MEMORANDUM *

Ken, Riley, and Sawyer Jaramillo appeal the district court’s final judgment in favor of the defendant Ford Motor Company. The Jaramillos sued Ford alleging that the company had defectively designed their Ford Explorer that rolled over when Angela Jaramillo, Ken’s wife and Riley and Sawyer’s mother, swerved to avoid a deer. The rollover accident resulted in the deaths of Angela and McKenna Jaramillo, Angela and Ken’s youngest daughter, and injury to Ken, Riley, and Sawyer. After a sixteen-day trial, the jury returned a defense verdict. We have jurisdiction under 28 U.S.C. § 1291 , and we reverse and remand.

The Jaramillos’ theory was that the Explorer was not reasonably safe as designed because the vehicle had a high center of gravity and a narrow wheel base that caused it to roll over on smooth, dry pavement. The Jaramillos contend that the district court erred by admitting over the Jaramillos’ objection comparative accident statistics that involved rollover accidents not just on smooth, dry pavement, but of all types. The district court admitted this evidence because the court determined the plaintiffs had opened the door by introducing similar evidence.

We review the district court’s evidentiary ruling for abuse of discretion. McEwin v. Crown Equip. Corp., 328 F.3d 1028, 1032 (9th Cir.2003). We will not reverse a jury verdict based on an evidentiary error unless the error was not harmless. Mukhtar v. Cal. State Univ., Hayward, 299 F.3d 1053, 1066 (9th Cir.2002) (en banc), amended by 319 F.3d 1053 (9th Cir.2003). In other words, “a party must demonstrate that the allegedly erroneous evidentiary ruling more probably than not was the cause of the result reached.” Jauregui v. City of Glendale, 852 F.2d 1128, 1133 (9th Cir.1988). Furthermore, “[i]f we are unable to say that the probabilities favor the same result and are unsure whether the error was harmless, a new trial is required.” Mukhtar, 299 F.3d at 1066-67 .

“Invited error occurs when the appellant opens the door to objectionable testimony by introducing it, rather than waiting for the appellee to do so.” Mukhtar, 299 F.3d at 1063 n. 6. The issue in this case is whether the objectionable evidence offered by Ford was sufficiently similar to evidence introduced by the Jaramillos earlier in the trial. We conclude that it was not.

Although the district court did not specify the evidence it relied on to conclude that the Jaramillos opened the door, Ford points to the Jaramillos’ introduction of a Consumer Reports article and their introduction of a videotaped deposition of Robert Pascarella, a Ford Design Analysis Engineer and the company’s 30(b)(6) witness, in which the Jaramillos questioned Pascarella about two Ford documents that were produced before the Explorer was de *78 signed. The Jaramillos utilized both the Consumer Reports article and the Ford documents to demonstrate that Ford had notice before it designed the Explorer that rollover accidents presented a significant risk. In doing so, the Jaramillos introduced statistics from Ford’s own documents that showed that Ford knew that light trucks and small utility vehicles rolled over more frequently than cars, and that the Ford Bronco II rolled over more frequently than some other SUVs.

Ford’s evidence that the Jaramillos object to is dissimilar to the background information introduced by the Jaramillos because the challenged evidence has nothing to do with what Ford knew before it designed the Explorer, but rather how the 1995-1999 Explorer performed relative to other vehicles on the road. Ford’s expert witness, Dr. Michelle Vogler, compared the 1995-1999 Explorer to different models of SUVs, sporty cars, small cars, full-size vans, pickups, midsize cars, premium sports cars, large cars, and small vans. Specifically, Dr. Vogler compared the Explorer to these vehicles based on the frequency with which they were involved in any type of accident, any type of rollover accident, and any type of fatal or fatal-severe accident. Additionally, Dr. Vogler compared these vehicles based on the frequency of occupant injury and fatality in all types of accidents and in rollover accidents, and the risk of fatalities in different vehicles depending on whether the occupants were belted or unbelted.

The Jaramillos’ introduction of Pascarella’s testimony did not open the door to these extensive comparative accident statistics. The Jaramillos did not introduce any comparative accident statistics that suggested the Explorer had a higher propensity to be involved in a rollover or any other type of accident than other SUVs or other types of vehicles. Pascarella’s testimony merely provided background information about what Ford knew before the Explorer was designed. Similarly, the Consumer Reports article’s discussion of the rollover rate of the Bronco II did not address the safety of the Explorer; it only provided background information about what Ford knew before Ford designed the Jaramillos’ vehicle. Accordingly, the Jaramillos did not open the door to Ford’s comparative accident statistics. 1

Ford alternatively contends that Dr. Vogler’s comparative accident statistics were relevant to custom within the automotive industry and to whether the harm of the Explorer’s design outweighed the harms that would have resulted from an alternative design. See Wash. Rev.Code § 7.72. Additionally, Ford argues that these comparisons would also help the jury determine whether the Explorer was unsafe to an extent that the ordinary consumer would not contemplate and to determine whether Ford had notice that there was a problem with the Explorer that should have led the automobile manufacturer to warn consumers. See id.

To be relevant, Ford’s comparative accident statistics must be based on accidents that occurred under circumstances similar to the Jaramillos’ accident, i.e. a rollover on smooth, dry pavement. “[E]very court of appeals ... agrees that when a plaintiff attempts to introduce evidence *79 of other accidents as direct proof of a design defect, the evidence is admissible only if the proponent demonstrates that the accidents occurred under circumstances substantially similar to those at issue in the case at bar.” Barker v. Deere and Co., 60 F.3d 158, 162 (3d Cir.1995). While this substantial-similarity doctrine has normally been applied to preclude other-accident evidence offered by plaintiffs, we have explained that the doctrine “rests on the concern that evidence of dissimilar accidents lacks the relevance required for admissibility under Federal Rules of Evidence 401 and 402.” Cooper v. Firestone Tire and Rubber Co., 945 F.2d 1103, 1105 (9th Cir.1991). Because this underlying concern is not limited to comparative accident statistics offered by plaintiffs, defendants like Ford that attempt to compare their products to others to show the relative safety of their designs or a lack of notice must show that their comparisons are based on accidents that are similar to the plaintiffs accident.

It is undisputed that Dr. Vogler’s comparative accident statistics were not limited to rollovers on smooth, dry pavement. Dr. Vogler’s comparative rollover statistics were irrelevant to the design defect at issue in this case: whether the Explorer was not reasonably safe as designed specifically because of its propensity to roll over on smooth, dry pavement. Whether the Explorer was safer than other vehicles in some other respect is inconsequential to the factual issue that was before the jury. Dr. Vogler’s testimony and exhibits should have been excluded under Fed.R.Evid. 402, and the district court abused its discretion in admitting this evidence.

Ford does not contend that the district court’s decision to admit Dr. Vogler’s testimony was harmless. Because the jury was allowed to consider the Explorer’s likelihood of being involved in other accident types, this evidence may have unduly prejudiced the jury’s view of whether the Explorer was not reasonably safe as designed. This was not an isolated error in the middle of a lengthy trial. Dr. Vogler testified, using graphic tables, for several hours about the relative safety of the Explorer and Ford’s counsel repeatedly emphasized the significance of her testimony in his closing argument. The evidence in this case was close. Under these circumstances, we are “unable to say that the probabilities favor the same result” if this evidence was excluded. Mukhtar, 299 F.3d at 1066 . Accordingly, we must reverse the judgment and remand for a new trial.

REVERSED AND REMANDED.

This disposition is not appropriate for publication and may not be cited to or by the courts of this circuit except as provided by Ninth Circuit Rule 36-3.

. Ford contends that the Jaramillos opened the door by introducing evidence of nine Explorer accidents. We fail to see how these rollover accidents involving the same model vehicle as the Jaramillos were driving could open the door to comparative accident statistics. We also reject Ford's arguments that the Jaramillos opened the door to this comparative risk evidence by stating in their opening statement that the Explorer "was dangerous to an extent that Angela in her research could never have contemplated.”