These cases raise important questions concerning the scope of § 8(e) of the National Labor Relations Act, 29 U.S.C. § 158(e), which prohibits “hot cargo” agreements. The primary issue is whether the construction industry proviso to § 8(e) renders lawful clauses in collective bargaining contracts forbidding the employers to subcontract work at any construction site to a firm not having a contract with the signatory union. Additional issues are whether unions may picket or strike to induce an employer to enter these agreements and whether the agreements, once entered, may be enforced by picketing or strikes. We conclude that the subcontracting clauses fall within the construction industry proviso and consequently do not violate § 8(e). We also conclude that unions may picket or strike to obtain such agreements, but not to enforce them. The orders of the National Labor Relations Board are accordingly enforced.
I.
Facts.
Two orders of the Board are under review here pursuant to 29 U.S.C. § 160(e) and (f). The first concerns a collective bargaining agreement between Oregon-Columbia Chapter of the Associated General Contractors of America, Inc. (AGC) and Local 701 of the International Union of Operating Engineers (Engineers). AGC is an association of some 200 employers engaged in construction in Oregon and southwest Washington. These employers have delegated their collective bargaining authority to AGC. AGC and Engineers entered into a collective bargaining agreement effective from June 1, 1975, through May 31, 1980, which contained the following clause:
Article VIII
Subcontractors and Other Employers.
Section 1. Employers shall not contract any work covered by this Agreement to be done at the site of the construction, alteration, painting or repair of a building, structure or other work to any person, firm or company who does not have an existing labor agreement with the Union covering such work.
*1305Subsequent clauses1 provide for a grievance procedure leading to arbitration and further provide that should the parties fail to comply with the arbitrator’s decision, then “[e]ither party may take such action as they deem necessary to enforce the findings .... ” The effect of the latter provision is to permit either party to employ self-help, such as strikes or lockouts, that would otherwise be prohibited by another article of the agreement.
Upon complaint by Pacific Northwest Chapter of the Associated Builders & Contractors, Inc., Engineers and AGC were charged with violating § 8(e)2 of the National Labor Relations Act, which prohibits agreements between a labor organization and an employer in which the employer agrees to refrain from using the products of another employer or to cease doing business with any other person. The case was submitted to the Board upon stipulated facts and the Board found that, while the agreement fell within the prohibition of § 8(e), it was exempted from the coverage of that provision by the construction industry proviso applicable to agreements “between a labor organization and an employer in the construction industry relating to the contracting or subcontracting of work to be done at the site of the construction, alteration, painting, or repair of a building, structure, or other work. . . . ” The Board also ruled that the clause authorizing enforcement of the subcontracting provision by self-help was invalid. All parties petitioned for review and the Board sought enforcement of its order.
It shall be an unfair labor practice for any labor organization and any employer to enter into any contract or agreement, express or implied, whereby such employer ceases or refrains or agrees to cease or refrain from handling, using, selling, transporting or otherwise dealing in any of the products of any other employer, or to cease doing business with any other person, and any contract or agreement entered into heretofore or hereafter containing such an agreement shall be to such extent unenforceable and void: Provided, That nothing in this subsection shall apply to an agreement between a labor organization and an employer in the construction industry relating to the contracting or subcontracting of work to be done at the site of the construction, alteration, painting or repair of a building, structure, or other work
The second decision of the Board concerns Woelke & Romero Framing, Inc., (Woelke), a framing subcontractor in the construction industry in Southern California. Woelke was a party to a collective bargaining agreement entered on July 5, 1974, with the United Brotherhood of Carpenters and Joiners of America, AFL-CIO (Carpenters), a group which included Locals Number 235 and 944 involved in this litigation. That agreement expired on June 15, 1977. From about June 3, 1977, to August 4, 1977, Carpenters and Woelke engaged in collective bargaining for the purpose of negotiating a successor agreement. They broke off negotiations when they reached an impasse over several issues.3 One such issue was the union’s demand for a clause in *1306the contract that prohibited Woelke from subcontracting work at any construction jobsite “except to a person, firm or corporation, party to an appropriate, current labor agreement with the appropriate union, or subordinate body signatory to this Agreement.” 4 In support of Carpenters’ demand for this contract provision, Locals 235 and 944 picketed Woelke’s construction sites, causing some work stoppages. The two locals were charged with violation of § 8(bX4)(i) and (iiXA) of the Act5 for picketing to require an employer “to enter into any agreement which is prohibited by section 8(e).”'
The case was submitted to the Board upon stipulation and the Board found that the proposed subcontracting agreement was not prohibited by § 8(e) because it fell within the construction industry proviso, and that picketing to obtain such a clause therefor did not violate § 8(b)(4). Woelke petitions for review of the Board’s order, and the Board seeks enforcement.6
Both decisions of the Board were reviewed by a panel of this court. 609 F.2d 1341 (9th Cir. 1979). On the authority of Connell Construction Co. v. Plumbers & Steamfitters Local Union No. 100, 421 U.S. 616, 95 S.Ct. 1830, 44 L.Ed.2d 418 (1975), the panel held that the subcontractor clauses were prohibited by § 8(e) and did not fall within the protection of the construction industry proviso because their applicability was not limited to jobsites where members of the signatory union were employed at some time during the construction project. Having found the present clauses unenforceable, the panel did not reach the questions whether picketing or striking either to obtain or enforce a lawful subcontractor clause would violate § 8(e). 609 F.2d at 1351. This court subsequently ordered rehearing before a limited en banc panel as provided by Public Law 95-486, 92 Stat. 1629, 1633 and Rule 25 of this court.
II.
Issues.
Four issues are presented for determination in these cases:
*1307(1) Are the subcontracting clauses “secondary” and therefore within the scope of § 8(e)?
(2) If the clauses are within the scope of § 8(e), are they exempted from its proscription by the construction industry proviso?
(3) If the subcontracting clauses are lawful, does it nevertheless violate § 8(bX4) for a union to picket or strike to induce an employer to agree to one?
(4) If the subcontracting clauses are lawful, does it nevertheless violate § 8(b)(4) to enforce them by picketing or strikes?
III.
Applicability of § 8(e).
The subcontractor clauses in the present case clearly require the signatory employer to cease or refrain from using the products of another employer or to cease doing business with another person. The clauses consequently fall within the literal language of § 8(e). In National Woodwork Manufacturers Association v. NLRB, 386 U.S. 612, 87 S.Ct. 1250, 18 L.Ed.2d 357 (1967), however, the Supreme Court held that § .8(e) is not to be read as expansively as its language seems to indicate. The Court pointed out that the purpose of Congress in enacting § 8(e) was to prohibit voluntary arrangements that reached the same result as coercive secondary boycotts prohibited by § 8(b)(4)(ii)(B) of the Act. The Court therefore held that an agreement between a union and an employer which prohibited certain kinds of subcontracting was permissible so long as the objective was to preserve work traditionally performed at the jobsite by employees of the primary employer. Such a “work preservation” clause was viewed as primary and therefore did not constitute an unfair labor practice.7 The test whether an agreement was primary and permissible was “whether the agreement or its maintenance is addressed to the labor relations of the contracting employer vis-a-vis his own employees.” Id. at 645, 87 S.Ct. at 1268-69. An agreement was secondary and prohibited when “the agreements and boycotts were tactically calculated to satisfy union objectives elsewhere.” Id. at 644, 87 S.Ct. at 1268.
By the National Woodwork test, union signatory clauses such as those in issue here are secondary in objective. They do not seek to preserve work for the employees of the contracting employer by limiting the amount of subcontracting. Nor do they seek to protect the wage and benefit standards of those same employees in their present employment by merely limiting subcontracting to third parties who maintain equivalent standards. Instead, these union signatory clauses permit unlimited subcontracting so long as the subcontractors have an agreement with the contracting union. Such clauses are properly viewed as secondary, for they are intended to satisfy union objectives involving employees and employers outside of the bargaining or work unit. Griffith Co. v. NLRB, 545 F.2d 1194 (9th Cir. 1976), cert. denied, 434 U.S. 854, 98 S.Ct. 171, 54 L.Ed.2d 124 (1977); NLRB v. National Maritime Union, 486 F.2d 907 (2d Cir. 1973), cert. denied, 416 U.S. 970, 94 S.Ct. 1993, 40 L.Ed.2d 559 (1974).
It is contended by Engineers that the objective of preserving work within the entire multi-employer bargaining unit of AGC is a primary goal and that this fact should take its subcontracting clause outside of the scope of § 8(e). This argument fails for two reasons. First, the clause in question does not limit subcontracting to firms that are members of AGC. Outsiders having an agreement with Engineers are eligible, even though they and their employ*1308ees are wholly foreign to the multi-employer bargaining unit. The clause therefore is not a primary work preservation clause. NLRB v. National Maritime Union, 486 F.2d at 914. Second, § 8(e) applies in any event to an agreement which attempts to preserve work for an entire multi-employer unit rather than for single employers’ employees. This court dealt with the issue in NLRB v. Joint Council of Teamsters No. 88, 338 F.2d 23, 28 (9th Cir. 1964):
Respondents seem to argue that [the clause] falls within the exception to section 8(e) because it preserves the job opportunities of employees within the multi-employer bargaining unit by prohibiting subcontracting to any employer outside that unit. But if this argument were accepted, the exception to section 8(e) would permit precisely what the section itself was intended to prohibit: An agreement by an employer to boycott another unless the latter entered into a union contract.
Accord, NLRB v. National Maritime Union, supra.
The union’s further contention that work preservation for the employees of each contractor was one of the objectives underlying the union subcontracting clause, does not compel us to alter our conclusion. Regardless of the union’s other objectives, if “an object of a clause is to aid union members generally rather than members of the unit, the object is secondary and unlawful.” National Maritime Union, 196 N.L. R.B. 1100, 1101 (1972), enforced, 486 F.2d 907 (2d Cir. 1973), cert. denied, 416 U.S. 970, 94 S.Ct. 1993, 40 L.Ed.2d 559 (1974) (emphasis supplied).
It is true that the primary-secondary distinction is particularly difficult to maintain in the construction industry, where contractors and subcontractors assemble their work forces for specific short-term jobs. There is a great deal of flexibility and interchangeability of employees between bargaining units. As a result, a subcontractor clause may well serve the interests of union members in the primary bargaining unit who may later find themselves working for a subcontractor. While we believe that fact to be of great significance for purposes of determining the scope of the construction industry proviso, we do not view it as determinative of the initial applicability of § 8(e) itself. See NLRB v: National Maritime Union, 486 F.2d at 913-14. The primary-secondary distinction between separate employers working the same job-site in the construction industry has been adhered to in the past, NLRB v. Denver Building & Construction Trades, 341 U.S. 675, 71 S.Ct. 943, 95 L.Ed. 1284 (1951), and we do not feel free to disregard it here. Moreover, the consequences of viewing union signatory clauses as primary would be not only to remove them from the scope of § 8(e) but also to remove picketing to enforce such agreements from the prohibition of § 8(b)(4)(ii)(B). For reasons stated in Section VI of this opinion, we believe that permitting coercive enforcement of union signatory clauses would be contrary to the intent of Congress.
We accordingly hold that the union signatory clauses in these cases are secondary and within the scope of § 8(e). Their validity consequently depends upon their eligibility for inclusion within the construction industry proviso to that section.
IV.
Scope of the Construction Industry Proviso to § 8(e).
In order to ascertain the scope of the construction industry proviso, it is necessary to examine its legislative history in the light of the practices existing in the industry at that time and the law then applicable. In so reviewing the origins of the 1959 Landrum-Griffin Act of which § 8(e) and its proviso were a part, we recognize that our initial conclusions as to its meaning are subject to any modifications or corrections that may be compelled by a fair reading of Connell Construction Co. v. Plumbers & Steamfitters Local Union No. 100, 421 U.S. 616, 95 S.Ct. 1830, 44 L.Ed.2d 418 (1975). Indeed, it is the effect of Connell that provides the primary ground of controversy in *1309the present litigation. We believe, however, that the Connell decision is best analyzed after a review of the state of the industry and the legislative and administrative history of the proviso governing it.
(A) Construction Industry Background.
The primary characteristics which differentiate the construction industry from most others are well stated in the Senate Report on the 1959 amendments:
The occasional nature of the employment relationship makes this industry markedly different from manufacturing and other types of enterprise. An individual employee typically works for many employers and for none of them continuously. Jobs are frequently of short duration, depending upon various stages of construction.
S.Rep.No. 187, 86th Cong., 1st Sess. 27, [1959] U.S.Code Cong. & Admin. News, pp. 2318, 2344; I Legislative History of The Labor-Management Reporting and Disclosure Act of 1959 423 (hereinafter Leg. Hist.). These peculiarities of the construction industry created special problems for both unions and employers long before 1959. Unions had to overcome difficulties of organizing temporary workers and of maintaining wage and benefit standards in unstable employment conditions. Employers required a ready source of skilled workers available for each new job, and had to be able to forecast labor costs accurately. To satisfy these needs, unions and employers hammered out a number of mutual accommodations. One such arrangement was the closed shop, which was prevalent in the industry by the 1920’s.8 A second was negotiation of area-wide agreements covering all construction work of a certain type, with multiple contracting employers agreeing to the standards set in those agreements.9 Of necessity, these contracts dealt with wages, hours and working conditions of employees yet to be hired for particular jobs by particular employers.10 A third device was the agreement to restrict subcontracting in various ways — to firms maintaining union standards, or to firms that were unionized,11 or to firms that agreed to bind themselves to contracts with the union signatory to the subcontracting clause.12
Some of these devices were not necessarily consistent with the organizational scheme of the National Labor Relations Act in its original form. The Act initially caused no disruption in the pattern of collective bargaining in the construction industry, however, because the National Labor Relations Board declined to exercise jurisdiction over that industry.13 The TaftHartley Amendments to the Act, passed in 1947, led to a reversal of the Board’s position. Those amendments prohibited striking or picketing to coerce an employer to refrain from dealing in the products of any other employer or to cease to do business *1310with any other person. § 8(b)(4)(A).14 Subsequent to the amendments, the Board’s General Counsel determined that the construction industry would be regulated by the Board.15
One effect of applying the amended Act to the construction industry was the invalidation of collective bargaining agreements entered before any employees were hired on the job. Guy F. Atkinson Co. and J. A. Jones Construction Co., 90 N.L.R.B. 143 (1950) , enforcement denied on other grounds, 195 F.2d 141 (9th Cir. 1952). The amended Act was also applied to prohibit a construction union from picketing a jobsite in order to force the general contractor with whom it had a collective bargaining agreement to cease doing business with a particular subcontractor on the job whose employees were non-union. NLRB v. Denver Building & Construction Trades Council, 341 U.S. 675, 71 S.Ct. 943, 95 L.Ed. 1284 (1951) . The picketing was found to violate § 8(b)(4)(A)’s prohibition of coercive secondary boycotts.
These applications of the amended National Labor Relations Act to the construction industry were sufficiently disruptive of existing practices that there was agitation for modification of the Act, particularly in regard to prehire practices and secondary a-ctivity. In 1954, President Eisenhower called for amendments to permit prehire agreements and to shorten to 7 days the time by which a nonunion worker could be required to join a union under a union-shop agreement. 100 Cong.Rec. Ill, [1954] U.S. Code Cong. & Admin. News 1531, 1532. Again in 1959, when calling for amendments to tighten the prohibition against secondary boycotts, the President added that provisions were needed “to make clear that secondary activity is permitted . . . under certain circumstances, against secondary employers engaged in work at a common construction site with the primary employer.” S.Doc. 10, 86th Cong., 1st Sess.; 105 Cong.Rec. 1297; I Leg.Hist. 82.
Accommodation to the special conditions of the construction industry was consequently an important consideration that found expression in the 1959 amendments to the Act. Those amendments, however, also embodied another major goal of Congress at the time — that of limiting “hot cargo” agreements.
In June 1958, the Supreme Court had decided Local 1976, United Brotherhood of Carpenters v. NLRB, (Sand Door), 357 U.S. 93, 78 S.Ct. 1011, 2 L.Ed.2d 1186, Sand Door stated that a union did not violate § 8(b)(4)(A) by securing the voluntary participation of an employer in a boycott of another employer’s product. Although the Court said it found no occasion to determine the legality of “hot cargo” clauses (by which an employer agreed to boycott nonunion firms) as such, id. at 107, 78 S.Ct. at 1020, Sand Door was widely interpreted as permitting them. The decision was seen as creating a loophole in the Act, and it contributed heavily to the move toward amending the secondary boycott provision of the Act.
*1311(B) The Landrum-GrUfin Amendments and § 8(e).
The 1959 revisions of the Act accomplished the congressional purposes in various ways. The clearest accommodation to the peculiar conditions of the construction industry was found in § 8(f),16 which authorized unions and employers in that industry to enter prehire agreements fixing terms and conditions of employment of workers yet to be hired. This provision amounts to an exception to §§ 8(a)(1), (2) and 8(b)(lXA), which make it an unfair labor practice for both the employer and the union to enter a collective bargaining agreement before the union, represents a majority of employees in the unit. International Ladies Garment Workers Union v. NLRB, 366 U.S. 731, 81 S.Ct. 1603, 6 L.Ed.2d 762 (1961); see NLRB v. International Association of Bridge, Structural & Ornamental Iron Workers, 434 U.S. 335, 344-45, 98 S.Ct. 651, 657-58, 54 L.Ed.2d 586 (1978). Section 8(f) also shortened to 7 days the period for nonunion workers to join the union under a union-shop agreement.
Conflicting policies were reflected clearly in the passage of § 8(e), which was intended to narrow or close the loophole for “hot cargo” agreements created by the Sand Door decision. The Senate bill prohibited such agreements only in the trucking industry,17 while the House bill outlawed them everywhere.18 The Conference Committee adopted the House version, but added the provisos creating exceptions for the garment and construction industries.19 The garment industry exception was unlimited as to time and place,20 while the construction industry proviso excepted agreements “relating to the contracting or subcontracting of work to be done at the site of the construction, alteration, painting or repair of a building, structure or other work....’’
*1312It is the nature and meaning of the construction industry proviso’s restriction to the jobsite that gives rise to the present controversy. The contractors assert that the proviso is intended to authorize subcontracting clauses in collective bargaining agreements only when union workers are present at the jobsite, so that the presence of nonunion workers there might lead to friction. The unions claim that the intent was to authorize subcontracting clauses whether or not union workers were at the jobsite, so long as the clauses refer to work to be performed at the construction site. Although the matter is not free from doubt, we believe that the legislative history favors the latter view.
(C) Legislative Discussion of the Construction Industry Proviso.
The legislative history regarding the construction industry proviso is much less exhaustive than that surrounding the garment industry proviso,21 but there are nevertheless indications of congressional intent in the records. Senator Kennedy, the chairman of the Conference Committee, said of the two provisos, that “[b]oth changes were necessary to avoid serious damage to the pattern of collective bargaining in these industries.” 105 Cong.Rec. 17899; II Leg.Hist. 1432. Unlike the garment industry, the construction industry was not exempted from the provisions of § 8(bX4) that prohibited striking or picketing to coerce a secondary boycott; it was merely exempted from § 8(e)’s prohibition of agreements between a union and an employer by which the employer would boycott others. The Conference Committee stated its intent in regard to the proviso as follows:
The committee of conference does not intend that this proviso should be construed so as to change the present state of the law with respect to the validity of this specific type of agreement relating to work to be done at the site of the construction project or to remove the limitations which the present law imposes with respect to such agreements.... To the extent that such agreements are legal today under section 8(b)(4) . . ., the proviso would prevent such legality from being affected by section 8(e).
H.R.Rep.No. 1147, 86th Cong., 1st Sess. 39, [1959] U.S.Code Cong. & Admin. News, p. 2511; I Leg.Hist. 943. Senator Kennedy analyzed the proposed § 8(e) as follows:
The first proviso under new § 8(e) of the National Labor Relations Act is intended to preserve the present state of the law with respect to picketing at the site of a construction project and with respect to the validity of agreements relating to the contracting of work to be done at the site of a construction project.
Agreements by which a contractor in the construction industry promises not to subcontract work on a construction site to a nonunion contractor appear to be legal today. They will not be unlawful under § 8(e)... .
It should be particularly noted that the proviso relates only to the “contracting or subcontracting of work to be done at the site of the construction.” The proviso does not cover boycotts of goods manufactured in an industrial plant for installation at the jobsite, or suppliers who do not work at the jobsite.
105 Cong.Rec. 17900; II Leg.Hist. 1433 (emphasis supplied).
As Senator Kennedy made clear, unions continued to be forbidden to picket or strike at a construction site for secondary objec*1313tives, which meant that the holding of Denver Building Trades, supra, was left in effect. On the other hand, the Conference Report and Senator Kennedy made it equally clear that unions and employers could enter subcontracting agreements as permitted by law prior to the enactment of the new § 8(e).
It seems more likely than not that the “present law” referred to by the Conference Committee did not restrict the agreements that Senator Kennedy said “appear to be legal” to jobsites where union workers were present. The Sand Door decision, supra, which supplied much of the impetus for § 8(e), had legitimized secondary boycotts the objectives of which clearly went beyond the mere avoidance of shoulder-to-shoulder friction between workers. It is true that the construction industry proviso narrowed the geographical scope of the Sand Door rule by introducing the jobsite limitation, but we do not interpret the words of the proviso as indicating an intent to narrow the substantive scope of the Sand Door rule on the jobsite.
It should be noted that shortly before enactment of the proviso, the District of Columbia Circuit had upheld a subcontractor agreement the terms of which were as broad as those presently in issue. Operating Engineers Local Union No. 3 v. NLRB, 266 F.2d 905 (D.C.Cir.), cert. denied sub nom. St. Maurice v. NLRB, 361 U.S. 834, 80 S.Ct. 86, 4 L.Ed.2d 76 (1959).22 While that case dealt with an interunion dispute, the terms of the secondary agreement were not restricted to that or any other situation of jobsite conflict. In upholding the agreement, the Court referred to the Board’s reasoning “that the preclusion of future work to a prospective subcontractor in the circumstances did not justify the Board in holding that the rights of unidentified employees of such prospective employer were adversely affected.” 266 F.2d at 908.
Moreover, a survey of major labor contracts in existence in 1959 revealed that more than 50 of them contained clauses restricting subcontracting to firms complying with the terms of the prime employer’s agreement. Lunden, Subcontracting Clauses in Major Contracts, 84 Monthly Labor Review 579 (Part I), 715 (Part II) at 715 (1961). While the survey does not adequately differentiate between union signatory clauses and union standards clauses, it is nevertheless clear that numbers of broad subcontractor agreements were in existence.23 The survey canvasses various limi*1314tations appearing in subcontracting clauses in the construction industry, but makes no mention of any restrictions to the jobsite or references to the problem of friction between union and nonunion workers there. Id. at 715-16. That there may not have been a consistent pattern of subcontractor agreements of the type in issue here is not crucial. What does emerge from the contemporaneous scene is that some such agreements existed and were legal and that Congress did not appear to intend a change in that condition.
V.
(A) The Connell Decision.
It was against this industrial and legislative background that the Supreme Court decided Connell Construction Co. v. Plumbers & Steamfitters Local Union No. 100, 421 U.S. 616, 95 S.Ct. 1830, 44 L.Ed.2d 418 (1975). Connell was an antitrust suit brought by a general contractor against a union that was picketing in order to coerce the contractor to enter an agreement to deal only with subcontractors having a collective bargaining agreement with the union. The union expressly disclaimed any interest in representing the general contractor’s employees. The picketing was successful and the contractor sought to annul the ensuing agreement on the grounds that it violated the antitrust laws.24 One of the defenses raised by the union was that the agreement could not be a violation of the antitrust laws because it was authorized by the proviso to § 8(e) of the National Labor Relations Act. The Supreme Court rejected that defense and held that broad subcontracting agreements applicable to every jobsite where the general contractor might hire subcontractors were invalid when entered outside of the context of a collective bargaining relationship between the union and the general contractor.
The employers who attack the subcontracting clauses in the present case argue that Connell’s language necessarily requires a conclusion that subcontracting clauses are permissible only when they serve the purpose of avoiding jobsite friction between union and nonunion workers employed at the same jobsite. So interpreted, Connell would invalidate the agreements in issue here at least until the presence of union workers at a particular jobsite triggered their application. We concede at the outset that the Connell opinion contains considerable emphasis on problems of shoulder-to-shoulder friction between union and nonunion workers at the jobsite, and additional emphasis on antagonism of the 1959 Congress to “top-down” organizing that is one of the effects of subcontractor agreements. Nevertheless, we believe that this reading of the Connell decision misconstrues the Supreme Court’s view of § 8(e) and gives too narrow a scope to the construction industry proviso. We think that the prevail*1315ing thrust of Connell is to the contrary and favors a broader view of the exemption. Our conclusion is supported in particular by the Connell opinion’s emphasis on the absence of a collective bargaining relationship, its references to legislative history and its characterization of the interests served by the proviso.25 Our view is reinforced by the adverse practical consequences that would follow from the narrower reading of the proviso. Each of these factors bears elaboration.
The first and most obvious distinction of Connell from the present cases is that the agreements in Connell arose outside of the context of a collective bargaining relationship. This fact was of course crucial for antitrust purposes, because the national policy of encouraging collective bargaining has served as the basis for a judicially created labor exemption from the antitrust laws.26 The absence of such an interest was pointedly referred to in Connell:
In this case, Local 100 had an interest in representing Connell’s employees. The federal policy favoring collective bargaining therefore can offer no shelter for the union’s coercive action against Connell or its campaign to exclude nonunion firms from the subcontracting market.
421 U.S. at 626, 95 S.Ct. at 1837.27 The focus of the Connell decision was therefore understandably upon the situation created by subcontracting clauses in the absence of a collective bargaining agreement with the prime contracting employer.
That focus on the part of the Court was not merely the product of the antitrust setting of Connell, however. The parties themselves had framed the issues toward the same end. The union contended that all subcontractor agreements were validated by the plain words of § 8(e). The employer responded by urging that “despite the unqualified language of the proviso, Congress intended only to allow subcontracting agreements within the context of a collective-bargaining relationship . . .,” but not those involving a “stranger” contractor. Id. at 627, 95 S.Ct. at 1837.28 The subject of *1316controversy, then, was the validity of subcontractor agreements outside of the collective bargaining context, and the Connell opinion must to a large degree be read in that light. The Court’s conclusion, reached after a discussion of the legislative history and policies behind § 8(e), was clearly limited to the situation before it:
We therefore hold that this agreement, which is outside the context of a collective bargaining relationship and not restricted to a particular jobsite, . .. may be the basis of a federal anti-trust suit. ...
Id. at 635, 95 S.Ct. at 1841. The holding of Connell, then, does not in the strict sense control the cases before us.
The Connell opinion did, however, discuss at length the reach of § 8(e) and its construction industry proviso, upon which the present cases turn. It therefore becomes necessary to interpret the language of Connell and apply it to the situation of the cases at bar where a collective bargaining relationship existed between the unions and the contractors with whom the agreements were sought.29 A major question is whether the presence or absence of a collective bargaining relationship, obviously significant in the antitrust context of Connell, remains pivotal when the issue is purely one of labor law.
We believe that there is a complex of factors affecting labor relations in the construction industry that renders the collective bargaining relationship crucial in the present cases and we find the underlying policies reflected in parts of the Connell opinion itself. We also believe that these factors exist at the construction jobsite but are by no means limited to the problem of shoulder-to-shoulder friction between union and nonunion workers there. These two themes — the nature of the collective bargaining relationship in the construction industry and the manner in which these characteristics are exhibited at the jobsite — are interwoven in Connell as they are in the present controversy. We will attempt, however, to discuss them separately even though each will necessarily intrude upon the other.
The legislative history already discussed should make it clear that the outstanding peculiarity of the construction industry is that the work force of any given employer — contractor or subcontractor — is highly flexible and grows or diminishes or even disappears depending on the temporary contracting or subcontracting arrangements at any particular jobsite. As Congress recognized in passing § 8(f), collective bargaining cannot even await the presence of workers on the job. Moreover, when there are workers already employed by a general contractor, the interests of those employees are not permanently or inevitably bound to that single employer. The employer has an option on any given project to use some of his own workers, to subcontract to an employer who will use those same workers, or to subcontract to one who will use others represented by the same union, a different union, or no union. Under those conditions the employees and their unions are faced with severe problems not only of maintaining wage levels, but of protecting benefits such as retirement pensions and paid vacations that depend upon continuity of employer contributions.
*1317When a union representing a contractor s employees seeks to restrict subcontracting to firms having a collective bargaining agreement with the union, the union in fact pursues a number of objectives some of which might fairly be characterized as primary and others as secondary. The welfare of the employees in the primary bargaining unit is clearly likely to be affected by an employer’s choice to hire a subcontractor who may himself employ those same employees. On the other hand, the union clearly has secondary goals that may include organizing the employees of a subcontractor who have no connection with the primary bargaining unit.
It is quite true that when a union seeks to serve the interests of employees in its bargaining unit not as employees of their existing employer, but as potential employees of some other employer, the arrangement is legally viewed as secondary. National Woodwork, 386 U.S. at 644-45, 87 S.Ct. at 1268; National Maritime Union, supra. We do not quarrel with this well-established rule, nor with the rule that when primary and secondary objectives are combined, an agreement is regarded as secondary for purposes of the National Labor Relations Act. NLRB v. Denver Building & Construction Trades Council, 341 U.S. at 689, 71 S.Ct. at 952, 95 L.Ed. 1284 (1951).
We do, however, find sufficient support in the legislative history of the 1959 Amendments to suggest that Congress was sensitive to both the primary and secondary considerations involved in subcontractor arrangements and wished not to interfere with these mixed union objectives when they could be achieved by agreement with the prime employer. Even though a con-
struction union s interest in subcontracting at the jobsite is secondary so as to bring it within § 8(e), it is far more direct and involves the union’s members in the primary bargaining unit far more than is the case when unions outside of the construction industry pursue secondary objectives.
The importance of this fact is that it renders the existence of a collective bargaining relationship highly significant. When a union represents employees of a prime contractor and seeks a subcontracting agreement, it has the interests of the employees in its primary bargaining unit at stake to at least some degree. In the Connell situation, the union is utterly without any objective that could be characterized as primary or direct. Because the union in Connell disclaimed any interest in representing the employees of Connell, all of its goals were necessarily secondary and could not be connected to any primary representational interest.
The significance of the restriction of the § 8(e) proviso to work performed at a job-site may now become more clear. It is at the jobsite that all of the special characteristics of the construction industry are manifested. It is true that one of these characteristics is that union workers are traditionally hostile to the presence of nonunion workers on the same job at the same time. But that is clearly not the only or even the major characteristic exhibited. It is at the jobsite where the general contractor can exercise the flexibility to use its own workers or those of a subcontractor, where the flexibility of subcontracting arrangements may lead to multiple employers hiring a few workers for short periods of time,30 and where working conditions applicable to one
*1318set of workers may very well affect others working on the same project.31
We believe that the Connell opinion is sensitive both to the interconnection of interests between employees of the various contractors and subcontractors and to the broad manner in which any or all of these factors may be exhibited at the jobsite. In dealing with the legislative history of the construction industry proviso, the opinion in Connell points out that the proviso was “explained only by bare reference to ‘the pattern of collective bargaining’ in the industry.” 421 U.S. at 628-29, 95 S.Ct. at 1838. The legislative history discussed above indicates, we believe, that the pattern of collective bargaining included broad subcontracting agreements. To the extent that the just-quoted comment in Connell dismisses the history regarding the pattern of collective bargaining, it may well be explained by the fact that Connell was dealing with a case not considered by Congress, in which there was no collective bargaining relationship.32
In any event, the subsequent language of the Connell opinion appears quite clearly to encompass the multiple considerations affecting the construction industry at the job-site. The opinion suggests that the construction industry proviso may have been adopted “as a partial substitute for an attempt to overrule” Denver Building Trades, 421 U.S. at 629, 95 S.Ct. at 1838. Denver Building Trades had outlawed the picketing of a contractor at a jobsite in order to force it to cease doing business with a nonunion subcontractor working there. The Connell opinion refers to congressional discussion of special problems in the construction industry related both to the § 8(e) proviso and to the attempt to overrule Denver Building Trades. That discussion, according to the Court, “focused on the problems of picketing a single nonunion subcontractor on a multiemployer building project, and the close relationship between contractors and subcontractors at the jobsite.” Id. 421 U.S. at 629-30, 95 S.Ct. at 1838. The Court’s footnote to this statement refers to legislative history that clearly goes beyond the problem of shoulder-to-shoulder conflict between union and nonunion workers at the jobsite. The first excerpt referred to is by Senator Morse whose statement includes the following:
The secondary boycott ban is based upon the concept of two employers — a primary and a secondary employer. The theory is that a neutral employer should not be *1319compelled to suffer because of a labor dispute the union has with somebody else, but the theory breaks down when the two employers are “allies.” Senator Taft said that the secondary boycott section “is not intended to apply to a case where the third party is, in effect, in cahoots with, or acting as a part of, the primary employer.” 33 The National Labor Relations Board and the courts have failed to keep Senator Taft’s concept in mind. The Board has considered subsidiaries to be separate employers. Other enterprises, substantially controlled by the same persons, have been considered separate employers, even though one manufactured a product incorporated into the finished article produced by the other.
The case of the building and construction industry represented probably the most flagrant injustice, where a general contractor is, in effect, entirely in control of the kind of labor relations taking place on a site which he runs. He lets subcontracts based upon price, responsibility, and the ability to handle labor relations. He lets those contracts, very well knowing the kind of labor relations which may exist within any of the subcontractor companies. He is not acting in the dark, Mr. President. He is not innocent of any unfair labor policies on the part of a subcontractor.
105 Cong.Rec. 17880-81; II Leg.Hist. 1425. Even where the sources cited by Connell deal with union workers at the job-site, they reflect a number of interests of the union that affect members of its own bargaining unit as well as the members in secondary employ. The memorandum by representatives Thompson and Udall is illustrative:
Where all the men are employed on the same project, the division into different trades or crafts, each with its own employer, must not be allowed to obscure their common interests — -they work side by side and the wages and working conditions of one trade affect all the others. This situation is utterly unlike the situation in true secondary boycotts, as where carpenters in Los Angeles refuse to work on nonunion doors made in Minnesota by men whose wages could not affect the wages of carpenters in Los Angeles. The illegality of such a boycott would not be affected by the amendment.
. . . [A] reason for outlawing secondary boycotts is that they drag neutral employers into disputes in which they have no interest. Where all the employers affected are bound together in the common work of putting up a single building, none is truly neutral. The undertaking is so integral that each is affected by and can influence the conduct of the others.
105 Cong.Rec. 15541; II Leg.Hist. 1577. These materials cited by the Connell Court suggest a number of reasons why Congress may have wanted to restrict the construction industry’s use of subcontracting agreements to work done on a jobsite. These sources do not limit themselves to and, indeed, place very little emphasis on, the problem of shoulder-to-shoulder friction between union and nonunion workers actually present at the jobsite.
The Connell opinion quotes National Woodwork Manufacturers, supra, to the effect that the construction industry proviso is “designed to allow agreements pertaining to certain secondary activities on the construction site because of the close community of interests there, but to ban secondary-objective agreements concerning nonjobsite work, in which respect the construction industry is no different from any other.” 421 U.S. at 630, 95 S.Ct. at 1839, quoting 386 U.S. at 638-39, 87 S.Ct. at 1265. It is clear that the Connell opinion views the jobsite *1320considerations of National Woodwork as going well beyond the problem of friction among workers,34 because Connell next refers to some courts that have suggested that the proviso serves an “even narrower function” — the alleviation of “the frictions that may arise when union men work continuously alongside nonunion men on the same construction site.” Id. quoting Drivers Local Union No. 695 v. NLRB, 361 F.2d 547, 553 (D.C.Cir.1966). The Connell opinion does not adopt this “even narrower” reading of the proviso, and indeed finds it unnecessary to treat the matter further because in the Connell case the union did not purport to represent any of the interests served by the proviso. 421 U.S. at 631, 95 S.Ct. at 1839.35
We believe that it would do some violence to the language of the Connell opinion to adopt the “even narrower” view of the construction industry proviso which the Court in Connell did not represent to be its own. Yet this consequence would follow from restricting the applicability of subcontractor agreements to those instances where shoulder-to-shoulder jobsite friction can occur. Instead, we think that it is more in accord with the language of Connell to permit those agreements to apply to any jobsite at which an employer with a collective bargaining relationship with the union is a contractor.
We concede that the language of the Connell opinion dealing with congressional antagonism to “top-down organizing” may well lean in a direction contrary to our views. We think it is countered by Connell's recognition of other vital concerns of collective bargaining in the construction industry reflected in the community of interest at the jobsite. There is some “top-down organizing” effect in any subcontracting agreement, including those which all parties must admit would be valid and within the proviso. When there are union workers on the job next to nonunion workers, the enforcement of a subcontracting clause is still likely to have a top-down organizing effect. The answer to this point is simply that Congress recognized that the community of interests on the jobsite justified the top-down organizational consequences that may attend the protection of legitimate collective bargaining objectives there. The proviso is, after all, an exception to the general thrust of the 1959 amendments.36
The Connell opinion’s condemnation of the “unlimited” top-down organizing weap*1321on with which it was confronted must be read in the context of the applicable factual and statutory framework. Because the union in Connell disclaimed any interest in organizing or representing Connell’s employees, its picketing was not limited to the 30 days established for organizational picketing under § 8(b)(7). It could therefore picket indefinitely to obtain its objective of a subcontracting agreement covering all subcontracting by Connell. If the subcontractor clause were held to be legitimate, there would be no limit at all to the coercive force the union could apply. The decision in Connell solved this problem by holding that the subcontracting agreement was an unlawful goal when it was sought outside of a collective bargaining context.
The situation is quite different when a union seeks a subcontractor clause in the context of a collective bargaining relationship. In that situation, the union’s coercive force is limited. If the clause is sought in the course of organizational or recognitional picketing, that picketing will be subject to the restrictions of § 8(b)(7). If the union already represents the contractor’s employees, the employer may resist any attempts to secure a subcontractor agreement by asserting its own bargaining position and presenting its own demands. The existence of a collective bargaining relationship therefore avoids the “unlimited” nature of the organizational tool involved in Connell, where a stranger contractor was powerless to resist unlimited coercion by any means other than signing the demanded subcontractor agreement.
Finally, in summarizing the proviso’s effect, the Connell Court stated:
[W]e think its authorization extends only to agreements in the context of collective bargaining relationships and, in light of congressional references to the Denver Building Trades problem, possibly to common-situs relationships on particular job-sites as well.
Id. at 633, 95 S.Ct. at 1840. While it might be susceptible to another reading, we interpret this language to mean that the existence of a collective bargaining relationship clearly brings a union signatory subcontracting agreement within the proviso tó § 8(e) and that in the absence of a collective bargaining relationship, it is possible that such an agreement may be valid where the union actually has workers on the job that may be threatened with shoulder-to-shoulder friction.37 We consequently believe that the quoted summary supports our conclusion that Connell does not deny the protection of the proviso to the subcontractor agreements in the present cases.
(B) Practical Consequences.
Our reading of Connell to permit broad subcontracting clauses within a collective bargaining relationship not only accords with the apparent thrust of the legislative history, it is supported by the severe practical limitations attending a contrary conclusion. Limitation of subcontracting agreements to sites where there may be shoulder-to-shoulder conflict between union and nonunion workers presents immense practical problems. If unions are confined to negoti*1322ating these agreements only when they find that there are union employees on the job-site, then they must inventory and negotiate over an incredible number of jobsites. AGC itself represents some 200 general contractors. The number of jobsites at which they contract is unquestionably a multiple of that figure. Jobsite by jobsite negotiation of subcontracting agreements would seem burdensome to the point of impossibility. It also appears to run counter to Congress’s accommodation in § 8(f) to the necessity in the industry of entering long term agreements applying to employees not yet hired for jobs not yet in existence. S.Rep.No. • 187, 86th Cong., 1st Sess. 28 (1959); I Leg.Hist. 424.
Even if general subcontracting agreements are permitted to be negotiated wholesale and then are triggered only when union workers appear on the job, the enforcement problems remain severe. The transitory nature of employment at each jobsite, as well as the difficulty of defining the applicable time span during which job-site conflict is threatened, present substantial problems.38 If a rule is adopted that the appearance of any union worker on the job triggers subcontracting agreements covering the entire project, it presents an especially strong incentive for anti-union discrimination by employers.
We are presented, then, with a number of practical and theoretical considerations supporting our decision that the present subcontracting agreements, arrived at in the context of collective bargaining relationships, are within the proviso of § 8(e). First, there are overwhelming practical difficulties in confining the subcontractor agreements to jobsites where actual union-nonunion or interunion conflict occurs among workers. Second, the legislative history as well as the actual conditions existing in the construction industry militate in favor of broad subcontractor clauses in collective bargaining agreements. Third, the Connell decision, which admittedly has language that might point to either result, may fairly be read to support the validity of subcontracting clauses when there is a collective bargaining relationship, regardless of the presence or absence of union workers at the jobsite.
Given these considerations, we reject a narrow interpretation of the proviso which will substantially change the existing practice in the industry. In addition, we believe that due regard should be given to the view of the National Labor Relations Board supporting these agreements.39 It is true that the position of the agency has not been an unwavering one,40 but the Board’s articulation of current labor policy is undoubtedly entitled to weight in our decision.
VI.
Additional Issues.
It is contended by AGC that even if the construction industry proviso authorizes *1323clauses requiring subcontractors to have agreements with unions in general, it does not permit clauses to require a collective bargaining agreement with a particular union. We note, however, that nothing in the language of the § 8(e) proviso suggests such a limitation, even though “particular-union” clauses were a part of industry practice at the time of its enactment. Two months prior to the creation of the proviso, the District of Columbia Circuit upheld a “particular-union” subcontractor agreement in a decision that was submitted to the House committee considering the 1959 amendments. Operating Engineers Local Union No. 3 v. NLRB, supra; 2 House Hearings 803-07. In the light of this history, we do not feel free to find in the proviso a limitation not written there by Congress. We are joined in this conclusion by the District of Columbia Circuit in Donald Shriver, Inc. v. NLRB, supra, 635 F.2d at 884-886. We also note that prohibition of “particular-union” subcontractor clauses would defeat at least some of the interests that we earlier found to be served by the construction industry proviso. One such interest is that of a union in protecting the wages and benefits of its members in the primary bargaining unit who may later work for subcontractors. That concern is shared by no other union, and is best served by a “particular-union” clause.
It is next contended, this time by Woelke, that even if a subcontractor agreement is protected by the construction industry proviso, it is an unfair labor practice for a union to picket or strike in order to induce a contractor to enter such an agreement. The legislative history of the 1959 amendments is in conflict on this point.41 The actual wording of § 8(b)(4)(ii)(A), however, prohibits coercion of any person “to enter into any agreement which is prohibited by [§ 8(e)].”42 The negative implication of this provision is surely that picketing or striking to induce entry into an agreement protected by the proviso to § 8(e) is lawful.
After the 1959 amendments, the Board originally took the position that such agreements had to be voluntarily entered, but its position was uniformly rejected by this and other courts of appeals. Construction Production & Maintenance Laborers Union, Local 383 v. NLRB, (Colson & Stevens Construction Co.), 323 F.2d 422 (9th Cir. 1963); Orange Belt District Council of Painters No. 48 v. NLRB, 328 F.2d 534, 537 (D.C.Cir. 1964); Essex County and Vicinity District Council of Carpenters and Millwrights v. NLRB, 332 F.2d 636 (3d Cir. 1964). The Board subsequently acquiesced in the courts’ view. Northeastern Indiana Building & Construction Council, 148 N.L.R.B. 854 (1964), remanded on other grounds, 352 F.2d 696 (D.C.Cir. 1965). We see no reason to disturb the present law and to read into the Act a limitation that runs counter to the suggestion of its plain words. Nor do we accept the contention that picketing or striking to obtain a subcontractor agreement violates the § 7 rights of employees of subcontractors; they are no more affected by a subcontractor agreement achieved by picketing than by one voluntarily entered.
The final issue concerns the effect of the provision in the AGC contract that permits each party to resort to self-help when the other party does not comply with an arbitration decision made pursuant to *1324the grievance procedure.43 The Board concluded that these provisions authorized picketing or strikes to enforce the subcontractor agreement, and held that the contract was consequently unlawful. The rule that picketing or strikes may not be used to enforce subcontractor agreements arises from the fact that the 1959 amendments exempted the construction industry from § 8(e)’s prohibition against secondary boycott agreements, but did not exempt it from § 8(b)(4)(ii)(B), which forbids economic coercion to enforce such a boycott. The legislative history makes clear Congress’s intent that subcontractor clauses entered pursuant to § 8(e) were to be enforced judicially rather than by self-help,44 and this court has followed that view. NLRB v. International Brotherhood of Electrical Workers, 405 F.2d 159, 163-64 (9th Cir. 1968), cert. denied, 395 U.S. 921, 89 S.Ct. 1772, 23 L.Ed.2d 237 (1969). We have also held that such clauses may be enforced by arbitration. La Mirada Trucking, Inc. v. Teamsters Local Union 166, 538 F.2d 286 (9th Cir. 1976).
Engineers argue, without pressing the point strongly; that self-help to enforce an arbitration award that in turn enforces a subcontractor agreement is not unlawful. The picketing or strike, according to the Union, is simply enforcing the arbitration award and not the underlying agreement. We believe that acceptance of this contention would too easily circumvent the clear intent of Congress and would do violence to the plain words of § 8(b)(4Xii)(B). Nor can we accept the suggestion of the Union that the AGC agreement must be construed to limit the self-help clause to self-help which is determined to be lawful. We must take the agreement to mean what it says. Operating Engineers, Local No. 701 v. NLRB, 578 F.2d 841, 842 (9th Cir. 1978); Donald Schriver, Inc. v. NLRB, 635 F.2d at 886-87.
VII.
Conclusion.
We find that provisions in a collective bargaining agreement forbidding the signatory employer to subcontract work at a construction site to firms that do not have a contract with the signatory union are within § 8(e) of the National Labor Relations Act, but are protected by the construction industry proviso to that section. They need not be limited to jobsites presenting the possibility of conflict between union and nonunion workers. We also find that unions may picket or strike to induce employers to enter such agreements but that, once entered, the agreement may not be enforced by self-help. Clauses authorizing self-help enforcement, directly or through arbitration decisions, are consequently invalid. We accordingly enforce the orders of the Board in their entirety.