dissenting in part:
When a traffic light simultaneously blinks “Stop” and “Go” everyone knows repairs are needed. If a motorist encountering the light proceeds ahead while another motorist pauses, it is unimaginable that both would be guilty of failing to heed the signal. The Board’s “law” governing pre-election wage increases is like the faulty traffic light and the Board’s enforcement of that “law” approaches the unimaginable. As the Board sees it, when yearly wage increases are the norm, and the one-year anniversary falls just before a representational election, employers who proceed to grant a raise on that date are illegally trying to influence the vote. The Board also believes that employers who hold back and let the date pass are just as guilty of an unfair labor practice. Board theory one is that a company giving the raise demonstrates its power over its employees, implicitly threatening them with the removal of benefits should they vote for the union; Board theory two is that a company postponing the wage increase until after the election unfairly attempts to scare its employees into voting against the union. Theory one has received the Supreme Court’s blessing, see NLRB v. Exchange Parts Co., 375 U.S. 405, 409, 84 S.Ct. 457, 11 L.Ed.2d 435 (1964), and it has also received severe criticism. See Derek C. Bok, The Regulation of Campaign Tactics in Representation Elections Under the National Labor Relations Act, 78 Harv. L. Rev. 38, 113 (1964); Robert A. Gorman, Basic Text on Labor Law: Unionization and Collective Bargaining 164 (1976). Regardless of how Board theory one is judged working alone, when it is considered in tandem with Board theory two it rises—or more accurately falls—to the level of arbitrariness. Many courts and administrative law judges have expressed exasperation with the Board’s Janus-faced doctrine. See, e.g., Pedro’s, Inc. v. NLRB, 652 F.2d 1005, 1008 n. 8 (D.C.Cir.1981); J.J. Newberry Co. v. NLRB, 645 F.2d 148, 151 (2d Cir.1981); Free-Flow Packaging Corp. v. NLRB, 566 F.2d 1124, 1130 (9th Cir.1978); NLRB v. Otis Hosp., 545 F.2d 252, 255 (1st Cir.1976); Osco Drug, Inc., 237 N.L.R.B. 231, 232-33, 1978 WL 7813 (1978). Why the Board does not call a halt to this nonsense is unfathomable. The Board plainly has the power do so. Through its Regional Directors, the Board can simply schedule elections at a time removed from the historical anniversary date for iinit wage increases.
Substantial evidence does not, in any event, support the Board’s finding that Perdue gave the wage increases on June 14, 1995, in order to influence the election scheduled for the next day, and thereby violated § 8(a)(1). Perdue acted on June 14 because the Board’s bewildering doctrine gave the company no other realistic option. Perdue took over the Dothan, Alabama plant early in 1995. In previous years, employees at the plant received their annual pay raises between June 1 and July I. In 1994, wage increases at Dothan were granted on June 15, the first day of the new pay period. We have held that employers may give wage increases prior to an election “in the normal course of [] business,” Pedro’s, 652 F.2d at 1008, and “in a manner and at a time in accord with past practice,” Allen v. NLRB, 561 F.2d 976, 981 (D.C.Cir.1977). There can be no dispute that it was Perdue’s “normal” practice to maintain an acquired company’s customary date for granting a pay raise. At Dothan, only one day naturally suggested itself as the customary date: June 14, 1995, which as in the 1994 Dothan raise, represented the first day of the new pay period. Only that date could clearly be considered “in the normal course of business” at Dothan, and in compliance—if not with the precedents of the Board—at least with the decisions of this court.
The obvious question is “What should the company have done differently?” One idea is that before the election Perdue should have announced (1) that it was withholding granting a wage increase on the customary date in order to avoid the appearance of attempting to “bribe” employees to vote against the union, but (2) that it would grant the raise after the election no matter what the outcome. But in terms of the effect on employees, there is only one difference be*840tween the company’s granting the wage increase on June 14 and giving a promise that the increase will occur on June 16, or some other time shortly after the election. The difference is not that the employees will feel less threatened by the company. The difference is that the announce-the-raise-but-give-it-later approach means employees will not get their raise on the customary date and thus will lose money. Who may the employees blame? Maybe the union. Better yet, the Board, and the courts who go along with the Board. As an alternative one might suppose that Perdue should have postponed the wage increase but made it retroactive to June 14. But by any measure there is no difference between announcing a wage increase on June 14, and announcing that a wage increase will be granted after the election, retroactive to June 14. In any event, both of these options, and others, erroneously place the burden on Perdue to alter its normal business practice to accommodate the union election. See Newport Div. of Wintex Knitting Mills, 216 N.L.R.B. 1058, 1058, 1975 WL 5400 (1975).
I wish also to explain why I would reject the Board’s conclusion that on June 13, two days before the election, Perdue relaxed Do-than’s attendance disciplinary policy in order to influence votes, and thereby violated § 8(a)(1). See Cooking Good Div., 323 N.L.R.B. No. 50, at 9-10, 1997 WL 156724 (Mar. 31, 1997). The evidence of any such change in policy is slim to nil. The Board relied chiefly on the testimony of employee Bryan Smith, who said he attended a meeting held by manager Tony Williams. At first, Smith could not recall Williams discussing such a policy change, nor could he remember the date of the meeting, or whether it was close to the election. Smith also testified that he had no knowledge of the attendance policy in place before the meeting. Only after reviewing his affidavit was Smith able to give June 13 as the meeting date. Smith then said Williams announced that the attendanee policy “changed” so that “when you missed a day, that would count as half an occurrence. And that if you missed two days, that was like, then you missed the whole day, so then you got write [sic] up....” J.A. 112. How Smith could take this as a “change” in policy when he admitted not knowing the existing policy is a mystery. Williams, on the other hand, flatly denied announcing any change to the attendance disciplinary policy at any time and Ed Scarborough, the human resources representative at Dothan, testified that Perdue first altered this policy on January 15, 1996—six months after the election.
Scarborough also testified that he was very familiar with the attendance disciplinary policy instituted by the previous employer, which he described as follows:
The first two times you are just verbally warned about being absent; the third time you was [sic] out was a written warning; the next time was a one-day suspension; the next time a three-day; and then termination.1
J.A. 276. This matches the description of the “change” Smith said he heard about at the June 13 meeting.
The Board also relied on a memo by Scarborough issued to employees on June 30, 1995. See Cooking Good Div., 323 N.L.R.B. No. 50, at 10, 1997 WL 156724. But the document merely describes the existing attendance policy established by the previous employer: “An associate that is late (2) two times has the equal of (1) one whole occurrence, which is the same as missing one full day and it may cause you to get a written letter of warning____” J.A. 339.
No one produced any written record of a policy change and not a single witness had a clear memory of any easing of the attendance policy before the election. If Perdue’s goal had been to influence the election, one would have expected it to broadcast the new policy *841loudly and clearly. Yet on this matter of importance to employees, there was no proof of any widespread knowledge among them. Given this state of affairs, there was no substantial evidence that Perdue made any policy change, let alone that it intended to influence the election.
. Scarborough described the new Perdue attendance policy implemented on January 15, 1996:
The first two times you are [absent], there is just nothing done about it, you are free; the third time you are out, you receive a written verbal [sic] warning; the next time you get a second warning; the third is a final warning; and then the fourth we give a three-day suspension pending investigation of our records to make sure that our records are correct. If they are correct, when they get the three-day suspension, they are term[inat]ed. They can work off an incident, though every 28 days with Perdue.
J.A. 277-78.