EASTERBROOK, Circuit Judge.

According to the complaint in this securities-fraud action, James Hofman, a stockbroker working for Prudential Securities, told 11 of his customers that Jefferson Savings Bancorp was "certain" to be acquired, at a big premium, in the near future. Hofman continued making this statement for seven months (repeating it to some clients); it was a lie, for no acquisition was impending. And if the statement had been the truth, then Hofman was inviting unlawful trading on the basis of material non-public information. He is a securities offender coming or going, see Bateman Eichler, Hill Richards, Inc. v. Berner, 472 U.S. 299, 105 S.Ct. 2622, 86 L.Ed.2d 215 (1985), as are any customers who traded on what they thought to be confidential information — if Hofman said what the plaintiffs allege, a subject still to be determined. What we must decide is whether the action may proceed, not on behalf of those who received Hofman's "news" in person but on behalf of everyone who bought Jefferson stock during the months when Hofman was misbehaving. The district judge certified such a class, invoking the fraud-on-the-market doctrine of Basic, Inc. v. Levinson, 485 U.S. 224, 241-49, 108 S.Ct. 978, 99 L.Ed.2d 194 (1988). Prudential asks us to entertain an interlocutory appeal under Fed.R.Civ.P. 23(f). For two reasons, this is an appropriate case for such an appeal, which we now accept. See Blair v. Equifax Check Services, Inc., 181 F.3d 832 (7th Cir. 1999).

First, the district court's order marks a substantial extension of the fraud-on-the-market approach. Basic held that "[b]ecause most publicly available information is reflected in market price, an investor's reliance on any public material misrepresentations, therefore, may be presumed for purposes of a Rule 10b-5 action." 485 U.S. at 247, 108 S.Ct. 978. The theme of Basic and other fraud-on-the-market decisions is that public information reaches professional investors, whose evaluations of that information and trades quickly influence securities prices. But Hofman did not release information to the public, and his clients thought that they were receiving and acting on nonpublic information; its value (if any) lay precisely in the fact that other traders did not know the news. No newspaper or other organ of general circulation reported that Jefferson was soon to be acquired. As plaintiffs summarize their position, their "argument in a nutshell is that it is unimportant for purposes of the fraud-on-the-market doctrine whether the information was `publicly available' in the . . . sense that . . . the information was disseminated through a press release, or prospectus or other written format". Yet extending the fraud-on-the-market doctrine in this way requires not only a departure from Basic but also a novelty in fraud cases as a class — as another court of appeals remarked only recently in another securities suit, oral frauds have not been allowed to proceed as class actions, for the details of the deceit differ from victim to victim, and the nature of the loss also may be statement-specific. See Johnston v. HBO Film Management, Inc., 265 F.3d 178, 185-92 (3d Cir. 2001). See also, e.g., Szabo v. Bridgeport Machines, Inc., 249 F.3d 672 (7th Cir. 2001). The appeal thus presents a novel and potentially important question of law.

Second, very few securities class actions are litigated to conclusion, so review of this novel and important legal issue may be possible only through the Rule 23(f) device. What is more, some scholars believe that the settlements in securities cases reflect high risk of catastrophic loss, which together with imperfect alignment of managers' and investors' interests leads defendants to pay substantial sums even when the plaintiffs have weak positions. See Janet Cooper Alexander, Do the Merits Matter? A Study of Settlements in Securities Class Actions, 43 Stan. L.Rev. 497 (1991); Reinier Kraakman, Hyun Park Steven Shavell, When Are Shareholder Suits in Shareholder Interests?, 82 Geo. L.J. 1733 (1994) Roberta Romano, The Shareholder Suit: Litigation Without Foundation?, 7 J.L. Econ. Org. 55 (1991). The strength of this effect has been debated, see Joel Seligman, The Merits Do Matter, 108 Harv. L.Rev. 438 (1994), but its existence is established. The effect of a class certification in inducing settlement to curtail the risk of large awards provides a powerful reason to take an interlocutory appeal.

Because the parties' papers have developed their positions fully, and the district court has set a trial date less than two months away, we think it best to resolve the appeal promptly, and thus we turn to the merits.

Causation is the shortcoming in this class certification. Basic describes a mechanism by which public information affects stock prices, and thus may affect traders who did not know about that information. Professional investors monitor news about many firms; good news implies higher dividends and other benefits, which induces these investors to value the stock more highly, and they continue buying until the gains are exhausted. With many professional investors alert to news, markets are efficient in the sense that they rapidly adjust to all public information; if some of this information is false, the price will reach an incorrect level, staying there until the truth emerges. This approach has the support of financial economics as well as the imprimatur of the Justices: few propositions in economics are better established than the quick adjustment of securities prices to public information. See Richard A. Brealey, Stewart C. Myers Alan J. Marcus, Fundamentals of Corporate Finance 322-39 (2d ed. 1998).

No similar mechanism explains how prices would respond to non-public information, such as statements made by Hofman to a handful of his clients. These do not come to the attention of professional investors or money managers, so the price-adjustment mechanism just described does not operate. Sometimes full-time market watchers can infer important news from the identity of a trader (when the corporation's CEO goes on a buying spree, this implies good news) or from the sheer volume of trades (an unprecedented buying volume may suggest that a bidder is accumulating stock in anticipation of a tender offer), but neither the identity of Hofman's customers nor the volume of their trades would have conveyed information to the market in this fashion. No one these days accepts the strongest version of the efficient capital market hypothesis, under which non-public information automatically affects prices. That version is empirically false: the public announcement of news (good and bad) has big effects on stock prices, which could not happen if prices already incorporated the effect of non-public information. Thus it is hard to see how Hofman's non-public statements could have caused changes in the price of Jefferson Savings stock. Basic founded the fraud-on-the-market doctrine on a causal mechanism with both theoretical and empirical power; for non-public information there is nothing comparable.

The district court did not identify any causal link between non-public information and securities prices, let alone show that the link is as strong as the one deemed sufficient (by a bare majority) in Basic (only four of the six Justices who participated in that case endorsed the fraud-on-the-market doctrine). Instead the judge observed that each side has the support of a reputable financial economist (Michael J. Barclay for the plaintiffs, Charles C. Cox for the defendant) and thought the clash enough by itself to support class certification and a trial on the merits. That amounts to a delegation of judicial power to the plaintiffs, who can obtain class certification just by hiring a competent expert. A district judge may not duck hard questions by observing that each side has some support, or that considerations relevant to class certification also may affect the decision on the merits. Tough questions must be faced and squarely decided, if necessary by holding evidentiary hearings and choosing between competing perspectives. See Szabo and, e.g., Isaacs v. Sprint Corp., 261 F.3d 679 (7th Cir. 2001).

Because the record here does not demonstrate that non-public information affected the price of Jefferson Savings' stock, a remand is unnecessary. What the plaintiffs have going for them is that Jefferson's stock did rise in price (by about $5, or 20% of its trading price) during the months when Hofman was touting an impending acquisition, plus a model of demand-pull price increases offered by their expert. Barclay started with a model devised by another economist, in which trades themselves convey information to the market and thus affect price. See Joel Hasbrouck, Measuring the Information Content of Stock Trades, 46 J. Fin. 179 (1991). Hasbrouck's model assumes that some trades are by informed traders and some by uninformed traders, and that the market may be able to draw inferences about which is which. The model has not been verified empirically. Barclay approached the issue differently, assuming that all trades affect prices by raising demand even if no trader is well informed — as if there were an economic market in "Jefferson Savings stock" as there is in dill pickles or fluffy towels. Hofman's tips raised the demand for Jefferson Savings stock and curtailed the supply (for the tippees were less likely to sell their own shares); that combination of effects raised the stock's price. Yet investors do not want Jefferson Savings stock (as if they sought to paper their walls with beautiful certificates); they want monetary returns (at given risk levels), returns that are available from many financial instruments. One fundamental attribute of efficient markets is that information, not demand in the abstract, determines stock prices. See Myron S. Scholes, The Market for Securities: Substitution Versus Price Pressure and the Effects of Information on Share Prices, 45 J. Bus. 179 (1972); Richard A. Brealey, An Introduction to Risk and Return from Common Stocks 15-18, 25-46 (2d ed. 1983). There are so many substitutes for any one firm's stock that the effective demand curve is horizontal. It may shift up or down with new information but is not sloped like the demand curve for physical products. That is why institutional purchases (which can be large in relation to normal trading volume) do not elevate prices, while relatively small trades by insiders can have substantial effects; the latter trades convey information, and the former do not. Barclay, who took the view that the market for Jefferson Savings securities is efficient, did not explain why he departed from the normal understanding that information rather than raw demand determines securities prices.

Data may upset theory, and if Barclay had demonstrated that demand by itself elevates securities prices, then the courts would be required to attend closely. What Barclay did is inquire whether the price of Jefferson Savings stock rose during the period of additional demand by Hofman's customers. He gave an affirmative answer and stopped. Yet it is not possible to prove a relation between demand and price without considering other potential reasons. Was there perhaps some truthful Jefferson-specific information released to the market at the time? Did Jefferson perhaps move with the market? It rose relative to a basket of all financial institutions, but (according to Cox's report) not relative to a portfolio of Midwestern financial intermediaries. Several Missouri banks and thrifts similar to Jefferson Savings were acquired during the months in question, and these transactions conveyed some information about the probability of a deal involving Jefferson Savings. If the price of Jefferson Savings was doing just what one would have expected in the presence of this changing probability of acquisition, and the absence of any Hofmaninduced trades, then the causal link between Hofman's statements and price has not been made out. By failing to test for and exclude other potential sources of price movement, Barclay undercut the power of the inference that he advanced.

Indeed, Barclay's report calls into question his belief that the market for Jefferson Savings stock is efficient, the foundation of the fraud-on-the-market doctrine. In an efficient market, how could one ignorant outsider's lie cause a long-term rise in price? Professional investors would notice the inexplicable rise and either investigate for themselves (discovering the truth) or sell short immediately, driving the price back down. In an efficient market, a lie told by someone with nothing to back up the statement (no professional would have thought Hofman a person "in the know") will self-destruct long before eight months have passed. Hofman asserted that an acquisition was imminent. That statement might gull people for a month, but after two or three months have passed the lack of a merger or tender offer puts the lie to the assertion; professional investors then draw more astute inferences and the price effect disappears. That this did not occur implies either that Jefferson Savings was not closely followed by professional investors (and that the market therefore does not satisfy Basic's efficiency requirement) or that something other than Hofman's statements explains these price changes.

The record thus does not support extension of the fraud-on-the-market doctrine to the non-public statements Hofman is alleged to have made about Jefferson Savings Bancorp. The order certifying a class is

REVERSED.