Appeal from the United States District Court for the Southern District of Indiana, Indianapolis Division. No. IP 95-C-1148-C-H/G--David F. Hamilton, Judge.
Before Coffey, Easterbrook, and Diane P. Wood,
The opinion of the court was delivered by: Easterbrook, Circuit Judge
In 1995 Indiana enacted a statute making the woman's informed consent a condition to an abortion. Ind. Code §16-34-2-1.1. Even though the text of this law is materially identical to one held constitutional in Planned Parenthood of Southeastern Pennsylvania v. Casey, 505 U.S. 833, 881-87 (1992), a federal district court issued a preliminary injunction preventing the statute from taking effect. A Woman's Choice-East Side Women's Clinic v. Newman, 904 F. Supp. 1434 (S.D. Ind. 1995). Two years later, the district court modified this injunction to permit the state to enforce most of the law, but it blocked enforcement of the requirement that information be provided "in the presence of the pregnant woman, [by] the physician who is to perform the abortion, the referring physician or a physician assistant" (§16-34-21.1(1)). See 980 F. Supp. 962 (1997). After four more years had passed, the judge held a trial and made permanent the injunction as modified in 1997. 132 F. Supp. 2d 1150 (2001).
By requiring information to be supplied "in the presence of the pregnant woman"--rather than by printed brochure, telephone, or web site--the statute obliges the woman to make two trips to the clinic or hospital. This raises the cost (both financial and mental) of an abortion. On the basis of studies concerning similar laws in Mississippi and Utah, the district court concluded that the higher cost will reduce by 10% to 13% the number of abortions performed in Indiana. Some of these women will travel to states that do not require two trips; others will forego an abortion; some who do have an abortion in Indiana will delay that procedure until the second trimester. These consequences show that the law creates an "undue burden" on abortion, the district judge held. Although by the time the district judge entered the permanent injunction we had concluded that the Mississippi study does not warrant condemnation of Wisconsin's law (which like Pennsylvania's requires two trips to the medical facility and a 24-hour wait), see Karlin v. Foust, 188 F.3d 446, 48488 (7th Cir. 1999), the district judge wrote that data from the Utah study, and a new analysis of the Mississippi data, require a different result. The judge also thought that experience in Indiana showing that the demand for abortion did not decline when information was provided on paper or over the telephone implies that the reduction in the number of abortions is attributable to higher cost (a bad reason) rather than to the statutory information (a valid reason).
Indiana's statute reads as follows:
An abortion shall not be performed except with the voluntary and informed consent of the pregnant woman upon whom the abortion is to be performed. Except in the case of a medical emergency, consent to an abortion is voluntary and informed only if the following conditions are met:
(1) At least eighteen (18) hours before the abortion and in the presence of the pregnant woman, the physician who is to perform the abortion, the referring physician or a physician assistant (as defined in IC 25-27.5-2-10), an advanced practice nurse (as defined in IC 25-23-1-1(b)), or a midwife (as defined in IC 34-18-2-19) to whom the responsibility has been delegated by the physician who is to perform the abortion or the referring physician has orally informed the pregnant woman of the following:
(A) The name of the physician performing the abortion.
(B) The nature of the proposed procedure or treatment.
(C) The risks of and alternatives to the procedure or treatment.
(D) The probable gestational age of the fetus, including an offer to provide:
(i) a picture or drawing of a fetus;
(ii) the dimensions of a fetus; and
(iii) relevant information on the potential survival of an unborn fetus; at this stage of development.
(E) The medical risks associated with carrying the fetus to term.
(2) At least eighteen (18) hours before the abortion, the pregnant woman will be orally informed of the following:
(A) That medical assistance benefits may be available for prenatal care, childbirth, and neonatal care from the county office of family and children.
(B) That the father of the unborn fetus is legally required to assist in the support of the child. In the case of rape, the information required under this clause may be omitted.
(C) That adoption alternatives are available and that adoptive parents may legally pay the costs of prenatal care, childbirth, and neonatal care.
(3) The pregnant woman certifies in writing, before the abortion is performed, that the information required by subdivisions (1) and (2) has been provided.
When the litigation began, plaintiffs challenged not only the requirement that advice be delivered in person but also the medical-emergency exception, which they deemed insufficient because it lacks details found in the Pennsylvania statute. The district court certified the medicalemergency issue to the Supreme Court of Indiana, whose interpretation, see A Woman's Choice-East Side Women's Clinic v. Newman, 671 N.E.2d 104 (Ind. 1996), satisfied the district judge. See 980 F. Supp. at 966. Plaintiffs then dropped this objection, leaving only the advice requirement as a ground of contention. *fn1
Indiana makes much of the fact that its statute has never been allowed to operate as written. It relies on United States v. Salerno, 481 U.S. 739, 745 (1987), for the proposition that, except in first amendment cases, a law may be held unconstitutional only when "no set of circumstances exists under which the Act would be valid." Yet in Stenberg v. Carhart, 530 U.S. 914 (2000), without so much as a mention of Salerno, the Court held invalid, in a pre-enforcement challenge, an abortion statute that might have been construed by the state courts to have at least some proper applications. This leaves us with irreconcilable directives from the Supreme Court. The Justices have insisted that courts lower in the hierarchy apply their precedents unless overruled, even if they seem incompatible with more recent decisions. See, e.g., State Oil Co. v. Khan, 522 U.S. 3, 20 (1997); Rodriguez de Quijas v. Shearson/American Express, Inc., 490 U.S. 477, 484 (1989). When the Justices themselves disregard rather than overrule a decision--as the majority did in Stenberg, and the plurality did in Casey--they put courts of appeals in a pickle. We cannot follow Salerno without departing from the approach taken in both Stenberg and Casey; yet we cannot disregard Salerno without departing from the principle that only an express overruling relieves an inferior court of the duty to follow decisions on the books. See also, e.g., Scheiber v. Dolby Laboratories, Inc., 293 F.3d 1014 (7th Cir. 2002) (following Brulotte v. Thys Co., 379 U.S. 29 (1964), even though it is incompatible with the rationale of more recent decisions). Troxel v. Greenville, 530 U.S. 57, 85 n.6 (2000), offers us a way out by calling the language in Salerno a "suggestion," an approach not essential to Salerno's judgment. Given the incompatibility between Salerno's language and Stenberg's holding, it is the language of Salerno that must give way.
Still, to say that a claim is justiciable does not mean that we must ignore the fact that enforcement has not commenced. Plaintiffs rely on predictions about what is likely to happen if Indiana's law were enforced as written. Because Indiana has been disabled from implementing its law and gathering information about actual effects, any uncertainty about the inferences based on other states' experience and how that experience would carry over to Indiana must be resolved in Indiana's favor. This, coupled with doubts about the role of predictions in constitutional analysis, turns out to be important, for reasons explained presently.
Casey stated, and Karlin reiterated, that an informedconsent statute may have effects that differ from the written terms, and that those effects could in principle demonstrate that an innocuous-appearing law actually imposes an undue burden on abortion. But neither decision explained how such factual arguments are to be evaluated: before implementation or after?, using what standards? Normally a court asked to say that a statute will have forbidden effects asks only whether a proper outcome is possible; it does not hold a trial--and, if a district judge nonetheless takes evidence and makes findings, the appellate court will re-examine matters with a heavy presumption favoring the law's constitutional application. See, e.g., Vance v. Bradley, 440 U.S. 93, 111 (1979); National Paint & Coatings Ass'n v. Chicago, 45 F.3d 1124 (7th Cir. 1995). One may say in response that these cases deal with rational-basis review, while abortion implicates fundamental rights. But laws that regulate, not abortion itself, but ancillary issues (such as informed consent), do not affect fundamental rights unless the ancillary rule creates an undue burden on the underlying right. How does the court handle factual disputes that bear on whether an undue burden has been created? It cannot simply assume that a fundamental right has been burdened; that begs the question.
Stenberg shows that the undue-burden standard must be applied at the level of logic, and to the nation as a whole, rather than one state at a time. Nebraska forbade use of "intact dilation and extraction" (D&X), a method of late-term abortion. Stenberg believed that this ban would have unacceptable consequences because it would induce physicians to steer clear of other procedures similar to the D&X. Nebraska's law therefore was held unconstitutional, as an undue burden on abortion, without the need for a trial. Meanwhile a trial had been held in Wisconsin, where the district judge found as a fact that the untoward consequences anticipated in Stenberg would not occur. Planned Parenthood of Wisconsin v. Doyle, 44 F. Supp. 2d 975 (W.D. Wis. 1999), affirmed under the name Hope Clinic v. Ryan, 195 F.3d 857 (7th Cir. 1999) (en banc), remanded, 530 U.S. 1271 (2000), decision on remand, 249 F.3d 603 (2001) (en banc). The Supreme Court vacated our decision without regard to the district court's findings; it was of the view (as we likewise had concluded, 195 F.3d at 872-73) that constitutionality must be assessed at the level of legislative fact, rather than adjudicative fact determined by more than 650 district judges. Only treating the matter as one of legislative fact produces the nationally uniform approach that Stenberg demands. This worked against the partial-birth-abortion laws in Stenberg but has worked in favor of other laws: the Court has held it constitutional to prevent nonphysicians from performing abortions, see Mazurek v. Armstrong, 520 U.S. 968 (1997), without factual inquiries into whether other medical professionals could do the job as safely, and how much prices may be elevated by a physician-only rule.
Findings based on new evidence could produce a new understanding, and thus a different legal outcome; the plurality implied this in Casey, as did we in Karlin. But if the issue is one of legislative rather than adjudicative fact, it is unsound to say that, on records very similar in nature, Wisconsin's law could be valid (as we held in Karlin) and Indiana's law invalid, just because different district judges reached different conclusions about the inferences to be drawn from the same body of statistical work. Because the Supreme Court has not made this point explicit, however, and because the undue-burden approach does not prescribe a choice between the legislative-fact and adjudicative-fact approaches, we think it appropriate to review the evidence in this record and the inferences that properly may be drawn at the preenforcement stage.
The district court found that the two-visit requirement in Mississippi and Utah reduced the number of abortions performed in those states by about 10% compared with neighboring states that do not require multiple visits. The judge also found that the number of abortions performed in Indiana has not declined because of the advice given to women under Ind. Code §16-34-2-1.1, though not necessarily in person (because that aspect of the statute has been enjoined). Indiana asks us to set aside these findings, but review under Fed. R. Civ. P. 52(a) is highly deferential, see Anderson v. Bessemer City, 470 U.S. 564 (1985), and we cannot say that the district court's findings are clearly erroneous. The studies' conclusions were hotly debated on both medical and statistical grounds, but the district judge dealt responsibly with these arguments pro and con, and his findings cannot be upset. But what happened in Mississippi and Utah, a question of historical fact on which appellate review is deferential, does not necessarily ordain what will happen in Indiana--or whether what is likely to happen in Indiana amounts to an "undue burden." That admixture of fact and law, sometimes called an issue of "constitutional fact," is reviewed without deference in order to prevent the idiosyncrasies of a single judge or jury from having farreaching legal effects. Only the findings of historical fact are sheltered by Rule 52(a). Thus our consideration of the studies' significance is not deferential. See Cooper Industries, Inc. v. Leatherman Tool Group, Inc., 532 U.S. 424 (2001) (constitutionality of punitive damages is reviewed de novo); Ornelas v. United States, 517 U.S. 690 (1996) (probable cause for a search or seizure is reviewed de novo in the absence of a warrant); Bose Corp. v. Consumers Union of United States, Inc., 466 U.S. 485, 499 (1984) (factual disputes that determine constitutionality under the first amendment are reviewed de novo). Cf. Los Angeles v. Alameda Books, Inc., 122 S.Ct. 1728 (2002) (without discussing the standard of review, all nine Justices freely substitute their judgment for that of the district and appellate judges on the significance of an empirical study in a constitutional suit). See also Henry P. Monaghan, Constitutional Fact Review, 85 Colum. L. Rev. 229 (1985).
By concluding that the empirical work had been carried out competently, the district judge established (for purposes of this litigation) that abortions dropped in Mississippi, compared to those in South Carolina, during the year after Mississippi enacted a statute requiring two visits. The authors of this study (and its replication in Utah) did not ask how Mississippi compares with Indiana. The study does not include a regression based on the sorts of variables, such as urbanization, income, average distance to an abortion clinic, average price of abortion, and so on, that might enable conclusions drawn from Mississippi to be extrapolated with confidence to other states. That is one reason why we held in Karlin that the Mississippi study was a poor basis for predicting what would happen in Wisconsin, which we thought more similar to Pennsylvania than to Mississippi. 188 F.3d at 485-86.
That shortcoming could have been fixed in one of two ways. First, the authors could have conducted a more comprehensive study, with additional variables and regression coefficients that would reveal their effects. That was not done. Second, the authors (or other scholars) could have gathered data from other states to test whether (and, if so, how) state-specific characteristics affect the results. That was not done either. What has happened in Pennsylvania, Wisconsin, and the other states whose informed-consent laws require two visits? Did Mississippi prove to be a better predictor of Wisconsin than Karlin anticipated, or was the outcome in Wisconsin dissimilar? This record is silent on these matters. Mississippi and Utah, two states with a history of hostility to abortion and very few abortion providers (implying long travel times), may be poor models for other states. Indianapolis has multiple abortion clinics; another in Fort Wayne serves the northeastern portion of the state; women in the northwest and southeast can use not only local providers but also those just across the state lines in Chicago and Louisville. So just as in Karlin the application of the Mississippi data (and now Utah's data) to a different state would be a leap of faith. Here is where the preenforcement nature of this suit matters.
Plaintiffs did try to deal with another problem identified in Karlin: that the original Mississippi study did not try to separate the raw costs of a two-visit requirement from the effects of the information that was provided during the first visit. 188 F.3d at 486-88. The Supreme Court's first two encounters with informed-consent statutes treated these laws as meddling in the physician-patient relation with no valid purpose, and no effect other than to heap pointless costs on women. See Akron v. Akron Center for Reproductive Health, Inc., 462 U.S. 416, 442-49 (1983); Thornburgh v. American College of Obstetricians and Gynecologists, 476 U.S. 747, 759-65 (1986). Casey overruled both of these decisions and held that states may try to persuade women not to abort their pregnancies. Maybe all the Mississippi study reveals is successful persuasion, we observed in Karlin. In this case the plaintiffs tried to separate the effect of information from the effect of making two visits. Since 1997 Indiana has been able to enforce the portion of its informed-consent statute requiring the provision of certain information to women who inquire about abortions. Yet the number of abortions has not declined. This shows, the district judge wrote, that the law lacks persuasive effect; and if a decline in abortions cannot be attributed to persuasion, then the cause must lie in some other and impermissible feature of the law.
Yet this assumes what is to be proven: that Indiana is like Mississippi and Utah, so that the number of abortions would decline 10% or more if the law were enforced as written. Maybe what Indiana's experience since 1997 shows is that Indiana differs from Mississippi and Utah and will not experience a substantial decline, with or without multiple visits. Or maybe what it shows is that presenting the information in person is critical to its persuasive effect. Our education system rests on the premise that information delivered orally, with an opportunity for give-and-take, "takes" better than information delivered exclusively in writing. Otherwise a university would simply mail a syllabus to the freshman class and ask the students to appear four years later for exams. So the fact that advice delivered in writing or over the phone is uninfluential need not imply that advice delivered in person will be uninfluential. Once again the fact that Indiana has been blocked from enforcing its law as written means that the record does not contain evidence needed for accurate assessment of that statute's effects.
Then there is an open question what the 10% reduction reflects. Let us suppose that abortions would decline 10% in Indiana if that state's law were fully enforced. What would the decline signify? One possibility is that many women who strongly want an abortion have been blocked by the cost (in money and time) of multiple visits to the clinic, or because the more times the woman must be absent the greater is the likelihood that an abusive parent, spouse, or partner would discover what the woman has planned and intervene notwithstanding the availability of the emergency bypass, which the Supreme Court of Indiana held to encompass any kind of threat to the woman's health or safety. See 671 N.E.2d at 108-09. Another possibility is that about 10% of all women who have abortions are on the fence between ending the pregnancy and carrying the pregnancy to term, so that even a modest cost tips the scales. If the former, then a twovisit rule might be deemed an undue burden; if the latter, the two-visit rule would not be an undue burden, for only a law that "has the purpose or effect of placing a substantial obstacle in the path of a woman seeking an abortion of a nonviable fetus" (Stenberg, 530 U.S. at 921; emphasis added; quoting from the plurality opinion in Casey) is an "undue" burden. This record does not permit (and the district judge did not make) an inference either way about the reason for the decline in Mississippi and Utah. Perhaps this shortcoming could be rectified by studying the effects of changes in out-of-pocket outlays or travel time as prices change, or clinics open, close, or move locations, but the studies in this record do not address the question.
Since 1992, when the plurality in Casey announced the "undue burden" standard, only two kinds of statute have flunked the test: a law forbidding the "intact dilation and extraction" (D&X) method of abortion (the subject of Stenberg) and a law requiring a woman to notify her husband before obtaining an abortion (discussed in Casey itself). Because the language used to describe the D&X also could be understood to prohibit other procedures that were common (and perhaps essential) to lateterm abortions, Stenberg concluded that the law would forbid abortions altogether for substantial numbers of women. The notification statute did not forbid abortions, but the Court feared that it would come to the same thing for those women whose husbands are likely to respond violently to the notice (if not to any contact from an estranged spouse). The plurality explained:
The spousal notification requirement is thus likely to prevent a significant number of women from obtaining an abortion. It does not merely make abortions a little more difficult or expensive to obtain; for many women, it will impose a substantial obstacle. We must not blind ourselves to the fact that the significant number of women who fear for their safety and the safety of their children are likely to be deterred from procuring an abortion as surely as if the Commonwealth had outlawed abortion in all cases. 505 U.S. at 893-94.
This record does not suggest that any woman in Indiana faces an obstacle of that magnitude in visiting a clinic twice. As we have stressed, Indiana's law has an emergency-bypass clause that has been authoritatively interpreted to cover any kind of physical or psychological risk to the woman. 671 N.E.2d at 108-09. Plaintiffs do not contend that this interpretation falls short of what Stenberg and Casey require for emergencybypass opportunities. It is accordingly difficult to see how the sort of outcome that doomed the spousal-notification rule could condemn Indiana's statute.
This is not to say that a two-visit requirement could not create a burden comparable to a spousal-notice requirement. Quoting the district court, Casey's plurality assumed that "for those women who have the fewest financial resources, those who must travel long distances, and those who have difficulty explaining their whereabouts to husbands, employers, or others, the 24-hour waiting period will be 'particularly burdensome.' " 505 U.S. at 886. But it held these considerations insufficient to condemn the Pennsylvania statute. All that the record in the current case shows is that these costs are positive and have some effect--something that the plurality in Casey assumed. Likewise in Mazurek the Court assumed that a statute preventing nurses and other skilled medical personnel whose training falls short of the M.D. from performing abortions would increase the expense (and thus, by the Law of Demand, reduce the number) of abortions; this again was held insufficient to show invalidity even on the assumption that one legislative purpose was to curtail abortion.
The record in this case does not show that a two-visit rule operates similarly to a spousal-notification rule by facilitating domestic violence or even inviting domestic intimidation. It shows nothing except a decline in the number of abortions in Mississippi and Utah--leaving open both the extent to which other states would experience the same effect and the reason why the effect occurs. This is not the sort of evidence that permits an inferior federal court to depart from the holding of Casey that an informed-consent law is valid even when compliance entails two visits to the medical provider. If Indiana's emergency-bypass procedure fails to protect Indiana's women from risks of physical or mental harm, it will be a failure in operation; it is not possible to predict failure before the whole statute goes into force.
Justice Souter reached a similar conclusion when denying a request to set aside a post-Casey decision enforcing Pennsylvania's statute. See Planned Parenthood of Southeastern Pennsylvania v. Casey, 510 U.S. 1309 (1994) (in chambers). Like the third circuit, whose decision, 14 F.3d 848 (1994), he declined to disturb, Justice Souter concluded that Casey itself had resolved the facial challenge to Pennsylvania's law. What remained was a challenge to the law in application, on a record showing how that law actually operated in Pennsylvania. 510 U.S. at 1311 & n.3. Just so in Indiana. For reasons we have given, what transpired in Mississippi need not portend what will happen in Indiana.
What is more, it would be incongruous to hold Indiana's informed-consent law invalid on the basis of studies covering Mississippi and Utah that (to the district judge's eyes) imply the un-constitutionality of the Mississippi and Utah statutes, while the laws continued to be implemented in Mississippi and Utah. Relying on Casey, the fifth circuit has allowed Mississippi to enforce its statute, see Barnes v. Moore, 970 F.2d 12 (5th Cir. 1992), and Utah's statute likewise has been sustained. See Utah Women's Clinic, Inc. v. Leavitt, 844 F. Supp. 1482, 1487, 1494 (D. Utah 1994), appeal dismissed in pertinent part for lack of jurisdiction, 75 F.3d 564 (10th Cir. 1995). No one has asked these courts to hold the Mississippi or Utah statute invalid on the basis of the local experience; and if these laws remain enforceable despite the consequences demonstrated in this record, it is difficult to see why Indiana's law should be unenforceable even though it is unclear whether similar effects would occur there. Indiana is entitled to an opportunity to have its law evaluated in light of experience in Indiana. And in ...