Dresser–Rand Company, Petitioner–Cross Respondent, v. National Labor Relations Board, Respondent–Cross Petitioner.

Reset A A Font size: Print

United States Court of Appeals, Fifth Circuit.

Dresser–Rand Company, Petitioner–Cross Respondent, v. National Labor Relations Board, Respondent–Cross Petitioner.

No. 15–60474

Decided: September 23, 2016

Before JOLLY, CLEMENT, and OWEN, Circuit Judges. Jeremy Daniel Kernodle, Haynes & Boone, L.L.P., Arthur Tracy Carter, Littler Mendelson, P.C., Dallas, TX, Alfred John Harper, III, Littler Mendelson, P.C., Houston, TX, for Petitioner–Cross Respondent. Linda Dreeben, Esq., Julie Brock Broido, Jeffrey William Burritt, Esq., Micah Prieb Stoltzfus Jost, Gregory Paul Lauro, National Labor Relations Board, Washington, DC, Rhonda P. Ley, National Labor Relations Board, Buffalo, NY, for Respondent–Cross Petitioner.

In 2007, Dresser–Rand and CWA Local 313 sought to negotiate the renewal of the union-members' employment contract but could not reach agreement. After a four-month strike, the parties had still not agreed on a contract. The union called off the strike and agreed to return to work without a contract. At that point, however, Dresser–Rand locked out the union employees. After a week, Dresser–Rand reversed course and allowed the union members to return to work.

The National Labor Relations Board held that the lockout violated the National Labor Relations Act. On appeal, the parties agree that whether the lockout violated the Act is determined by Dresser–Rand's motivation for the lockout:  “if the lockout was motivated by antiunion animus, it was illegal; if the lockout was not motivated by antiunion animus, it was permissible. The NLRB found that the lockout was motivated by antiunion animus based on actions that Dresser–Rand took after the lockout ended.

Dresser–Rand petitions for review of the NLRB order, arguing that 1) its post-lockout conduct was lawful; 2) even if the post-lockout conduct was not lawful, this conduct was not, itself, motivated by animus; and 3) even if the later conduct was motivated by animus, it does not show that the lockout was also motivated by animus.

We hold that much of the later conduct did not violate the Act and the conduct that did violate the Act was not motivated by animus. Accordingly, these violations do not establish that the lockout was motivated by antiunion animus. We therefore grant in part and deny in part Dresser–Rand's petition to deny enforcement of the NLRB order;  correspondingly, we deny in part and grant in part the NLRB's cross-application for enforcement.

I.

Dresser–Rand, a company headquartered in Texas, operates multiple factories across the country. One of those factories, in Painted Post, New York, employs a workforce comprised of members in the Industrial Division of the Communications Workers of America, Local 313 Union who are covered under a collective-bargaining agreement. In 2007, the then-current collective bargaining agreement was scheduled to expire. Anticipating this expiration, the parties began early meetings in April. The early talks did not produce an agreement, so the parties began regular negotiations in July. By August 4, the negotiations had still produced no agreement and the collective-bargaining agreement expired. The union called for a strike and, initially, all members of the union participated in the strike. This strike would ultimately last for over four months.

During the strike, Dresser–Rand attempted to continue production. To do so, it initially hired temporary replacements and, later, hired permanent replacements. About a month into the strike, thirteen of the strikers made an unconditional offer to return to work. They crossed the picket lines and resumed work; we, like the parties, refer to these employees as the “crossovers.” Dresser–Rand and the union continued to bargain during the strike, meeting three times.

On November 19, the union called off the strike and made an unconditional offer to return to work. On November 23, Dresser–Rand rejected this offer and instituted a lockout, an employer negotiating tactic to pressure unions into a contract by withholding work until agreement is reached. During this lockout, Dresser–Rand continued to operate with a small workforce consisting primarily of permanent replacements hired during the strike. All employees who had participated in the strike were precluded from working at Dresser–Rand during the lockout. On advice of counsel, Dresser–Rand also locked out the crossovers, along with the other strikers, because it believed that failing to lockout the crossovers would violate labor law.1 When it did so, Dresser–Rand informed the departing crossovers that they would be hired back as soon as it could legally do so. In the meantime, Dresser–Rand informed the union that it would end the lockout if the union agreed to Dresser–Rand's bargaining demands.

Six days into the lockout, Dresser–Rand changed course and agreed to lift the lockout. Dresser–Rand announced this decision on Thursday, November 29. Shortly after this announcement, the union told Dresser–Rand that the union would tell members to show up for work on Friday at 7:00 AM. Dresser–Rand replied that 7:00 would not work and that Dresser–Rand needed to assess its manpower needs. The parties spoke on the phone, and discussed negotiating a recall procedure. Without waiting for this negotiation, however, Dresser–Rand informed the crossovers that they could return to work.

Dresser–Rand attempted to negotiate with the union over a process for returning strikers to work, but the union was mostly non-responsive. Specifically, late Friday afternoon, Dresser–Rand informed the union that it would recall striking employees based on seniority and job performance; Dresser–Rand also proposed that it contact employees based on this list beginning on Sunday. The union representative responded briefly via email, saying that he was traveling but would have access to email;  he neither agreed nor objected to Dresser–Rand's recall proposal or timeline. On Sunday, Dresser–Rand sent the union the names of non-crossover strikers, listed in the order Dresser–Rand proposed to recall them. Without hearing either agreement or disagreement from the union, Dresser–Rand began recalling employees that same day.

One employee, Kelvin Brown, was not on the recall list. Dresser–Rand determined that Brown had engaged in picket-line misconduct and thus decided not to recall him, effectively firing him.2

After the recall was implemented, Dresser–Rand unilaterally changed its policy regarding paid lunch breaks for weekend overtime work. Dresser–Rand previously had a longstanding policy of providing paid twenty-minute lunch breaks to any employees working weekend overtime shifts of at least seven hours. This policy was not memorialized in any contract. After the recall, Dresser–Rand ended this policy. It did not discuss this change with the union.

In late April 2008, a recalled employee was attending a department meeting. This employee, Marion Cook, stated that “there were too many salaried workers and too many scabs for it to be safe to work.” Dresser–Rand subsequently suspended Cook without pay for “violating common decency or morality on company property” based on his use of the word “scab.”

In August and September, Dresser–Rand recalled the last of the striking employees (those lowest on the recall list). Dresser–Rand informed twenty-three of these employees that they had lost their accrued paid vacation because they had not worked at least 900 hours in the preceding twelve months. (These employees had been striking, locked out, or waiting on the recall list for that period.)

The employees worked without a contract until 2009, when they finally reached an agreement.

Subsequently, the union filed various unfair labor practice charges against Dresser–Rand which were tried before an administrative law judge in a nine-day trial in 2009. The ALJ determined that Dresser–Rand illegally locked out employees, illegally recalled employees after the lockout, illegally disciplined two strikers after the lockout, and illegally effectuated new employment terms after the strike. In 2012, the NLRB affirmed this determination. That affirmance was invalidated by NLRB v. Noel Canning, –––U.S. ––––, 134 S.Ct. 2550, 189 L.Ed.2d 538 (2014), and remanded to the NLRB. In 2015, a divided panel of the NLRB affirmed again. Dresser–Rand filed a petition for review in our court, requesting that we deny enforcement of the NLRB order;  the NLRB filed a cross-application seeking to enforce the order. The union intervened on behalf of the NLRB.3

II.

We uphold the NLRB's factual findings “if [they are] supported by substantial evidence on the record considered as a whole.” Mobil Expl. & Producing U.S., Inc. v. NLRB, 200 F.3d 230, 237 (5th Cir. 1999). “Motive is a factual matter” in NLRB proceedings and thus is subject to appellate review as a factual finding. NLRB v. Mini–Togs, Inc., 980 F.2d 1027, 1032 (5th Cir. 1993). Legal questions, such as questions of contractual interpretation, are reviewed de novo. Strand Theatre of Shreveport Corp. v. NLRB, 493 F.3d 515, 518 (5th Cir. 2007).

III.

The NLRB argues that Dresser–Rand violated Sections 8(a)(3) and 8(a)(1) of the National Labor Relations Act in multiple ways. We address each argument in turn.4

A. The preferential recall of crossovers.

It is an unfair labor practice for an employer “by discrimination in regard to hire or tenure of employment or any term or condition of employment to encourage or discourage membership in any labor organization.” 29 U.S.C. § 158(a)(3). The NLRB argues that Dresser–Rand violated this section by first recalling crossovers before recalling full-term strikers, thereby allegedly retaliating against full-term strikers for their union activity.

The Supreme Court has addressed a similar stratagem:  in TWA, the Court held that an employer, who had hired crossovers during a strike, was not required to dismiss those crossovers when the strike ended and more senior employees returned to work; instead, the employer could permissibly leave those crossovers in place, effectively granting them preference above all full-term strikers. Trans World Airlines, Inc. v. Indep. Fed'n of Flight Attendants, 489 U.S. 426, 109 S.Ct. 1225, 103 L.Ed.2d 456 (1989) (“TWA”).5 Thus, if Dresser–Rand had not instituted a lockout, TWA would clearly govern and Dresser–Rand could permissibly have retained the crossovers in preference to the full-term strikers.

Here, however, Dresser–Rand did lockout employees, including the crossovers. According to the NLRB, that difference is decisive. The NLRB argues that, once the crossovers were locked out, the positions they had filled during the strike were vacant and thus Dresser–Rand had a duty to fill those positions without providing preference to the crossovers. In support of this position, the NLRB cites Peerless Pump Co., 345 NLRB 371 (2005). In Peerless Pump, the Board held that the company could not grant recall preference to employees who had agreed to crossover but had not yet returned to work. Id. at 376.

Dresser–Rand, on the other hand, cites to NLRB v. Delta–Macon Brick & Tile Co., 943 F.2d 567 (5th Cir. 1991). Delta Macon held that:

when the company temporarily lays off a permanent striker replacement ․ [this temporary layoff does not create a vacancy that must be made available to strikers] so long as the laid-off, permanent striker replacement has a reasonable expectancy that he will be recalled by the company following the layoff. This is so because an employer may promise permanent striker replacements that if they are subsequently laid off the layoff will not create vacancies that the employer must fill by recalling the strikers whom the replacement workers had replaced and who by definition will have greater seniority, having been hired earlier. When laid-off, permanent striker replacements have an objectively reasonable expectancy of recall, the employer may recall them ahead of the unreinstated strikers.

Id. at 572 (citations and quotations omitted) (emphasis added). Dresser–Rand further argues that the crossovers had a “reasonable expectancy of recall” because Dresser–Rand told them that they would be recalled immediately after the lockout ended.

Accordingly, we must decide whether locking-out crossovers created a job vacancy (which then must be filled by returning strikers) or whether it constituted a temporary layoff from the job (in which case the crossovers would have priority over the returning strikers). We hold that, on the facts of this case, the crossovers were more similar to temporarily-laid-off replacement workers. Here, the crossovers occupied a job position (as crossovers) for a period of months before they were locked out for only a few days. Moreover, the crossovers were told, before being locked out, that they would resume their then-current positions as soon as the lockout ended. Thus, recalling them first simply allowed them to return to the position they had been permanently performing.6 We therefore hold that substantial evidence does not support the NLRB's conclusion that Dresser–Rand violated labor law by recalling the crossovers before recalling the current strikers. In short, the jobs occupied by crossovers were not vacant positions to be filled by returning strikers.

B. Unilateral recall procedure.

The Board next argues that Dresser–Rand violated labor law by implementing a recall procedure without properly collaborating with the union about the details of that procedure. It is fundamental that employers are generally required “to bargain collectively with the representatives of [their] employees.” 29 U.S.C. § 158(a)(5). A union may waive the right to bargain but will not “be held to have waived bargaining over a change that is presented to it as a fait accompli.” Gulf States Mfg., Inc. v. NLRB, 704 F.2d 1390, 1397 (5th Cir. 1983). The NLRB argues that Dresser–Rand did not bargain with the union but instead presented the union with a fait accompli method of recalling employees and then implemented that method without input from the union. According to the Board, Dresser–Rand did this despite knowing that the union would prefer a different recall procedure.

Dresser–Rand responds that it attempted to contact the union over a several-day period and that the union's failure to respond to Dresser–Rand's suggestions could reasonably be interpreted as agreement to the procedure—especially when the union knew that time was of the essence.

We do not believe that the Board's conclusion is supported by substantial evidence. Instead, the record is plain that time was of the essence as Dresser–Rand began the recall. It made its intentions clear to the union representatives, and kept the union informed at each step. The parties had agreed that recall negotiations would go on through their lawyers. The union agreed that the recall should begin as soon as possible and stated its position that the workers be recalled by seniority. The union had ample opportunity to make another counterproposal after it received notice of Dresser–Rand's proposal. In sum, the evidence does not reasonably suggest that Dresser–Rand failed to bargain or that it stated its terms as a fait accompli. Under the well-understood, urgent circumstances, the union had a reasonable, albeit brief, opportunity to bargain and suggest counter-terms, but it failed to do so. See N.L.R.B. v. Pinkston–Hollar Const. Servs., Inc., 954 F.2d 306, 310 (5th Cir. 1992) (“[W]hen an employer notifies a union of proposed changes in terms and conditions of employment, it is incumbent upon the union to act with due diligence in requesting bargaining. Any less diligence amounts to a waiver by the bargaining representative of its right to bargain.”) (citations omitted); Nabors Trailers, Inc. v. N.L.R.B., 910 F.2d 268, 273 (5th Cir. 1990) (“[T]here is no violation of the [NLRA], even in the absence of an impasse, if the employer notifies the union that it intends to institute the change and gives the union the opportunity to respond to that notice.”).

We therefore hold that substantial evidence does not support the NLRB's conclusion that Dresser–Rand refused to bargain with the union over the recall procedure.

C. Failing to recall Brown based on the picket-line incident.

The NLRB argues that Dresser–Rand violated labor law by failing to recall Kelvin Brown. The parties agree that Dresser–Rand could not permissibly refuse to recall Brown unless he had engaged in serious misconduct sufficient to justify dismissing him. See, e.g., Newport News Shipbuilding & Dry Dock Co. v. N.L.R.B., 738 F.2d 1404, 1408 (4th Cir. 1984) (“An employer who refuses or delays the reinstatement of strikers who have engaged in a protected strike violates § 8(a)(3) and (1) of the Act unless he can show ‘legitimate and substantial business justifications' for his actions. A showing that an employee engaged in serious strike misconduct constitutes such justification.”) (citations and quotations omitted). They disagree, however, about the seriousness of Brown's misconduct.

Dresser–Rand discharged Brown after police arrested him for disorderly conduct and after a police officer stated that Brown “pretended like he got hit by the [van bringing in workers].” The fact that an arrest was made does always not establish, as a matter of law, that discharge was proper. See Newport News, 738 F.2d at 1411 (4th Cir. 1984) (analyzing the propriety of discharging an employee on the merits, even though the employee had been charged by the police). Nonetheless, police involvement does provide some evidence that the misconduct was serious. According to Dresser–Rand, this misconduct was severe enough to justify discharging Brown because it put Brown and others in danger.

The Board responds, however, that Brown's conduct did not result in harm to anyone, did not injure anyone, and seems not to have delayed any employees significantly. Thus, according to the NLRB, Brown's misconduct was minor enough that he could not have been permissibly discharged.

While we must accept the Board's factual findings so long as they are supported by substantial evidence, “[t]he Board's decision on whether or not strike misconduct is serious enough to deny reinstatement is a legal conclusion that we, of course, are free to accept or reject.” Newport News, 738 F.2d at 1408. See also, e.g., N.L.R.B. v. E–Systems, Inc., 642 F.2d 118, 122 (5th Cir. 1981) (finding that “substantial evidence” supported the Board's factual determinations but rejecting the Board's “legal appraisal of the incident”). Further, other circuits have drawn the “serious misconduct” line at “conduct that is intended to threaten or intimidate nonstrikers.” Newport News, 738 F.2d at 1408 (citations and quotations omitted). Accord, e.g., N.L.R.B. v. W.C. McQuaide, Inc., 552 F.2d 519, 528 (3d Cir. 1977) (“The test is whether the misconduct is such that, under the circumstances existing, it may reasonably tend to coerce or intimidate employees in the exercise of rights protected under the Act.”); Richmond Recording Corp. v. N.L.R.B., 836 F.2d 289, 295 (7th Cir. 1987) (“Strike misconduct becomes serious if it reasonably tends to coerce or intimidate.”).

Brown placed his body in front of a van bringing in non-strikers and pretended that he was hit by the van. His conduct was clearly meant to intimidate or coerce the non-strikers in the van. And it was egregious enough to compel police involvement. Although it is not clear from the record whether the van was moving at the exact moment that Brown moved in front of the van, the van was attempting to move through a dense crowd of picketers. Brown's conduct could have injured himself and caused substantial disruption. In short, Brown's misconduct was “sufficiently serious to disqualify him for employment,” E–Systems, Inc., 642 F.2d at 121, and consequently, Dresser–Rand did not violate the NLRA by refusing to recall Brown.

D. Denying strikers their accrued vacation benefits

Before the strike, the 2004 collective-bargaining agreement established that any employee who worked at least 900 hours in a calendar year “shall at the end of such year be entitled, irrespective of any subsequent occurrence, to a minimum vacation with pay in the following calendar year.” Restated:  An employee is entitled to no vacation in the first year of his employment; if in his first year he works at least 900 hours, he is then entitled to vacation in his second year;  similarly, he is entitled to paid vacation in his third year only if he worked 900 hours in his second.

Dresser–Rand concluded that this language meant that vacation pay for striking employees (who struck in August, 2007) did not vest under this agreement, because these employees did not reach the end of 2007 under that agreement. Instead, in November 2008—at the end of the strike—they were governed by the implemented agreement, which entitled them to paid vacation only if they worked 900 hours in the previous twelve months. Thus, Dresser–Rand concluded that those employees who, between the strike, the lockout, and waiting on the recall list, were out of work for more than twelve months were not entitled to paid vacation in 2009 because they had not worked 900 hours in the preceding year, 2008—they had been striking, locked out, or awaiting recall for much of that year. This argument is a plausible reading of the 2004 agreement and of the implemented agreement;  indeed, if we were interpreting those agreements on a blank slate, we might read it similarly.

According to the Board, we are not interpreting this provision de novo. Instead, the Board argues, we must interpret the 2004 agreement with reference to a previous arbitration that resolved the meaning of the same contractual language. This arbitration had concluded that, if “an employee ․ worked the requisite number of hours to be eligible for vacation in [one] calendar year [,] [then] upon recall in a subsequent calendar year [that employee will] be immediately eligible to take vacation.” In other words, if an employee was eligible for paid vacation in the year a strike began, then he would be immediately eligible for paid vacation when recalled—even if he had not worked the required number of hours in the calendar year preceding his recall. Under this interpretation, the recalled strikers should have immediately been eligible for the vacation they accrued under the 2004 agreement when they returned. The Board notes that “it is the arbiter's construction [of the agreement] which was bargained for [and thus] ․ an arbiter's award does not merely help explain the parties' agreement; ‘the arbitrator's interpretation is the parties' agreement.’ ” (quoting Penn. Am. Water Co., 359 NLRB No. 142 (2013)) (citation omitted) (emphasis added). Thus, according to the Board, Dresser–Rand erred in denying the accrued vacation.

Dresser–Rand, however, argues that this “thirty-year-old arbitration agreement ․ [is] simply inapposite” because it “did not address the facts here [ ] where the existing agreement was replaced by new terms.” We agree. The new implemented agreement went into effect before the end of the year. Thus, the 2004 agreement was no longer in effect and the fact that the vacation time was frozen under that agreement is irrelevant. Accordingly, we hold that Dresser–Rand did not commit an unfair labor practice in this respect.

E. Changing the paid-lunch-break policy without bargaining

Finally, Dresser–Rand argues that it did not violate labor law when it changed its policy of providing paid lunch breaks for weekend shifts of over seven hours. The parties agree that paid lunch breaks were not required under the terms of the 2004 agreement and were not required under the implemented agreement. The Board argues, however, that Dresser–Rand was bound by past practice and could not unilaterally change that practice—even if that practice was not mandated by the agreement. To support this claim, the Board cites Bonnell/Tredegar Indus., Inv. v. NLRB, 46 F.3d 339, 344 (4th Cir. 1995). But Bonnell does not support this claim. Instead, Bonnell says only that

collective bargaining agreements may include implied as well as express terms. An employer's established past practice can become an implied term of a collective bargaining agreement. [P]arties' collective agreement includes both the specific terms set forth in the written agreement and any well established practices that constitute a ‘course of dealing’ between the carrier and employees. Past practices rise to the level of an implied agreement when they have ripened into an established and recognized custom between the parties.

Id. (citations and quotations omitted). This holding does not establish, as the Board argues, that a “course of dealings” can persist from one agreement to the next. Moreover, in Bonnell, the employer followed a longstanding custom in the first year governed by an agreement, and then changed that policy the following year—a year governed by the same agreement. Dresser–Rand convincingly argues that “[w]hatever past practice had been grafted onto [the 2004 agreement] ended when the 2004 agreement expired and was replaced by the implemented terms.”

This argument is especially strong given the months-long strike, the intensive bargaining over a new contact and, ultimately, a new contract with no reference supporting the practice. The justification for treating past practice as an implied term of an agreement is that parties can come to rely on a practice and can reasonably expect that practice to continue over time. When, however, the relationship between the parties is disrupted by a lengthy strike and a new, hard-fought contract, the parties cannot reasonably expect all gratuitous practices to pick up as though there had been no interruption. Here, the union had little to no reliance interest in unwritten customs that had built up around the old contract—and thus these prior customs were not silently adopted as a part of the new contract regime. This conclusion is further bolstered by the changes in Dresser–Rand's practices regarding weekend work more generally. Accordingly, we hold that Dresser–Rand's treatment of lunch breaks did not violate the NLRA.

F. The lockout

Finally, we consider whether the lockout itself was illegal. The NLRB concedes, as it must, that lockouts can be legal. Indeed, lockouts are an important negotiation tool for employers, and the Supreme Court has explicitly reaffirmed the importance of protecting their viability. NLRB v. Brown, 380 U.S. 278, 284, 85 S.Ct. 980, 13 L.Ed.2d 839 (1965). As both parties agree, this lockout was impermissible only if it was motivated by antiunion animus, instead of the desire to negotiate.

The NLRB has argued in this appeal that it can show animus by pointing to post-lockout conduct motivated by animus. As discussed above, however, substantial evidence does not support the finding that the post-lockout conduct was motivated by animus and the Board has not offered any further evidence to show that the lockout was motivated by animus. Accordingly, we hold that substantial evidence does not support a finding that the lockout was in violation of the National Labor Relations Act.

IV.

In conclusion, we hold that the following NLRB findings were not supported by substantial evidence: the finding that the lockout was illegal, the finding that Dresser–Rand's treatment of crossovers was illegal, the finding that Dresser–Rand illegally failed to negotiate a recall procedure, the finding that Dresser–Rand illegally failed to recall Brown, the finding that Dresser–Rand's modification of the lunchbreak policy was illegal, and the finding that Dresser–Rand's treatment of vacation time was illegal. Accordingly, the Board's order is denied enforcement in those respects. Dresser–Rand does not challenge the NLRB finding that Cook was improperly suspended. Thus, the Board's order is due to be enforced in that respect.

Accordingly, Dresser–Rand's petition is GRANTED IN PART and DENIED IN PART and the NLRB's cross-application is correspondingly DENIED IN PART and GRANTED IN PART, as stated in the preceding paragraph.

FOOTNOTES

1.   Dresser–Rand believes that Local 15, Int'l Bhd. Elec. Workers v. NLRB (“Midwest Generation”), 429 F.3d 651, 661 (7th Cir. 2005) established that employers must lockout every employee who participated in a strike, even crossovers. The Board argues that the law in this area is “unsettled,” and thereby concedes, at a minimum, that Dresser–Rand would have risked violating labor law if it had not locked out the crossovers.

2.   The alleged misconduct had occurred back in September, before the lockout. Near the picket line, a collision occurred between two vans carrying crossovers and replacement workers. A Dresser–Rand security guard stated that Brown had, at some point, “jumped onto the front of the van or pretended like he got hit.” As a result of this incident, Brown was arrested for and convicted of disorderly conduct. Dresser–Rand investigated the incident but did not interview Brown before deciding to omit his name from the recall list.

3.   Because the union and the NLRB advance similar arguments, we do not distinguish between them except where necessary.

4.   Dresser–Rand's petition does not challenge the NLRB's finding that Dresser–Rand violated labor law by suspending Cook for saying “scab.” We therefore do not address this issue.

5.   TWA was decided under a different statutory scheme but the parties agree that it is controlling precedent here. See Encino Tarzana Reg. Med. Ctr., 332 NLRB 914 (2000).

6.   We expressly decline to address the scenario of a substantially longer lockout, where recalling crossovers might appear to be more of an (impermissible) reward for crossing over and less of a practice of keeping current employees in positions they had recently occupied.

E. GRADY JOLLY, Circuit Judge: