Opinion for the Court filed by Circuit Judge WALD.
WALD, Circuit Judge:
Following a mail-in election during which a unit of home health aides voted in favor of union representation, the National Labor Relations Board (“NLRB” or “Board”) certified the elected Union, Local 670, Retail, Wholesale and Department Store Union (“Union”), as the unit’s collective bargaining representative. When the employer, Kwik Care Ltd., d/b/a VIP Health Care Services (“employer” or “VIP”), refused to bargain with the unit, the Union filed an unfair labor practices charge with the Board. Finding that VIP had violated § 8(a)(1) and (5) of the National Labor Relations Act, 29 U.S.C. §§ 151, et seq., (“NLRA” or “Act”), the Board ordered VIP to bargain with the Union. Kwik Care Ltd., 316 N.L.R.B. No. 98, 1995 WL 79319 (Feb. 24, 1995) (“Final Order”); Appendix (“A-”) 297.
VIP now petitions this court for review, requesting that we deny enforcement of the Board’s order because the mail-in election was improper. Specifically, the employer claims: (1) that the Board abused its discretion in ordering a mail-in election rather than a manual one; (2) that even if the Board were justified at the outset in ordering the election-by-mail, it should have rejected the results based on election complications; and (3) that the Board should have invalidated the vote in favor of representation based on the Union’s distribution of two improperly designated sample ballots which could have misled voters into thinking that the sample ballot represented the official position of the Board rather than propaganda of the Union.
Given the specific characteristics of the bargaining unit eligible to participate in this election, we find that the Board acted well within its broad discretion when it ordered a mail-in election rather than a manual one. We also find that the Board properly determined that the Union’s sample ballot constituted legitimate propaganda since it would not tend to mislead voters into thinking it had been produced by the Board. Accordingly, we deny the employer’s petition for review and grant the Union’s cross-application for enforcement of the Board’s order.
I. BACKGROUND
A. The Hearing
The Union initially petitioned the NLRB in August 1993 to represent a unit of home health care workers employed by VIP. Those employees who provide personal living assistance or health care to patients in adult care facilities or in the patients’ homes were eligible to vote in the representation election. [1124]*1124These workers live and work at locations scattered around New York, and are supervised out of five VIP offices in Westchester, Rockland, Nassau, Queens and the Bronx. Although weekly paychecks are disbursed to workers (or designated family members or co-workers) in the company offices one day each week, a significant number of the employees (14%, by the employer’s estimate) work either 24-hour shifts or 12-hour shifts during daylight hours and cannot make special trips to the office on payday. As an alternative to in-person check pick-ups, the employer mails payment to these long-shift workers and any other employees who elect to receive their checks by mail. According to the employer, approximately 20% of the unit employees receive their checks by mail. Testimony of Cynthia Rosen, VIP Director, at 93 (Sept. 8, 1993) (“Rosen Testimony”); A — 16. Of the remaining 80% of unit workers, whose paychecks are picked up each week in the VIP offices, the record is unclear how many send a friend or relative as an emissary and how many make the trek in person. The employer acknowledged that workers in group homes will sometimes send just one person to retrieve all of the paychecks, id. at 126; A-68, and VIP can conclude only that “more than half’ retrieve the paychecks themselves. Id. at 94; A-47.
The employer, urging a manual election, claimed that only 14% of the unit workers (those who work either 12- or 24-hour shifts) would have trouble making it into the offices to vote. Rosen Testimony at 99; A-52 (“[W]e asked [the directors] just to give us who was unable to vote during [the proposed] times. That’s how we came up with that. And that’s what those were, those were the live-ins and the 12 hours.”). Rosen testified that the employer would attempt to supply “replacement” workers for these employees, to allow them time to participate in the election, but could not, at the time of the hearing, guarantee coverage for all workers who might request it. Id. at 153; A-92. Rosen also opined that the employees working in group care facilities, many of whom send a representative to pick up their checks, would be able to travel to the offices to vote. “I know that some of the employees in the adult care facilities work fewer than eight hours,” she testified. “Therefore, I know that they could come to pick up their own check. That’s what I know.” Id. at 138; A-80. Rosen’s testimony, however, appears to have been based on the assumption that the employer considered any worker who had at least one hour off-duty during the time the polls would be open to be “able” to vote in a manual election. See id. at 96-102; A-49-55. For example, someone who worked a 12-hour shift at night would be deemed available to vote during the day. Id. at 100-01; A-53-54.
During her testimony, Rosen also noted that a subset of employees might work other jobs when not on duty for VIP, and that some have reported that their availability is limited due to “other responsibilities].” Id. at 114-15; A-63-64. Additionally, she testified that many of the workers depend on public transportation for their mobility. Id. at 126; A-68.
Noting the scattered location of the workers’ job sites, and the difficulty many employees might face traveling to the offices for a manual election, the Board’s Regional Director determined over VIP’s objections that the use of mail-in ballots would offer the best opportunity for the maximum number of employees to cast a vote. Kwik Care Ltd., 29-RC-8198, Decision & Direction of Election, at 12 (Dec. 14, 1993) (“Direction of Election”); A-103, 114.
B. The Ballots
On February 8, 1994, the Board sent each of the 860 eligible VIP employees a mail ballot kit which, by agreement of the parties, was printed in four languages (English, Spanish, Russian, and Haitian/Creole). Each kit contained — in all four languages — a Notice of Election, a ballot, and detailed instructions on how to east a vote. The Notice contained a warning (in large, bold, capital letters on the English and Spanish versions) informing voters that the Board would remain a neutral party in the election:
WARNING: THIS IS THE ONLY OFFICIAL NOTICE OF THIS ELECTION AND MUST NOT BE DEFACED BY ANYONE. ANY MARKINGS THAT [1125]*1125YOU MAY SEE ON ANY SAMPLE BALLOT OR ANYWHERE ON THIS NOTICE HAVE BEEN MADE BY SOMEONE OTHER THAN THE NATIONAL LABOR RELATIONS BOARD, AND HAVE NOT BEEN PUT THERE BY THE NATIONAL LABOR RELATIONS BOARD. THE [BOARD] IS AN AGENCY OF THE UNITED STATES GOVERNMENT, AND DOES NOT ENDORSE ANY CHOICE IN THE ELECTION.
Notice of Election (emphasis in original); A-149.
The voting instructions informed the employees that they had to mark the appropriate box on the ballot (and that if they made any other markings, their ballot would not count) and sign the return envelope (if they failed to sign the envelope, their ballot would not count). The instructions also included a checklist that said:
YOUR BALLOT WILL BE VOID AND WILL NOT BE COUNTED UNLESS you:
1. Mark on the ballot only in the place provided for marking; do not identify yourself on the ballot.
2. Return the ballot in the same envelope which you received for that purpose.
3. Sign your name in your own handwriting on the outside of that envelope after the word “Signature,” so that your name can be checked against the eligibility list.
Instructions (emphasis in original); A-L6. Additionally, the form stated that the ballots had to be mailed in time to be “RECEIVED” by the Board by March 8, 1994. Voters were instructed that if they had any questions about the election they should contact the Regional Director, whose address and telephone number were provided on the form.
C. The Union’s Altered Sample Ballot
The same day the Board sent out the voter kits, the Union mailed to all eligible workers a one-page letter and a sample ballot altered to show support for the Union. Kwik Care Ltd., 29-RC-8198, Supplemental Decision on Objections and Certification of Representation (Aug. 1, 1994), App. D, F (“Supp.Dec.”); A-175, 209, 211. Each envelope bore the Union’s name and return address, and contained the two unattached sheets of paper. The letter, which clearly designated the Union as its author, unquestionably constituted legitimate partisan propaganda urging support of the Union. The sample ballot, however, contained no facial reference to the Union, and the employer claims that it thus had the potential to mislead voters as to its origin.
The sample ballot resembled the official ballot distributed by the Board in the voter packets, but with certain significant differences. The official ballot posed the question, “Do you wish to be represented for purposes of collective bargaining by [the Union]?” and had at the bottom two boxes, one marked “yes” and the other “no.” Official Ballot; A-L12. The Union’s sample ballot posed the same question, but bore a heading which read, “THE BALLOT WILL LOOK LIKE THIS,” and had an X marked in the “yes” box. Additionally, the sample ballot did not have the Board’s seal on it, as did the official ballot (and the official “sample ballot” included in the Notice of Election). The following day, the Union again mailed out the altered sample ballot, this time stapled to another propaganda letter on Union letterhead urging the employees to vote in favor of representation. Again the materials were enclosed in a clearly-marked Union envelope.
D. The Election and Subsequent Proceedings
On March 8, the day originally scheduled for the vote count, the parties agreed to extend the election for an additional week. The delay resulted in part from the discovery that some of the return envelopes had been stamped with the wrong bar code, which caused those ballots to be routed through an NLRB office in Philadelphia. Supp.Dec. at 18-14; A-187-88. By March 15, the day the ballots were ultimately counted, 72% of the eligible workers had returned ballots to the Board. A final tally of the votes revealed that 560 of the 613 votes the Board received were valid; 43 were void (most of them either for failure to sign the envelope or for bearing extraneous markings); and 10 were [1126]*1126challenged. Id. at 3; A-177. Of the valid ballots, 295 supported the Union and 265 opposed it. Id. The valid ballots represented the views of 66% of the eligible voters. Id. at 13; A-187.
The employer objected to the election, but the Regional Director overruled the objections and certified the Union as the unit’s bargaining representative. Id. at 24-25; A-198-99. The company filed a petition for review which the Board denied. Kwik Care Ltd., Order, 29-RC-8198 (Jan. 6, 1995) (“Order”); A-217a. When the employer, which continued to object to the Union’s certification, refused to bargain with the unit, the Union filed an unfair labor practices charge with the Board. The Board granted summary judgment to the NLRB’s General Counsel, finding that VIP had violated § 8(a)(1) and (5) of the NLRA by refusing to recognize the Union,1 and ordered the employer to bargain with the Union. Final Order; A-297. VIP then petitioned this court for review and the Union cross-applied for enforcement of the Board’s order.
II. Discussion
A. Election by Mail
First, VIP challenges the Board’s initial decision to hold this representation election by mail rather than manually at the VIP offices.
As a general matter, the Board enjoys broad discretion in its administration of representation elections, and the party challenging the Board-certified results of an election carries a heavy burden. See, e.g., C.J. Krehbiel Co. v. NLRB, 844 F.2d 880, 882 (D.C.Cir.1988); J. Ray McDermott & Co., Inc. v. NLRB, 571 F.2d 850, 854 (5 Cir.1978); Amalgamated Clothing Workers v. NLRB, 424 F.2d 818, 827 (D.C.Cir.1970).
In this case, the employer has failed to meet this tough standard of review. Even though postal elections generally inspire lower participation than on-site elections, see Petitioner’s Brief at 25-26 (citing Memorandum from Richard Roth to William Gould (June 23, 1994); A-L55-57), there is no reason to believe that this assumption would prove correct in the particular case at hand. In fact, the Board had every reason to believe that the voter participation rate would have been significantly lower if it had opted for a manual election for VIP’s employees. The same memorandum cited by the employer to tout the advantages of on-site elections also acknowledges that “[m]ail ballots are useful ... [when] there is no central place where' the[ ] employees normally report to.” Id. In other words, the employer’s own evidence suggests that mail ballots are useful in situations like this one, where the voters’ work descriptions support the conclusion that voter participation might be even lower in a manual election.
Similarly, the Board’s internal Casehan-dling Manual, which provides nonbinding guidance for the agency’s staff members, see NLRB Casehandling Manual, Part Two, Representation Proceedings, Purpose of Manual (“Manual”) (“The guidelines included ... are not intended to be and should not be viewed as binding procedural rules.”), can be read to support the Regional Director’s decision in this case. Most importantly, the Manual acknowledges the Director’s discretion on this issue: “The Regional Director should decide whether, or to what extent, mail voting should be employed.” Manual at § 11336. And even though the Manual acknowledges that mail elections, when challenged by one party, should be limited “to those circumstances that clearly indicate the infeasibility of a manual election,” id., the Manual also explicitly states that “[p]articu-larly ... where eligible voters are scattered because of their duties, the possibility [of mail balloting] should be explored.” Id. Thus, in cases like this, the Regional Director was instructed to explore the possibili[1127]*1127ty of mail balloting and exercise his discretion in ordering an election.
In this case, a manual election would have required many of the workers to make a special trip to one of the employer’s offices. A significant number of the employees (including, at the very least, the estimated 14% who work 24r-hour shifts or 12-hour daytime shifts) would have had to coordinate with replacement workers who could assume their duties while they went to vote, and many members of the bargaining unit would have had to spend significant time and effort navigating the public transportation system in order to participate in this election. And even those with duty-free hours or short shifts might well face other circumstances preventing their participation in a manual election. As noted, some of the employees hold second jobs and have “other responsibilities” that might bar them from spending their single off-duty hour traveling to the VIP office to cast a vote.
Instead of ordering a manual election, the Board sensibly opted for a mail-in ballot, which gave every employee the opportunity to cast a vote from home after reviewing the material sent out in the ballot kits. We find that the Board acted within its discretion when it decided to forego a manual election under these particular circumstances. Given that only 72% of the qualified voters exerted the minimal effort required to check a box and return the ballot in a pre-stamped envelope, we are unconvinced by the employer’s argument that a greater percentage of employees would have spent their off-duty hours in time-consuming travel across town in order to vote in a manual election. To be sure, some employees visit the VIP headquarters regularly to pick up paychecks and could have been asked to vote at that time. But there is no evidence that, relative to the mail-in ballots, the possibility of increased participation from those site-visiting employees would offset predictably lower participation by those who never have other cause to cheek in at headquarters.
Therefore, we conclude that the Board’s decision was not unreasonable.
B. Election Complications
VIP has suggested several “problems” with the mail-in election, but we do not believe that any of them warrant our vacating either the results of the election or the Board’s subsequent orders.
1. Ambiguous Instructions
First, VIP argues that the Board’s instructions were unduly confusing, since language on the ballot itself directed voters to drop the vote in a ballot box, even though, in a mad election, there was no box available. The instructions (and checklist) accompanying the ballot, however, advised voters to return the ballot in the enclosed envelope. VIP argues that these conflicting directions caused confusion, which was further exacerbated by the fact that the ballot kits included materials in four different languages.
Responding to VIP’s concern, the Board pointed to four separate instructions in the ballot kit materials advising voters to mail in their ballots, and concluded that these directions provided voters with “ample notice concerning the proper balloting procedure.” Supp.Dec. at 11; A-185. Moreover, the Board took note of the invitation on the instructions for the voters to call the Regional Office with any questions. Thus, even if the ballot initially confused a voter, that voter could still have determined the proper procedure to follow.
There is also no reason the Board should have assumed that the multilingual nature of the voting kit had the potential to confuse any employees, since each voter could reasonably be expected to read the instructions in her language of choice and disregard the extraneous material. In fact, the ballots appeared in four, languages precisely because the parties had all agreed at the outset that the linguistic diversity would provide the best means of making sure that all the workers fully understood the election rules.
2. Invalid Ballots
Second, VIP claims that the large number of invalid ballots (43 “voided” ballots and 11 late ones) demonstrates that the Board’s procedure confused VIP’s employ[1128]*1128ees. Central to this objection is the employer’s assumption that the problems rendering the ballots invalid would have been avoided by holding a manual election. VIP points to 29 ballots which were voided because the voters did not sign the envelopes; 11 which arrived after the March 15 tally date; 14 voided because they had extraneous writing on them; and 5 voided because the voters mailed in stubs instead of ballots.2 Petitioner’s Brief at 22.
The Board determined that the instructions clearly required a signature on the envelope, and that the failure of some people to follow the explicit directions should not invalidate the election results. The Board also noted that the voters had been instructed several times to mail their ballots early enough to guarantee a March 8 arrival date. The 11 late ballots were received more than a week after that deadline. Supp.Dec. at 14; A-188.3 We cannot conclude that the Board acted unreasonably in assuming that an employee who failed to make the effort to return a pre-stamped envelope within the month-long election period might well have also failed to make the sacrifice of time, energy and money that a manual vote would have required.
The employer claims that the 14 mis-marked ballots would have been successfully cast if the election had been a manual one, since the voters would immediately have been able to ask the Board agent supervising the election to replace the ballots with new ones. Petitioner’s Brief at 22-23. First, we note that mismarked ballots are not unique to mail-in elections. See, e.g., NLRB v. Precise Castings, Inc., 915 F.2d 1160, 1163 (7th Cir.1990), cert. denied, 499 U.S. 959, 111 S.Ct. 1581, 113 L.Ed.2d 646 (1991) (discussing an on-site election, and noting that “Ballots are not always read. The number of spoiled or invalid ballots attests to this.”). More importantly, however, the voters who mismarked their ballots most likely did not know that they had erred. Had they been aware of their mistakes, they could have called the regional director and requested new ballots, just as they would have requested new ballots from an on-site election supervisor in a manual election.
In sum, although this election was far from smooth, see supra at 1127-1128 & n. 3, the Board’s ruling that the improperly cast ballots should not invalidate this election cannot be held to be unreasonable.
C. The Sample Ballot
The parties agree that the Union’s distribution of altered ballots should invalidate the election only if “the document ha[d] the tendency to mislead employees into believing that the Board favors one party’s cause.” SDC Investment, Inc., 274 N.L.R.B. 556, 557, 1985 WL 45800 (1985) (“SDC”). The Union argues that the Board’s recent decision in Brookville Healthcare Center, 312 N.L.R.B. 594, 1993 WL 391269 (1993) (“Brookville”), establishes that, as a matter of law, altered ballots do not have the tendency to mislead voters, since the Board’s revised Notice of Election form bears the disclaimer quoted above, supra, at 1124, proclaiming Board neutrality in all elections. VIP argues that Brookville does not govern the question presented in this case, and cites two opinions the employer claims mandate reversal of the Board’s determination that the sample ballot constituted legitimate propaganda.
1. The SDC/3-Day Blinds Test
For years, the Board has used a two-part analysis for evaluating the propriety of an altered ballot distributed by a party involved in a scheduled election. First, under this test, if the face of the sample ballot clearly identifies the source of any alterations or partisan comments, then the use of that ballot as propaganda is justified. For example, in this case, VIP had itself distributed a sample ballot with an X marked in the “no” box, to show a vote against the Union, but since the employer’s distribution, unlike the [1129]*1129Union’s, clearly stated on its face that it had been “Prepared by VIP,” the ballot passed the first prong of the SDC test and has never been challenged.
Second, if the sample ballot does not reveal on its face the source of any alterations, then the Board will examine “the nature and contents of the document,” as well as “the circumstances of [its] distribution,” 3-Day Blinds, 299 N.L.R.B. 110, 1990 WL 122544 (1990), to determine “whether it was likely to give employees the misleading impression that the Board favored [one side] in the election.” Id.
In 3-Day Blinds, the employer had distributed by hand a photocopy of the Board’s official sample ballot, marked with a vote in favor of the employer. The altered ballot was printed on the same light green paper used for official Board-produced election documents. At the bottom of the page, the employer had written “THIS IS HOW TO MARK YOUR BALLOT TO GIVE THE NEW OWNERS A CHANCE.” Agents of the employer had distributed the ballot in person, at which time the agents identified themselves as the new owners and urged the employees to vote “no” in the upcoming election. In addition to the altered ballot, the employer distributed an unattached form guaranteeing the workers that they would be free from retaliation based on their vote. This “guarantee handbill,” unlike the ballot, was clearly marked as having been produced by the employer.
The Board in 3-Day Blinds determined that the ballot’s facial reference to the “new owners” failed sufficiently to identify it as the employer’s partisan propaganda. Because the sample ballot did not identify its origin on its face, the Board moved on to the second prong of the analysis and concluded that the circumstances surrounding its distribution had the potential to mislead voters into thinking the Board, rather than the employer, might have issued the ballot. The Board noted that the form had been “handed out separately and distinctly” from any other material identifying the employer as its source; that it looked like an official document; and that the contrast between the ballot and the clearly-partisan guarantee handbill had the potential to contribute to the perception that the ballot had been produced by a different source. Id. at 111-12. Even though the employer had distributed the altered ballots, “distribution of a document by a party does not necessarily mean that the party, rather than the board, prepared the document.” Id. at 112.
VIP argues that the circumstances in this case make the Union’s ballot even more misleading than the employer’s in 3-Day Blinds: for example, the Union’s ballot here bore no markings suggesting its source, whereas in 3-Day Blinds the form urged readers to give “the New Owners” a chance. Additionally, VIP claims that the Union’s altered ballot had the potential to mislead voters because it was not physically attached to the partisan letter (at least in the first mailing) and because the partisan letter did not explicitly refer to the altered ballot. Despite certain similarities to 3-Day Blinds, the Board determined that the Union’s ballot in this ease would not tend to mislead voters. Even though it might not have been stapled to the Union letter, it was enclosed in the same envelope with clearly partisan propaganda written on Union letterhead, and the envelope itself was marked with the name of the Union. Supp.Dec. at 21-23; A-195-97. Moreover, the altered ballot is white whereas the official ballot is yellow, and the Board seals do not appear on the facsimile. Compare SDC, supra, 274 N.L.R.B. at 558 (“[T]he presence of the Board seal at the top and the official Board sample ballot on the reverse side did much to make the document appear to be official.”). Although the Union’s failure to identify itself as the author of the document gives us pause, we cannot conclude that the Board acted unreasonably when it determined that voters would understand that the sample ballot had been altered by the Union.
2. The Brookville Test
In its order denying VIP’s request for review of the decision to certify the results of the election, the Board noted an additional rationale for the conclusion that the Union’s sample ballot would not mislead voters. Citing Brookville, supra, 312 N.L.R.B. 594, the [1130]*1130Board found that “the warning language concerning the Board’s neutrality and the unauthorized marking of sample ballots ... makes it extremely unlikely that the Petitioner’s facsimile ballot gave employees the impression that the Board endorsed the Petitioner in this election.” Order; A-217a. Although VIP correctly argues that the situation at hand is distinguishable from that in Brookville, the difference does not invalidate the Board’s reliance on the Brookville rationale as support for its decision here.
The partisan alteration at issue in Brook-ville had been made on the sample ballot that appeared on a copy of the Board’s then-newly-revised Notice of Election, containing the firm declaration of Board neutrality quoted supra, at 1124. Here, by contrast, the alteration appeared on a sample ballot sent out to voters under separate Union cover. Thus, the holding in Brookville does not directly apply to this case; the Board there held that “in all cases involving the defacement of a revised notice, the SDC Investment analysis no longer is required,” since the new warning on the Notice is sufficient “to preclude a reasonable impression that the Board favors or endorses” one side in the election. 312 N.L.R.B. at 594.
Although the earlier case is not directly on point, it was not unreasonable for the Board here to rely on Brookville to support the assumption that participants in the VIP mail election would have read the Notice — including the statement that “ANY MARKINGS THAT YOU MAY SEE ON ANY SAMPLE BALLOT ... HAVE BEEN MADE BY SOMEONE OTHER THAN THE [BOARD],” A-149 (emphasis added) — and would therefore have known that the altered ballot constituted the Union’s partisan propaganda. This court need not decide if the Brookville analysis by itself would support the Board’s finding, since the Board only relied on Brookville to bolster its reasonable determination that the Union’s mailing satisfied the SDC/3-Day Blinds standard.
For the foregoing reasons, the employer’s petition for review is denied and the Board’s cross-application for enforcement of the Board’s order is granted.
So ordered.