I join fully in Chief Justice Lucas’s lead opinion. I write separately to emphasize that rewriting of statutes and initiatives to salvage their constitutional validity, while within our authority, is a task to be undertaken sparingly and cautiously. As Justice Mosk points out, and as this court has often recognized, the judicial role in a democratic society is fundamentally to interpret laws, not to write them. The latter power belongs primarily to the people and the political branches of government; the courts should exercise the power only in compelling circumstances, and only where we can be certain that our duty to implement the legislative intent requires us to act. The present case, I believe, is not one in which we can act with the requisite certainty; consequently, respect for the limitations of the judicial role demands we refrain from exercising the extraordinary power to rewrite an unconstitutional law.
I
The considerations of comity and prudence that have governed judicial action in the delicate area of rewriting unconstitutional legislation cannot be reduced to a simple formula. In some circumstances, the nature of the constitutional flaw and the history of the legislation clearly command salvage by rewriting, rather than total invalidation. In other circumstances, reformation is unworkable and unwarranted, and invalidation is the only course the courts may take.
Courts hesitate least to rewrite a law when all that is called for is to “construe” a vague or overbroad law by specifying or narrowing its application. Two justifications are apparent for judicial willingness to impose such a narrowing construction, even if it requires, in effect, insertion of limiting language into a law enacted without such limitation. First, a narrowing reformation decreases rather than increases the number of cases in which the statute will apply. In most circumstances a court can reasonably *676assume the legislative body that decreed the statute’s application to a broad set of cases would prefer it remain applicable to a subset of those same cases, rather than be invalidated completely.
Second, that a court faced with an unconstitutionally overbroad statute may limit its scope by a narrowing construction is a well-accepted principle of constitutional review. (See, e.g., United States v. 12 200-Ft. Reels of Film (1973) 413 U.S. 123, 130, fn. 7 [37 L.Ed.2d 500, 507, 93 S.Ct. 2665]; Welton v. City of Los Angeles (1976) 18 Cal.3d 497, 505-506 [134 Cal.Rptr. 668, 556 P.2d 1119].) Consequently, we may presume legislative bodies are aware of the possibility broad statutory language touching on constitutional interests will, if held unconstitutional, be subject to narrowing in the courts.
If construction to save a vague or overbroad statute is at one end of the reformation continuum, at the other are those cases in which the proposed rewriting involves substitution of a new and substantially different scheme for that created by the legislative body. (See, e.g., Blair v. Pitchess (1971) 5 Cal.3d 258, 283 [96 Cal.Rptr. 42, 486 P.2d 1242, 45 A.L.R.3d 1206] [declining to reform unconstitutional claim and delivery law that could not be repaired without “completely redrafting its provisions”]; Dillon v. Municipal Court (1971) 4 Cal.3d 860, 870-871 [94 Cal.Rptr. Ill, 484 P.2d 945] [declining to construe parade licensing ordinance as imposing ministerial duty to issue permit on stated conditions, where ordinance on its face imposed no such duty and appeared to grant officials uncontrolled discretion].) Under these circumstances reformation is truly “a practice we strive to avoid.” (United States v. Treasury Employees (1995) 513 U.S____[130 L.Ed.2d 964, 987, 115 S.Ct. 1003].)
When the task of saving a statute from unconstitutionality has involved more than severance of a clause or insertion of narrowing language, courts have properly proceeded with caution. One important consideration is whether the court can be confident it possesses sufficient expertise in the subject matter to draft an enforceable and practical law. Not surprising, therefore, is that the primary example of such redrafting in California jurisprudence, In re Edgar M. (1975) 14 Cal.3d 727 [122 Cal.Rptr. 574, 537 P.2d 406], involved a rule for operation of the justice system, an area in which this court could claim both expertise and general authority. (See also United States v. Thirty-Seven Photographs (1971) 402 U.S. 363, 372 [28 L.Ed.2d 822, 831-832, 91 S.Ct. 1400] [justifying addition of constitutionally required time limits to statute by observation that “we possess as much expertise as Congress” in determining appropriate time limits for prosecutorial and judicial functions].)
In some cases—for example, when a statute is extended by eliminating an unconstitutional exception—rewriting an unconstitutional law may require *677implementing one legislative policy while allowing another to go uneffectuated. The propriety of such reformation has been said to depend on “the intensity of commitment to the residual policy and ... the degree of potential disruption of the statutory scheme” that elimination of the exception would cause. (Welsh v. United States (1970) 398 U.S. 333, 365 [26 L.Ed.2d 308, 333, 90 S.Ct. 1792] (conc. opn. of Harlan, J.).)1
Thus, the historical commitment to the residual policy has been deemed especially important. If the substance of the law after rewriting is of longstanding and fundamental nature, an inference naturally arises the legislative body would strongly prefer the law to remain in force despite a change in operation, as by extension to some undesired applications. (See Welsh v. United States, supra, 398 U.S. 333, 365-366 [26 L.Ed.2d 308, 333-334] (conc. opn. of Harlan, J.) [long tradition of exempting conscientious objectors is compelling reason to repair conscientious objector statute, even though repair requires extension to those the statute had excluded]; Speegle v. Board of Fire Underwriters (1946) 29 Cal.2d 34, 46-47 [172 P.2d 867] [court could sever unconstitutionally vague exception from the Cartwright Act, leaving the rest of the act intact, in part because the fundamental policy against combinations in restraint of trade had existed at common law and been codified in the act before the exception was created].)
II
The lead opinion demonstrates that neither the reformation proposed by petitioners, nor that devised by Justice Baxter, would effectuate all three of the key policy judgments embodied in Proposition 73’s contribution limits. (Lead opn., ante, at pp. 664-670.) I agree that fact alone is a sufficient reason *678for not attempting to salvage the initiative by rewriting it. The considerations outlined above, in my view, also militate against attempting judicial repair of Proposition 73.
Both the proposed reformations would require this court not merely to sever or limit provisions of the initiative, but actually to replace the sections held unconstitutional with substantially different provisions that were never considered by the voters. The unconstitutional provisions lie at the heart of the statutory scheme created by the initiative. We are thus faced not with the task of correcting minor, peripheral flaws to save an otherwise valid statute, but with rewriting the law’s central operational provisions. On the reformation continuum sketched earlier, the proposed repairs fall at the far extreme of practices courts properly seek to avoid; only the most compelling circumstances should move the court to step so far outside its normal role, and even then, only when the court can be confident its actions effectuate the fundamental policy objectives of the legislative body.
In his concurring and dissenting opinion, Justice Baxter emphasizes that the voters who enacted Proposition 73 wanted campaign contribution limits. This court, however, cannot accurately assess the intensity of that residual policy, divorced from any consideration as to the appropriate amounts and timing of the contributions allowed. The voters enacted a scheme quite specific on the amounts and timing of allowable contributions; this court, in my view, has no grounds on which to presume they cared only that limits be imposed, whatever the limits might be. In the absence of any direction in the initiative itself, we cannot be confident the electorate wanted the judiciary to rewrite the central provisions of the law, should those provisions prove invalid.
Were we to undertake the task of reformation, we would have for guidance no historical commitment to contribution limits. Nor can we say invalidation would disrupt an existing operational scheme on which parties not before us are depending. The subject matter, finally, is complex and falls outside the area in which an appellate court can claim special expertise or thorough knowledge. Were we, therefore, to undertake to rewrite Proposition 73 in order to save it, we would be assuming legislative powers, but without the compelling circumstances that have occasionally led courts to do so in the past. I join the majority of my colleagues in declining to make such an unwarranted departure from our normal judicial duties.
Where the legislative scheme involves provision of government benefits, the disruption caused by invalidation of the statute may be the most important consideration. Aware that “[a] judgment invalidating a benefit statute takes something of value from a class not before the court” (Ginsburg, Some Thoughts on Judicial Authority to Repair Unconstitutional Legislation (1979) 23 Clev. St. L.Rev. 301, 321), courts are especially willing to stretch their interpretive powers, or to forthrightly rewrite a law by eliminating an unconstitutional exception, where the alternative is invalidation of an established benefit program. (See, e.g., Califano v. Westcott (1979) 443 U.S. 76, 92 [61 L.Ed.2d 382, 395, 99 S.Ct. 2655] [extending statute in part because alternative would “terminate needy families’ entitlement to statutory benefits”].) Conversely, we chose invalidation over extension of an automatic survivorship benefits provision in circumstances where invalidation would not “result in any substantial disruption of the overall workers’ compensation scheme, or any unfair hardship to an employee’s survivors.” (Arp v. Workers’ Comp. Appeal Board (1977) 19 Cal.3d 395, 410 [138 Cal.Rptr. 293, 563 P.2d 849].) The balance between the excised exception and the residual law can sometimes be measured numerically: “When a member of a relatively small class seeks access to a right or benefit enjoyed by a substantially larger class, it seems a fair guess that the legislature would prefer to preserve what it authorized even if the preservation requires enlarging the benefited class.” (Ginsburg, supra, 28 Clev. St. L.Rev. at p. 318.)