EASTERBROOK, Circuit Judge.
Trapshooters at Boxhorn’s Big Muskego Gun Club compete in leagues. The Club, located in Muskego, Wisconsin, has facilities for both competitive and individual trapshooting. Before and after shooting, the competitors may use the Club’s restaurant and tavern. The Club also sells guns, shells, and the other paraphernalia of trapshooting.
Between 1971, when Dick and Lois Bennett bought the Club, and 1978, the Club’s business grew about 15% per year. The Club was remodeled in 1974 to handle more customers, and in 1977 the Bennetts decided to enlarge the Club again. The Club hired Glancey Co. as the general contractor, paying it a fixed fee for the project. Gerald Glancey, the alter ego of Glancey Co., hired subcontractors. The subcontractors handling about 66% of the work had collective bargaining agreements with craft unions. Durski Electrical Contractors Co., which Glancey hired to do some electrical work, did not. Neither did Glancey, who personally performed carpentry work.
Forty-two teams of shooters at the Club in 1977-78 were sponsored by unions represented in the Milwaukee County Labor Council. When someone noticed that nonunion workers were remodeling the Club, Ralph Gondek, the business agent for Local 494 of the International Brotherhood of Electrical Workers, asked Dick Bennett to use union workers exclusively. The two did not get along well. Bennett said that the project was more than half completed and that he would not switch contractors; Gondek suggested that Bennett consider his enlightened self-interest; they parted unpleasantly.
[1018]*1018The Council’s Recreation Committee, displeased with the thought that its requests might be taken lightly, withdrew its sponsorship of the 42 teams of shooters. It put the Club’s name on the Council’s “We Do Not Patronize” list. Later it dispatched a team, led by Gondek, to the entrance of the Club. For four days in late July or early August 1978 these people distributed 2,000 handbills reading:
PLEASE DO NOT PATRONIZE BIG MUSKEGO GUN CLUB
The Milwaukee County Labor Council has withdrawn 42 teams in protest against Mr. Bennett’s decision to use non-union construction workers.
Our request to use Union Workers was scoffed at by Mr. Bennett.
You can help by withdrawing your team from the Big Muskego Gun Club.
Electricians Local Union 494 IBEW
These handbills reached almost all of the Club’s customers. The Council’s newspaper printed a photograph of the handbill in its July-August issue. And after holding a trial the district court found that the handbillers engaged in picketing. 620 F.Supp. 1073, 1076 (E.D.Wis.1985). The district court found almost everything the Unions did to be unlawful, and it awarded the Club more than $230,000 in damages under §§ 301 and 303 of the Labor-Management Relations Act, 29 U.S.C. §§ 185 and 187.
I
The defendants (the Unions) dispute the district court’s findings of fact, principally the conclusions that picketing occurred and that the Club was not the primary employer in the dispute. None of the findings is clearly erroneous.
The four handbillers stationed themselves at the entrance to the long side road leading to the Club. One blocked each incoming car, forcing it to slow down or stop. The handbiller then offered the driver of the car a handbill. No one was impeded from entering the Club if he chose. 620 F.Supp. at 1076. Several witnesses (including four unaffiliated with the Club) testified that they saw people carrying picket signs. One whose testimony the court believed said that the sign read
“AFL-CIO Job Action Local 494.” The district judge found that picketing occurred at least some of the time. The handbillers all denied picketing, and they point out that (a) the adverse witnesses differed substantially in their descriptions of the pickets, their signs, and their patrolling, and (b) the printing bill for their activity did not include the cost of printing placards. This is not enough to show that the findings are clearly erroneous. Nothing compels printers to bill for handbills and placards on a single invoice, if the Unions used new (instead of generic) placards at all. And witnesses often disagree fundamentally. There were two possible views of the evidence. The district court’s view therefore prevails. Anderson v. City of Bessemer City, 470 U.S. 564, 105 S.Ct. 1504, 1511-13, 84 L.Ed.2d 518 (1985).
The conclusion that this was secondary pressure also stands. If the Club were the “employer” of the non-union laborers, the picketing would be primary. The district court concluded, however, that Glancey was the employer. The Club hired Glancey for a fixed price, and Glancey then decided to hire Durski Co. and Glancey personally to do about a third of the work. This decision was not the Club’s to make, and without the right to control the apportionment of work between union and nonunion labor the Club was not the primary disputant. See NLRB v. Enterprise Ass’n of Steam Pipefitters, 429 U.S. 507, 97 S.Ct. 891, 51 L.Ed.2d 1 (1977); NLRB v. Denver Building Council, 341 U.S. 675, 71 S.Ct. 943, 95 L.Ed. 1284 (1951). The district court was entitled to find that the general contractor, not Bennett, controlled who would do the work, making the Club a neutral, 620 F.Supp. at 1076-77. This is therefore a case of secondary pressure.
II
A union’s request that people not patronize an employer is governed by § 8(b)(4) of the National Labor Relations Act, 29 U.S.C. § 158(b)(4), which states that it is an unfair labor practice for a union or its agents:
(ii) to threaten, coerce, or restrain any person engaged in commerce or in an industry affecting commerce where ... an object thereof is—
(B) forcing or requiring any person to cease using, selling, handling, trans[1019]*1019porting, or otherwise dealing in the products of any other producer, processor, or manufacturer, or to cease doing business with any other person, ... Provided, That nothing contained in this clause (B) shall be construed to make unlawful, where not otherwise unlawful, any primary strike or primary picketing;
... Provided further, That for the purposes of this paragraph (4) only, nothing contained in such paragraph shall be construed to prohibit publicity, other than picketing, for the purpose of truthfully advising the public, including consumers and members of a labor organization, that a product or products are produced by an employer with whom the labor organization has a primary dispute and are distributed by another employer, as long as such publicity does not have an effect of inducing any individual employed by any person other than the primary employer [to cease work].
The Unions asked the Club’s shooters to cease patronizing the Club, for the purpose of putting pressure on the Club to put pressure on Glancey. So there are two questions: first whether the approaches to the consumers — the picketing, handbilling, photograph, and We Do Not Patronize list —“threaten, coerce, or restrain any person” within the meaning of § 8(b)(4)(ii); second, whether any of the activities is saved by the publicity proviso (the Proviso) in the “provided further” clause, which applies to publicity other than picketing.
The picketing asked customers not to patronize the Club as a whole. It was therefore coercive under NLRB v. Retail Store Employees, 447 U.S. 607, 100 S.Ct. 2372, 65 L.Ed.2d 377 (1980) (Safeco). It would have been deemed coercive under NLRB v. Fruit & Vegetable Packers, 377 U.S. 58, 84 S.Ct. 1063, 12 L.Ed.2d 129 (1964) (Tree Fruits) as well, because the Unions did not ask the customers to limit their boycott to the “product” — the restaurant and tavern at the Club — of the employers with which the Unions had a primary dispute.
Safeco indicates that the handbilling, too, is coercive or restraining under § 8(b)(4)(h) when designed to close the whole business. See also Soft Drink Workers v. NLRB, 657 F.2d 1252, 1263-68 (D.C.Cir.1980); Florida Gulf Coast Building Trades Council, 273 N.L.R.B. No. 172 at 4 (Jan. 15, 1985). The Unions wanted to stop anyone from using the Club, which might have induced the Club to change its construction contractor. The Proviso applies to handbilling, but only if the handbill “truthfully advis[es] the public ... that a product or products are produced by an employer with whom the labor organization has a primary dispute and are distributed by another employer”. The district court found, 620 F.Supp. at 1078, and the parties agree on appeal, that the Club “distributes” a “product” of Glancey and Durski. See also Edward J. DeBartolo Corp. v. NLRB, 463 U.S. 147, 155-57, 103 S.Ct. 2926, 2931-33, 77 L.Ed.2d 535 (1983). We need not decide whether this is correct.
The Proviso cannot protect the Unions’ handbilling. The Unions argue that their handbill was true, but the Proviso does not apply to truth in the abstract. It protects only publicity on a particular subject, and then only if the publicity is truthful. “The only publicity exempted from the prohibition is publicity intended to inform the public that the primary employer’s product is ‘distributed by’ the secondary employer.” DeBartolo, 463 U.S. at 155, 103 S.Ct. at 2932. The Unions’ handbill does not state that the Club is distributing the product of Glancey and Durski, with whom the Unions have a dispute. It therefore is not protected by the Proviso, even if all the statements in the handbill are true. (The district court found that they are not, however. 620 F.Supp. at 1078-79.) Imagine a handbill that stated in huge letters: “THE UNIONS DEMAND THAT YOU NOT PATRONIZE BIG MUSKEGO GUN CLUB”. The handbill would be truthful, but it would not be protected by the Proviso, because it does not tell the customers that the dispute is with an identified primary employer. The handbills in this case suffer from the same problem.
The district court found that the photograph of the handbill in the Unions’ newsletter, and the inclusion of the Club on the “We Do Not Patronize” list, also violated the statute. 620 F.Supp. at 1079. Neither the photo nor the list identifies the primary dispute, so if the case turned on the application of the Proviso the photo and the list would be treated just like the hand-[1020]*1020billing. The court did not separately discuss how the list and the photo might be deemed threatening or coercive, however, and as an original matter it is hard to see how they could be. Words and pictures work by persuading rather than by coercing. Whatever a customer makes of a glowering handbiller, or a person taking down his name at the door (the Club alleged that the Unions used this tactic, but the district court did not make a finding on the point), few feel threatened by a list in a newspaper. Tammany Hall felt “threatened” by Thomas Nast’s political cartoons, but the threat was a figure of speech for good political persuasion;1 the threats and coercion of which § 8(b)(4) speaks must be more direct. (We put aside any questions that would have arisen if the Unions had printed the names of those who patronized the Club, calling for ostracism. Cf. the “List of Scabs” in Old Dominion Letter Carriers v. Austin, 418 U.S. 264, 94 S.Ct. 2770, 41 L.Ed.2d 745 (1974).)
We grant that the statutory language refers to threatening or coercing the secondary employer — the “person engaged in commerce or in an industry affecting commerce” (§ 8(b)(4)(H)). The question is whether pressure comes home to the secondary employer. But this does not mean just any kind of pressure. Tree Fruits held that the pressure generated by a boycott of a single product in a multi-product store does not amount to threats or coercion under the statute. Similarly, the parties agree that the pressure generated by the Unions’ decision to withdraw their sponsorship of trapshooting teams is not statutory coercion, threat, or restraint. We suppose that if Gondek had held a press conference to explain what the Unions were doing, this too would have been outside the statute, even if the press had omitted Gondek’s identification of the primary dispute (or if Gondek had not made this clear). So too with a long article in the newsletter calling Bennett a dirty name, of which there are many in the colorful vocabulary of the labor movement, and concluding with a call for the faithful to “recognize their duty as loyal members.” For essentially the same reasons, the photo and list are not covered as threats, coercion, or pressure.
Three additional considerations support this conclusion. First, picketing, patrolling, and on-the-spot handbilling have been viewed as particularly effective forms of pressure. People may be induced by the presence of others to act in a way they otherwise would not, and the effect (the restraint or coercion) on the secondary employer will be greater. The degree of effect is a continuum, as the difference between Safeco and Tree Fruits demonstrates. Words and photos in a union newsletter fall closer to the slight effect of Tree Fruits than to the shutdown threatened in Safeco.
Second, the legislative history of § 8(b)(4)(ii)(B) and the Proviso, from both 1947 (when they were enacted) and 1959 (when they were substantially amended) show that Congress was concerned about picketing the secondary employer’s place of business and about blocking delivery of “hot cargo”, not about publicity in newsletters. The Supreme Court summarized the history in Tree Fruits, Safeco, and NLRB v. Servette, Inc., 377 U.S. 46, 84 S.Ct. 1098, 12 L.Ed.2d 121 (1964). It would be pointless to rehash the debates. The Court in Tree Fruits summed up the legislative history as showing that Congress wanted to deal with “an ‘isolated evil’ believed to require proscription of peaceful customer picketing at secondary sites” (377 U.S. at 63, 84 S.Ct. at 1066). See also id. at 71, 84 S.Ct. at 1070. None of the legislative history discussed in those cases, and none we have found, suggests the slightest legislative concern with publicity away from the secondary employer’s place of business.
Third, Servette, 377 U.S. at 55, 84 S.Ct. at 1104, stressed that Congress had a “profound ... concern that the unions’ freedom to appeal to the public for support of their case be adequately safeguarded.” Concern about both prudential and constitutional issues raised by a broad reading of coercion under § 8(b)(4)(H) led to the enactment of the Proviso. See Tree Fruits, 377 U.S. at 69, 84 S.Ct. at 1069. The legislation must be understood in light of the decision to [1021]*1021leave leeway to ask customers to do what customers have a right to do — withhold their patronage for any reason that strikes their fancy. In at least one case the Labor Board has held that the publication of a handbill in a newsletter can be covered by § 8(b)(4)(ii). Hospital & Service Employees Union, 263 N.L.R.B. 996 (1982). The Ninth Circuit remanded this case for further consideration, especially in light of the first amendment. Hospital and Service Employees Union v. NLRB, 743 F.2d 1417, 1427-28 (1984). Constitutional hurdles to regulating the contents of newsletters— even commercial ones, see Pacific Gas & Electric Co. v. Public Utilities Commission, — U.S. —, 106 S.Ct. 903, 89 L.Ed.2d 1 (1986); Central Hudson Gas & Electric Corp. v. Public Service Commission, 447 U.S. 557, 100 S.Ct. 2343, 65 L.Ed.2d 341 (1980) — make us chary of concluding that Congress meant any such thing. See also Lowe v. SEC, 472 U.S. 181, 105 S.Ct. 2557, 86 L.Ed.2d 130 (1985) (construing the Investment Advisers Act, in light of the first amendment, not to require the registration of investment newsletters).
It is not that we think Congress wants to make life easy for courts by allowing them to evade hard constitutional questions. Often Congress wants to press right to the limit of its constitutional power, and surveying those limits is what courts are for. See Henry J. Friendly, Benchmarks 210-11 (1967). Courts may not trim back statutes to keep them out of a danger zone; Congress can have as much danger as it desires. Cf. CFTC v. Schor, — U.S.—, 106 S.Ct. 3245, 3252, 92 L.Ed.2d 675 (1986). But when the legislative history shows that a statute was written carefully to avoid a particular issue, we take the drafters at their word. They asked for a construction that would give unions leeway to publicize and avoid potential collision with the first amendment. Here, as in Lowe, such a construction means a conclusion that the contents of newsletters are not within the prohibition. (It would be a different matter if the handbill were a copy of the newsletter containing a copy of the handbill, just as it is a different matter when an attorney advertises in person. See Ohralik v. Ohio State Bar Ass’n, 436 U.S. 447, 98 S.Ct. 1912, 56 L.Ed.2d 444 (1978). So, too, it might be different if the printed word threatened retaliation against those who used the Club. Cf. NLRB v. Village IX, Inc., 723 F.2d 1360, 1367-68 (7th Cir.1983). We mention these possible variations to make clear the limits and hazards of generalization.)
It does not matter that the photo and the list may be deemed part of a consolidated campaign to put pressure on the Club. The withdrawal of sponsorship of the teams also is part of the campaign; a press conference would be too; yet neither is coercion within the meaning of the statute. A union’s devices may be assessed cumulatively, but for the reasons we have expressed the things to cumulate are the sorts of pressure brought to bear at the secondary employer’s place of business.
Ill
The conclusion that the statute does not reach the contents of the Unions’ newsletter makes the first amendment issue much simpler. The Unions state that hand-billing and picketing for the purpose of providing information to the Club’s customers is protected speech. The handbills are speech, but the process of picketing and handbilling is more than speech. It is a combination of speech and economic or social pressure, an implied threat of ostracism or injury (economic or physical) to those who take the dare. That is why the process is covered in the first place as threat, coercion, or restraint. The Supreme Court has held repeatedly that activities covered by § 8(b)(4)(ii) and not sheltered by the Proviso are not sheltered by the first amendment either. International Longshoremen’s Ass’n v. Allied International, Inc., 456 U.S. 212, 226-27, 102 S.Ct. 1656, 1664-65, 72 L.Ed.2d 21 (1982); Safeco, 447 U.S. at 616, 100 S.Ct. at 2378 (Powell, J., joined by Burger, Stewart & Rehnquist, JJ.), 618-19 (Stevens, J.). See also Soft Drink Workers, 657 F.2d at 1267-69.
IV
The district court awarded more than $230,000 in damages. The court first concluded that the Unions’ activity cost the Club 416 of its 1,892 members. 620 F.Supp. at 1080. Then it held that each member was worth about $90 to the Club per year: $450 in gross receipts less about $370 in variable costs (shot, targets, and so on). Id. at 1080-81. The Club had been growing at about 15% per year until July [1022]*10221978, and the court reckoned that this “lost” group of members also would have grown at 15% per year (had it stayed), because existing members bring in new ones. So the court awarded the Club $90 X 416 for the first year after the handbilling, $90 X 478 for the second year (416 x 1.15 = 478), $90 X 550 for the third year, and $90 X 632 for the fourth year, all augmented by prejudgment interest. The court cut off further counting as of July 1982, four years after the handbilling stopped, on the ground that after that the award would be too speculative. 620 F.Supp. at 1081-82.
The Unions did three things that harmed the Club. They picketed and passed out handbills at the Club, they put items in their lists and newsletters, and they withdrew sanction from their leagues. Only the first of these was unlawful, and the damages must be computed to prevent penalizing lawful conduct. Teamsters v. Morton, 377 U.S. 252, 260-62, 84 S.Ct. 1253, 1258-59, 12 L.Ed.2d 280 (1964). The district court concluded that it was “impossible” (620 F.Supp. at 1080) or not “feasible” (id. at 1082) to separate the effects of the lawful and unlawful components of the conduct, but it did not explain why. We may assume that if the Unions’ own behavior made it impossible to find the damages attributable to the illegal portion of its conduct, the Union must bear the consequences. E.g., Mead v. Retail Clerks Ass’n, 523 F.2d 1371, 1376-79 (9th Cir.1975) (analogizing suits under § 303 of the LMRA to suits under the antitrust laws). We may even assume something that is highly debatable (and on which we express no view): that when similar legal and illegal acts have a common objective it is appropriate to award damages stemming from both. Abreen Corp. v. Laborers’ International Union, 709 F.2d 748, 759 (1st Cir.1983), cert. denied, 464 U.S. 1040, 104 S.Ct. 702, 79 L.Ed.2d 167 (1984). Still it is usually possible to separate the effects, at least roughly, when there are different kinds of acts at different times. E.g., MCI Communications Corp. v. AT & T, 708 F.2d 1081, 1161-64 (7th Cir.), cert. denied, 464 U.S. 891, 104 S.Ct. 234, 78 L.Ed.2d 226 (1983); Frito-Lay, Inc. v. Teamsters, 623 F.2d 1354, 1362-65 (9th Cir.), cert. denied, 449 U.S. 1013, 101 S.Ct. 571, 66 L.Ed.2d 472 (1980); Federal Prescription Service, Inc. v. Amalgamated Meat Cutters, 527 F.2d 269, 279-80 (8th Cir.1975). That describes this case.
The Unions withdrew their sanction from the league in mid-July 1978. The league, which comprised 252 shooters, had its last meet on July 17,1978. So far as the record shows, the 42 teams never returned. The picketing did not start until late July or early August. (The district court did not find when it started, but a police log suggests August 8.) The Club insists that it did not notice a decline in the number of shooters until the picketing started. This is consistent with a conclusion that the league stopped play on July 17, because, we were told at oral argument, that was the last date it had planned to compete; the absence of these shooters therefore would have been “normal” until the league did not start up again in the fall. The district court did not make findings about the 252 league members. When did their season end? Would they have returned in the fall but for the picketing? Was the withdrawal of the Unions’ sponsorship, rather than the picketing, responsible for their departure? If so the Club is not entitled to damages based on the profits these 252 shooters would have supplied. We do not say that the district court must subtract the loss from these shooters; it is essential, however, that the court determine when and why they departed. It also may be possible (though it will be more difficult) to decide whether the remaining 164 members departed because of the picketing or because of the Unions’ newsletter. It may be appropriate to conclude that those who left at the time of the picketing did so because of the picketing, and that those who withdrew later did so because of the newsletter. On the other hand, this line may be too thin. Either way, the district court must make the necessary findings of fact rather than offer the bare conclusion that it is “impossible” to link effects and causes. See Nemmers v. United States, 795 F.2d 628, 634 (7th Cir.1986).
Because the damages must be recomputed on remand, we express some misgivings about the methodology the district court employed. It assumed that existing patrons generate new ones, so that damages grow at an exponential rate from year to year. The district court recognized that [1023]*1023after a while the method becomes unreliable; in a few more years 15% compounded growth would produce a stupendous and quite implausible sum. But the need to cut off the damages just when they are getting big suggests that the method is a little fishy. It assumes the conclusion: does the effect of a boycott grow or diminish over time?
A different way to look at things assumes that once the picketing ends people slowly trickle back. For example, the Club loses 400 net members in 1978-79, 300 in 1979-80, 200 in 1980-81, and so on. This is a routine phenomenon — consider what happens to an airline after a strike, as it rebuilds its base of customers and within a year may be larger than before. When the Club lost 416 members, it had space available and could have attracted new members without continually adding facilities. The evidence suggests that something of this sort happened. Although the Club lost 22% of its members in the summer of 1978, its gross revenues did not go down. It simply experienced an end to its period of growth. The missing members were made up somehow, and because the Club had limited space it is not appropriate to assume that it would have enjoyed the 416 members, plus the new members they brought in (more than 200 in four years, at 15% compounded growth), plus the people who actually joined and prevented the Club’s revenues from falling.
We do not insist that the district court assume that the effect of the picketing declines. The nature of damages must be determined in light of the facts of the case. But whether the damages increase or decrease with time must be resolved, not assumed, and the parties should focus on whether the Club was able to replace the 416 missing members in some other way (or whether any of them came back; the Club apparently knows the identities of its members). Any method that projects an exponential increase in damages must explain why the Club could not make alternative use of its space and why the 15% rate of growth would continue. The Club would become congested, which would retard growth or compel it to incur new costs to expand, costs the district court’s calculation did not include. A constant rate of growth assumes no change .in general economic conditions and no competition. Was there competition from new clubs or new sports? Did other sports clubs grow at undiminished rates in 1978-82? And it assumes that there is an inexhaustible base of potential trapshooters. Such predictions rarely hold up. A simple projection based on the rate of growth of the bowling industry during its peak years in the 1950s and 1960s would have led to the conclusion that by 2000 every adult in the United States would spend most of the day bowling.2 And if the Club should continue growing at 15% per year until 2030, every resident of Wisconsin would be a member. Rates of growth are subject to moderating influences, and a computation of damages that assumes exponential growth must take these influences into account.
The judgment is affirmed to the extent that it holds the picketing and handbilling to be violations of § 8(b)(4)(ii)(B). The remainder of the judgment is vacated, and the case is remanded for further proceedings consistent with Part IV of this opinion. No costs.