concurring in part and dissenting in part.
I concur in Sections I. and II. A. of the majority opinion. I cannot join Section II.B, however, because the majority misapplies the “direct is direct” approach to bar TMTA from coverage under Hartford’s crime shield policy for a direct loss resulting from employee theft.
The only policy construction issue Hartford presented to our court is this: Does the policy’s indirect loss exclusion require denial of coverage for Tyler’s theft of TMTA’s money. We summarily and correctly dispose of that issue in footnote 5 of the majority opinion by holding that Hartford failed to negate coverage through the “indirect loss” exclusion. That decision resolves the case in favor of coverage. Yet the majority goes on to explain why it thinks Hartford is not required to provide fidelity coverage despite TMTA’s payment of premiums and the inapplicability of the indirect loss exclusion. It does so by employing an erroneous construct.
*679A fidelity bond, like the crime shield policy at issue, provides coverage to the insured for a first-party loss; in other words, immediate financial loss to the insured resulting directly from employee theft or dishonesty — for example, embezzlement. See Retail Ventures, Inc. v. Nat’l Union Fire Ins. Co., 691 F.3d 821, 828 (6th Cir.2012); Vons Cos. v. Fed. Ins. Co., 212 F.3d 489, 492 (9th Cir.2000). A liability policy, by contrast, requires the insurer to pay the insured’s obligation to a third party for some act committed by the insured or its employee. See Lynch Prop., Inc. v. Potomac Ins. Co., 140 F.3d 622, 629 (5th Cir.1998). Although a fidelity policy may cover the loss of third-party property in the possession of the insured who acts as bailee or trustee, the policy does “not serve as liability insurance to protect employers against tortious acts committed against third-parties by their employees.” Id.; Retail Ventures, Inc., 691 F.3d at 829 n. 4.
To preserve the purpose of fidelity bonds, the insurance industry inserted into the standard coverage clause the phrase, “loss resulting directly from.” Retail Ventures, Inc., 691 F.3d at 829. The industry adopted this measure to prevent the insured from transforming first-party fidelity coverage into third-party liability insurance. See Vons Cos., 212 F.3d at 491-93. The majority carefully identifies a split between those courts that follow the insurance industry’s preferred “direct is direct” approach to include only first-party claims within fidelity coverage and those courts that employ a proximate cause approach to include coverage for third-party claims within fidelity coverage as well. I do not dispute that such a split exists, for our court has recognized it twice before in recent opinions without joining either side of the debate. Retail Ventures, Inc., 691 F.3d at 828-29; First Defiance Fin. Corp. v. Progressive Cas. Ins. Co., 688 F.3d 265, 270-71 (6th Cir.2012). I find the majority’s declaration of allegiance to the “direct is direct” approach not just in error but also particularly problematic in light of our court’s recent questioning of “whether the distinction between direct cause and proximate cause is a meaningful one.” First Defiance Fin. Corp., 688 F.3d at 270 (citing Staub v. Proctor Hosp., — U.S.-, 131 S.Ct. 1186, 1192, 179 L.Ed.2d 144 (2011) (observing that proximate cause requires only some direct relation between injury asserted and injurious conduct alleged and excludes only those links that are too remote, purely contingent, or indirect)).
My more immediate concern is, however, that the majority unties the “direct is direct” approach from its mooring and uses it in a completely inappropriate way: to analyze the relationship between TMTA, the Agency, and Tyler. According to the majority’s reasoning, if Tyler had not committed theft, the commissions would have flowed through the Agency before finally reaching TMTA’s bank account. Because the money did not flow directly from Tyler to TMTA, but passed through the Agency, the majority applies the “direct is direct” approach, then construes that approach to mandate the conclusion that the fidelity policy bars coverage.
This analysis distorts the fundamental purpose of the “direct is direct” approach, which is to distinguish between first- and third-party claims and restrict fidelity coverage to first-party claims, such as we have here. TMTA, the insured, paid Tyler, its employee, to sell insurance policies to TMTA members. TMTA owned the commissions earned on the sale of those policies, but Tyler embezzled the money for his own use. The loss of money TMTA suffered resulted directly from the theft by its own employee. We have no need to decide whether this policy would cover a *680third-party claim for damages arising from Tyler’s misconduct, because that fact pattern is not presented. Moreover, this flawed analysis certainly should not be employed to deny coverage when correct application of the “direct is direct” concept to these facts requires coverage.
To decide in this case whether the Agency’s existence has any impact on coverage, we must inquire into what the insurance contract means by “loss ... which results directly from theft by an employee.” Because the case comes to us under diversity jurisdiction, we must look to Michigan law to interpret the meaning of the policy. See Erie R.R. Co. v. Tompkins, 304 U.S. 64, 58 S.Ct. 817, 82 L.Ed. 1188 (1938); Retail Ventures, Inc., 691 F.3d at 825-26 (deciding how Ohio courts would interpret similar policy). If the Michigan Supreme Court has not decided an issue pertinent to our analysis, we must “ascertain the state law from all relevant data,” which include the state’s intermediate appellate court decisions, unless we are convinced by other persuasive information that the Michigan Supreme Court would decide otherwise. Garden City Osteopathic Hosp. v. HBE Corp., 55 F.3d 1126, 1130 (6th Cir.1995) (quoted case and internal quotation marks omitted).
The Michigan Court of Appeals has held that the word “direct” means “immediate” or “proximate” cause, “as distinct from remote or incidental causes.” Acorn Inv. Co. v. Mich. Basic Prop. Ins. Ass’n, No. 284234, 2009 WL 2952677 (Mich.Ct.App. Sept. 15, 2009) (unpublished per curiam). In Acorn Inv. Co., thieves removed copper piping and other fixtures from a rental property, causing extensive flooding in addition to the loss of the piping and fixtures. The property owner sought coverage for the flood damage under a named perils policy for “direct physical loss” caused by vandalism. Id. at *2. The Michigan Court of Appeals inclusively defined the word “direct” to encompass both immediate and proximate causes, thereby finding coverage under the policy for flood damage. Id. See also Universal Image Prods., Inc. v. Fed. Ins. Co., 475 Fed.Appx. 569, 573-74 (6th Cir.2012) (relying on Acorn Inv. Co. to discern Michigan law on the meaning of the phrase “direct physical loss or damage”). There is no reason to believe that the Michigan Supreme Court would adopt a different definition of the word “direct.” Thus, to insure seamless application of Michigan law, we should follow our own lead in Universal Image Prods., Inc. and apply Acorn Inv. Co. to the case before us.
Doing so is consistent with our recent opinion in Retail Ventures, Inc. In that case, our court looked to Ohio law to discern the meaning of the identical phrase “loss resulting directly from” included within a crime policy. Retail Ventures, Inc., 691 F.3d at 824-25, 826-28. Considering the pertinent endorsement and exclusions in that policy, the court determined that the phrase “resulting directly from” did not unambiguously limit coverage to loss resulting “solely” or “immediately” ’ from the theft itself. Id. at 830-31. Noting that the Ohio Court of Appeals had applied a proximate cause standard to determine whether there was a “direct loss” under other kinds of first-party coverage, our court concluded that the Ohio Supreme Court “would apply a proximate cause standard to determine whether the loss was covered.” Id. We undertook a comparable analysis in Union Planters Bank, N.A. v. Cont’l Cas. Co., 478 F.3d 759, 764 (6th Cir.2007), where we relied on Tennessee law holding that the definition of “direct cause” incorporates the overlapping concepts of “proximate cause” and “efficient cause” in order to find coverage under a crime policy for loss “resulting directly from” forgery. See also Scirex Corp. v. Fed. Ins. Co., 313 F.3d 841, 850 *681(3d Cir.2002) (holding that dishonesty of nurses employed by Scirex Corporation rendered medication studies worthless so that nurses’ conduct “proximately, and therefore directly, caused Seirex’s losses.”).
Tyler’s theft of commissions owned by TMTA — even if not characterized as immediate due to the existence of the Agency — was certainly the proximate cause of TMTA’s loss. See Acorn Inv. Co., 2009 WL 2952677, at *2. TMTA formed the Agency as a mere conduit for the payment of commissions in order to comply with Michigan insurance law, which forbids TMTA from selling insurance directly to consumers. As the district court explained, the Agency, which is a single-member limited liability company controlled by TMTA, accumulates no income, pays no taxes, owns no property, and has no employees. TMTA absorbs all of the Agency’s income and losses, and pays all of the taxes. The Agency’s income is, by law, TMTA’s income. In addition, TMTA paid Tyler as its employee. Tyler’s gains came only at TMTA’s expense. See Gonzalez Design Eng’g, Inc. v. Citizens Ins. Co., No. 187413, 1997 WL 33350547, at *2 (Mich.Ct. App. Apr. 15, 1997). This is because, as we have expressly recognized, “[embezzlement is a zero-sum game. For the employee to win, the employer must lose.” Fed. Deposit Ins. Corp. v. St. Paul Fire & Marine Ins. Co., 942 F.2d 1032, 1036 (6th Cir.1991). TMTA suffered a loss of money which resulted immediately or proximately from Tyler’s theft of the commissions owed to TMTA. Consequently, TMTA is entitled to coverage under the policy for that loss.
“The purpose of insurance is to insure against a loss.” Fed. Ins. Co. v. Hartford Steam Boiler Inspection and Ins. Co., 415 F.3d 487, 495 (6th Cir.2005). We should not construe a policy to defeat coverage unless the policy language requires it. Id. The language of the crime shield policy, properly construed, does not defeat coverage; it authorizes coverage. Accordingly, for the reasons stated, I would reverse the district court’s grant of summary judgment in favor of Hartford and remand with an instruction to enter judgment in favor of TMTA for the limits of the policy.