OPINION
M. SMITH, Circuit Judge:
Defendant Rogelio Lemus appeals his conviction for possession with intent to distribute more than 50 grams of methamphetamine. Because we conclude that insufficient evidence supported the jury’s quantity determination, we reverse in part and remand for resentencing pursuant to the lowest quantity category under the sentencing guidelines.
FACTS AND PRIOR PROCEEDINGS
I. The Investigation
In early May of 2011, FBI informant Ana Montano was dispatched to a bar to meet with Defendant Rogelio Lemus. Le-mus, seeing Montano’s gang tattoo, volun
teered that he was a member of the same gang, and asked Montano if she knew the clique to which he belonged. Montano told him that she was looking for somebody who could supply ounce-quantities of methamphetamine. Lemus responded that he had a pound for sale.
On May 16, 2011, Montano made a recorded call to Lemus. She stated that she wanted to buy two ounces. Lemus responded: “Just two? ... I’m going to tell the guy, because, well, you know, the bags have to be torn up, you understand?” On May 18, 2011, Montano and Lemus arranged to meet to carry out the sale and agreed to a price for the two ounces, but Lemus was delayed by the absence of his associate. When he finally arrived, Le-mus, consistent with his initial offer of a pound and his earlier hesitancy to “tear up” the bags, but contrary to his agreement earlier that day to sell a smaller quantity, stated that he was unable to sell the methamphetamine in ounce quantities because they only sold it by the pound. Lemus offered to give Montano a sample, but Montano demurred, saying that her buyer would not trust that the sample was the same quality as the pound.
After the meeting, FBI agents followed Lemus to his house, and were able to identify him from the motor vehicle records for his truck. The agents did not conduct a traffic stop, and did not obtain a search warrant to search for drugs. No drugs were seen or observed on the date of the meeting, and Montano did not believe that Lemus had the pound of methamphetamine in his truck during the meeting, although she believed that he had it that day.
At his post-arrest interview, Lemus denied involvement in drug trafficking, and claimed that he often joked on the phone about drugs. At trial, the government did not produce the drugs or present any testimony that someone saw Lemus in possession of a substance that appeared to be methamphetamine.
DISCUSSION
I. Sufficiency of the Evidence
A. Standard of Review
In reviewing a conviction for sufficiency of the evidence, we ask whether, “after viewing the evidence in the light most favorable to the prosecution,
any
rational trier of fact could have found the essential elements of the crime beyond a reasonable doubt.”
United States v. Nevils,
598 F.3d 1158, 1163-64 (9th Cir.2010) (quoting
Jackson v. Virginia,
443 U.S. 307, 319, 99 S.Ct. 2781, 61 L.Ed.2d 560 (1979)). Under this two-step inquiry, we therefore first consider the evidence presented at trial in the light most favorable to the prosecution, and second, determine whether the evidence so viewed is adequate to allow any rational trier of fact to find the essential elements of the crime beyond a reasonable doubt.
Id.
at 1164.
B. The Elements of the Offense
Lemus was convicted of violating 21 U.S.C. § 841(a)(1), which prohibits,
inter alia,
possession of a controlled substance with intent to distribute. The jury found that he possessed at least 50 grams of methamphetamine, subjecting him to the penalty specified in 21 U.S.C. § 841(b)(1)(A)(viii). To violate this statute, actual possession is not required: constructive possession also suffices. “The term ‘constructive possession’ does not connote a legal fiction. Rather, the term simply reflects the common sense notion that an individual may possess a controlled substance even though the substance is not on his person at the time of arrest.”
United States v. Disla,
805 F.2d 1340, 1350 (9th Cir.1986).
Constructive possession means “the exercise of ‘dominion and control,’ ” and “may be demonstrated by direct or circumstantial evidence that the defendant had the power to dispose of the drug.”
Id.
(citing
United States v. Amaro, 422
F.2d 1078, 1080 (9th Cir.1970);
Arellanes v. United States,
302 F.2d 603, 606 (9th Cir.1962)).
[O]ne having a working relationship or a sufficient association with those having physical custody of the drugs so as to enable him to assure their production, without difficulty, to a customer as a matter of course may be held to have constructive possession. But a casual facilitator of a sale, who knows a given principal possesses and trades in narcotics but who lacks the working relationship with that principal that enables an assurance of delivery, may not be held to have dominion and control over the drug delivered and cannot be said to have possession of it.
Hill v. United States,
379 F.2d 811, 814 (9th Cir.1967) (quoting
United States v. Jones,
308 F.2d 26, 30-31 (2d Cir.1962)).
“Constructive possession may also be proven by the defendant’s participation in a ‘joint venture’ to possess a controlled substance.”
Disla,
805 F.2d at 1350 (citing
United States v. Valentin,
569 F.2d 1069, 1071 (9th Cir.1978)). “[Cjoordi-nated activity among the defendants raises a reasonable inference of a joint venture.”
United States v. Smith,
962 F.2d 923, 930 (9th Cir.1992) (quoting
United States v. Hernandez,
876 F.2d 774, 778 (9th Cir.1989) (alteration in original)). “In addition to association, the government must also establish that the defendant had a role in directing or planning the acquisition or transportation of the drugs.”
Id.
C. Reliance on Lemus’s Statements
Lemus argues that without corroboration, his own inculpatory statements made during the offense cannot support a conviction. Lemus relies upon
United States v. Valdez-Novoa,
780 F.3d 906, 922 (9th Cir.2015). However,
Valdez-Novoa
involved a defendant’s confession, not contemporaneous statements.
Id.
And,
Valdez-Novoa
framed the rule as directed to confessions: “the contemporary iteration of the common law
corpus delicti
rule” is that “[although the government may rely on a defendant’s confession to meet its burden of proof ... in order to serve as the basis for conviction, the government must also adduce some independent corroborating evidence.”
Id.
Lemus relies on out-of-circuit authority to argue that we should extend the corroboration requirement from statements made during a confession to those made during the commission of the crime, citing among others
United States v. Bryce,
208 F.3d 346, 356 (2d Cir.1999) and
United States v. Baggett,
890 F.2d 1095-97 (10th Cir.1989). We need not address this contention, however, because the conviction here does not rest solely on a bald statement by Lemus that he had methamphetamine to sell. He also took steps consistent with that statement, including agreeing on price, arranging a sale date, collaborating with his associate, and, especially, offering a sample. These actions would sufficiently corroborate Lemus’ initial incriminatory statements, even if we were to extend the rule.
D. The Sufficiency of the Statements to Support the Conviction
The evidence presented at trial is sufficient to sustain the conviction under a constructive possession theory because a reasonable jury could have concluded that Lemus (1) had “a working relationship or a sufficient association with those having physical custody of the drugs so as to
enable him to assure their production, without difficulty, to a customer as a matter of course,”
Hill,
379 F.2d at 814, or (2) that he engaged in a joint venture, with a “role in directing or planning the acquisition or transportation of the drugs.”
Smith,
962 F.2d at 930.
At the first meeting between Lemus and Montano at the bar, Lemus said that he had one pound of “crystal” (methamphetamine) that he was willing to sell. On the day of the intended sale, Montano and Lemus arranged to meet, and Lemus instructed Montano that the price would be $1,100 for each ounce. Lemus had to delay the meeting because his associate, someone who has never been identified, was not yet in the area.
In a phone call explaining why it was taking so long to arrive, Lemus told Mon-tano that the methamphetamine was “in a city nearby.” When he finally arrived, he instructed Montano to come down, but without the money, because “we are going to explain something here.” Montano asked: “Do you have it or don’t you?” Lemus responded that “[tjhere was a misunderstanding, I will explain it to you.”
When Montano reached the car, Lemus stated that he could only sell by the pound, not by the ounce, but that Lemus had a sample to give Montano if she wanted. Lemus instructed Montano to tell her buyer that he dealt in “a pound and up,” that Lemus had “a picture” he could show and that he would bring a “sample” of “what I have in pounds” that Montano could assess for quality. Montano declined the sample, explaining that her buyer did not want a sample because of the concern that the pound would not be of the same quality. The meeting then ended.
After his arrest, Lemus told the interviewing FBI agent that remarks he made over the phone about drugs were jokes. Given the content and context of the recorded calls introduced as evidence, a reasonable jury would have no difficulty rejecting that explanation. These were not offhand, playful remarks made and received in jest. Instead, they appear to be a continuous and serious attempt to arrange a drug transaction.
Lemus had constructive possession via “the power to dispose of’ a pound of methamphetamine,
Disla,
805 F.2d at 1350, although due to the policies of his distribution structure, he lacked the power to dispose of the drug in smaller, ounce-sized quantities. While this limitation shows that Lemus lacked unfettered discretion to dispose of the drugs, that is consistent with his participation in a joint venture in which he lacked sole decision-making authority, but in which he played a role directing the transportation of the drugs. “[Vjiewing the evidence in the light most favorable to the prosecution,”
Nevils,
598 F.3d at 1164, a rational trier of fact could have found beyond a reasonable doubt that Lemus possessed methamphetamine with intent to sell it. However, Lemus was convicted of possessing more than 50 grams of methamphetamine, which affects the applicable penalty. We turn now to an assessment of the evidence introduced concerning that quantity determination.
E. The 50 Gram Quantity
Lemus argues that the quantity finding is unsupported by the evidence because there was no drug seized that could be tested for purity to determine whether it contained at least 50 grams of methamphetamine. To prove that the unobserved pound of substance at issue contained at least 50 grams of methamphetamine, the government offered an FBI agent’s testimony concerning the range of purity of methamphetamine previously purchased by the FBI. The agent testified that only
four of the approximately 30 controlled purchases made in the Los Angeles area by the FBI from 2008 to 2014 were of below 90% purity.
A pound contains approximately 453 grams. Thus, for there to be 50 grams of methamphetamine in a pound of material, it would need to be slightly over 11% pure. We have previously upheld convictions requiring proof of at least 50 grams of methamphetamine in the absence of the drugs themselves, and hence, in the absence of purity testing, where the jury could infer that Lemus’s methamphetamine was at least as pure as some actual methamphetamine that could be used for comparison. Thus,
United States v. Maciel
upheld a conviction for conspiracy to possess with intent to distribute 50 grams or more of methamphetamine in the absence of the actual drugs. 461 Fed.Appx. 610, 615-16 (9th Cir.2011).
The defendant in
Maciel
offered to supply a pound, and the lowest purity of methamphetamine that had been seized from the stash house involved in the conspiracy was 12%, a pound of which would exceed 50 grams.
Id.
at 615-16. This case differs significantly from
Maciel,
however, because the drugs used for comparison here were not connected to Lemus, but instead were from other purchases in the Los Angeles area.
We find analogous authority in the sentencing context, where courts are often called upon to estimate drug quantities. The standards applicable to such estimates are not identical to those applicable to jury findings; they are in some ways more flexible, and in other ways, more restrictive. For purposes of sentencing, quantity need only be proven by a preponderance of the evidence, but “the information which supports an approximation must possess sufficient indicia of reliability to support its probable accuracy,” and the district court must err on the side of underestimating the quantity.
United States v. Kilby,
443 F.3d 1135, 1141 (9th Cir.2006) (quoting
United States v. Culps,
300 F.3d 1069, 1076 (9th Cir.2002)).
In
Kilby,
the district court based its approximation on tablets of “Foxy” “seized in two unrelated cases from different parts of the country,” with no evidence of a common supplier or evidence that Foxy tablets are always the same approximate size, and where the two samples had quite different weights.
Id.
at 1142. While noting that we have repeatedly approved approximations “based on facts specific to the defendant’s case,”
Kilby
held that the district court’s approximation was insufficiently reliable.
Id.
at 1141-42. By contrast,
United States v. Flores,
725 F.3d 1028 (9th Cir.2013) approved an estimate employing an average pill size consistent with those typically sold by the conspiracy at issue and by the defendant.
Id.
at 1035-37.
This ease is similar to
Kilby
in that the comparison methamphetamine came from cases that were not tied to Lemus, but different in that the comparison methamphetamine came from Los Angeles, not elsewhere in the country. Unlike in
Flores,
here there was no evidence of actual drug quantity from other arms of a conspiracy, and indeed, Lemus was not charged with conspiracy.
These sentencing determination cases are instructive in a general sense, but this case involves a factual determination reached by the jury. In that sense,
Ma-ciel,
although non-precedential, remains the closest ease, and again, unlike here,
the comparison drugs in
Maciel
were from the conspiracy at issue.
It would be a bridge too far to allow a jury to extrapolate from comparison drugs that were not from activity related to the defendant or a conspiracy in which the defendant is involved. A 90% level of purity would more than suffice to support the jury’s quantity determination, if adequately connected to the drugs concerning which Lemus had constructive possession. However, the government failed to include evidence connecting that purity level to Lemus. Viewing the evidence in the light most favorable to the government, no reasonable factfinder could have determined beyond a reasonable doubt that Lemus possessed more than 50 grams of methamphetamine.
Because the drug quantity finding fails based on insufficient evidence, the government may not retry that issue, and instead must seek resentencing based solely on the basic possession conviction, i.e., under the lowest quantity category in Federal Sentencing Guidelines § 2D1.1. “The Double Jeopardy Clause forbids a second trial for the purpose of affording the prosecution another opportunity to supply evidence which it failed to muster in the first proceeding.”
Burks v. United States,
437 U.S. 1, 11, 98 S.Ct. 2141, 57 L.Ed.2d 1 (1978);
see also United States v. Vera,
770 F.3d 1232, 1250 (9th Cir.2014) (holding that retrial did not violate the Double Jeopardy Clause where reversal based on trial error, distinguishing reversal for insufficient evidence).
II. The District Court’s Decision Not to Declare a Mistrial
A. Legal Standard
We review a district court’s denial of a motion for a mistrial for abuse of discretion.
United States v. Dorsey,
677 F.3d 944, 954 (9th Cir.2012). A cautionary instruction from the judge is generally sufficient to cure any prejudice from the introduction of inadmissible evidence, and “is the preferred alternative to declaring mistrial when a witness makes inappropriate or prejudicial remarks; mistrial is appropriate only where there has been so much prejudice that an instruction is unlikely to cure it.”
United States v. Escalante,
637 F.2d 1197, 1202-03 (9th Cir.1980). A decision to not declare a mistrial will be reversed only if the improper comment, viewed in the context of the entire trial, more likely than not materially affected the verdict.
Dorsey,
677 F.3d at 954.
B. The Remark and the District Court’s Response
Before trial, Lemus moved to exclude any evidence of his membership in MS-13 or any other gang. The district court granted the motion in part, ruling that while evidence of gang membership would be somewhat prejudicial, the prejudice did not outweigh the evidence’s probative value, specifically, as to why “the two parties engaged in a drug transaction, not knowing each other very well at all.” However, the court excluded the name of Lemus’s claimed gang as irrelevant. The district court made clear that because “gang membership is incredibly prejudicial,” its use at trial would be limited to a particular purpose, and that trial would not be opened up to examine the acts of the MS-13 gang.
Despite that ruling, and despite the prosecutor asking
“[wjithout telling us the name of the gang,
for how long had the informant been a member of the gang before she began cooperating with the FBI?”, the agent responded “Um, I don’t know the exact number of years, but, um, the informant became a member of MS-13 — whereupon the defense immediately objected. The court sustained the objection, instructed the jury to disregard the answer, and directed the prosecutor to ask another question. At the break, the defense moved for a mistrial, arguing that a
curative instruction would only highlight the issue.
The government argued that the testimony was “an altogether unfortunate habitual reference ... by someone who makes the reference day in and day out in his work and slipped into it when made.” The government opposed a mistrial, but acknowledged that granting one was within the court’s discretion, and that the court was in a position to assess what remedy was required. The court ruled that its immediate instruction to disregard the testimony, and the instruction it would provide after closing arguments concerning the limited relevance of the gang evidence, were sufficient to cure any prejudice.
Later, one juror informed the court clerk that while she understood the court’s instruction to disregard the testimony, she started thinking about it, and believed that she had read about the case, and knew “all about it.” Counsel discussed with the court that there may have been press about the takedown that occurred in this case, and the defense stated that one of the articles referred to Lemus with his gang moniker. The court questioned the juror outside of the presence of the other jurors, and the juror reported that she had seen reports covering the sting operation which discussed the global nature of this gang and “the sort of violence” attached to it.
The juror confirmed that she would be able to follow the court’s previous instruction to disregard information about the particular gang, and would be able to follow the forthcoming limiting instruction. The juror confirmed that she did not think that the press reports were relevant to anything, but felt compelled to inform the court that she remembered them. The juror stated that the press reports would not affect her view of the case at all, and that she would disregard them and base her verdict only on the law and the court’s instructions. Based on that discussion, the court was satisfied that the juror would disregard anything she had read about the gang.
“When the court strikes testimony and gives ... an instruction [to disregard it], there is a strong presumption that the jury has followed the court’s instruction.”
United States v. Pavon,
561 F.2d 799, 803 (9th Cir.1977). “The trial court is in the best position to determine whether an incident merits a mistrial.”
United States v. Gardner,
611 F.2d 770, 777 (9th Cir.1980). That is because “the district court had the opportunity to see and hear the witnesses and to weigh their testimony. It was in a far better position than we to judge the effects of the incidents.”
United States v. Love,
535 F.2d 1152, 1157 (9th Cir.1976).
Here, the district court was well aware of the prejudicial impact that disclosure of the particular gang could have, and previously determined that the risk of prejudice outweighed any probative value of that information. When the FBI agent nonetheless mentioned the name of the gang, the district court immediately sustained the objection and ordered the jury to disregard it. When one juror reported that she had read something about the gang in question, the district court carefully examined the juror to ensure that she could disregard that' information in reaching a verdict. Finally, the court gave a closing instruction limiting the jury’s use of the gang information solely to under
stand the relationship between Lemus and Montano.
Having observed the jury and all of the testimony, and having given considered thought to this issue in light of the entire proceeding, the district court determined that a mistrial was not necessary. The district court did not abuse its discretion in so determining.
CONCLUSION
AFFIRMED in part, VACATED in part, and REMANDED. We VACATE the 50 gram finding and the sentence, which was entered pursuant to 21 U.S.C. § 841 (b)(1)(A)(viii). We REMAND for re-sentencing for violation of 21 U.S.C. § 841(a)(1) only.