Defendant Graves was charged with first-degree murder on the basis of the fatal shooting of Frank Stephens. At the close of the prosecution’s case, defense counsel moved for a directed verdict of acquittal regarding the first-degree murder charge. The trial court denied the motion and instructed the jury that it could find defendant guilty of first-degree murder, second-degree murder, or voluntary manslaughter, or not guilty. The jury returned a verdict of voluntary manslaughter (thereby acquitting defendant of the first- and second-degree murder charges).

Defendant argued in his appeal to the Court of Appeals that he was entitled to a new trial because the trial court had erred in submitting the first-degree murder charge to the jury. Defendant’s argument was *479premised on the following statement from People v Vail, 393 Mich 460, 464; 227 NW2d 535 (1975):

[W]here a jury is permitted consideration of a charge unwarranted by the proofs there is always prejudice because a defendant’s chances of acquittal on any valid charge is substantially decreased by the possibility of a compromise verdict. For this reason it is reversible error for a trial judge to refuse a directed verdict of acquittal on any charge where the prosecution has failed to present evidence from which the jury could find all elements of the crime charged. [Emphasis added.]

The Court of Appeals agreed with defendant’s contention that there had been insufficient evidence of premeditation and deliberation to support the first-degree murder charge and therefore held that Vail required defendant be granted a new trial on the manslaughter charge. 224 Mich App 676, 681; 569 NW2d 911 (1997).

However, Judges Griffin and Bandstra urged this Court to overrule Vail.1 Judges Griffin and Bandstra stated that if they were not bound by Vail, they would join the majority of other jurisdictions in rejecting the harsh and unrealistic position that actual prejudice may be presumed by the mere possibility of a compromise verdict. Id. at 679. We granted the prosecution’s application for leave to appeal to consider whether we should overrule Vail. 456 Mich 905 (1997).2

*480Defendant argues that we should not overrule Vail because it has been a part of this state’s jurisprudence since People v Stahl, 234 Mich 569, 572; 208 NW 685 (1926). While the Vail rule has been longstanding, it has not always been the rule in Michigan, and the rule has not been uniformly supported. See, e.g., People v Robinson, 228 Mich 64, 72; 199 NW 622 (1924), which stated that if the court erroneously instructed the jury regarding murder, the manslaughter verdict showed that the error was without prejudice. More recently, this Court implicitly criticized the Vail rule as being unduly speculative in People v Johnson, 427 Mich 98, 116, n 15; 398 NW2d 219 (1986).

In any event, the standards we apply in deciding whether to overrule a case do not rest upon the length of time that the rule has been in effect. It is true of course that we do not lightly overrule a case. This Court has stated on many occasions that “[u]nder the doctrine of stare decisis, principles of law deliberately examined and decided by a court of competent jurisdiction should not be lightly departed.” People v Jamieson, 436 Mich 61, 79; 461 NW2d 884 (1990). 3 Further, as aptly stated in McEvoy *481v City of Sault Ste Marie, 136 Mich 172, 178; 98 NW 1006 (1904):

Before this court overrules a decision deliberately made, it should be convinced not merely that the case was wrongly decided, but also that less injury will result from overruling than from following it.

When it becomes apparent that the reasoning of an opinion is erroneous, and that less mischief will result from overruling the case rather than following it, it becomes the duty of the court to correct it. Attorney General ex rel Barnes v Midland Co Bd of Supervisors, 178 Mich 513, 518; 144 NW 883 (1914). Although we respect the principle of stare decisis, we also recognize the common wisdom that the rule of stare decisis is not an inexorable command. With that principle in mind, we believe that our reexamination of the so-called Vail rule, and the subsequent decision to overrule it, is proper because it is informed by a judgment premised on prudential and pragmatic considerations.

We first note that Vail is a rule of automatic reversal. As this Court recently reiterated in People v Belanger, 454 Mich 571, 575; 563 NW2d 665 (1997), “[r]ules of automatic reversal are disfavored, for a host of obvious reasons.”4 Indeed, the automatic *482reversal rule of Vail is inconsistent with this Court’s modem harmless-error jurisprudence.

As explained in People v Grant, 445 Mich 535; 520 NW2d 123 (1994), and People v Mateo, 453 Mich 203; 551 NW2d 891 (1996), errors are to be classified as constitutional or nonconstitutional and preserved or unpreserved.

If the error is constitutional, it must be further classified as structural or nonstructural. If the constitutional error is structural, reversal is automatic. People v Anderson (After Remand), 446 Mich 392, 404-405; 521 NW2d 538 (1994). However, if the constitutional error is not a structural error, a defendant’s conviction should be affirmed if the reviewing court is satisfied that the error is harmless beyond a reasonable doubt. Id. at 406.

If the error is a preserved error, but not constitutional in its nature, the Court must determine whether the error was harmless. This Court grappled in Mateo and People v Gearns, 457 Mich 170; 577 NW2d 422 (1998), with the proper test to apply in such circumstances. In Mateo, six members of this Court agreed that some level of assurance less than the “harmless beyond a reasonable doubt” standard should be utilized in reviewing preserved nonconstitutional error.5 Id. at 216. In Geams, four justices indicated a preference for the “highly probable” test, i.e., they will not reverse a conviction if it is highly probable that the *483nonconstitutional error did not affect the judgment. Id6

Thus, depending on the nature of the error, it is appropriate to (1) automatically reverse, (2) reverse unless the error is harmless beyond a reasonable doubt, or (3) (pursuant to the preference of four justices in Geams) reverse unless it is highly probable that the error did not affect the verdict.

Every member of this Court is now on record as holding that preserved nonconstitutional error, such as occurred here, should be reviewed under a lesser standard than the harmless beyond a reasonable doubt standard.7 8Obviously, when Mateo rejected the beyond a reasonable doubt standard for review of preserved nonconstitutional errors, it implicitly rejected the much higher automatic reversal standard. Thus our overruling of Vail is consistent with the post-Vaii developments in this Court’s harmless-error jurisprudence, i.e., the automatic reversal rule of Vail is inconsistent with Mateo and Geams.6

*484We also find that Vail is flawed in that it does not discuss or even acknowledge MCL 769.26; MSA 28.1096, the statutory statement of harmless error which establishes that a defendant should not be granted a new trial on the ground of misdirection of the jury unless it appears after examination of the entire record that the error resulted in a miscarriage of justice. We have applied this statute in a number of contexts in recent years. See People v Hall, 435 Mich 599, 612, n 10; 460 NW2d 520 (1990).9

Moreover, Michigan is one of only a handful of states that requires reversal under the circumstances of this case. Anno: Modem status of law regarding cure of error, in instruction as to one offense, by conviction of higher or lesser offense, 15 ALR4th 118, part III, §§ 21-24, pp 164-177. The vast majority of states apply a harmless-error analysis. Id.10

*485We find the reasoning supporting the majority position persuasive. First, as Judge Griffin’s excellent opinion points out, Vail is premised on the unwarranted assumption that jurors do not follow their instructions not to compromise their views. 224 Mich App 679-680. Moreover, an examination of the legal underpinnings of Vail reveals that the stem judicial response of automatic reversal is based on a supposition that can only be equated with rough guesswork regarding what happened in the jury room. The jurors in the case at bar received the standard instruction not to compromise their views (“[N]one of you should give up your honest opinion about the case just because other jurors disagree with you or just for the sake of reaching a verdict”). The court gave further instructions after a period of deliberation wherein the court told the jury “the verdict to which you agree must of course be your own verdict as a result of your own convictions and not on mere acquiescence and the conclusion of your fellow jurors . . . .” It is also the case that the jurors were polled after rendering their verdict and each juror affirmed a verdict of manslaughter.

The idea that courts of review should automatically undertake a reconstmction of the jury’s deliberative process anytime the jury is presented with an erroneous instruction, even where there is no record evidence of jury compromise, is flawed both in principle and in practice. Automatic reversal represents, at best, a rough guess at the psychological dynamic at work during jury deliberations, guesswork that seemingly concludes that at least one juror argued for acquittal on all charges, while another juror argued for conviction on the erroneous charge, and those *486arguments were not fairly heard, fairly considered, and reasonably rejected by other jurors. Such a conclusion is unduly speculative.

It is well established that jurors are presumed to follow their instructions. People v Hana, 447 Mich 325, 351; 524 NW2d 682 (1994). The presumption of prejudice in Vail is inconsistent with the presumption that the jury followed its instructions not to compromise. See People v Johnson, 427 Mich 116, n 15, where this Court made the following relevant comments:

There is no basis on this record to assume that the jury’s verdict was a product of compromise. There is simply no more reason for assuming that jurors have compromised on a verdict when there is an erroneous charge than there is to believe they have simply reached a middle ground when several instructions are correctly given. If there was error in allowing the first-degree murder charge to go to the jury, the jury corrected that error by acquitting defendant of that charge and returning a proper verdict of second-degree murder. Most courts agree that a proper verdict of second-degree murder cures an error in instructing a jury on first-degree. . . . While Michigan cases have not agreed, . . . any other conclusion is based on judicial speculation that jurors who have acquitted the defendant have compromised their views despite an express direction from the trial court to the contrary. [Citations omitted.]

The Vail rule gives the jury far less credit than is warranted.

Further, the Vail rule also overlooks the fact that the error is cured when the jury acquits the defendant of the unwarranted charge. We are persuaded by the view that a defendant has no room to complain when he is acquitted of a charge that is improperly submitted to a jury, as long as the defendant is actually con*487victed of a charge that was properly submitted to the jury. Such a result squares with respect for juries. Further, not to adopt this view is to countenance a misuse of judicial resources by automatically reversing an otherwise valid conviction.11

On the basis of this record, we are satisfied that it is highly probable that the error did not affect the verdict. Although it might have been error to submit the first-degree murder charge to the jury, it is undisputed that a second-degree murder charge was properly submitted to the jury. The jury acquitted defendant not only of the first-degree murder charge, but also of the supported second-degree murder charge. Where a jury acquits a defendant of an unwarranted charge (first-degree murder) and a lesser included warranted charge (second-degree murder) before convicting of a still lesser charge (voluntary manslaughter), we find that it is highly probable that the erroneous submission of the unwarranted charge did not affect the ultimate verdict.12 There is no basis on this record to find that it was a product of juror compromise.

If, however, sufficiently persuasive indicia of jury compromise are present, reversal may be warranted *488in certain circumstances. That is to say, a different result may be reached, under a Geams review, where the jury is presented an erroneous instruction, and: 1) logically irreconcilable verdicts are returned, or 2) there is clear record evidence of unresolved jury confusion, or 3) as the prosecution concedes in the alternative, where a defendant is convicted of the next-lesser offense after the improperly submitted greater offense.

It is for these reasons that we overrule Vail and find that any error in submitting the first-degree murder charge to the jury was harmless in light of the fact that the jury returned a manslaughter conviction.

Because the Court of Appeals reversed on the basis of Vail, without reaching defendant’s other issues, id. at 681, we remand this case to the Court of Appeals for consideration of the other issues Graves raised in his appeal.

(dissenting). I would decline to overrule People v Vail, 393 Mich 460; 227 NW2d 535 (1975); therefore, I respectfully dissent.

I agree with the Court of Appeals:

Viewing the evidence in a light most favorable to the prosecution, we find insufficient evidence to establish the elements of premeditation and deliberation. There were no witnesses to the shooting and the weapon was never found. The only direct evidence linking defendant to the shooting was defendant’s confession that, because of fear inspired when the man who had just robbed him at gunpoint turned to again approach defendant’s vehicle, defendant crouched down and fired his gun through his car window. There was no evidence of a prior relationship between defendant and *489the victim, or that the victim or anyone else was defendant’s intended target. Under these circumstances, a reasonable factfinder could not find premeditation or conclude that defendant had time to take a “second look” before firing his weapon. Accordingly, the trial court erred in denying defendant’s motion for a directed verdict with regard to the charge of first-degree murder. [224 Mich App 676; 569 NW2d 911 (1997).]

The next question that must be resolved is whether the error of the trial court was harmless. The majority in this case holds that the error was harmless, and it overturns established precedent that has proved to be workable under People v Vail. In Vail, the defendant was charged with first-degree murder for his participation in a shooting that resulted in the death of the victim. The defendant was found guilty of voluntary manslaughter by a jury, and the Court of Appeals affirmed. The defendant appealed in this Court, contending that there was insufficient evidence for the jury to find him guilty of first-degree murder and that it was error to deny his motion to dismiss the count. This Court held that the trial judge erred in refusing to grant the defendant’s motion for a dismissal of the first-degree murder count where the proofs were insufficient to establish the elements of premeditation and deliberation.

This Court stated:

[W]here a jury is permitted consideration of a charge unwarranted by the proofs there is always prejudice because a defendant’s chances of acquittal on any valid charge is substantially decreased by the possibility of a compromise verdict. For this reason it is reversible error for a trial judge to refuse a directed verdict of acquittal on any charge where the prosecution has failed to present evi*490dence from which the jury could find all elements of the crime charged. [Id. at 464.]

The precise holding of Vail has been part of this state’s jurisprudence dating as far back as 1926. See People v Hansen, 368 Mich 344; 118 NW2d 422 (1962); People v Marshall, 366 Mich 498; 115 NW2d 309 (1962); People v Stahl, 234 Mich 569; 208 NW 685 (1926). For the reasons that follow, I agree with our prior holdings that where there is no evidence of first-degree murder, it is improper for the trial court to submit the charge to the jury.

In Stahl, supra at 572, we stated:

When twelve jurors agree on amount or degree generally there must be composition of views. Here the jurors to determine degree were required improperly to compose their views between the major charge of murder in its degrees, and manslaughter. Defendant testified; if truly, he was innocent. The case was serious, sad. If the murder feature had been omitted from the instructions, and the case submitted on the theory of manslaughter, it cannot now be said with certainty that the jury would have reached the same result.

Further, in People v Gessinger, 238 Mich 625, 628; 214 NW 184 (1927), this Court stated:

[I]t is evident to most practitioners of experience that it would be much easier to secure an acquittal if the defendant were only charged with the lesser offense than it would be were he charged with all three offenses. The tendency of jurors is to compromise their differences. Where there is only one charge they are obliged to meet the question squarely by yes or no, or disagree, but where the charges are three, the juror who thinks there should be no conviction, and the juror who thinks that a conviction should be had of the greater offense are quite hable to agree upon a conviction of the lesser offense.

*491However, the majority argues that Vail should be overruled because it represents the minority view among other states that have considered the issue. Whether the minority or majority view, I would continue to adhere to the principles laid out in Vail because I believe that justice requires us to do so.

Recently, in People v Gearns, 457 Mich 170; 577 NW2d 422 (1998), a majority of justices held that the test for harmlessness should be the “highly probable” standard. Under that standard, “ ‘[u]nless the appellate court believes it highly probable that the error did not affect the judgment, it should reverse.’ ” People v Mateo, 453 Mich 203, 219; 551 NW2d 891 (1996), quoting Traynor, The Riddle of Harmless Error (Ohio State Univ Press, 1970), pp 34-35.1 In this case, I cannot say that there was a “high probability” that the submission of the first-degree murder charge did not affect the jury’s verdict.

A rule requiring reversal where a first-degree murder charge is erroneously submitted to the jury under *492Vail is consistent with the highly probable standard for nonconstitutional error under Geams. Because I cannot say that there is a high probability that the erroneous submission of the charge to the jury did not affect its verdict, I would uphold Vail. While the majority’s history of the development of harmless error law in Michigan is interesting, it does not change the fact that this Court should adhere to the rule articulated in Geams. Because the Geams highly probable test is consistent with the rule requiring reversal in Vail, and is consistent with Michigan precedent dating as far back as 1926, I would respectfully decline to overrule Vail.

In Price v Georgia, 398 US 323; 90 S Ct 1757; 26 L Ed 2d 300 (1970), a man was charged with murder and found guilty of the lesser included offense of voluntary manslaughter. The prosecution sought and obtained reversal of the defendant’s initial conviction. The defendant was retried on both the murder charge and the voluntary manslaughter charge. The second trial also resulted in a conviction of voluntary manslaughter. The defendant appealed on double jeopardy grounds. The United States Supreme Court held that it was error to retry the defendant on the murder count because he already had been acquitted of that charge in the first trial.

In addressing the harmlessness of the second trial, the prosecution argued that, because the defendant was convicted of the same crime at both the first and the second trials, therefore, he suffered no harm. However, the Supreme Court rejected this contention, stating:

[W]e cannot determine whether or not the murder charge against petitioner induced the jury to find him guilty of the *493less serious offense of voluntary manslaughter rather than to continue to debate his innocence. See United States ex rel Hetenyi v Wilkins, 348 F2d 844 (CA 2, 1965), cert denied Mancusi v Hetenyi, 383 US 913 (1966). [Price, supra at 331-332.]

The Court held that the error was not harmless.

I recognize that this case dealt with double jeopardy, while Graves does not. However, I cite it merely to show that the United States Supreme Court rejected the prosecution’s harmless error argument because it could not determine whether the jury compromised its verdict. I agree. Because I cannot state that there is a “high probability” that the submission of the first-degree murder charge did not affect the jury’s verdict, I would find that the error requires reversal.