concurring in part and dissenting in part:
I join section one of the majority’s opinion, but dissent from its holding in *892section two. The majority’s holding that the district court.abused its discretion in determining that the U.S.S.G. § 2Ll.l(b)(5) enhancement applied to Torres-Flores’ offense is based on evidence not in the record and conflicts with our' normal procedure for reviewing Guidelines sentences.
I.
Torres-Flores concealed Marquez-Cruz in a earved-out compartment located beneath the rear passenger seat of an extended cab pick-up. Torres-Flores’ counsel estimated, that the rear bench seat, which was- placed over Marquez-Cruz’s body, weighed approximately forty to fifty pounds (not accounting for the speakers and the child safety seat that rested on top of the bench seat during Marquez-Cruz’s transportation). The bench seat was not bolted to the floor of the pick-up and Marquez-Cruz was placed in the compartment without any restraints. Photographs of the vehicle revealed protruding metal edges located at the sides of the carved-out compartment:
After examining color photographs of the vehicle and hearing extensive argument from the parties, the district court set forth detailed factual findings, which the majority does not dispute.
[WJhen you get down to common sense, it’s very simple: that there was an area that tvas carved out of the bottom. It’s not smoothed out. It’s not padded. There are some edges, and you can’t tell how sharp the edges are, but there are edges if you look in Exhibit H, that color photograph, and Exhibit I. The person is in that compartment. He pulls the rug over him, and then he pulls the seat back over him. The seat is not bolted in place. And so moving back and forth, just the motion of the car runs the risk of the seat popping out and moving for the very reason that [the defense] argue[s] that [Marquez-Cruz] can lift the seat up with great ease. I mean that’s just one thing.
But one has to look at if there’s a collision, and collisions happen, what’s going to happen? The car is not going to be— the vehicle’s not going to be in perfect shape where metal is not going to be bent. Metal bends. Things happen. Fires occur. Explosions occur. And this person is in a situation where he can’t get out unless, even taking the facts in the light most favorable to [the defense], he pulls the rug back — assuming he wasn’t injured and his head didn’t hit one of the sides of the compartment here and he’s still conscious — he has to pull the rug back, push the seat up, and then somehow get over the seat. The seat’s now pushed against the driver’s seat. Then get out of the vehicle when the clock is ticking and seconds are ticking and this vehicle can explode from a collision or there’s other fire that happens in collisions. I wouldn’t want to be in that situation.
Common sense tells me that that’s a dangerous, a serious risk of personal injury or even death to ride in a vehicle that way. And I think that’s what’s meant.
The Dixon case is a really specific case that dealt with hatchbacks where one could get out very easily. This is not something that someone can get out to the degree of speed that one can get out with a hatch-back. Again, if you look at the pictures, you see that there is uneven metal, it’s not padded, and the person is leaned against that uneven metal. And this is a bad, bad situation. And it is one that the defendant was aware of. It was his compartment and clearly fits within intentionally, recklessly creating a sudden risk of death or serious bodily injury. *893The facts are by proof beyond a reasonable doubt. No one’s really disputing the facts. Whether you have to force the front seat up. or force it are not really as irrelevant if you don’t have to force it up; There[] are still a lot of obstacles to getting out of this vehicle in a crash, assuming that one can do it. [The defense] assume[s] [that in case of an accident,] ... the vehicle’s ... going to be in the same condition that [it] is in now; it’s not. If it’s from the rear, especially from the sides, that person is going to be compressed. His head is going to be pushed against the compartment without any padding. He’s not restrained at all. This is not a good situation.
So I hold the six-level increase is appropriate.
The district court’s conclusion is entirely consistent with the plain language of the Guidelines. Section 2L1.1 addresses offenses involving the smuggling, transportation, or harboring of unlawful aliens. Subsection (b)(5) of that provision directs the sentencing court to increase the defendant’s offense level “[i]f the offense involved intentionally or recklessly creating a substantial risk of death or serious bodily injury to another person.” U.S.S.G. § 2Ll.l(b)(5) (2005). Application Note 6 further explains that § 2Ll.l(b)(5) applies to “a wide variety of conduct,” including: (1) “transporting persons in the trunk or engine compartment of a motor vehicle,” (2)“carrying substantially more passengers than the rated capacity of a motor vehicle,”, and (3) “harboring persons in a crowded, dangerous, or inhumane condition.” U.S.S.G. § 2L1.1 cmt. m 6 (2005).1 As the majority recognizes, the plain language of § 2Ll.l(b)(5) applies to transporting passengers in a way that “could increase the risk that an accident, if it should occur, would cause injury or death.” See maj. op. at 890.
Our case law has echoed the expansive language of the application note. Our pri- or, decisions have held that “[n]o precise formula undergirds the determination of what constitutes substantial risk.” United States v. Carreno, 363 F.3d 883, 890 (9th Cir.2004), vacated and remanded on other grounds, 543 U.S. 1099, 125 S.Ct. 1000, 160 L.Ed.2d 1000 (2005). In this spirit, we have emphasized that district courts should “carefully evaluate] the cumulative effect of the interrelated factors supporting the enhancement” and “assess the degree of risk created by the totality of the defendant’s conduct.” Id. The Fifth Circuit has expressly held that transporting aliens in a manner that exposes them to a substantial risk of serious bodily injury in’ the event of an ’accident is sufficient to uphold a § 2Ll.l(b)(5) enhancement. See United States v. Zuniga-Amezquita, 468 F.3d 886, 890 (5th Cir.2006).
The district court made the common sense determination that transporting an alien unrestrained by a seatbelt, crammed into a hidden compartment, and placed under a forty 'or, fifty pound bench seat which impeded his ability to exit posed serious dangers in the event of an accident. The conclusion that Torres-Flores’ conduct created “a substantial risk of *894death or serious bodily injury to another person” was undoubtedly correct.2
II.
The majority overturns the district court’s straightforward application of the Guidelines on the ground that the likelihood of an accident was not significant enough to merit a 6-level enhancement in the offense level. Maj. op. at 890-91. I disagree with this conclusion for two reasons.
First, the majority bases its conclusion that an accident was “highly unlikely,” and therefore the district court erred in applying the enhancement, on the majority’s sua sponte review of highway statistics that were not part of the record, and are at best inconclusive. See maj. op. at 891 & n. 9. Extrapolating from general National Highway Traffic Safety Administration statistics, the majority derives the surprisingly precise determination that the likelihood of Torres-Flores getting into an accident “was probably close to 0.03%.” The majority explains that it “calculate[s] that probability based on the defendant’s plan to drive 135 miles from the San Ysidro checkpoint to Marquez-Cruz’s destination in Los Angeles, and the frequency of accidents in March 2005-203 crashes for every 100 million miles driven.” Maj. op. at 891 n. 9.
The statistical information provided by the National Highway Traffic Safety Administration, and the majority’s factual extrapolation from that information, were not in the record before the district court and should not affect our review. Appellate courts generally consider only the “record before the trial judge when his decision was made.” Kirshner v. Uniden Corp. of Am., 842 F.2d 1074, 1077 (9th Cir.1988) (internal quotation marks omitted). Moreover, “[i]t is rarely appropriate for an appellate court to take judicial notice of facts that were not before the district court.” Flick v. Liberty Mut. Fire Ins. Co., 205 F.3d 386, 392 n. 7 (9th Cir.2000). Even if such traffic statistics were an appropriate subject of judicial notice, see Fed.R.Evid. 201(b) (requiring that facts be “not subject to reasonable dispute”), it is inappropriate to rely on such data where the government has had no notice or opportunity to respond. Cf. Fed.R.Evid. 201(e) (“A party is entitled upon timely request to an opportunity to be heard as to the propriety of taking judicial notice and the tenor of the matter noticed.”). Moreover, we have previously expressed doubt about relying on general statistics as proof of an ultimate issue in an individual case. See Haugen v. Brosseau, 351 F.3d 372, 389 (9th Cir.2003) (relying on general statistics of the dangers of high-speed car chases was improper, given that the statistics were not supplied by the parties, or responded to by the party adversely affected; also noting the Supreme Court’s rejection of this general statistical approach “to prove dangerousness in an individual case”).
More important, the results of the majority’s extrapolation are highly questionable. For example, the majority’s analysis fails to take into account statistics showing that because Torres-Flores was a twenty-five year-old male, he was over 36 % more likely to be in a crash than the general population. Nat’l Highway Traffic Safety Admin., Traffic Safety Facts 2005 98 (involvement rate of 7,354 compared to 5,398). The majority’s calculation also fails to take into account other factors that could affect the probability of Torres-Flores getting into an accident, including *895the effect of driving a light truck, driving where the speed limit was 55 miles per hour or higher, driving near the city of Chula Vista, and driving in the state of California. See id. at 17, 51, 148, 176. It is a truism that statistics can be used to prove anything, and absent specific information for drivers transporting concealed aliens from San Ysidro to Los Angeles (information apparently not provided by the National Highway Traffic Safety Administration), the statistics relied on by the majority are simply not an appropriate basis for determining that the district court abused its discretion.
Second, I question the majority’s analytical approach. The majority concludes that because the § 2L1.1(b)(5) enhancement resulted in a 6-level increase in defendant’s offense level, the district court abused its discretion in determining that the § 2Ll.l(b)(5)' enhancement was applicable to the facts of the case. Maj. op. at 890-91. But our determination as to whether a district court was correct in applying the law to the facts should not vary based on the size of the sentencing enhancement that results from this analysis. Torres-Flores himself concedes that “the guidelines do not suggest that the effect of the enhancement in terms of levels of increase , to the base offense level should be factored into the substantive factual analysis.” Although the disproportionate impact of an otherwise applicable sentencing factor may raise a need for additional procedural protections in fact-finding, see, e.g., United States v. Jordan, 256 F.3d 922, 926 (9th Cir.2001), Torres-Flores does not raise this issue. Under our procedure, if the sentencing factor is applicable by its terms; a court then turns to consider whether the sentence suggested by the advisory Guidelines is unreasonable in light of the § 3553(a) factors.3 Here there is no basis to conclude that Torres-Flores’ within-Guidelines sentence is unreasonable. See Rita v. United States, — U.S. -, 127 S.Ct. 2456, 168 L.Ed.2d 203 (2007).
The district court correctly applied U.S.S.G. § 2Ll.l(b)(5) in determining that the transportation of Marquez-Cruz in an unprotected compartment with uncovered metal edges, unrestrained by a seatbelt, and prevented from ready egress by forty to fifty plus pounds of material “created a substantial risk of death or serious bodily injury.” Therefore, I would affirm.
. Indeed, the language of Application Note 6 to U.S.S.G. § 2L1.1 is broad enough to cover the precise conduct in this case. The application note explains that the “wide variety of conduct” meriting the enhancement includes “harboring persons in a crowded, dangerous, or inhumane condition.” U.S.S.G. § 2L1.1 cmt. n. 6 (2005). This would include Torres-Flores’ conduct in concealing Marquez-Cruz ■in a condition the district court correctly found to be “dangerous.” See Webster's New World College Dictionary 647 (4th ed.2005) (defining to "harbor” as "to serve as, or provide, a place of protection to; shelter or house; conceal or hide”).
. As the majority notes, there is an intracir-cuit conflict regarding whether we review the district court’s application of the Guidelines to the facts of this case de novo or for an abuse of discretion. See maj. op. at 891 n. 11. I would affirm the district court under either standard.
. We follow "a two-step procedure for reviewing sentences imposed following the date the Supreme Court issued its opinion in Booker." United States v. Mix, 457 F.3d 906, 911 (9th Cir.2006) (citing United States v. Cantrell, 433 F.3d 1269, 1279-81 (9th Cir.2006)). First, we determine whether the district court properly considered and applied the applicable Sentencing Guidelines. Cantrell, 433 F.3d at 1279-81. If the district court did not err in applying the Sentencing Guidelines, we review the sentence for reasonableness in light of the factors set forth in 18 U.S.C. § 3553(a). Cantrell, 433 F.3d at 1280. The majority seemingly conflates these two steps.