A Woman's Choice-East Side Women's Clinic v. Scott C. Newman, Prosecuting Attorney for Marion County Indiana, on Behalf of a Class of Prosecutors

DIANE P. WOOD, Circuit Judge,

dissenting.

In today’s opinion, the majority disregards the standards that were established by the Supreme Court in Planned Parenthood of Southeastern Pennsylvania v. Ca *705 sey, 505 U.S. 833 , 112 S.Ct. 2791 , 120 L.Ed.2d 674 (1992), for evaluating laws that impose burdens on a woman’s right to seek an abortion, and it brushes aside the painstakingly careful findings of fact the district court made in support of the limited preliminary injunction it issued against Indiana’s so-called informed consent law, Ind.Code § 16-34-2-1.1. The careful reader of the majority’s opinion will see that the majority regrets the fact that the Supreme Court held in Stenberg v. Carhart, 530 U.S. 914 , 120 S.Ct. 2597 , 147 L.Ed.2d 743 (2000), that pre-enforcement challenges of abortion statutes, like the one presently before us, are permissible. Nevertheless, Stenberg is the law of the land and we must follow its direction, including its endorsement of the constitutional standards governing abortion legislation first articulated by the Casey plurality. See Stenberg, 530 U.S. at 921 , 120 S.Ct. 2597 . That direction is by no means the opaque mess the majority accuses the Supreme Court of creating. In my view, the Court has not left us with “irreconcilable directives” nor has it put courts of appeals “in a pickle.” Ante at 687. At the most, if we were reviewing legislation in some field unrelated to abortion (or speech), we might be faced with the problems the majority describes. As for abortion regulation, the Court’s guidance is crystal clear. In the end, the majority concedes that Stenberg governs, which ought to be enough for present purposes to lead to an affirmance of the district court’s grant of the injunction.

When one follows the analytical path outlined in Casey, it becomes clear that the district court did not abuse its discretion when it concluded that one narrow requirement of Indiana’s law had to be enjoined. In support of that conclusion, the court found that in the particular circumstances faced by Indiana women, and on the basis of the expanded factual record that the Supreme Court invited in Casey, the law’s requirement that women receive certain advice “in the presence” of “the physician who is to perform the abortion, the referring physician or a physician assistant” (§ 16-34-2-1.1(1)) amounts to an unconstitutional “undue burden” on the abortion decision. I would affirm the district court’s decision.

Before turning to the areas of disagreement that lie between the majority and me, it is important to point out that we also share some areas of agreement. First, it is clear that Indiana’s requirement to furnish the statutory information “in the presence of the pregnant woman” (instead of, for example, mailing written materials, having a telephone conversation, or visiting a local doctor who Is neither the referring physician nor the physician who will perform the procedure) is one that raises the cost of obtaining an abortion. This is because the “presence” rule normally necessitates two trips to the abortion facility.

Second, the majority notes, and I agree, that there are both unconstitutional ways in which costs may be raised and constitutional ones: an increased cost is unconstitutional if it is has the purpose or effect of forcing some women to give up their constitutional right to choose abortion; it is constitutional if it genuinely furthers the state’s legitimate interest in persuading women not to select abortion when faced with an unwanted pregnancy.

Third, I agree with the majority that the standard of review for constitutional or legislative facts is more searching than the one we use for historical facts. That does not mean, however, that we may disregard the district court’s findings of historical fact. To the contrary, the Supreme Court has emphasized that we owe deference to such findings even in constitutional cases. See Ornelas v. United States, 517 U.S. 690, *706 699 , 116 S.Ct. 1657 , 134 L.Ed.2d 911 (1996) (“a reviewing court should take care both to review findings of historical fact only for clear error and to give due weight to inferences drawn from those facts by resident judges and local law enforcement officers”). Were we to abandon that rule, many constitutional matters would receive far less restrained review than we presently give them: from possible violations of the Fourth Amendment, to the voluntariness of confessions, to the First Amendment protection accorded to public employee speech.

I

Turning now to the way in which we should resolve this appeal, it is useful to begin with some reminders about what Casey held. (For ease of exposition, I refer to Casey alone rather than to “the Casey standard as endorsed in Stenberg,” since the latter formulation, while more accurate, is needlessly cumbersome.) First, Casey dictates how to draw the line between permissible state regulation and unconstitutional regulation:

Numerous forms of state ' regulation might have the incidental effect of increasing the cost or decreasing the availability of medical care, whether for abortion or any other medical procedure. The fact that a law which serves a valid purpose, one not designed to strike at the right itself, has the incidental effect of making it more difficult or more expensive to procure an abortion cannot be enough to invalidate it. Only where state regulation imposes an undue burden on a woman’s ability to make this decision does the power of the State reach into the heart of the liberty protected by the Due Process Clause.

505 U.S. at 874 , 112 S.Ct. 2791 . The opinion later elaborates on the undue burden standard:

A finding of an undue burden is a shorthand for the conclusion that a state regulation has the purpose or effect of placing a substantial obstacle in the path of a woman seeking an abortion of a nonviable fetus. A statute with this purpose is invalid because the means chosen by the State to further the interest in potential life must be calculated to inform the woman’s free choice, not hinder it. And a statute which, while furthering the interest in potential life or some other valid state interest, has the effect of placing a substantial obstacle in the path of a woman’s choice cannot be considered a permissible means of serving its legitimate ends.

Id. at 877 , 112 S.Ct. 2791 .

Applying this standard, the Court struck down the Pennsylvania statute’s spousal consent requirement and the record-keeping requirement relating to spousal notice; it upheld the statute’s parental consent requirement (which contained the necessary one-parent and judicial bypass provisions), the medical emergency provisions, the rest of the record-keeping requirements, and the “informed consent” requirement. Knowing both what failed the new test and what passed it gives litigants a roadmap of the kind of claims that are likely to succeed, and the kind of evidence they must present. It also gives us concrete guidance on the critical questions now before us: (1) under the Casey test, must the statute create an “undue burden” for every single woman, or is it enough that it create an undue burden for some women; (2) to what extent are we dealing with empirical, fact-specific issues, and to what extent with “legislative” issues; and (3) how must the statute allow for flexible compliance with the state’s broader goals?

The first question — how many women must be affected — is really another way of putting the question about facial chai- *707 lenges that the majority addresses. Ante at 687. In this connection, despite its disclaimers, one is left with the strong impression that the majority is applying either United States v. Salerno, 481 U.S. 739 , 107 S.Ct. 2095 , 95 L.Ed.2d 697 (1987), or something very close to it. In essence, it holds that a state statute like the one before us now would be unconstitutional only if there was “no set of circumstances” under which it was valid — by which it seems to mean that not a single woman in Indiana would find the law’s burdens tolerable. This is an impermissible back-door application of Salerno . Worse yet, it assumes the answer to the question before us: whether the system Indiana wants to put in place will unduly burden Indiana women. Since the pertinent part of the statute has never gone into force, the majority indulges in the presumption that the law imposes no burden at all. But this presumption is found nowhere in our jurisprudence, at least for laws implicating fundamental constitutional rights. Furthermore, this methodology is inconsistent with Casey.

Part V-C of Casey addressed the spousal notification requirement of the Pennsylvania law, also under circumstances in which enforcement had not yet begun. The district court had found, and the Supreme Court accepted, that “[t]he vast majority of women consult their husbands prior to deciding to terminate their pregnancy. ...” 505 U.S. at 888 , 112 S.Ct. 2791 . We can assume, therefore, that the spousal notification requirement was not an undue burden, or any kind of burden, for that “vast majority” of women; they are already doing what the statute specified. But the Court went on to consider the plight of women who were not already consulting the putative father: the 2,000,-000 women a year who are the victims of severe assaults by their male partners. As for those women — “the victims of regular physical and psychological abuse at the hands of their husbands,” id. at 893 , 112 S.Ct. 2791 , matters were different. The Court found, based on “[t]he limited research that has been conducted with respect to notifying one’s husband about an abortion, although involving samples too small to be representative,” id. at 892 , 112 S.Ct. 2791 , that the spousal notification requirement was “likely to prevent a significant number of women from obtaining an abortion.” Id. at 893 , 112 S.Ct. 2791 . Later, to underscore the point, it reiterated that “[t]he analysis does not end with the one percent of women upon whom the statute operates; it begins there.” Id. at 894 , 112 S.Ct. 2791 .

That takes us to the second critical question: whether the reduction in abortions performed is the result of the law’s persuasive force or the consequence of the impermissible placement of obstacles in the path of a woman’s right to choose. One may assume, for the sake of argument, that fewer women in Indiana will forego an abortion than did women in Mississippi, according to the studies in the record. One may further assume that a larger percentage of women in Indiana who forego an abortion will do so because they were persuaded by the law’s informational requirements, contrary to their sisters in Mississippi. No matter: under Casey, our focus must be on those women who, like those affected by the spousal notification requirement in Pennsylvania, will forego the abortion because of the burden, and not because of persuasion. 1

*708 I cannot imagine a more resounding repudiation of the Salerno approach than the Casey opinion gave. The majority opinion in Stenberg makes it clear that this was no accident or oversight. We must therefore look at the effect of the “in the presence” requirements on the Indiana women upon whom the statute operates: the approximately 10% (as the record suggests and as the district court found) who will no longer be able to obtain abortions under the new regime. (Note that the 10% number could be off by an order of magnitude and we would still be required to enjoin this part of the law if it affected “only” 1%, the number presumptively affected by the spousal notification rule in Pennsylvania.)

But, the majority responds, the Supreme Court in Casey upheld something almost exactly like the Indiana “in the presence” requirement when it found that Pennsylvania’s informed consent rules passed muster. Informed consent at the most general level, of course, was not the issue either in our case or in Casey; under the injunction the district court entered, every Indiana woman is furnished with the information the state deems helpful, and when she shows up for the abortion procedure the doctor can once again assure herself that the patient’s consent is informed. Our concern is with the specific way in which the state wants the information to be transmitted. 2

The majority suggests that Casey has already answered this question, insofar as it addressed a regulatory regime with a similar “two visit” rule. But a look at the Casey opinion shows that the Court was not writing so broadly; to the contrary, the Court took great pains not to rule on informed consent/two-visit rules either in general as a matter of fact or as a matter of law. Instead, it explicitly limited its holding to the record before it. It stated that there was “no evidence on this record that requiring a doctor to give the information as provided by the statute would amount in practical terms to a substantial obstacle to a woman seeking an abortion.” Id. (emphasis added). There is no reason to treat the phrases “on this record” and “in practical terms” as casual insertions. The Court thought that the waiting period question was a close one, particularly because it would often translate into a two-visit requirement. The Pennsylvania district court had not made the necessary findings of fact to show that a two-visit requirement would amount to an undue *709 burden (largely because that court had applied the old trimester test from Roe v. Wade, 410 U.S. 113 , 93 S.Ct. 705 , 35 L.Ed.2d 147 (1973), and had invalidated the rule for other, less factually sensitive, reasons). If there could be any doubt remaining on the question whether the Court was restricting its ruling to the record before it, the following passage from Casey should set it to rest:

And the District Court did not conclude that the waiting period is such an obstacle even for the women who are most burdened by it. Hence, on the. record before us, and in the context of this facial challenge, we are not convinced that the 24-hour waiting period constitutes an undue burden.

505 U.S. at 887 , 112 S.Ct. 2791 .

Casey, therefore, establishes the following guidelines for the present case: (1) we must evaluate the Indiana law based on those upon whom it is operating, which is to say the set of women who will be burdened by the “in the presence” requirement; (2) if there were no evidence before the court tending to show that this requirement, like the spousal notification rule considered in Casey, is “likely to prevent a significant number of women from obtaining an abortion,” 505 U.S. at 893 , 112 S.Ct. 2791 , then we would be required to uphold Indiana’s rule on this facial challenge; but (3) since there is evidence that is at least as reliable — if not much more so — than the evidence on which the Casey opinion relied in evaluating the spousal notification rules, we must look at what that evidence shows, deferring to the district court’s findings of historical fact, just as the Supreme Court did in Casey. Cf. Ornelas, 517 U.S. at 699 , 116 S.Ct. 1657 . .1 now turn to the evidence before the district court.

II

Initially, it is necessary briefly to consider what we mean by the term “fact” and how a fact may be established. The majority has tried to explain how and why it is reversing the district court, even while it accepts such critical findings of fact as (1) abortions dropped in Mississippi after enactment of a two-visit rule, as compared to those in South Carolina, which did not have a two-visit rule; and (2) the number of abortions performed in Indiana has not declined because of the advice given to women pursuant to the statute. Even though these facts support the district court’s finding, the majority argues that the ultimate finding of an “undue burden” cannot be sustained, largely because the Supreme Court did not find such a burden in Casey for a similar rule, nor did this court in Karlin v. Foust, 188 F.3d 446 (7th Cir.1999), the decision upon which the majority places most of its reliance. With respect, I believe this approach confuses two fundamentally different inquiries: the first concerns the way in which a certain fact must be established, and the second asks, whether this fact will logically vary from- case to case or if, once properly established, it is “legislative” in nature such that it cannot be questioned over and over again.

' Casey, as the preceding discussion makes clear, was focused on the first of those questions. The Court there decided that the existence of an “undue burden” had not been established on the record then before it. It left the door open, however, for later parties to present more evidence that would cure the gaps in the record that existed. This point can be illustrated by an analogy to new drug approval. Suppose a pharmaceutical company approaches the Food and Drug Administration with an application for approval of a new drug, Alpha. Naturally, it sub *710 mits supporting information to the agency. If, however, the FDA deems that information insufficient, it will reject the application. This does not mean that the company cannot re-apply later, after it conducts more clinical studies or otherwise cures the deficiencies in the earlier record. Based on a fully supported application, the FDA will decide whether Alpha should be approved as safe and effective for the designated uses. Our situation is exactly the same. We now have in this case the “reapplication” for a finding whether a rule that requires two visits to the clinic (here, Indiana’s “presence” requirement) constitutes an undue burden. Are there, in other words, women for whom this rule has the “effect of placing a substantial obstacle in the path of ... seeking an abortion of a nonviable fetus”? 505 U.S. at 877 , 112 S.Ct. 2791 .

In order to answer that question, we must evaluate the evidence presented in this particular case. Before doing so, however, it is also useful to note where the concept of “legislative facts” — on which the majority relies — legitimately applies in this case, and where it does not. Skipping over the crucial question about the way in which facts must be established, the majority treats this case as one in which the factual record is identical to the record in Karlin and then assumes that if there was nothing unconstitutional about a two-visit rule in Karlin there can be nothing unconstitutional here. Burdens are burdens, no matter what state a court is considering. Furthermore, reasons the majority, that is how the Supreme Court treated efforts to regulate the late-term abortion procedure at issue in Stenberg , and thus it must be the way to treat all facts relating to the abortion issue.

With all due respect, the majority has failed to take into account significant differences in the record that was compiled in this case, as compared with the record before the Karlin court, and it has made assumptions about inter-state differences that are unsupported in this record (and, I suspect, unsupportable). Once again, I discuss the particular evidence in the present record later. What is important here is to recognize that this evidence is highly pertinent to the case. A central reason why the majority treats this as a “failure of proof’ case, to the extent it does, is that it assumes that studies done in Mississippi, or Utah, or North Carolina, have nothing to do with Indiana. This assumption is mysterious. What we are considering, after all, is a simple matter of human reactions to sets of incentives or disincentives: will a particular measure be seen as a disincentive at all; if so, will the obstacle be a mere inconvenience, or will it effectively ban a particular option? In the field of economics, we assume that people will react in similar and predictable ways to incentives. (And sometimes it takes more than one study to ascertain what the incentive effects of a particular measure are, even if, once understood, those effects are presumed to be universal.) Consistently with that well-accepted proposition, there is every reason here to assume that Indiana women will react to proven incentives and disincentives in the same way women from other states (e.g., Mississippi) have been shown to respond. The Law of Demand is based on generalized assumptions about human behavior and rationality, and there is no reason to waste time trying to prove that people in one area are exceptions to these rules. The Supreme Court relied on the same idea in Stenberg : faced with high uncertainty about which procedures were legal and which were not, coupled with draconian penalties for an incorrect guess, it was safe to assume that all doctors, everywhere, would err on the side of caution and refuse altogether to perform certain kinds of late-term abor *711 tions. Maybe the Court should have carved out an exception for doctors in places famous for attracting gamblers, like Las Vegas or Atlantic City; but for obvious reasons it did not.

The majority acknowledges that under the Law of Demand, higher prices for abortion will decrease the number demanded, but (as it also appears to recognize) that is not the difficult question here. It is instead whether the observed increase in price caused by the “presence” or two-visit rule is a permissible one under the undue burden analysis. See City of Los Angeles v. Alameda Books, Inc., - U.S. -, 122 S.Ct. 1728, 1742 , 152 L.Ed.2d 670 (2002) (Kennedy, J., concurring) (reduction in demand is merely the first step of an analysis on the permissibility of a regulation: the crucial step is whether that decrease was achieved through allowable governmental action). That is precisely the issue that the Court identified in Casey as an empirical point, where a different result was possible on a more complete record. It is unclear at best to me why the two judges in the majority on this panel think that they know better than the district court judge, who heard all the testimony and weighed all the evidence, what the answer is to the question whether a critical number of Indiana women would experience the “in the presence” rule as such a significant burden that it would effectively prevent them from exercising their constitutionally protected choice. Instead of respecting the district court’s extensive work, the majority finds flaws with the evidence on which the court based its factual findings. It thinks, for instance, that the evidence in the record should have taken into account factors like degree of urbanization, average distance to abortion clinics, and income levels. Ante at 689.

But this simply leads to the majority’s second misunderstanding about the legal significance of the differences between Mississippi and Indiana that these factors might reveal. At best, studies incorporating these variables at a greater level of detail will indicate that there are some women in Indiana for whom the “presence” rule is not a problem, just as there were many women in Pennsylvania who did not anticipate any problem with spousal notification. Surely the majority does not think that every woman in Indiana lives close to a clinic; like all states, Indiana has significant rural areas and significant numbers of people living far from a reproductive health services facility. (There are 11 abortion clinics in Indiana, see Indiana Family Institute, Fact Sheet: Abortion in Indiana, at http://www.hoosier-family.org/FactSheetl3.html, covering a territory of some 36,000 square miles, see U.S. Census Bureau, State and County QuickFacts, at http://quickfacts.cen-sus.gov/qfd/states/18000.html. That adds up to one abortion clinic on the average for almost every 3,300 square miles. And, needless to say, it is quite unlikely that these clinics are distributed with perfect geographical regularity; to the extent the clinics are concentrated around major cities like Indianapolis, that means that other women in rural Indiana will live substantial distances away from the nearest facility.) At most, the details the majority demands might suggest that more Indiana women can withstand the burdens of the Indiana statute than their counterparts in Mississippi could. But the question is not, for example, whether Indiana women as a group live closer to abortion clinics. It is whether an Indiana woman living 60 miles away from a dime in Indiana who cannot afford (either financially, socially, or psychologically) to make two visits, will respond the same way a Mississippi woman living 60 miles away from a clinic in Mississippi with similar constraints did. To repeat, Casey made it clear that the set of *712 women we must consider are those who are burdened by the law, and it found 1% enough to justify striking down the spousal notification rule. Maybe 10% of the women in Mississippi have that problem and “only” 3% of women in Indiana do. No matter. The district court was quite reasonable to find that women in Indiana are like all other people and that their responses will be the same as those of women elsewhere.

Or it could be that the majority thinks that women in Indiana are more likely to be persuaded by the “presence” requirement than are women in Mississippi, so that the decrease in abortions due to the requirement could be attributed to the constitutionally permissible persuasive force of the law. Again, however, all the previous criticisms apply: the question is not whether more women in Indiana are persuaded than are women in Mississippi (bearing in mind that there was no evidence before the district court indicating why that should be the case). It is instead whether a sufficient number of Indiana women (something akin to the 1% of Casey) are not so persuaded, and yet are among those who will be forced to forego their right to choose.

The majority rejects wholesale the relevance of the Mississippi studies (and several other studies) by implying that the district court clearly erred in its decision that Indiana women would react to burdensome two-visit requirements in the same way, and for the same reasons, as Mississippi women did. But even in this context, it offers no reason at all to believe that Indiana women are so idiosyncratic, nor in my view could it. The Supreme Court has consistently endorsed the use of studies from other states or areas- — shared experience is exactly how the “laboratories” in the several states ought to work. See, e.g., Renton v. Playtime Theatres, Inc., 475 U.S. 41, 51-52 , 106 S.Ct. 925 , 89 L.Ed.2d 29 (1986) (in the First Amendment context, no requirement that “a city, before enacting such an ordinance, [ ] conduct new studies or produce evidence independent of that already generated by other cities, so long as whatever evidence the city relies upon is reasonably believed to be relevant to the problem that the city addresses.”).

Ill

Turning now to the evidence demonstrating that the Indiana “presence” rule indeed constitutes an undue burden, we find detailed and meticulous findings from the district court to support that proposition. This evidence was entirely competent to support the district court’s decision; Casey’s discussion of the spousal notification rule makes it clear that evidence on undue burden does not have to meet some heightened standard of perfection. To the contrary, the Court there relied on “limited research that has been conducted with respect to [the issue at hand, there notifying one’s husband about an abortion], although involving samples too small to be representative,” 505 U.S. at 892 , 112 S.Ct. 2791 , to support its conclusion about spousal notification.

We have more than zero, and less than perfection, when it comes to information about the burden the Indiana statute places on the women affected by the two-visit rule. The majority, in effect, has not only demanded perfection; it also wants a showing that some number of Indiana women significantly larger than the number the Court accepted in Casey are unduly burdened by the law. Every time the plaintiffs come back with more studies and more information (as they have surely done here, in comparison with the record they created in Karlin), the majority raises the bar higher and tells them to come back another day — even though this court *713 has specifically held that “the biases of one study in one case may be avoided or reversed in the next.” Mister v. Illinois Cent. Gulf R.R. Co., 832 F.2d 1427, 1437, n. 3 (7th Cir.1987) (Easterbrook, J.).

In this case, we have evidence, and importantly we have significant new evidence that has been developed since Karlin , which answers precisely the kinds of questions that Karlin directed future plaintiffs to address. The distinct court relied on this evidence to conclude that the “in the presence” part of the Indiana law would indeed impose an undue burden on enough Indiana women seeking abortions that this part of the statute had to be enjoined. The Supreme Court has consistently endorsed the use of the type of evidence that was presented here, and has analyzed it in the “factual context of each case in light of all the evidence presented by both the plaintiff and the defendant.” Bazemore v. Friday, 478 U.S. 385, 400 , 106 S.Ct. 3000 , 92 L.Ed.2d 315 (1986). Key among this evidence before the court was a new (post-Karim ) study published in the Journal of the American Medical Association (JAMA), one of the most highly respected journals in the medical field, and one that meets any conceivable standard for peer-review. Cf. Daubert v. Merrell Dow Pharms., Inc., 509 U.S. 579 , 113 S.Ct. 2786 , 125 L.Ed.2d 469 (1993); Robinson v. California, 370 U.S. 660 , 667 n. 9, 82 S.Ct. 1417 , 8 L.Ed.2d 758 (1962) (relying on findings in JAMA); Budd v. California, 385 U.S. 909 , 912 n. 3, 87 S.Ct. 209 , 17 L.Ed.2d 138 (1966) (For-tas, J., dissenting from denial of certiorari) (same).

The JAMA study is a time-series and regression analysis designed to assess the effect of the Mississippi law on the abortion and birth rates of Mississippi residents in two ways: first, through a retrospective analysis of those rates before and after the passage of the statute, and second, through a comparison between Mississippi and two similar states, Georgia and South Carolina, neither of which had a “presence” requirement in effect at the relevant time but were otherwise similar in the relevant respects. Regression analy-ses are an important tool in much of social science research, as well as in law. See McCleskey v. Kemp, 481 U.S. 279, 293-94 , 107 S.Ct. 1756 , 95 L.Ed.2d 262 (1987) (discussing their role in Title VII of the Civil Rights Act of 1964 cases and in sentencing context); see also Bazemore v. Friday, 478 U.S. at 398-401 , 106 S.Ct. 3000 (additional regression analyses conducted in response to criticisms and suggestions by the district court, all of which confirmed, and some of which even strengthened, the study’s original conclusions; further concluding that multiple-regression analysis need not include every conceivable variable to establish a party’s case; and chiding the court of appeals because it “failed utterly to examine the regression analyses in light of all the evidence in the record”); Arlington Heights v. Metropolitan Housing Dev. Corp., 429 U.S. 252 , 266 n. 13, 97 S.Ct. 555 , 50 L.Ed.2d 450 (1977) (discussing jury selection context). 3

*714 The difficulty here is that there is no single independent variable that will show “undue burden.” The only way to prove that a particular part of a law is imposing an undue burden on abortion choice is to hypothesize that it might constitute an undue burden, and then to show that no other reasonably related variable satisfactorily explains the drop in abortion rates (i.e., the phenomenon that is being tested)' — essentially a process of elimination. This is a methodology we have approved before. See In re Oil Spill by Amoco Cadiz, 954 F.2d 1279 (7th Cir.1992) (per curiam). In that case, this court endorsed the use of simple linear regressions as a way to draw out all other plausible explanations leaving only the hypothesized one as a reason for what part of a loss in business was attributable to a massive oil spill. Id. at 1320 . We expressed satisfaction there with this way of getting “a better grip on the relation between dependent and independent variables” and reaching “an inference of causation, and of the size of the effect.” Id. Indeed, the latest pronouncement by the Supreme Court in First Amendment matters endorses a study by the city of Los Angeles aimed at demonstrating the connection between its ban on multiple-use adult establishments and its interest in reducing crime — a connection that can only be shown by ruling out some (but clearly not all) of the other potential independent variables that could have contributed to the effect on crime. See City of Los Angeles v. Alameda Books, Inc., supra, - U.S. -, 122 S.Ct. 1728 , 152 L.Ed.2d 670 (plurality opinion). Dealing with the nature of such studies, the plurality concluded that a party “does not bear the burden of providing evidence that rules out every [other] theory.” Id. at 1735. The study, however imperfect, was respected for its probative value precisely because of the lack of evidence presented by the other side to rebut its finding. As in this case, once the point was made, the opponent had to produce concrete evidence on the other side; it was not enough merely to point out an alleged imperfection in the study. Id. (respondents did “not offer a competing theory, let alone data” to counter the city’s assertions).

That is exactly what the JAMA study did. It was tailored to explore the question (unanswered in the record in Karlin) whether the decrease in abortions in Mississippi was an effect not of the persuasive power of the law, but rather of its burdensome qualities. And it showed that the latter explanation was the correct one. The principal outcome measures in the study were birth rates, abortion rates, the percentage of late abortions, and the percentage of abortions performed outside the state. The researchers found that the resident abortion rates declined 12% more in Mississippi than they did in South Carolina after the passage of the law, and they declined 14% more in Mississippi than they did in Georgia. Limited to Caucasian adults, abortions declined 22% more in Mississippi than in South Carolina and 20% more in Mississippi than in Georgia. Abortions performed after the 12-weeks gestation mark increased 39% more in Mississippi than in either South Carolina or Georgia. The percentage of abortions performed out-of-state increased 42% more among women in Mississippi relative to women in South Carolina.

The JAMA study also showed that in the 12-month period after the law took effect, the total rate of abortions for Mississippi residents decreased by approximately 16%; the proportion of Mississippi residents traveling to other states for an abortion increased by about 37%; and the number of second trimester abortions increased by some 40%. The study concluded that these statistics “suggest that Mississippi’s mandatory delay statute was *715 responsible for a decline in abortion rates and an increase in abortions performed later in pregnancy.” The researchers who conducted the study testified before the district court that the only salient difference among Mississippi, Georgia, and South Carolina for these purposes was that only Mississippi had a two-trip requirement. Otherwise, the laws regulating abortions in the three states were functionally the same — and no other statistically significant events had taken place in the various states.

The district court realized that it needed to address one final, but critical, question of fact: were declines observed in abortions in Mississippi because women in Mississippi had been persuaded to forego abortions, or were the declines because the new law was impermissibly burdening the right to seek an abortion — precisely the inquiry mandated by Karlin . The court found that the latter explanation was the correct one, for several reasons, all amply based on the evidence in the record. First, the “persuasive power” hypothesis was severely undercut by the evidence showing that Mississippi women were leaving the state to have their abortions elsewhere and having more second-trimester abortions. Those women quite evidently had not been persuaded to carry their pregnancy to term; they aborted their pregnancies, but they did it outside the state of Mississippi or at a later and riskier time. (And certainly we must assume that the Mississippi legislature was trying to persuade women to forego abortion, rather than to persuade them to travel out-of-state for the procedure or to postpone it to a riskier time.) Second, the court looked at evidence from Indiana itself that showed no changes in abortion rates from the information standing alone (a part of the statute that the court permitted to take effect, stripped only of the “presence” requirement). The court discussed other evidence as well, and I commend its thorough analysis, particularly its evaluation of similar studies conducted in Utah and Louisiana. In the end, its conclusion was that the “sum of this evidence, and the absence .of evidence of any persuasive effect, shows convincingly that the predicted reduction in abortion rates would result not from persuasion but from restrictions posing a substantial obstacle for some women’s ability to obtain abortions.” As mentioned before, even if the relative numbers of women who were persuaded versus burdened are different in Mississippi and Indiana, the study conclusively reveals that a significant number of Indiana women will be unduly burdened by the “presence” requirement. And those women — not those who are persuaded, or unaffected, or not even contemplating an abortion — are those on whom the majority should have concentrated under Casey and Stenberg .

Whether this court is looking at the record de novo, under an abuse of discretion standard, or merely for clear error, the district court’s findings should stand. I find nothing in the majority’s speculation that comes close to refuting the evidence upon which the court relied. See City of Los Angeles , - U.S. at-, 122 S.Ct. at 1735 (disapproving of fact that the “Court of Appeals simply replaced the city’s theory ... with its own” and that its analysis “implicitly requires the city to prove that its theory is the only one that can plausibly explain the data”). Only by disregarding key points such as the number of women who willingly undertook the burden of seeking an abortion out-of-state, where they could have the entire procedure accomplished in one visit, rather than staying in-state and enduring the two-visit burden, can the majority come to the result it does.

IV

Finally, although not necessary to my analysis, it is worth noting that Casey said *716 that “[a] finding of an undue burden is a shorthand for the conclusion that a state regulation has the purpose or effect of placing a substantial obstacle in the path of a woman seeking an abortion of a nonviable fetus” 505 U.S. at 877 , 112 S.Ct. 2791 (emphasis added). The majority considers only the “effect” part of that disjunctive test, perhaps thinking that this court’s dismissal of the “purpose” half in Karlin was binding on the Supreme Court. I am under no such illusion. I believe, therefore, that it is appropriate to take a brief look at the purpose Indiana offered for this regulation, to see if it might either help to save the statute or to condemn it.

The district court found that “[tjhere is no evidence tending to show how the ‘in the presence’ requirement actually furthers the state’s legitimate interests in maternal health or in protecting potential life.” It said this, importantly, after the plaintiffs had made an extensive prima facie showing that the statute furthered neither legitimate interest; in the sense of a burden of production, the court was concerned that Indiana had offered nothing to the contrary. (Indiana argues strenuously that the district court imposed an impermissible shift in the ultimate burden of proof, but it is clear from the court’s opinion as a whole that it did no such thing; it was simply addressing the evidentiary vacuum on Indiana’s side in the face of the plaintiffs’ evidence.)

Indeed, my search of the legislative history of the Indiana statute reveals no reason whatsoever for imposing a two-visit requirement for the dissemination of the required information. Acting as if it were conducting rational basis review, the majority speculates that some Indiana legislator might have thought that absorption of information occurs more effectively when it is transmitted in person. Maybe so, but that does not explain why the state could not simply have said that at the point of checking into the clinic, the previously transmitted information must be reiterated in person. The change from an oral communication to an “in the presence” requirement was added by a floor amendment in the House of Representatives that was marked by scant debate. After some members of the House suggested that the “in the presence” requirement was intended as an obstacle to abortion, its chief sponsor stated instead that the concern was that unless the information was given in person, “how do you know this is the doctor or the person you’re supposed to be talking to.... I would think you would want to talk personally to the person who may be performing that and know ... that they are the person indeed that they [say they] are.” A special concern with impostor doctors, or generally with practitioners not being who they say they are over the telephone is not a problem that is specific to abortion — or at least there were no such findings other than a statement that this possibility (of talking to an impostor) “is very dangerous, especially when you’re talking about the symptoms and consequences of an abortion.” Literally the only other scrap of evidence from the legislature seems to reflect a fear that women would receive the information while they were under sedation on an operating table. But that cannot be what the legislature really feared, for the simple reason that this concern is already addressed in Indiana’s law of “informed consent,” which cannot be given by persons already under anaesthesia. See, e.g., Culbertson v. Mernitz, 602 N.E.2d 98, 103 (Ind.1992) (endorsing the American Medical Association’s 1992 Code of Medical Ethics with respect to necessary consent, and rejecting as invalid consent given “where the patient is unconscious or otherwise incapable of consenting”). I would be surprised if many Indiana doctors were in the habit of *717 obtaining consent for medical procedures from unconscious or drugged patients; they would risk loss of their medical license if they did, whether they were performing an appendectomy, knee surgery, a vasectomy, a prostate removal, or an abortion. The law of informed consent is consistent across the spectrum of surgeries and procedures: if no reason is given why heightened consent is needed in the abortion context, then this cannot be accepted as the reason for the “in the presence” requirement.

V

For all these reasons, I believe that the majority has seriously mis-applied the Casey test. It has substituted its own factual assumptions for evidence that is in the record; it has failed to focus on the women for whom that statute will create problems; and it seems to think that the Casey Court was not serious when it emphasized the lack of evidence in the record before it, by implying that the result in Casey dictates the result here. I respectfully dissent.

. In fact, the parallel with the spousal notification requirement in Casey runs deeper than the mere finding of an undue burden. Were the Supreme Court to have reasoned along the same lines as the majority of the panel, it would have concluded that the 1% of women *708 who forego abortions because of the notification requirement have in fact merely changed their minds — been "persuaded” — after the consultation with their husbands that would not have occurred but for the notification requirement. The Court focused on those women who were not persuaded, but who instead were forced to carry their pregnancy to term because of the risk of violence or abuse from their male partner. Similarly, this court should focus on those who are not persuaded because of the "presence” requirement, but for whom this requirement is close to the equivalent of a flat prohibition on abortion.

. As the majority acknowledges, Indiana’s law has been construed to have an emergency by-pass provision that covers any kind of physical or psychological risk to the woman from any of its provisions, including presumably the "presence" requirement. But that does not distinguish it from Pennsylvania's statute, which also relieved a physician of compliance with the informed consent rules "if he or she can demonstrate, by a preponderance of the evidence, that he or she reasonably believed that furnishing the information would have resulted in a severely adverse effect on the physical or mental health of the patient.” 18 Pa. Cons.Stat. § 3205(c) (1990). The existence of such a statute was not enough to convince the Court that the spousal consent requirement was permissible. By the same token, the existence of a similar safety valve in the Indiana statute is not enough to save the otherwise burdensome "presence” requirement, for the reasons I explain below.

. It is worth reviewing for a moment what this kind of study does. A regression takes a dependent variable (here, the decrease in abortions or abortion rates) and tests it against a number of independent variables. The independent variables are chosen before the regression is run; here, for example, they included factors like the opening of new clinics, changes in marriage rates, changes in the percentage of the population living in metropolitan areas, the increased availability of contraceptives, and changes in per capita income. Once the data is collected, the researcher runs a regression on each variable, which shows the effect of that variable, in isolation, on the dependent variable. If a variable is found to have no meaningful correlation to the dependent variable, it is discarded.