A jury convicted Louis Manzella of one count of possession of cocaine with intent to distribute it, in violation of 21 U.S.C. Sec. 841(a)(1), and one count of conspiracy to distribute cocaine and to possess it with intent to distribute it, in violation of 21 U.S.C. Sec. 846. The judge sentenced Manzella to prison for two and a half years on each count, the sentences to run concurrently. Manzella appeals, arguing that there was insufficient evidence to convict him on either count and that the judge should have given an instruction on entrapment.

Lopez, an undercover agent of the Drug Enforcement Administration, began negotiating with Ernest Rizzo to buy a kilogram of cocaine. At a meeting with Tom Apuzzo, Manzella told Apuzzo that Rizzo would need a new source for the cocaine that he wanted to sell to Lopez. Later the same day, Rizzo told Lopez on the phone that Manzella "has to bring the guy [with the cocaine] to me." Lopez and Rizzo agreed to meet the next day at a restaurant to complete the sale. When Lopez arrived at the appointed time, Manzella was there with Rizzo. Manzella told Lopez that he would make a phone call and the cocaine would arrive in 15 minutes. Manzella went to make the call and came back and said the cocaine was en route. When it did not arrive Manzella became angry. Eventually he received a phone call at the restaurant and was overheard saying, "The package won't be ready?" Manzella then told Lopez that the people who had the cocaine couldn't do the deal. Another meeting was set up--this time in a conversation between Lopez and Manzella--but was later postponed because Manzella was unable to procure the cocaine. During this period Manzella was negotiating with Apuzzo, who had found a possible source for the cocaine, Richard Weiss. Eventually Manzella told Lopez that the deal was set for the next day and that he should call Rizzo. Rizzo and Lopez then arranged another restaurant meeting. At the meeting Apuzzo showed up at last with cocaine that Weiss had supplied him, and Apuzzo and Rizzo were then arrested. At trial Manzella testified that he had sought a source for the cocaine out of friendship for Rizzo, and not for money.

The evidence was ample to show that Manzella conspired with Rizzo and Apuzzo to sell cocaine to Lopez. It is true that, so far as appears, only one sale of cocaine was made or even contemplated and that neither the buyer nor the seller can be guilty of conspiracy to sell cocaine, or to possess with intent to sell it, merely by virtue of the relationship established by the sale. United States v. Bascaro, 742 F.2d 1335, 1359 (11th Cir.1984); United States v. Roth, 777 F.2d 1200, 1205 (7th Cir.1985) (dictum). If a crime is so defined that an agreement is required to commit it, the presumption is that that minimum agreement cannot also be punished as a conspiracy. Iannelli v. United States, 420 U.S. 770, 779-84, 95 S. Ct. 1284, 1290-93, 43 L. Ed. 2d 616 (1975). The theory of punishing the conspiracy on top of the substantive crime is that a conspiracy is, on average anyway, more dangerous than a one-man crime. A conspiracy involves more people and can therefore commit more crimes; and it can do so more efficiently, by exploiting the division of labor and by arranging concealment more effectively--sometimes through suborning law enforcers. If the crime is defined to require plural actors, the punishment is presumably suited to whatever greater dangerousness the minimum plurality poses; and to add on a punishment for conspiracy would be duplicative--unless there are additional actors, id. at 782 n. 15, 95 S. Ct. at 1292 n. 15; United States v. Carrascal-Olivera, 755 F.2d 1446, 1451 (11th Cir.1985); United States v. Collazo, 732 F.2d 1200, 1206 (4th Cir.1984), and there were here. This was not a simple deal between Rizzo as seller and Lopez as buyer. On the seller's side it was a four-cornered transaction involving Rizzo, Manzella, Apuzzo, and Weiss. A kilo of cocaine is a large quantity and the number of people required to get it to market illustrates how conspiracy makes criminal activities possible that one person would be incapable of. Manzella's role as broker probably was essential and certainly was helpful to the consummation of the transaction. He was as much a part of the conspiracy as a real estate broker is a part of the deal to sell a house. It makes no difference that Manzella was not present at the sale; he had played a significant role, though his negotiations with Apuzzo, in bringing about Weiss's sale to Lopez.

But in the analogy to the real estate broker we encounter the greatest weakness in the government's case--the lack of evidence that Manzella possessed cocaine; for remember that he was convicted for possession as well as conspiracy. A real estate broker does not possess the house he is trying to sell, or a loan broker the money of the bank whose funds he is trying to get for his customer; and while a stock broker often will have custody of his customer's stocks, rarely will he possess the stocks before buying them for the customer. Manzella knew where the cocaine was. He arranged (at first unsuccessfully) for it to be transported to the place of sale. But he never had control over it.

There is, it is true, a doctrine of constructive possession, under which a person can be convicted for possessing cocaine though he does not possess it in a literal sense. The doctrine creates a legal fiction to take care of such cases as that of a drug dealer who operates through hirelings who have physical possession of the drugs. It would be odd if a dealer could not be guilty of possession, merely because he had the resources to hire a flunky to have custody of the drugs. Of course he might be guilty of many other things, not only conspiracy but also violation of the drug "kingpin" statute, 21 U.S.C. Sec. 848, which carries the heaviest noncapital penalties in the federal criminal code; but he would also be guilty of possession with intent to distribute. See, e.g., United States v. Caspers, 736 F.2d 1246, 1249 (8th Cir.1984); United States v. Caballero, 712 F.2d 126, 130-31 (5th Cir.1983).

But as this example suggests, the essential point is that the defendant have the ultimate control over the drugs. He need not have them literally in his hands or on premises that he occupies but he must have the right (not the legal right, but the recognized authority in his criminal milieu) to possess them, as the owner of a safe deposit box has legal possession of the contents even though the bank has actual custody. United States v. Shackleford, 738 F.2d 776, 785 (7th Cir.1984); United States v. Samad, 754 F.2d 1091, 1096 (4th Cir.1984). Mere association with those who possess the drugs is not good enough. Id.; United States v. Tolliver, 780 F.2d 1177, 1182 (5th Cir.1986). There is no evidence that Manzella had possession in this sense. His was the role of a finder, a broker, a bringer together of seller and buyer, much as in United States v. Gallagher, 565 F.2d 981, 984-85 (7th Cir.1977), and United States v. Jackson, 526 F.2d 1236 (5th Cir.1976), both cases where the defendant's conviction on a theory of constructive possession was reversed. Manzella was furious when his principals in the first transaction pooped out but there is no suggestion that he thought that by failing to show up with the drugs as arranged the people who had the drugs had stolen them from him, as would have been the case if he had been the possessor and they the custodians. He had no right to control Weiss's cocaine. Weiss, and later Apuzzo, had the possessory interest in that cocaine. Manzella was again just a broker, a finder working for Rizzo.

The line between broker and seller in the murky depths of the underworld is of course a fine one, and many cases have upheld convictions of middlemen for possession in circumstances superficially like those in the present case. But we can find no case in which the evidence was so weak. In United States v. Esdaille, 769 F.2d 104, 105, 108-09 (2d Cir.1985), the defendant was identified to the buyer of the cocaine as a supplier, led the buyer to an apartment, and emerged a short time later with the cocaine; this was held to be enough evidence that he constructively possessed the cocaine in the apartment. United States v. Davis, 679 F.2d 845, 848, 852-53 (11th Cir.1982), is similar; the evidence supported an inference that the defendant was part of a group of conspirators who jointly controlled the cocaine. There is no indication that Manzella had any control, individually or jointly, over the cocaine that he tried to obtain for Rizzo. In United States v. White, 660 F.2d 1178, 1180, 1182 (7th Cir.1981), where we acknowledged that the evidence of defendant Rogers' constructive possession was much weaker than that relating to his codefendant White, Rogers was present in White's apartment when the sale of heroin was made and accompanied the buyer to her car to get the money for it, so again the evidence pointed to a conspiracy in which the coconspirators jointly controlled the drugs. In United States v. Virciglio, 441 F.2d 1295 (5th Cir.1971), the defendant collected, and retained half of, the purchase price of the machine gun that he was accused of constructively possessing. He had actual possession of many other illegal guns. The court said that, jointly with others, he had "the intention and the power to exercise dominion and control over the machine gun." Id. at 1298. The same cannot be said about Manzella in regard to Weiss's cocaine.

Admittedly there is language in some cases which if taken literally would allow Manzella to be convicted of constructive possession; for he "instigated the sale, negotiated the price, and caused the drug to be produced for the customer," United States v. Felts, 497 F.2d 80, 82 (5th Cir.1974), and he "likely had some appreciable ability to guide the destiny of the drug," United States v. Staten, 581 F.2d 878, 883 (D.C.Cir.1978) (footnote omitted). But in each of these cases the evidence of constructive possession was stronger than in this case. If the quoted language were taken literally, every broker would be guilty of constructive possession, a result inconsistent with the fundamental proposition that constructive possession requires the right or power to control. Some cases hold that ability to assure delivery is enough, in combination with association with the actual possessors, to convict of constructive possession. See, e.g., United States v. Weisser, 737 F.2d 729, 732 (8th Cir.1984); United States v. Wells, 721 F.2d 1160, 1162 (8th Cir.1983). We have no quarrel with this formulation; he who can assure the delivery of a good controls it. But given the difficulties Manzella had in arranging delivery, it can hardly be inferred that he could assure delivery at the time and place of his choosing; the extent of his association with the actual possessors was also more limited (so far as the record shows) than in the cases we have cited.

It could be argued that Manzella was an aider or abettor of Apuzzo's possession, in that he tried, eventually successfully, to put Apuzzo in possession of drugs that Apuzzo could deliver to Rizzo for sale to Lopez. See United States v. Collins, 779 F.2d 1520, 1530-31 (11th Cir.1986); United States v. Harris, 713 F.2d 623, 626 (11th Cir.1983). Aiding and abetting does not have to be separately charged in the indictment. United States v. Galiffa, 734 F.2d 306, 315 n. 11 (7th Cir.1984). But as no mention of aiding and abetting was made at trial, and no instruction on aiding and abetting was given, it comes too late to save the conviction.

The government seeks to save Manzella's conviction for possession by invoking the doctrine of Pinkerton v. United States, 328 U.S. 640, 66 S. Ct. 1180, 90 L. Ed. 1489 (1946), whereby one conspirator can be held liable for the crimes of another if committed in furtherance of the conspiracy--but only if the jury is instructed in this theory of guilt. See United States v. Wozniak, 781 F.2d 95, 97 (7th Cir.1985); United States v. Bascaro, supra, 742 F.2d at 1363. The idea behind the Pinkerton doctrine is that the conspirators are each other's agents; and a principal is bound by the acts of his agents within the scope of the agency. Whether as an original matter this doctrine of commercial transactions should have been transposed to the criminal law--whether it really adds anything besides complication, given the possibility of basing liability on aiding and abetting--are not questions open to us to reexamine. The only question for us is the adequacy of the charge.

In United States v. Zabic, 745 F.2d 464, 474-75 (7th Cir.1984), we upheld the following Pinkerton instruction:

If you find that the defendant is guilty of conspiracy as charged in Count One, you may also find the defendant guilty of a substantive offense as charged in any other count of the indictment, provided that you find that the essential elements of that count as defined in these instructions have been established beyond reasonable doubt, and provided that you also find beyond reasonable doubt,

First, that the offense defined in the substantive count was committed pursuant to the conspiracy, and

Second, that the defendant was a member of the conspiracy at the time the substantive offense was committed.

Under the conditions just defined a defendant may be found guilty of a substantive count even though he did not participate in the acts constituting the offense as defined in the substantive count. The reason for this is that a co-conspirator committing a substantive offense pursuant to a conspiracy is held to be the agent of the other co-conspirators.

See also United States v. DeLuna, 763 F.2d 897, 917 n. 5 (8th Cir.1985). The instruction we have quoted contains every element of the Pinkerton doctrine, arrayed in an order calculated to maximize the likelihood that the jury will grasp this complicated concept.

The district judge in the present case gave a drastically compressed version of this instruction (so compressed that the government's lawyer on appeal said that no Pinkerton instruction had been given at all!):

If you find that the possession of the 1,041 grams of cocaine with intent to distribute was committed in furtherance of the conspiracy and that the defendant was a member of the conspiracy at the time the substantive offense was committed, then you are instructed that a co-conspirator committing an offense in furtherance of the conspiracy is the agent of the other conspirators.

This was meant to convey the idea that if the jury found that Apuzzo (who was not a defendant) had possession of the cocaine with intent to distribute it, and had it as part of the conspiracy with Rizzo, Weiss, and Manzella, the jury should deem Apuzzo the agent of Manzella and convict Manzella as the principal. This indeed is what the Pinkerton doctrine permits but we think the judge must make clearer to the jury what they are being asked to decide. The jury must be made to focus on the coconspirator's act, on whether it is a crime, on whether the coconspirator's guilt of this crime was proved beyond a reasonable doubt, and on whether it was committed in furtherance of the conspiracy in which the defendant participated. The judge's telescoping of these elements in this case was so drastic that we think it improbable in the extreme that the jury actually considered, as an alternative to finding Manzella guilty of constructive possession, finding Apuzzo guilty of possession and attributing his crime to Manzella under the Pinkerton doctrine. The instruction does not say whose possession the judge is thinking about (i.e., Apuzzo's, not Manzella's) and does not even tell the jury that the principal (Manzella) can be guilty of his agent's (Apuzzo's) substantive offense (possession). It is not so much that the instruction is unduly favorable to the government, which in fact proposed a much fuller instruction, but that it fails to present the Pinkerton doctrine in an intelligible form to the jury.

Although Pinkerton instructions less complete than the one we quoted from Zabic have been upheld, see, e.g., United States v. Galiffa, supra, 734 F.2d at 313; United States v. Molina, 581 F.2d 56, 60 n. 7 (2d Cir.1978); United States v. Sperling, 506 F.2d 1323, 1341 n. 27 (2d Cir.1974), all at least state clearly that the defendant can be convicted of a substantive crime committed by his coconspirator in furtherance of the conspiracy. Here all the judge told the jury was that if it found possession in furtherance of the conspiracy, one conspirator can be an agent of the others. This enigmatic and incomplete utterance, in which seemingly unrelated propositions are placed in a causal sequence, was not elaborated into a theory under which Manzella could be punished for Apuzzo's possession. Manzella may well have been guilty under Pinkerton but he had a right to have his guilt pronounced by a jury, not by us, and the form of the charge did not direct the jury's attention to this distinct theory of his violation of the substantive count.

Last, we must decide whether the judge's refusal to give an instruction on entrapment was justified. As a defense to a criminal prosecution "entrapment" means the government's inducing a person to commit a crime who was not predisposed to commit it--in other words, who would not have committed it but for the particular inducement that the government held out. See United States v. Russell, 411 U.S. 423, 433-35, 93 S. Ct. 1637, 1643-44, 36 L. Ed. 2d 366 (1973); United States v. Navarro, 737 F.2d 625, 634-35 (7th Cir.1984). The significance of this qualification is that if the inducement merely affects the timing of the offense--inducing the criminal to commit it at a time and in a place where the government can easily apprehend him and make a case against him--punishing the criminal will, or at least may, reduce the crime rate, by taking out of circulation a person who, had he not been caught, would have committed the same crime, only in different circumstances, making it harder to catch him. But if the inducement was so great that it tempted the person to commit a crime that he would not otherwise have committed, punishing him will not reduce the crime rate; it will merely deflect law enforcement into the sterile channel of causing criminal activity and then prosecuting the same activity.

The inducer ordinarily is a government agent posing as a criminal. That was Lopez's role in this case. But Lopez's inducement of Manzella was distinctly secondary. Lopez induced Rizzo to sell cocaine and it was Rizzo who induced Manzella to assist him in procuring the cocaine for Lopez; Manzella's own testimony is that he was acting to help out Rizzo. Rizzo was not a government agent; he wasn't even an informer, as in Sherman v. United States, 356 U.S. 369, 78 S. Ct. 819, 2 L. Ed. 2d 848 (1958). There is no defense of private entrapment. United States v. Busby, 780 F.2d 804, 806-07 (9th Cir.1986). Private entrapment is just another term for criminal solicitation, and outside the narrow haven created by the defense of necessity or compulsion, the person who yields to the solicitation and commits the solicited crime is guilty of that crime. All crime is a yielding to temptation, the temptation to obtain whatever gains, pecuniary or nonpecuniary, the crime offers. The temptation is a cause of the crime but not a cause that exonerates the tempted from criminal liability if he yields, just as poverty is not a defense to larceny. Cause and responsibility are not synonyms.

One can imagine, though only with difficulty, circumstances in which the government might be said to have entrapped the defendant even though no government agent dangled the bait directly in front of his nose. Suppose Rizzo had no propensity to deal in illegal drugs but Lopez offered him (somehow credibly) a million dollars for an ounce of cocaine. Suppose Rizzo then turned around, as Lopez knew he would, and offered $500,000 to Manzella, who let us assume for the moment also had no proclivities for drug dealing, but like our fictional Rizzo was overwhelmed by the temptation to embark on this criminal activity. Would only Rizzo have a defense of entrapment? Could the government be allowed to do indirectly, through Rizzo as an unwitting agent, what it would not have been allowed to do in face-to-face dealings between Lopez and Manzella--namely, induce Manzella to commit a crime to which he was not predisposed and then prosecute him for the crime? Or should the government's lack of control over persons who are not its agents be a complete defense to any effort to use the conduct of such a person as the basis for a defense of entrapment by someone with whom the government has not dealt directly?

These fascinating questions have divided the other circuits. See the catalog of cases in United States v. Leroux, 738 F.2d 943, 947-48 (8th Cir.1984), and compare United States v. McLernon, 746 F.2d 1098, 1108-09 (6th Cir.1984), with United States v. Cruz, 783 F.2d 1470 (9th Cir.1986). Ours has not yet considered whether there is such an animal as "vicarious entrapment." The district judge, however, answered no. He thought there was enough evidence of predisposition to warrant an instruction if vicarious entrapment could be a defense; but since it could not (in his view), he refused to give the instruction anyway.

We disagree that there was enough evidence, and therefore need not decide whether vicarious entrapment can ever be a defense. At Rizzo's request, Manzella made persistent and eventually successful efforts to arrange a sale of cocaine. The fact that by his own account he did this for friendship rather than for money, far from suggesting a yielding to overwhelming temptations dangled (at one remove) by the government, indicates a predisposition to engage in drug offenses by brokering illegal drug deals. If the government had never approached Rizzo another potential customer might well have done so and Rizzo would have had the same problem in completing the deal--finding a new source for cocaine--as he had with Lopez, and would have turned to Manzella and completed the deal with Manzella's aid. The government did not lure Manzella into the drug business through extraordinary inducements; he was there already, and merely waiting for new opportunities to ply his trade. There is no evidence that Lopez offered more than the market price for cocaine; and this in itself implies that if Lopez hadn't entered the market someone else would have offered Rizzo the same price, precipitating the same crimes only in circumstances where the chance of apprehending Rizzo and his coconspirators, including Manzella, would have been diminished. There was no factual basis for an instruction on entrapment. United States v. Nations, 764 F.2d 1073, 1079-80 (5th Cir.1985).

To summarize, we affirm Manzella's conviction for conspiracy but reverse his conviction for possession. Since the error in instructions was a trial error not related to the sufficiency of the evidence, and since we do not find that there was insufficient evidence to convict Manzella of possession under the theory of the Pinkerton case, the government may if it wants retry him for possession without encountering the bar of double jeopardy. Whether it does this or not, however, Manzella is entitled to be resentenced, if he wants, for conspiracy, even though he received concurrent sentences. See, e.g., United States v. Shively, 715 F.2d 260, 269 (7th Cir.1983). The judge may have imposed a heavier sentence on the conspiracy count--erroneously believing that the defendant had properly been convicted of possession as well--than he would have imposed had it been the only count.

AFFIRMED IN PART, REVERSED IN PART, AND REMANDED.