Judge, concurring:
Although I fully agree with the majority that the Board decision should be vacated and the matter remanded for consideration of the June 26, 2000, VA medical report as “evidence in the file” at the time of the death of the veteran, Frederick C. Castellano, I write separately to address several *162concerns, including my disagreement with the majority’s attempt, although through dicta, to deem Hayes v. Brown, 4 Vet.App. 353 (1993) overruled.
A. Evidence in the File at Date of Death
Pursuant to 38 U.S.C. § 5121(a), a veteran’s surviving spouse may receive accrued benefits consisting of “periodic monetary benefits ... to which [the veteran] was entitled at death under existing ratings or decisions, or those based on evidence in the file at date of death ... and due and unpaid.” 38 U.S.C. § 5121(a) (emphasis added). In Hayes, the Court noted that although section 5121(a) permits only evidence in the file at date of death, section 5121(c) “appear[ed] to contradict, or at least qualify, that provision” by permitting a claimant to submit information to complete the application. 4 Vet.App. at 360. Given this ambiguity, the Court stated that “the Secretary ha[d] wide latitude in establishing department policy as to what postdate-of-death evidence may be considered.” Id. Noting that the regulatory framework established to implement section 5121(a), (c) was “confusing at best,” see 38 C.F.R. § 3.1000(d)(4)(i) (1992),9 the Court nevertheless stated that it was “clear” that § 3.1000(d)(4), “on its face, applie[d] only to evidence in the file at date of death and the death certificate and, d[id] not, by its terms contemplate the submission of any post-death evidence.”10 Id. at 358. However, the Court also recognized that “[s]uperimposed on this regulation ... [wa]s a confusing array of provisions of the VA Manual, M21-1 ... scattered amongst two chapters, regarding what post-date-of-death evidence [wa]s acceptable.” Id. at 360 (citing VA Manual, M21-1, paras. 5.25, 27.08). The Court held that “[t]o the extent that these Manual provisions affect what post-date-of-death evidence may be considered, they have ‘ “the force of law” ... [as they affect] a substantive right [of appellant to accrued benefits,] and ... placement in a procedural [M]anual cannot disguise [the] true nature [of these provisions] as ... substantive rule[s].’ ” Id. (quoting Fugere v. Derwinski, 1 Vet.App. 103, 107 (1990)).
What is clear from the Court’s decision in Hayes, is that both 1992 M21-1 provisions discussed by the majority, M21-1, part VI, para. 5.25(a), (b) (1992), were found to be “substantive rules,” creating substantive rights, because they affected what post-date-of-death evidence could be considered under § 5121(a). Thus, absent adherence to the APA notice-and-comment process and specific notice to the public of intent to revoke these substantive provisions, neither provision could be overturned. Fugere, 1 Vet.App. at 109-10 (setting aside the Secretary’s rescission of an M21-1 provision for failing to follow the *163APA notice-and-comment process), aff'd, 972 F.2d 331 (Fed.Cir.1992).
As relevant to this appeal, paragraph 5.25(a) provided that “[ejvidence ‘in file’ includes ... [‘[rjeports of hospitalization, treatment or examinations authorized by VA’], even if such reports are not reduced to writing or are not physically placed in file until after death.” M21-1, pt. VI, para. 5.25(a) (1992). Paragraph 5.25(b), entitled “[e]vidence [essentially Complete,” provided that
[t]he provisions of 38 C.F.R. § 3.1000(d)(4) permit favorable action under 38 U.S.C. § 5121 on the basis of the evidence as a whole when the evidence is essentially complete and in file at the time of death.... The cited regulations also provide for the acceptance of evidence after death for verifying or corroborating evidence “in file” at death.
M21-1, pt. VI, para. 5.25(b) (1992). Because paragraph 5.25(b) explicitly referenced the Secretary’s regulation, and paragraph 5.25(a) did not, my colleagues state that paragraph 5.25(a) does not “purportf ] to be interpreting a regulation.” See section III.A.2.b, ante at 154. In addition, because they perceive paragraph 5.25(b) to directly conflict with the Secretary’s pre2002 regulation, they conclude that the Secretary was permitted to amend the M21-1 provision without adhering to the notice and comment process. They assert that since Hayes issued, “it has been clarified that a properly promulgated regulation trumps an M21-1 provision or other VA directive that plainly is erroneous or inconsistent with regulation.” Section III. A.2.b, ante at 154 (citing Smith v. Shinseki, 647 F.3d 1380 (Fed.Cir.2011) and Haas v. Peake, 525 F.3d 1168, 1197 (Fed.Cir.2008)).
I disagree with the majority’s classification of the two Manual M21-1 paragraphs because it is clear to me that both paragraph 5.25(a) and paragraph 5.25(b) reflect the Secretary’s determination regarding what post-date-of-death evidence is to be considered “evidence in the file.”11 That paragraph 5.25(b) mentioned the regulation, and paragraph 5.25(a) did not, does not alter the fact that both paragraphs represented the Secretary’s interpretation of 38 U.S.C. § 5121 and the evidentiary requirements for an accrued benefits claim, which the Secretary has conceded he had failed to establish by regulation until the 2002 regulatory change. The Secretary stated in his 2002 notice of proposed rulemaking that the then-current version of 38 C.F.R. § 3.1000(d)(4) failed to define the term “ ‘evidence in the file’ ” but instead permitted VA “in certain instances ... [to] accept identifying, corroborating[,] or verifying information from the death certificate and evidence submitted with the claim for accrued benefits to support prima facie evidence already in the file.” See Evidence for Accrued Benefits, 67 Fed.Reg. 9638, 9639 (Mar. 4, 2002) (emphasis added). Moreover, the Federal Circuit’s recent discussion of Hayes in Hyatt v. Shinseki recognized this Court’s “perceived conflict between subsections (a) and (c) of § 5121” and that VA clarified this “confusion” by regulation in 2002. 566 F.3d 1364, 1370 (Fed.Cir.2009) (citing Evidence for Accrued Benefits, 67 Fed.Reg. 9638, 9639 (Mar. 4, 2002); Evidence for Accrued Benefits, 67 Fed.Reg. 65,707 (Oct. *16428, 2002)). Thus, I believe that until 2002, when the Secretary formally published his intent to delete from the M21-1 those provisions that were inconsistent with his proposed definition of “evidence in the file at the date of death,” both paragraph 5.25(a) and paragraph 5.25(b) remained valid substantive rules, which were applicable to the appellant’s claim for accrued benefits filed in June 2000.
I also disagree with the majority’s reliance on Smith, Haas, and Fournier v. Shinseki, 23 Vet.App. 480, 487-88 (2010), for the proposition that “a properly promulgated regulation trumps an M21-1 provision ... that plainly is erroneous or inconsistent with regulation.” See section III.A.1, ante at 151. Although it is true that the Federal Circuit concluded that Mr. Haas could not rely on a 1991 M21-1 provision that was inconsistent with the Secretary’s regulation — Haas is distinguishable because the Secretary’s regulation, which was promulgated after a notice- and-comment process, went into effect in July 2001, one month before Mr. Haas filed his claim for disability compensation benefits. 525 F.3d at 1196. Because the Secretary had “formally taken a position that excluded Mr. Haas from the scope of the regulation” before his claim was filed, the Federal Circuit concluded that (1) Mr. Haas could not rely on the 1991 M21-1 provision, which was inconsistent with the later regulation, and (2) it was unnecessary for the Agency to conduct parallel rule-making to amend the Agency’s “interpretative statement” contained in the M21-1. Id. at 1196-97.
Unlike the regulation in Haas, which was formally promulgated after the M21-1 provision had already been in effect, the pre-2002 version of 38 C.F.R. § 3.1000(d)(4) had been in effect since at least 1970, thus pre-dating the M21-1 provision at issue here, and, as acknowledged by the Secretary in his 2002 notice of proposed rulemaking, failed to give meaning to “evidence in the file at the date of death” and the ambiguity created by 38 U.S.C. § 5121(a), (c). See 38 C.F.R. § 3.1000(d)(4) (1970); 67 Fed.Reg. at 9639. As discussed earlier, the Court in Hayes recognized the ambiguity and confusion surrounding what constituted “evidence in the file at date of death” and held that, to the extent paragraph 5.25 “affected what post-date-of-death evidence may be considered” it had the “ ‘force of law’ ” and created a “substantive right.” Hayes, 4 Vet.App. at 360. In this circumstance, VA’s regulation, which had been effect since at least 1970, cannot “trump,” or otherwise invalidate, a subsequently created “substantive right.”
The majority’s reliance on Smith and Fournier, both supra, is similarly misplaced because neither case involved a Manual M21-1 provision that created a substantive right. See Smith, 647 F.3d at 1384-85 (noting that VA’s interpretative statement contained in its adjudication Manual was consistent with VA’s TDIU regulation, which did not require the agency to consider the availability of work when determining entitlement to TDIU); Fournier, 23 Vet.App. at 487 (holding that the Manual M21-1 provision at issue did “not establish or alter the criteria for benefits but only illuminates a suggested procedural practice for VA adjudicators”). In Fournier, the Court also discussed the Federal Circuit’s decision in Haas as indicating that “when an Agency’s interpretation of a statute is dear from existing regulations, any discrepancy between M21-1 and the regulation ‘does not confer any rights’ on a claimant.” 23 Vet.App. at 487 (2010) (emphasis added). As noted above, prior to 2002, the Secretary’s interpretation of his statutory authority was anything but clear. See Hyatt, 566 F.3d at 1370; cf. Hayes, 4 Vet.App. at 361 (stating *165that “VA, at the earliest possible opportunity, may wish to clarify its policy as to what evidence submitted after death may be considered”); 67 Fed.Reg. at 9639.
For these reasons, I cannot agree with my colleagues’ conclusion that the May 8, 2000, amendments to paragraph 5.25(b) were properly effectuated without adherence to the APA notice-and-comment process. Rather, I would conclude that the 1992 M21-1 provisions, specifically paragraph 5.25(a) and paragraph 5.25(b), remained valid substantive rules in accordance with the Court’s decision in Hayes until 2002 when VA formally rescinded its Manual M21-1 provisions as “inconsistent” with the 2002 amendment to 38 C.F.R. § 3.1000(d)(4). See 67 Fed.Reg. at 9640, 65,707.
In addition to noting my fundamental disagreement with the majority’s analysis on this point, I would be remiss if I neglected to point out that the majority’s discussion concerning paragraph 5.25(b) is not necessary to its decision and, therefore, in my opinion, amounts to nothing more than dicta. See Black’s Law Dictionary 465 (7th ed. 1999) (defining “judicial dicta” as “[a]n opinion by a court on a question that is directly involved, briefed, and argued by counsel, and even passed on by the court, but that is not essential to the decision”); see also Pelegrini v. Principi, 18 Vet.App. 112, 130 (2004) (Ivers, J., concurring in part and dissenting in part) (noting that “ ‘[t]he danger of ... dicta is that, although theoretically and technically not binding, practically, they give the appearance of carrying the cloak of judicial acceptance’ ” (quoting Lasovick v. Brown, 6 Vet.App. 141, 153 (1994))). Simply put, once my colleagues concluded that the Board clearly erred when it determined that the June 2000 report did not constitute a “VA-authorized report of treatment” such that it was error not to deem the report “evidence in the file at date of death,” it was unnecessary to consider, only to reject, the appellant’s alternative argument that the report should be considered because it serves to “summarize, corroborate[,] and verify evidence in file at the date of his death.” Appellant’s Brief (Br.) at 11-13 (citing Hayes, 4 Vet.App. at 360-61 and noting that the Board failed to address this argument). Because discussion of paragraph 5.25(b) was not essential to the Court’s disposition, it is axiomatic that the majority’s “holding” that Smith and Haas overturned Hayes is not binding.
To the extent that the majority relies on its analysis of the Secretary’s May 2000 change to paragraph 5.25(b), see M21-1, pt. VI, para. 5.06(d) (Change 73-May 8, 2000) (stating that “[t]he cited regulation provides for the acceptance of evidence ... when considered in connection with the identifying, verifying, or corroborative effect of the death certificate”), to remand the question whether the veteran’s private nursing home records may be deemed evidence in the file at the date of death pursuant to paragraph 5.06(d), I also note that this argument was not raised by the parties. Instead, the appellant argued that the nursing home records were “treatment records created prior to the veteran’s death and it is entirely unclear why such relevant treatment records were not considered part of the veteran’s file.” Appellant’s Br. at 12 n. 6 (citing Hayes, 4 Vet.App. at 360-61, for the proposition that “hospitalization and treatment records may be considered constructively part of the veteran’s file on the date of death for the purposes of adjudicating an accrued benefits claims even if not placed in the veteran’s file prior to death”). The Board’s reasons were unclear because the Board summarily stated that “records from Silver Ridge Village outside the period for which VA authorized and contract*166edfor treatment at that facility ... cannot be considered in the context of the ... appeal.” R. at 8 (emphasis added). I do not believe that the appellant’s “reasons or bases” allegation of error reasonably raised any alternative basis for consideration of the records outside the authorized period, and as argued by the Secretary, the appellant has not identified any record that the Board purportedly failed to consider that weighs in favor of her claim. Secretary’s Br. at 16-17. The majority’s discussion in this regard can only be viewed as a misguided attempt to render its discussion of paragraphs 5.25(b) (1992) and 5.06(d) (2000) relevant and essential to the Court’s disposition in order to reach a desired result — deeming Hayes overruled.
B. The Majority’s Review of the Board’s Findings
As noted above, I agree that the June 26, 2000, report should be deemed evidence in the file at the time of the veteran’s death and that this matter should be remanded to the Board for consideration of the report. I also agree that the basis for admitting the report is, as conceded by the parties, that it constitutes a VA-authorized report of treatment. However, I write separately on this point to express my concern regarding the majority’s broad interpretation of the Secretary’s M21-1 provision, which permitted “[r]eports of hospitalization, treatment!,] or examinations authorized by VA” to be considered “evidence ‘in file’ ... even if such reports are not reduced to writing or are not physically placed in file until after death.” M21-1, pt. VI, para. 5.25(a) (1992); see also M21-1, pt. VI, para. 5.06(c) (Change 73-May 8, 2000). My concern is that my colleagues’ conclusion that a VA-authorized report of treatment may be deemed evidence in the file, “no matter what form it took at the moment of the veteran’s death,” section III.A.3.a.(l), ante at 156, may suffer from the law of unintended consequences. That is, although it may not seem unreasonable, under the facts of this case, to deem a VA report of treatment that is reduced to writing five weeks after the veteran’s death evidence in the file, the majority’s holding appears to permit a VA report of treatment, that is requested and written many years after a veteran’s death, to be deemed evidence in the file at the date of death. Such a broad interpretation of the Secretary’s Manual M21-1 may have the unintended effect of opening Pandora’s box, possibly leading to a development of evidence in an accrued benefits claim that is not contemplated by statute. My concerns are augmented by the fact that the Secretary’s current Manual M21-1 contains a similar provision that requires VA reports of hospitalization, examination, or treatment to be placed in the claims file “even if they are not reduced to writing ... until after death,” and therefore, the majority’s holding may not be limited only to those cases that have been pending prior to the Secretary’s 2002 amendment to 38 C.F.R. § 3.1000(d) and rescission of paragraph 5.06(c). See M211MR, pt. VIII, chs. 3, l.f (2010).
Rather than broadly construe the Secretary’s Manual M21-1 provision, I would have reviewed the Board’s factual findings concerning when the report was requested and prepared and concluded that the Board clearly erred when it failed to afford the appellant the benefit of the doubt. See 38 U.S.C. § 5107(b) (“When there is an approximate balance of positive and negative evidence regarding any issue material to the determination of a matter, the Secretary shall give the benefit of the doubt to the claimant.”); Mariano v. Principi, 17 Vet.App. 305, 313 (2003) (the Court reviews the Board’s application of the benefit of the doubt rule under the “clearly erroneous” standard); 38 C.F.R. § 3.102 (2011) (“When, after careful consideration of all *167procurable and assembled data, a reasonable doubt arises regarding service origin, the degree of disability, or any other point, such doubt will be resolved in favor of the claimant.”). In ■ short, I agree with the appellant that the Board’s conclusions concerning the preparation of the report are highly speculative and are not based on any affirmative evidence in the record.
In its decision, the Board found that the June 2000 treatment report was not requested or prepared prior to the veteran’s death and, therefore, concluded that it could not be considered as part of the accrued benefits claim. The Board observed that the veteran died on May 18, 2000, and that the report, dated June 26, 2000, indicated that it was prepared at the request of the veteran’s wife and veteran’s service representative (VSR). The Board also acknowledged that the report was written in the present tense, which indicated that the physician thought the veteran was still alive. The Board stated that there was no evidence to indicate when the psychiatrist prepared the report, but expounded that assuming the veteran’s wife requested the report, it was “highly unlikely ” that the psychiatrist would have been asked to prepare the report prior to the veteran’s death and have waited more than five weeks to reduce it to writing. R. at 10. Thus, the Board concluded that the benefit of the doubt was “not implicate[d]” in the absence of any evidence that the psychiatrist prepared the report before the veteran died. Id.
To the extent the VSR may have requested the report, the Board stated that “it is reasonable to assume that the VSR would have delayed the [June 15, 2000, hearing] until the psychiatrist’s report was received,” had it been requested before the veteran’s death. Id. (emphasis added). Based on the foregoing, the Board found that “the VA psychiatrist’s June 2000 report was not requested until after the representative’s Informal Hearing Presentation, dated June 15, 2000 — almost one month after the Veteran had died — and so, a fortiori, the psychiatrist’s report could not have been prepared prior to his death.” R. at 11.
“The Board’s task [is] to make findings based on evidence of record — not to supply missing facts.” Beaty v. Brown, 6 Vet.App. 582, 536-37 (1994) (holding that the Board’s finding that “ ‘with eye protection, these occupations [“farmwork and heavy equipment operation”] would still be feasible’ ha[d] no evidentiary basis in the record” and was clearly erroneous); see also Tucker v. West, 11 Vet.App. 369, 374 (1998) (holding that the Board clearly erred when it relied on its own medical speculation and noting that “[b]are assertions without identifying the underlying factual basis for the conclusion are not sufficient”). In this case, what is clear from the face of the report is that it is dated June 26, 2000, five weeks after the veteran’s death, and that the veteran’s wife and VSR requested the report. However, the fact that the report postdates the veteran’s death is not dispositive because the Secretary’s Manual M211 provision clearly contemplates that evidence may be deemed in the file even though it is not reduced to writing until after death. Thus, as conceded by the Board, it is “possible” that the report was “prepared” prior to the veteran’s death, but not reduced to writing until after death. R. at 10. This possibility is bolstered by the fact that the VA doctor repeatedly referred to the veteran in the present tense, strongly suggesting that the physician believed that the veteran was alive when the report was prepared. See, e.g., R. at 333 (“The patient’s current GAF (Global Assessment of Functioning) is scored as 25.”) and (“The patient is totally and permanently disabled and now requires total nursing home care for supervi*168sion of his basic activities for daily living. He may benefit from the use of medications to control his emotional outbursts, confusion and disorientation. His prognosis is guarded.”) (emphasis added). On the other hand, there is no factual basis in the record to support the Board’s conjecture that it is “highly unlikely” that it would take the physician more than five weeks to reduce a report to writing nor is there any basis for the Board’s assumptions as to the VSR’s thought process in preparing for and submitting an informal hearing presentation on June 15, 2000.
Once we remove the Board’s speculative determinations concerning the timing of the preparation of the report and consider the evidence in the record, it is clear that the Board erred when it found that the benefit of the doubt was “not implicated.” This is precisely the scenario in which the determination as to when the report was prepared is “too close to call” and the benefit of the doubt should have been afforded to the appellant. See generally Ortiz v. Principi, 274 F.3d 1361, 1365 (Fed.Cir.2001) (“[T]he benefit of the doubt rule may be viewed as shifting the ‘risk of nonpersuasion’ onto the VA to prove that the veteran is not entitled to benefits.”). Thus, I would have concluded that the Board clearly erred when it failed to afford the appellant the benefit of the doubt to conclude that the VA psychiatrist prepared the report prior to the veteran’s death and, therefore, the report should be deemed “evidence in the file” even though it was not reduced to writing until after the veteran’s death.
C. Retroactive Application of 38 C.F.R. § 3.1000(d)(4)
The majority holds that the Board’s application of § 3.1000(d)(4) was erroneous “because the regulation was not intended to be retroactive and its retroactive application has an impermissible retroactive effect.” Section III.A.3.a.2, ante at 156. Although I generally agree that the Secretary’s 2002 amendment to § 3.1000 was not intended to apply retroactively and that any such application would have an impermissible retroactive effect, I do not agree with the majority’s analysis of the Princess Cruises factors. See section III. A.3.a.2, ante at 155-59; Princess Cruises, Inc. v. United States, 397 F.3d 1358, 1362-63 (Fed.Cir.2005). I also do not believe that my colleagues’ Princess Cruises analysis is necessary to the Court’s disposition because the Secretary has not maintained that the regulation applies retroactively. See generally Evidence for Accrued Benefits, 67 Fed.Reg. 65,707 (Oct. 28, 2002) (indicating that the Secretary’s amendment to § 3.1000 is effective “November 27, 2002”); Secretary’s Br. at 14-15 (citing 38 C.F.R. § 3.1000 (1999) as the applicable version of the regulation). Assuming the Board applied the current version of the regulation, I agree that such application was erroneous.
Accordingly, for the foregoing reasons, I concur in the result reached by the Court, but must respectfully disagree with those portions of the majority’s opinion discussed above. The Board’s decision should be vacated and the matter remanded for the Board to consider the June 26, 2000, VA medical report.
. In 1992, the Secretary's regulation, provided as follows:
(4) Evidence in the file at date of death ... will be considered to have been met when there is on file at the date of the veteran’s death:
(i) Notwithstanding § 3.200(b) evidence, including uncertified statements, which is essentially complete and of such weight as to establish service connection or degree of disability for disease or injury when substantiated by other evidence in file at date
of death or when considered in connection with the identifying, verifying, or corroborative effect of the death certificate.
38 C.F.R. § 3.1000(d)(4)(i) (1992).
. Contrary to the majority’s view, the Court in Hayes explicitly stated that "the exceptions provided for by ¶ 5.25(a), (b) are not set forth in ... § 3.1000(d)(4)” and, therefore, did consider whether para. 5.25(a) and (b) could stand in direct contravention of the regulation. Id. at 360; see section III.A.2.b, ante at 154.
. Indeed the plain language of paragraph 5.25(a) reveals that it is the Secretary's interpretation of what constitutes "evidence in the file.” It states: "Evidence in File at Date of Death. Evidence 'in file’ includes the following....” M21-1, pt. VI, para. 5.25(a) (emphasis added); see also M21-1, pt. VI, para. 5.23 (1992) (stating that "[e]ntitlment to [accrued benefits] must be based on evidence in file at date of death” and referencing para. 5.25(a)).