguided, were not animated by malice or reckless disregard for the rights of its employees. A rule categorically excluding such evidence from the punitive damages inquiry would therefore be inappropriate.
However, before we vacate the award of punitive damages in this case and remand for a new trial, it is incumbent upon us to consider whether Ameritech was genuinely prejudiced by the exclusion of its evidence regarding the CBA and prospective arbitration over the discharge of Amos. Fed. R.Evid. 103(a)(1), Fed.R.CivP. 61; see Mason v. Southern Illinois Univ., 233 F.3d 1036, 1042-43, 1047 (7th Cir.2000); Jones v. Lincoln Elec. Co., 188 F.3d 709, 724-25 (7th Cir.1999), cert. denied, 529 U.S. 1067, 120 S.Ct. 1673, 146 L.Ed.2d 482 (2000); DeBiasio v. Illinois Central R.R., 52 F.3d 678, 685 (7th Cir.1995), cert. denied, 516 U.S. 1157, 116 S.Ct. 1040, 134 L.Ed.2d 188 (1996). Unless there is a significant chance that the evidence might have affected the jury’s verdict, then any error in its exclusion was harmless. Id.
For a number of reasons, I find it wholly implausible to suggest that the jury might have reached a different decision as to punitive damages had the excluded evidence been admitted. First, Ameritech proffered the evidence in order to explain why it did not fire Amos on two specific occasions. Yet, the EEOC’s case for punitive damages was not founded exclusively or even primarily on the company’s failure to fire Amos on these occasions; it rested instead on compelling proof of a long-term pattern of half-hearted and ineffectual responses to Amos’ unrelenting harassment. Second, close inspection makes clear that the probative worth of the excluded evidence is diminutive as compared with the prejudice posed by a protracted inquiry into matters that ultimately are collateral to Ameritech’s culpability. See Fed. R.Evid. 403. It strikes me as probable that on remand, the district court will exclude the evidence on that basis. Third, as Judge Easterbrook has demonstrated so persuasively, any difficulty that the CBA and arbitration posed to discharging Amos was of the company’s own making. And to the extent that evidence concerning the CBA and the arbitration process might explain why the company was hesitant to fire Amos, it does not explain the company’s failure to take other measures that might have addressed the harassment. Finally, a key premise underlying Ameri-tech’s theory — that it was better for the women of Ameritech not to have dis*539charged Amos in December 1992 or June 1993, because it left no obstacle to discharging him later (after he had harassed more of his co-workers) — effectively turns Title VII on its head.
1.
One point that must be made at the outset is that Ameritech’s CBA/arbitration evidence supplies at best an incomplete answer to why the company tolerated Amos’ harassment for such a long period of time. Ameritech offered the evidence in order to explain the company’s decision not to fire Amos on two particular occasions — in December of 1992, following the Wentland incident, and again in June 1993, following the Wolter incident. See Ameri-tech’s En Banc Br. at 17-18, 21, 31-32, 34-35. But Amos, of course, harassed a number of women in addition to Wentland and Wolter, and the company had opportunities to address his harassment on other occasions and through means other than terminating him. It is not accurate, therefore, to say that the exclusion of the CBA/arbitration evidence as to these two incidents may have “disabled]” Ameritech from presenting its best answer as to why the harassment persisted. See ante at 528.
In fact, as I have pointed out before, there was no need at all to further explain the company’s response to the first of these incidents. E.E.O.C. v. Indiana Bell Tel. Co., 214 F.3d 813, 829-30 (7th Cir. 2000) (Rovner, J., concurring in part and dissenting in part). The jury was told that because the company did not promptly suspend Amos once Wentland reported that Amos had touched her, the CBA gave Ameritech only 30 days in which to decide whether and how to discipline Amos for having touched her stomach. No witness expressed the belief, and the EEOC’s attorneys never argued, that the company could have fired Amos once that deadline expired. As far as the jury knew, the time limit was written in stone, so that once the deadline passed without action, the company’s hands were tied. One can hardly say that Ameritech was prejudiced by the exclusion of additional evidence to the effect that an arbitrator likely would have enforced the deadline. What the EEOC argued was not that the company should have risked arbitration by firing Amos beyond the 30-day deadline, but that the company inexcusably missed the deadline in the first instance.
In practical terms, then, Ameritech’s need for the CBA/ arbitration evidence was limited to the Patti Wolter incident. Ameritech would have established that the CBA required “just cause” to fire an employee, that it had lost a number of previous arbitrations regarding discharged employees, one of which involved a harasser, and that an arbitrator likely would not have viewed the “you look so sexy today” note that Amos had written to Wolter as sufficiently serious to warrant his discharge. I accept for the present that this evidence might have persuaded the factfin-der that the company’s decision to suspend, rather than discharge, Amos in response to the note was at worst negligent, rather than indicative of malice or reckless indifference to the rights of those whom Amos harassed. But in view of the scope of the EEOC’s case against Ameritech, I find it difficult to believe that the proffered explanation for the decision not to discharge Amos was likely in any way to have altered the jury’s decision to impose punitive damages.
The evidence bespeaking Ameritech’s reckless indifference to the rights of its female workers was anything but marginal. Long before any of the three claimants in this case complained of misconduct by Amos, he had engaged in the full range of *540harassment — from offensive remarks to intimate, unwelcome bodily contact — and Ameritech had already tried (unsuccessfully) suspension and other measures short of discharge to keep the harassment from recurring. Yet, notwithstanding the ample notice of Amos’ proclivities, the company’s course of action during the period of time in which Amos harassed the claimants remained passive, wavering, and ineffectual. Consider the following:
1. By the time the Civil Rights Act of 1991 took effect, Ameritech was well aware that it had a serial harasser on its hands. The complaints in his file indicated that he had been making inappropriate remarks to his female co-workers, touching them, and exposing himself to them since 1975. By 1991, twelve women had complained about Amos’ behavior. The company also knew that a variety of disciplinary measures had already proven unsuccessful in stopping Amos. In 1990, when six women filed a joint complaint regarding Amos’ conduct (he had rubbed his penis against five of them), the company had suspended Amos for two weeks, stripped him of a sales award, and warned him (as it had in 1988) that he would be terminated if his misbehavior persisted. It did persist.
2. In February 1992 — nine months before the incident that caused Debbie Wentland to complain to Ameritech — co-worker Jennifer Rice complained that Amos had made inappropriate remarks about her physique, rubbed her neck, and rubbed his body against hers. Despite Amos’ lengthy history of engaging in such conduct, Ameritech did not fire him, as it had twice before promised it would do. Nor did it suspend him, as it had in 1990. In fact, the company did not take any punitive action at all against Amos. Amos himself suggested that he be transferred out of the small business office, where 95 percent of his co-workers were women. But his supervisor rejected even that suggestion. The company’s sole action was to admonish Amos that he was not to have any physical contact with his co-workers — as if he were simply confused about the boundaries of acceptable behavior — and to warn him (yet again) that he faced suspension or termination if he failed to comply. Not surprisingly, given that he had been touching his co-workers inappropriately for more than fifteen years, Amos did not comply with the no-contact order.
3.In November 1992, Amos approached pregnant-co-worker Debbie Wentland, patted her stomach, and remarked, “Oh, so you are going to be a mom.” Amos had been touching Wentland, and touching himself in front of her, for some time; but this event finally prompted her to complain. When Wentland reported Amos’ behavior, Ameritech failed to suspend him immediately, an omission that restricted to 30 days the period of time that the company had to act on the complaint. EEO Coordinator Monica Sharp recommended to her superiors within that time frame that Amos be discharged, finding that Amos had violated the no-touch rule. Yet, because Labor Relations Manager Joyce Leek was on vacation, Sharp’s recommendation sat unread on Leek’s desk until after the 30-day period had expired. At that point, the company decided it could take no action *541against Amos. It simply warned him, for the fourth time, that further misconduct could result in suspension or termination. Predictably, Amos did engage in further misconduct.
4. On April 15, 1993, Lori Everts filed an EEOC charge asserting that Amos had been sexually harassing her for two years. Without conducting any investigation of its own into Everts’ charge (consistent with its protocol for external complaints, the company instead simply cooperated with the EEOC’s inquiry), an attorney in Ameritech’s legal department summarily concluded that the charge had no merit. It seems that just a few months before Everts filed the charge, Sharp had been looking into Amos’ conduct and had attempted to speak, largely without success, to Everts and another employee, Patricia Black, about Amos’ behavior. Black’s union representative had constantly interrupted her interview and allowed her to say very little; and Everts had refused to say anything at all, explaining that as a chief union steward it would be inappropriate for her to help the company investigate a union member. The refusal to cooperate with Sharp’s inquiry had prompted the company to warn the union’s president that it would seek to hold the union hable if Amos struck again. The company’s frustration with the union, as well as its decision to leave the investigation of Everts’ charge to the EEOC (and so to avoid any charge of interference or retaliation), are understandable. Given Amos’ history, however, it is surprising, to the say the least, that the company would so quickly dismiss Everts’ charge as meritless. If Ameritech had been truly interested in preventing further harassment, Everts’ charge should have raised alarm bells. The company, however, did nothing. The results were, by this time, foreordained. Having been given a pass by his employer, Amos went on to harass Everts for several months after she filed her charge.
5. In June 1993, Patti Wolter complained after Amos left a note on her desk that read, “Patti, you look so sexy today.” Remarks akin to this one had prompted a number of previous complaints from Amos’ coworkers, and often these remarks were precursors to the unwelcome bodily contact and display of Amos’ genitals with which the company was also well familiar. Still, the company did not fire Amos, believing that termination could not be justified on the basis of this incident, notwithstanding the wealth of complaints that had preceded it. Instead, the company opted to suspend him for the maximum period of time possible — 30 days — although suspension had already proven ineffective. Beverly Gray, a supervisor in Amos’ division, herself doubted that suspension would have any effect on Amos’ behavior. And, indeed, it did not.
6. Not until March of 1994 did Ameri-tech finally take the decisive action with which it had been threatening Amos (to no avail) for years. For months, Amos had been touching Wendy Pollard’s hair, staring at her with his belt unbuckled, telling her he was obsessed with her, and displaying photos of lingerie-clad and topless women to her. Being new to the small business unit, Pollard bore *542all of this in silence until one evening, as she was moving her work station to the “pod” at which Amos also worked, she discovered Amos with his pants open, masturbating. Nearly 20 years after another woman at Ameritech complained of similar behavior, the company at last suspended him with dispatch, conducted an investigation, and fired him.
Against this backdrop, Ameritech’s case for the admission of the CBA/arbitration evidence is myopic, in the sense that it seeks to explain the company’s decision not to fire Amos on just two occasions. What so convincingly demonstrates the company’s indifference to the rights of its female workers, however, is not its failure to fire Amos on any particular occasion, but the whole of its meandering, passive, and predictably ineffective response to Amos. The proffered CBA/arbitration evidence does not explain why the company missed the 30-day deadline for discipline vis ... vis Debbie Wentland’s complaint. It does not explain why the company’s own monitoring of Amos did not progressively increase with each confirmation it received that Amos, despite many prior warnings, was continuing to sexually harass his colleagues. It does not explain why the company did not steadily ratchet upward the disciplinary response to Amos as it received successive complaints of harassment. It does not explain why, after Am-eritech threatened Amos with discharge in 1988 and 1990 if he continued to harass his co-workers, it took more than three years, and several more complaints of harassment, for the company to remove him. It does not explain why the company, knowing that Amos shifted his unwelcome attentions from one worker to the next, left him in a division staffed almost entirely by women and where his misdeeds apparently went unnoticed by anyone except his victims. The excluded evidence, in short, supplies no answer to how an employee could have harassed at least seventeen women over a period of many years except in a work environment in which sexual harassment was tolerated. Ameritech’s invocation of the CBA, and its predictions of failure in arbitration, are a rather large red herring.
2.
Even if we believe that the district court’s stated rationale for excluding certain evidence was wrong, we may deem the error harmless if the evidence might have been properly excluded on other grounds. See, e.g., United States v. White, 222 F.3d 363, 372 (7th Cir.2000). As Judge Easter-brook himself suggests, the district judge retains the authority, under Rule 403, “to keep the trial within sensible bounds.” Ante at 527. Looking more closely at the nature of the evidence that Ameritech proposes to introduce, as well as its purpose in offering the evidence, I am not convinced that its probative worth is sufficiently strong to compel its admission, and thus to warrant a retrial.
a.
A substantial problem with the CBA/arbitration evidence is the many collateral questions it raises. Ameritech would like to demonstrate, for example, that it has an unenviable record in arbitration — Everts (who by the time of trial, was vice-president of the union) would testify that the company had lost every arbitration with which she had been involved — in order to explain why it was likely that a “premature” discharge would have resulted in Amos’ reinstatement. That assertion, of course, invites comparison to the facts underlying the prior arbitrations and predictions as to whether an arbitrator might *543have been more or less likely to find that the company had “just cause” under the CBA to fire Amos, particularly in view of his extensive history of harassment. Further, as I pointed out in my first dissent, a ruling requiring the reinstatement of a harasser may be challenged in court on public policy grounds. 214 F.3d at 831-32. The EEOC will be free to explore that possibility as well in order to counter the impression that Ameritech inevitably would have been bound by an arbitrator’s negative ruling regarding Amos. Ameri-tech also suggests, however, that if arbitration resulted in a finding that Amos had been terminated without just cause, the arbitrator, in addition to ordering Amos reinstated, might also have ordered the company to expunge Amos’ disciplinary record, a step that would have made it more difficult to fire Amos if and when his harassment recurred. I find it hard to believe that an arbitrator would in this way hamstring a company’s ability to deal with a recidivist harasser, or that a court would find such relief consistent with the terms of Title VII. More to the point, I find it hard to imagine how a jury could reasonably sort out all of the various possibilities. I understand, of course, that the jury need not determine whether, in fact, an arbitrator (and a court) would have sustained Amos’ discharge; its task will be to assess whether Ameritech’s delay in dealing decisively with Amos rose to the level of reckless indifference. But in order to determine whether Ameritech’s professed concerns had some genuine basis, the jury inevitably will have to evaluate the likelihood of the scenarios about which Ameritech claims to have worried. The possibility for a lengthy, complex sideshow on collective bargaining and the arbitration process seems quite real.
b.
The true allure of the CBA/arbitration evidence, of course, lies in its potential to shift the blame for the delay in firing Amos from Ameritech to the arbitrator, who (the theory goes) was likely to have stood in the way had the company attempted to discharged him earlier. Yet, as Judge Easterbrook has so persuasively explained, the CBA is a contract of Ameri-tech’s own making, and the arbitrator likewise is not an external force beyond the company’s control — he exercises only such power as the parties convey to him, and to that extent he speaks for the employer. Ante at 522; Eastern Associated Coal Corp. v. United Mine Workers, 531 U.S. 57, 121 S.Ct. 462, 467, 148 L.Ed.2d 354 (2000). This is not to say that an employer will necessarily agree with the arbitrator’s ruling in any particular case. It is to say, however, that the employer cannot shirk responsibility for the outcome and effects of arbitration. If a company previously has lost a series of arbitrations arising from its efforts to deal with wayward employees, including those who engage in sexual harassment, those losses do not entitle the company to simply throw up its hands when faced with another harasser. As a party to the collective bargaining agreement which delegates authority to the arbitrator, the employer has the ability to seek whatever contractual changes are necessary to deal with harassers appropriately. See ante at 521-22. And if the CBA and hostile arbitrators remain in the way of discharging or otherwise disciplining a harasser, the employer still is not relieved of the obligation to take whatever unilateral measures are within its power (reassignment, for example) to keep him from inflicting harm on his co-workers. See ante at 522.
Recall the hypothetical truck driver, known to be a menace, who strikes a pedestrian while on company business. Ante at 522. Judge Easterbrook is right to as*544sert that the company could not justify its failure to fire the driver by claiming that an arbitrator would have reinstated him. As my colleague points out, even if an arbitrator had ordered the driver reinstated, the company, knowing of his incompetence, remained obliged to address the danger he posed to the public at large. Ante at 522, 523. Short of discharge, the company might have retrained and tested him, monitored him for drug and alcohol use, put another driver in the truck with him, and, if need be, posted him to a desk job unless and until satisfied that he could drive safely. Compensating for the results of arbitration in this way would surely have been costly to the company, but the cost is one attributable to the company itself — either it was not arbitrating effectively, or it had voluntarily bound itself to a dispute resolution system that unduly restricted its ability to deal with errant drivers.
For these very reasons, I doubt the probative worth of the excluded evidence regarding the CBA and the possibility of arbitration to Ameritech’s defense against an award of punitive damages. Even if the company can show that an arbitrator would not have sustained a decision to fire Amos in December 1992 or in April 1993, that begs the question why the company otherwise left Amos in a position that enabled him to continue harassing his female co-workers in precisely the same ways he had been harassing them for years. It also leads one to wonder why, if indeed Ameritech did have a poor record in arbitration, the company did not seek changes to the CBA that would have removed whatever restraints arbitrators were placing on the company’s ability to rid itself of recalcitrant harassers.
Restated as a pure concern about the monetary cost of discharging Amos prematurely, see ante at 524-25, Ameritech’s theory of defense remains a double-edged sword. A financial judgment that the precautionary measures necessary to prevent injuries from occurring are not cost-justified can came back to haunt the company, as Ford learned in the Pinto litigation. See Grimshaw v. Ford Motor Co., 119 Cal.App.3d 757, 174 Cal.Rptr. 348, 384 (1981). Even if the jury had known that Ameritech risked liability for a backpay award had it discharged Amos in December 1992 or April 1993, I cannot believe that its assessment of the company’s culpability would have been different. No one but the company itself was responsible for missing the 30-day disciplinary deadline in December 1992, when by its own account it had a perfectly legitimate basis for terminating him. The possibility that a belated discharge on that occasion, or a timely discharge in April 1993 (when his behavior was less egregious), might have proved costly for Ameritech, hardly seems likely to have elicited the jury’s sympathy, when the company had been putting up with Amos’ misconduct for years — and years.
3.
Perhaps the most appealing argument that Ameritech seeks to make with the arbitration evidence is that it was more prudent for the company to await a firm basis for discharging Amos — i.e., one that an arbitrator was likely to approve — than to risk having him reinstated. A premature discharge followed by reinstatement, the theory goes, would demoralize Amos’ co-workers and make it all the more difficult for the company to discharge him in the future.
Implicit in this theory, however is the acknowledgment that Amos inevitably would strike again. Gone is any pretense that the disciplinary measures attempted until that point might have succeeded in stopping the harassment. The argument, *545in fact, rests on the premise that the harassment would recur, that consequently there would be future opportunities to discharge Amos, and that it was better to await an instance of harassment that an arbitrator likely would agree was severe enough to constitute just cause for a discharge. What the argument also assumes then, is that women harassed in the interim, in ways that the company did not think an arbitrator would find sufficiently egregious to warrant Amos’ discharge, would simply have to sacrifice their right to work unmolested for the greater good of the other women who worked at Ameritech. It is as if the company said to the three claim ants in this case, “Look, we know Amos might be harassing you. But we’ve got these rules that make it hard for us to get rid of him. And we don’t want to fire him, only to have some arbitrator tell us to take him back. Better to get rid of him once and for all. So, get back to us when he does something really bad.” For the women who, like Debbie Wentland, Lori Everts, and Wendy Pollard, suffered the harassment that foreseeably occurred while Ameritech awaited a rock-solid case for discharging Amos, the theory reflects a reckless indifference to the right to do one’s job without having to encounter a coworker’s penis.
4.
In sum, the evidence supporting an award of punitive damages in this case was strong, and the likely impact of the evidence regarding the CBA and Ameritech’s prospects in arbitration is modest, at best. The harassment in this case spanned a period of nearly twenty years. By December of 1992, when the first of the three claimants in this case complained, the company itself was convinced that Amos lacked either the willingness or the ability to stop harassing his co-workers. In the words of the company’s EEO coordinator, Monica Sharp, Amos just “didn’t seem able to control himself at Indiana Bell.” But, having already missed a number of opportunities to deal with Amos decisively, the company inexcusably allowed yet another opportunity to pass by missing the deadline for discharging Amos. As a result of that mistake, and the company’s consistently passive approach to dealing with Amos, three more women would fall victim to his harassment, bringing the reported total to seventeen. Not until he did something so over the top as to expose himself and masturbate at his workstation was the company finally willing to remove him from the workplace.
Ameritech’s proffered CBA/arbitration evidence offers little, if any, explanation for the company’s reactive and complacent responses to Amos’ persistent harassment. It purports to explain only two of the missed opportunities to deal with Amos, and in actuality is only necessary as to one of the incidents. Even then, the evidence is offered to explain only why the company did not fire Amos; it supplies no explanation as to why the company did not take a variety of other measures to protect its female workers from a familiar and unrepentant harasser. In short, the evidence raises as many questions as it answers. I find it implausible that the company was genuinely prejudiced by the exclusion of this evidence. The district court should not be put to the burden of conducting another nine-day trial, nor should the victims of Amos’ harassment be made to relive a shockingly hostile work environment yet again.