DIANE P. WOOD, Circuit Judge.
Salvatore Lorefice was convicted on federal charges of mail fraud and wire fraud, in violation of 18 U.S.C. §§ 1341 and 1348, for his participation in a particularly unsavory insurance fraud scheme. The scheme involved insuring critically ill individuals under employee group life policies, without their knowledge, and then collecting when they died. On this appeal, he asserts both that his conviction was tainted by trial errors of various kinds and that his sentence was not properly calculated under the Sentencing Guidelines. Although we agree that some errors may have been committed in the trial, we have concluded that none requires reversal; wé also find that the district court did not err in its sentencing determinations. We therefore affirm in all respects.
I
Lorefice and his wife Bonnie owned a travel agency called Corporate Travel Consultants Corporation (“CTC”); Loref-ice served as CTC’s Chairman of the Board. Noordin Lakhani was the accounting director, Chief Financial Officer, and insurance expert for the company; he reported to Lorefice. From December of 1990, continuing through at least June 1994, Lorefice and Lakhani hatched and carried out a scheme to defraud five different insurance companies: Metropolitan Life Insurance Company (“Metlife”), Mutual Benefit Life Insurance Company (“Mutual”), Royal Maccabees Life Insurance Company (“Royal”), Hartford Life and Accident Insurance Company (“Hartford”), and Benefit Trust Insurance Company (“Trustmark”). The plot involved the purchase of group life insurance policies for persons allegedly employed by CTC. This kind of policy was attractive because group policies usually do not require individual medical examinations as a prerequisite for coverage. The insurance companies instead protected themselves in other ways: to be covered, the person had to be a current, full-time employee, and the policies prohibited “stacking,” or the purchase of several policies on the same risk leading to a cumulative level of coverage in excess of that which a single company would underwrite without medical examinations or other, more rigorous scrutiny.
The key to the system Lorefice and Lakhani devised was to find individuals who were critically ill and thus likely to die soon, to list them as employees of CTC, and (unbeknownst to the “insured” individual) to designate Lorefice, his daughters, or in one case Lakhani as the beneficiary of the policy. Certain arrangements were necessary to accomplish this. First, Lo-refice and Lakhani submitted a false census of employees
(i.e.
a list of all full-time employees, with names, titles, and dates of birth) to each insurer. In some instances, people included in the census were not CTC employees at all; in other instances, they were
bona fide
employees, but they did not know that the policies existed. For example, Lorefice made himself the beneficiary on multiple policies for Rosina Lorefice, his mother, age 95; Anna Orofi-no, his godmother, who was 89; and Joe Lorefice, his brother. None was a current
CTC employee. Lorefice also made himself the beneficiary on policies for CTC employees Mabel Barsema, Joel Timmel, and Steve Young, and he made his daughters the beneficiaries on policies for CTC employee Jeffrey Metcalf. None of the “insureds” knew about the coverage. Barsema was over 70 years old and had recently suffered a stroke and a heart attack, while Timmel, Young, and Metcalf were HIV positive. Lakhani was the beneficiary on a policy for his father Akbarali Lakhani, who was not an employee of CTC and who was in poor health. Second, Lo-refice and Lakhani took steps to make sure that each insurance company was unaware of the existence of the others.
Two of the individuals covered by these policies (Rosina Lorefice and Jeffrey Met-calf) died before the scheme was discovered. Following their deaths, Lorefice collected $600,000 in death benefits, from Metlife, Mutual, Royal, and Hartford. The plot unwound, however, before anyone else died. On June 12, 1997, a grand jury indicted both Lorefice and Lakhani for four counts of mail fraud and three counts of wire fraud. Lorefice pleaded not guilty and went to trial, at which his theory of defense was that Lakhani, who handled all dealings with the insurance companies, defrauded the insurers without Lorefice’s knowledge. For his part, Lakhani, after initially pleading not guilty, changed his plea to guilty on Count One (mail fraud) and agreed to testify for the government on the eve of trial. The jury returned a verdict of guilty on January 20, 1998; on August 26, 1998, the district court sentenced Lorefice to 57 months’ imprisonment, three years’ supervised release, and a $50,000 fine, and it ordered that he make restitution in the amount of $600,000.
II
Lorefice presents four arguments for our consideration, three of which relate to his conviction and the last to his sentence. First, he claims that the prosecutors violated his right to a fair trial by improperly asserting that both an F.B.I. agent and Lorefice’s own lawyer held the opinion that he was guilty as charged. Second, he argues that the district court improperly redacted two paragraphs from the indictment without resubmitting it to the grand jury, which he believes hampered his defense. Third, he argues that he was deprived of his right to a fair trial when one of the jurors was exposed to a newspaper article implying that Lorefice was implicated in an unrelated corruption probe. Finally, he claims that his sentence should be reduced for two reasons: the district court used an improper figure for “intended loss” under U.S.S.G. § 2F1.1, and the court should have reduced his total offense level by three levels under U.S.S.G. § 2X1.1, the “attempt, solicitation, or conspiracy” guideline. We address these points in turn.
A.
Improper Opinion Evidence
The most egregious instance of improper opinion evidence, according to Lorefice, occurred when F.B.I. Special Agent Ray Ruebenson was testifying. Agent Rueben-son had been actively involved in the investigation of Lorefice and Lakhani from its inception in June 1994. During the government’s case-in-chief, he testified about numerous documents the F.B.I. had subpoenaed. Lorefice then testified, at which point the prosecutor asked him why he had purchased a paper shredder. Lorefice’s explanation was that he had bought the shredder in anticipation of the relocation of CTC’s offices. Wanting to rebut this claim, the government called Agent Rue-benson back to the stand, aiming to show that Lorefice had bought the shredder shortly after he learned that the F.B.I. was investigating him, for the purpose of destroying evidence. The government elicited the following testimony from Rue-benson about the meeting during which Lorefice learned of the F.B.I. investigation:
PROSECUTOR: What was the general nature of the information that you re
vealed during that meeting to the defendant and his attorney?
AGENT RUEBENSON: I showed him the insurance documents that I had and my belief that they were fraudulent.
Lorefice’s lawyer immediately objected and moved to strike. The district court granted the motion to strike and added a standard curative instruction to the jury telling them to disregard the opinion evidence. Later on, Lorefice’s lawyer moved for a mistrial based on Ruebenson’s testimony, but the district court denied that motion, both because it thought the comment was not prejudicial in light of the rest of the testimony before the jury and because it believed the curative instruction was sufficient.
The second instance of improper opinion evidence occurred during the prosecution’s closing argument, when the AUSA alluded to the testimony of defense witness Steven McMullen. Lorefice had hired McMullen, an attorney, to help him obtain death benefits after one of his claims on Metcalfs policy was denied. McMullen testified that he and Lorefice ultimately decided not to file a lawsuit against the company because “we became fearful that Mr. Lak-hani had lied with respect to the insurance information on Mr. Metcalf.” This testimony was designed to support Lorefice’s contention that he knew nothing about the shady dealings, and that the entire scheme had been masterminded and carried out by Lakhani. At the end of the trial, during her closing argument, the AUSA made the following comment to the jury: “Well, you know, ladies and gentlemen, that eventually Steve McMullen learned there was a scheme to defraud, and the defendant was involved.” Again, defense counsel objected promptly, but the court overruled the objection after the AUSA argued that she believed the jury could infer from the fact that no lawsuit was filed that McMullen thought Lorefice was involved in the fraud. The court also took the precaution of instructing the jury that “what McMullen thought isn’t the issue here; it is what you think.”
This court uses' a two-part analysis to assess claims that improper prose-cutorial comments have infringed a defendant’s right to a fair trial.
United States v. Brisk,
171 F.3d 514, 524 (7th Cir.1999);
United States v. McClinton,
135 F.3d 1178, 1188 (7th Cir.1998). First, we decide whether the comments were in fact improper. If not, then the claim obviously fails. If they were improper, we ask whether they had the effect of denying the defendant a fair trial. In evaluating the effect of prosecutorial misconduct, there are five factors that we assess, namely, (1) the nature and seriousness of the misconduct; (2) whether the conduct of the defense counsel invited the prosecutor’s remarks; (3) whether the trial court’s instructions to the jury were adequate; (4) whether the defense was able to counter the improper arguments through rebuttal; and (5) the weight of the evidence against the defendant. See
United States v. Rivera,
153 F.3d 809, 814 (7th Cir.1998). Implicit in the five-factor analysis is the recognition that not every trial error requires reversal; under Federal Rule of Criminal Procedure 52(a), we must disregard “[a]ny error, defect, irregularity or variance which does not affect substantial rights.... ” Although one of the alleged errors here came from an F.B.I. agent rather than the prosecutor, we find this test appropriate for assessing both claims.
Looking first at Ruebenson’s unfortunate comment, we agree with the district court that it crossed the line of propriety and was correctly stricken and made the subject of a curative instruction. This is so because “prosecutorial statements constituting a representation of personal belief as to the defendant’s guilt are improper.”
United States v. Haddon,
927 F.2d 942, 947 (7th Cir.1991). Agent Rue-benson was not the prosecutor, of course, but from a legal standpoint, misconduct by the investigating law enforcement agents
is indistinguishable from misconduct by the prosecuting attorneys. See,
e.g., United States ex rel. Smith v. Fairman,
769 F.2d 386, 391 (7th Cir.1985) (prosecutor is responsible for failure to produce exculpatory information under the control of the police, even though prosecutor was not aware that the evidence existed);
Wedra v. Thomas,
671 F.2d 713, 717-18 n. 1 (2d Cir.1982) (“[T]he knowledge of a police officer may be attributable to the prosecutor if the officer acted as an arm of the prosecution.”);
cf. Kyles v. Whitley,
514 U.S. 419, 437, 115 S.Ct. 1555, 131 L.Ed.2d 490 (1995) (“[T]he individual prosecutor has a duty to learn of any favorable evidence known to the others acting on the government’s behalf in the case, including the police.”);
Giglio v. United States,
405 U.S. 150, 154, 92 S.Ct. 763, 31 L.Ed.2d 104 (1972) (prosecutor is responsible for nondisclosure of communications between another assistant U.S. attorney and a prosecution witness even if made without the prosecuting attorney’s knowledge).
The more difficult question is whether the comment was so deeply prejudicial that counsel’s motion for a mistrial should have been granted, or if the district court’s actions were enough under the circumstances to allow this trial to stand. There is no way to read (or hear) Ruebenson’s statement about his “belief that [the documents] were fraudulent” as anything but an opinion from a law enforcement officer that the defendant was guilty of the offenses charged, contrary to the rule expressed in
Haddon.
Although it is a closer call, we also think that the AUSA improperly suggested in her closing argument that McMullen held the opinion that there was a scheme to defraud, and that Lorefice was part of it. First, McMullen had made no such statement; to the contrary, he had expressed concern about Lakhani’s truthfulness about Metcalf. Second, the inference that McMullen dropped the suit because he thought Lorefice was guilty is a stretch: McMullen testified that he and Lorefice together decided to drop the suit because they concluded that Lakhani had lied. We agree with Lorefice’s contention that the AUSA’s statement in effect invited the jury to infer that McMullen, his own lawyer, thought that he was guilty.
Because these statements were improper in the first instance, we turn to the
Rivera
analysis to see whether they were sufficiently prejudicial to deprive Lorefice of a fair trial. In short, they were not. They were two isolated instances in a trial otherwise replete with evidence about Lo-refice’s involvement in the fraudulent scheme. The trial judge gave prompt and clear instructions to the jury in both instances, and we presume that the jury followed the court’s instructions.
McClinton,
135 F.3d at 1189. Even though defense counsel did not invite either misstep, Lorefice’s lawyer could have addressed Agent Ruebenson’s comment in closing argument if that had seemed strategically wise. We are satisfied on the record as a whole that these errors were harmless, and thus that they do not justify the drastic step of setting aside the result of the trial.
B.
Redacted Indictment
Immediately prior to the trial, the district court allowed the government to redact paragraphs 55 and 58 from the indictment, without resubmitting it to the grand jury. Those two paragraphs alleged that Lorefice had written two checks to Lakhani, each in the amount of $20,000, and that Lakhani had deposited the checks in his bank account. In granting the motion to strike, the court accepted the government’s explanation that those paragraphs were “mistaken allegations.” The prosecutor explained that they had initially believed that the two payments were kickbacks from Lorefice to Lakhani, but later they discovered that the $40,000 represented legitimate commission payments. The court allowed the defense to question Lakhani about the two payments, but it denied defense counsel’s request to intro
duce evidence about the redaction itself because it was “immaterial.”
There was nothing wrong in the district court’s actions. An indictment may be altered without resubmission to the grand jury as long as the alteration makes no material change and there is no prejudice to the defendant. See
United States v. Mankarious,
151 F.3d 694, 700 (7th Cir.1998);
United States v. Graffia,
120 F.3d 706, 710-11 (7th Cir.1997). Indeed, in
United States v. Miller,
471 U.S. 130, 105 S.Ct. 1811, 85 L.Ed.2d 99 (1985), the Supreme Court expressly held that to drop from an indictment allegations that are unnecessary to an offense clearly contained within it does not unconstitutionally amend the indictment.
Id.
at 144, 105 S.Ct. 1811, overruling part of
Ex parte Bain,
121 U.S. 1, 7 S.Ct. 781, 30 L.Ed. 849 (1887). Here, the redaction had the effect of narrowing the indictment. When this is done to an indictment charging a conspiracy or a scheme to defraud, resubmission is not required.
Graffia,
120 F.3d at 710-11. Furthermore, the deletion of those two paragraphs made no material change in the indictment and even Lorefice is hard put to find any prejudice from the alteration. If anything, the redaction reduced the burden on Lorefice’s defense, because he no longer had to defend against the allegation that he gave kickbacks to Lak-hani. There is also no merit to Lorefice’s argument that the redaction was a proper subject for exploration at trial. The underlying facts were available to him to use in attacking Lakhani’s credibility, insofar as he might have wanted to suggest that Lakhani was lying when he claimed not to have gained financially from the scheme. Evidence about the government’s drafting decisions concerning the indictment is not relevant for the jury’s consideration, and it is potentially confusing. The district court sensibly prohibited exploration of the redaction at the trial.
C.
Juror Groll
The next issue on appeal concerns information not introduced into evidence but which, through no misconduct on the part of the prosecution, made its way to one of the jurors from an outside source. The day after Lorefice’s trial began, the
Chicago Tribune
ran an article about the proceedings. The article described the opening statements, and, in its final paragraph, it stated, “The charges against Lo-refice stemmed from a lengthy corruption probe of former longtime Oakbrook Terrace Mayor Richard Sarallo and developer Robert Krilich.” Notwithstanding this journalistic association, the Oakbrook Terrace probe had nothing to do with the case against Lorefice and was not admissible as evidence. Accordingly, the district court polled the jurors to find out if any of them had read the article. Juror Groll replied that he had.
The court thereupon excused the rest of the jurors and questioned Juror Groll individually. The judge first asked him what he recalled about the article, and Juror Groll responded, “Very little. I scanned it ... [and] remembered some of the names involved.” In response to a follow-up question, Juror Groll explained that he did not learn anything more from the article than he already knew from opening statements. The judge then asked “Is there anything about the article or your recollection of it that would affect your attitude toward the case or toward the merits of the case or toward any of the parties to the case beyond what you had already learned from the opening statements of counsel and whatever evidence had been received.” Juror Groll replied that “it could reflect on the final outcome.” With this confusing statement on the record, the court probed further, clarifying that Juror Groll had learned nothing more than what he knew from opening statements and that nothing in the article would affect his ability to be a fair and impartial juror.
The court ultimately denied defense counsel’s motion to dismiss Juror Groll, for the following reason:
I don’t think there’s any prejudice, and I think that the information in the last paragraph is not as damaging as you think, even if he did read it and remembered it. The fact that this investigation grew out of a so-called “corruption probe” doesn’t make the case any more damaging against this defendant than it would be if it had arisen out of thin air. The allegations that are made against this defendant in this case are about as damaging as anything can be by way of a fraud investigation.
Last, the court commented about Juror Groll, “What he did was bear out my invariable experience: People read articles and very little of it remains with them. That’s what happened. Motion denied.” At the request of defense counsel, the judge instructed Juror Groll not to discuss the article with the other jurors, and there is no allegation that Groll failed to heed this instruction.
We review the district court’s decision whether to dismiss a juror on the ground that the juror was exposed to potentially prejudicial influences for abuse of discretion.
Cf. United States v. Berry,
92 F.3d 597, 600 (7th Cir.1996) (district court’s decision to grant or deny a mistrial on account, of prejudice possibly resulting from jury’s exposure to information not in evidence is reviewed for abuse of discretion);
United States v. Bruscino,
687 F.2d 938, 940 (7th Cir.1982)
(en banc)
(same). Granted, Juror Groll’s responses to the court’s questions were to a degree contradictory. Nevertheless, the district judge was in the best position to assess what the juror meant, to consider what might have been confusing, and to weigh the juror’s ultimate statement that he could be fair and impartial despite having seen the
Tribune
article. See
Berry,
92 F.3d at 600;
Bruscino,
687 F.2d at 941. Furthermore, the district court correctly observed that the information in the article about the Oakbrook Terrace corruption investigation paled in comparison with the fraud scheme Lorefice was accused of operating. This is not a case where the outside information suggested that a defendant was engaged in particularly heinous conduct, which might have overshadowed the charges actually before the jury. Under the circumstances, we see no abuse of discretion in the trial court’s decision to allow Juror Groll to remain on the jury and to deny defense counsel’s motion to dismiss him.
D.
Sentencing
Mail fraud and wire fraud offenses fall under U.S.S.G. § 2F1.1, which establishes a base offense level of 6, and then requires that level to be increased depending on the amount of the loss involved. Application Note 8 discusses the way in which loss should be calculated, incorporating by reference the methodology set forth in U.S.S.G. § 2B1.1 (the guideline for larceny, embezzlement, and other forms of theft). That Application Note states that “[a]s in theft cases, loss is the value of the money, property, or services unlawfully taken.... ” It then explains that “Consistent with the provisions of § 2X1.1 (Attempt, Solicitation, or Conspiracy), if an intended loss that the defendant was attempting to inflict can be determined, this figure will be- used if it is greater than the actual loss.”
In this case, there are four potential numbers that might be relevant. The actual loss to the insurance companies, everyone agrees, was $600,000, the amount they paid out. The district court decided that the amount of loss Lorefice intended to inflict was $4,200,000, which represents the face value of all the policies he and Lakhani obtained on the lives of Rosina Lorefice, Anna Orofino, Joe Lorefice, Jeffrey Metcalf, Mabel Barsema, Joel Tim-mel, and Steve Young, the policies on which they had made Lorefice or his daughters beneficiaries. The $4,200,000 therefore does not include the face value of the policy on Lakhani’s father, for which Lakhani was the sole beneficiary. Loref-ice suggests a third figure, $1,000,000, which is the amount he attempted to collect following the deaths of Rosina Loref-ice and Jeffrey Metcalf. He argues that he might never have collected another
cent, even if the F.B.I. had not entered the picture, because the other four people might not have died, he might not have been able to complete the claim forms, he might not have used the mails or wires in order to collect, and so on. A fourth possible figure, about which we have no evidence, is the amount of insurance on the entire employee complement, given the fact that these were group policies.
Lorefice concedes that the district court correctly refused to use $600,000 as the amount of loss. If the court had used the $1,000,000 figure, then under § 2Fl.l(b)(l)(L), it would have added 11 to the base offense level of 6. Instead, using the $4,200,000 figure, the court added 13 levels to the base level of 6, following § 2Fl.l(b)(l)(N). (We note that the government did not argue that the face amount of the policies on the entire employee complement was the proper measure of the intended loss, even though it is clear that the existence of multiple policies alone, regardless of the identity of the beneficiaries,. was an act that defrauded the insurance companies. Any argument based on this theory has therefore been waived.) In essence, Lorefice is making an argument about the legal meaning of the term “intended loss,” and we therefore review the district court’s decision
de novo. United States v. Bonanno,
146 F.3d 502, 508-09 (7th Cir.1998) (“A district court’s calculation of the intended loss under U.S.S.G. § 2Fl.l(b)(l) is a finding of fact reviewed for clear error, while its construction of the sentencing guidelines is reviewed
de novo”).
When the focus of relevant conduct is on a defendant’s intent, rather than her successful execution of a scheme, the fact that it might be difficult for her to carry out a plan is irrelevant. See,
e.g., United States v. Coffman,
94 F.3d 330, 336-37 (7th Cir.1996). That is the case with § 2F1.1. See
United States v. Stockheimer,
157 F.3d 1082, 1090 (7th Cir.1998). Here, the district court made a fact finding, not clearly erroneous, that Lorefice’s actions, including his decision to continue to pay premiums on the policies, show that he intended to collect their face values. The fact that the “insureds” had not yet died makes no difference, because future events that are beyond the defendant’s control do not affect his intent.
Bonanno,
146 F.3d at 510. Although Lorefice relies on the Fifth Circuit’s decision in
United States v. Lghodaro,
967 F.2d 1028 (5th Cir.1992), for the proposition that the intended loss in an insurance fraud case is the amount the defendant actually sought to claim (here, the $1,000,000), we do not read
Lghodaro
so narrowly. There, the defendant was arguing that his offense level should have been based on the $9,074.45 he was actually paid, rather than the $58,816.07 for which he filed false claims. Intended loss, he asserted, should be used only if actual loss was difficult to determine. Not surprisingly, the Fifth Circuit rejected that position and held that the full $58,816.07 was the proper amount to use. Nothing in the opinion indicates that the government had argued that the ' face value of the policies was a more proper measure of intended loss, and thus
Lghodaro
has nothing to say on the latter point.
Last, we consider Lorefice’s argument that, even if the district court properly used the $4,200,000 figure as his intended loss, he was guilty of only a “partially completed offense” and thus was entitled to the three-level reduction provided by U.S.S.G. § 2X1.1. One reaches § 2X1.1 via Guideline 2F1.1, Application Note 8, which cross-references the commentary to § 2B1.1. Application Note 2 to § 2B1.1 in turn says that “[i]n the case of a partially completed offense ..., the offense level is to be determined in accordance with the provisions of § 2X1.1.” The problem with this argument is in its initial premise, which is that the mail and wire fraud for which Lorefice was convicted were only partially completed. This court faced an identical argument in
Coffman, supra,
where a wire fraud scheme was interrupted before any concrete harm at all had been inflicted. The defendants there , argued too that
they should be entitled to the adjustment under § 2X1.1, but we disagreed with the following explanation:
Here the fraud had a real victim in its" sights but was interrupted before it could do any harm. That makes this a clear case of attempted fraud, where the intended loss is the relevant benchmark rather than the actual loss, by definition zero because it was only an attempt.... We add that because the sentence is to be based on the intended rather than the actual loss, the three-level downward adjustment for attempts, § 2Xl.l(b)(l), is inapplicable, ... — wholly apart from the inappropriateness of making such an adjustment when the substantive offense, mail or wire fraud, is a crime of attempting rather than attaining.
Coffman,
94 F.3d at 337 (internal citations omitted). Here too, because the substantive offense was complete with the mail and wire fraud, there was no occasion to apply the downward adjustment of § 2X1.1.
The judgment of the district court is Affirmed.