Portland Federal Employees Credit Union (Portland) appeals from the district court’s entry of summary judgment in favor of Cumis Insurance Society, Inc. (Cu-mis) on the question whether a fidelity bond issued by Cumis covers certain losses Portland suffered because of its employees’ fraud, dishonesty, or lack of faithful performance. The parties dispute whether these losses qualify as losses of “property” as defined by the fidelity bond. The district court had diversity jurisdiction under 28 U.S.C. § 1332 . We have jurisdiction pursuant to 28 U.S.C. § 1291 . We reverse and remand.

This case involves two separate cases brought by Portland, consolidated in the district court and tried together. The facts are undisputed for purposes of this appeal. In case number 86-134 (the loan claim), Hoekstra, Portland’s treasurer-manager, disbursed a loan of $62,100.00 to another employee. Disbursement of a loan this large without prior approval from Portland’s credit committee violated Portland’s policies and bylaws as well as Oregon state law. In addition, the loan was secured by real property for which the appraisal was inaccurately high. Thus, the loan exceeded the collateral’s value by an impermissible amount, again in violation of Portland’s policies. When the loan defaulted, Portland foreclosed, but sale of the collateral failed to recover the full amount of the outstanding balance on the loan. This action seeks recovery of the difference.

In case number 85-1897 (the building claim), Portland’s losses and damages stem from Hoekstra’s activities in connection with the construction of a new credit union building. Part of Portland’s losses result from Hoekstra’s execution of a consultant agreement to build, furnish, and equip the new building. This consultant agreement selectively ignored directives from Portland’s Board, and was never shown to the Board. Other losses arose from Hoeks-tra’s purchase of land for the building and his execution of contracts for interior work and general work without Board knowledge or approval. As a result of these improper contracts, Portland acquired a building which it did not need, and which is worth less than it paid.

According to Portland, Chapman, its legal counsel, is also partly responsible for some losses on the building claim. Chapman allegedly gave a legal opinion that the contracts mentioned above were binding, an opinion allegedly motivated by improper activity upon his personal loan (presumably in collusion with Hoekstra).

In addition, Portland suffered losses through interest payments on an allegedly illegal insider loan from the Oregon Corporate Central Credit Union (OCCCU) secured by Hoekstra. Although proceeds from OCCCU loans could be used only for liquidity needs arising from Portland’s loan demands and share withdrawals, Hoekstra used the proceeds to meet progress billings for the building project. Portland also suffered $2,000,000 in losses to its reserves. Portland estimates its loss on the building claim at $10,065,431.00.

For purposes of this summary judgment motion, Cumis conceded that the cause of Portland’s losses satisfied the requirements of its coverage. In other words, for present purposes Cumis concedes that the claims resulted from fraud, dishonesty, or failure of an employee to perform his duties well and faithfully. The question is whether the Cumis fidelity bond covers the type of loss suffered by Portland.

Both cases were assigned to a magistrate, who consolidated them. Cumis filed motions for summary judgment contending that the type of loss was not covered by the fidelity bond. The magistrate ruled that Portland’s losses were not covered and recommended that Cumis’s summary judgment motions be granted. Pursuant to 28 U.S.C. § 636 (b)(1), the district court conducted a de novo review of the magistrate’s findings and recommendations. Portland filed a memorandum stating its objections to the magistrate’s findings and recommendations, and Cumis filed a response. The district court agreed with and adopted the magistrate’s findings and recommendations, and entered an order and judgments granting Cumis’s summary judgment motions. Portland then appealed.

We first address the threshold question whether we have jurisdiction to entertain this appeal. This depends on whether Portland’s appeal was timely.

Portland suggests that we construe its “Petition for Leave to Appeal” as a timely notice of appeal. Cumis points out that Portland filed the wrong document with the wrong court.

Fed.R.App.P. 3(c) provides that “[a]n appeal shall not be dismissed for informality of form or title of the notice of appeal.” Under Rule 3(c) “we are required to broadly construe the notice of appeal provisions of [Fed.R.App.P.] 4(a) ... [and] we have discretion, where the interests of substantive justice require it, to disregard irregularities in the form or procedure for filing a notice of appeal.” San Diego Committee Against Registration and the Draft (CARD) v. Governing Board, 790 F.2d 1471, 1474 (9th Cir.1986), citing Cel-A-Pak v. California Agricultural Labor Relations Board, 680 F.2d 664, 667 (9th Cir.), cert. denied, 459 U.S. 1071 , 103 S.Ct. 491 , 74 L.Ed.2d 633 (1982).

We construe Portland’s petition as a timely notice of appeal filed in the district court on the date it was filed with us, pursuant to Rule 4(a)(1). Cumis suggests that even so construed, however, Portland’s notice of appeal is without force because Portland agreed to withdraw the “Petition for Leave to Appeal” following the prebriefing conference. Closer examination reveals that Portland did not simply agree to withdraw the “Petition.” Rather, Portland agreed to withdraw its petition contingent upon this court’s decision to treat its untimely notice of appeal as timely. In effect, Portland has never withdrawn its petition. We have jurisdiction.

Ill

We review the district court’s summary judgment de novo. Poland v. Martin, 761 F.2d 546, 547 (9th Cir.1985). We apply Oregon substantive law. Under Oregon law, the parties’ rights and obligations under the terms of the fidelity bond are first to be determined by the contract itself. The primary consideration is to ascertain the intent of the parties. First Far West Transportation, Inc. v. Carolina Casualty Insurance Co., 47 Or.App. 339 , 614 P.2d 1187, 1190 (1980). We look within the “four corners” of the document and read the contract as a whole. Denton v. International Health & Life Insurance Co., 270 Or. 444 , 528 P.2d 546, 549 (1974).

Whether a contract is ambiguous is a question of law, reviewed independently. Mortgage Bancorporation v. New Hampshire Insurance Co., 67 Or.App. 261 , 677 P.2d 726, 728 (1984). This is a threshold inquiry. See Adams v. Northwest Farm Bureau Insurance Co., 40 Or.App. 159 , 594 P.2d 1256, 1258 (1979). The construction of an unambiguous contract is a question of law. Timberline Equipment Co. v. St. Paul Fire & Marine Insurance Co., 281 Or. 639 , 576 P.2d 1244, 1246 (1978).

The district court decided that the fidelity bond was unambiguous, and held that coverage was precluded because Portland's losses did not qualify as "money" as defined in the fidelity bond. The bond contains a definition of "property" that encompasses money and securities among other items. Following the definition of "property" are detailed definitions of "money" and "securities":

(Emphasis added.) The court held that Portland's loan claim involved only "a loss from a loan, an item not listed in the definition of money." Portland's building claim involved economic losses or economic disappointment but likewise involved no loss of money.

On appeal, Portland challenges the district court's decision that its losses are not covered by the fidelity bond. Portland does not renew its claim that its losses are covered under a directors and officers liability policy attached as an endorsement to the fidelity bond.

We agree with the district court's determination that the definition of money is unambiguous and that Portland's loss on the loan claim does not qualify as loss of money. Portland's loan claim does not involves loss of currency, coin, bank notes, Federal Reserve Notes, revenue stamps or postage stamps. Where a word or phrase is defined in an insurance policy, that definition is binding on the court. Waterman Steamship Corp. v. Snow, 222 F.Supp. 892, 896 (D.Or.1963) (Waterman), affd, 331 F.2d 852 (9th Cir.1964) (applying Oregon law).

We cannot substitute a more expansive definition of money in order to find coverage. See id. No case of which we are aware, from Oregon or any other jurisdiction, has interpreted loss of "money," so defined, to include the loss Portland has suffered. Portland's reliance on the court's expansive interpretation of "money" in Hooker v. New Amsterdam Casualty Co., 33 F.Supp. 672, 673 (W.D.Ky.1940), is misplaced, because the bond in that case nowhere defined the term "money." Other cases cited by Portland as establishing coverage rely on contractual terms other than "money." Portland advances numerous policy contentions and arguments concerning construction of insurance contracts to show why the fidelity bond should be construed to cover this loss, but even if we were to agree with every argument, under Oregon law we would still be bound by a contractual definition of "money" which precludes Portland's claim.

Portland also raises the issue of whether its losses are "property" within the meaning of the bond. This issue was raised in Portland's "Supplemental Memorandum in Opposition to Defendant's Motion for Summary Judgment," filed as an objection to the magistrate's findings and recommendations. Portland points out that, although not discussed by the magistrate, the definition of property includes among other items "securities ... in which the Insured has a pecuniary interest or which are held in the Insured in any capacity." "Securities" are further defined in the fidelity bond as:

We observe that “securities” is a broad term and that securities may embrace intangible as well as tangible interests. Portland argues that the loan claim involves the use of checks and mortgages, both terms included in the definition of securities.

Portland also draws our attention to exclusion (e), which provides that the fidelity bond does not cover:

Portland suggests that, read together with other provisions, this exclusion indicates that losses from loans (or any transactions in the nature of or amounting to loans) are covered if the loss is caused by the type of behavior indicated in Insuring Clause A. In other words, bad business decisions are not covered, but loans are covered when made because of a Portland employee’s fraud, dishonesty or failure to perform his duties well and faithfully.

Cumis responds that under Oregon law, an exclusion cannot be used to create coverage. See Progressive Casualty Co. v. McManus, 83 Or.App. 582 , 732 P.2d 932 , 934 n. 5 (1987). In any event, Cumis argues, the terms of this exclusion indicate that coverage is determined solely in Insuring Clause A.

We believe that an Oregon court would hold that the fidelity bond does cover the type of loss involved in Portland’s loan claim. It probably would conclude that this loss falls under the fidelity bond’s definition of securities in either of two ways. First, the loss involves a check “in which

[Portland] has a pecuniary interest or which [is] held by [Portland] in any capacity.” The record discloses that Hoekstra disbursed the loan through a “check request form” and execution of a check payable to the Portland employee. Portland suffered a loss at the moment when, through Hoekstra’s improper conduct, the check was exchanged for collateral of lesser value. Alternatively, in light of the fraudulent or dishonest appraisal of the house which served as collateral, the loss also could be characterized as loss to property in which Portland has a pecuniary interest, property being defined as “[mortgages and instruments in the nature of mortgages upon real estate ... and upon interest therein, and assignments of such mortgages, ... and all other instruments similar to or in the nature of the foregoing.” We believe the Oregon court would reach this conclusion by reading the contract as a whole, including exclusion (e). Although that exclusion cannot alone create coverage, read together with Insuring Clause A it supports the conclusion that coverage exists. It would make no sense for exclusion (e) to contain the language “except as [the loan loss] may be covered by Insuring Clause A” unless losses from loans such as that involved here may indeed be covered by Insuring Clause A.

This conclusion does not necessarily mean that every dollar of Portland’s loan claim is entitled to coverage. Exclusion (h) excludes “[a]ny indirect or consequential loss.” It may be that some of Portland’s itemized losses will be excluded under exclusion (h). The district court based its decision rejecting coverage on its interpretation of “property” and “money,” and therefore did not reach the question whether part or all of Portland’s loss was direct as opposed to indirect or consequential. We leave this determination for the district court on remand to consider in the first instance.

We turn now to Portland’s building claim. The building claim involves several types of loss. The district court, without distinguishing among the numerous components of Portland’s claim, held that cover age was precluded because Portland had merely suffered economic "disappointment" or "losses" as a result of the contract decisions involved, but had not experienced loss of "money."

The record does not indicate the precise form of payment used by Hoekstra to purchase the land for the building, nor does it indicate the form of payment through which Hoekstra or Portland disbursed payment pursuant to the consultant agreement and other allegedly improper contracts. It is probable that the form of payment in each case was one of those enumerated in the definitions of "money" or "securities." We believe an Oregon court wQuld hold that Portland suffered a loss at the moment when (through Hoekstra's fraud, dishonesty or lack of faithful performance) currency, a check, a money order, a promissory note or a similar instrument was exchanged for land, goods or services of lesser value. If the transaction resulting in the building loss was consummated or facilitated by any such instrumentalities, the building claim would fall within the definition of a covered security loss.

The parties ardently dispute whether "intangible" losses are covered under the fidelity bond's definition of "property" and "securities." Portland points to two major cases which construe a bond containing similar language to cover intangible losses. See First National Bank of Decatur v. Insurance Co. of North America, 424 F.2d 312, 315 (7th Cir.) (Decatur) cert. denied, 398 U.S. 939 , 90 S.Ct. 1844 , 26 L.Ed.2d 272 (1970); Clarendon Bank & Trust v. Fidelity & Deposit Co. of Maryland, 406 F.Supp. 1161, 1169 (E.D.Va.1975) (Clarendon). Cumis attempts to distinguish these cases because there, the definition of property included "evidences of debt" and "rights," terms not included in the fidelity bond here. The district court agreed with this distinction and held that intangible losses were not covered.

Aside from the inclusion of these terms, the definition of property in Clarendon and Decatur was virtually identical to the definition of money and securities here. Those cases may therefore be distinguished only if "evidence of debt" and "rights" are so different in character from the other terms listed in the definition that they alone could have justified other courts' con-elusion that intangible property was covered. We do not find such a great difference. Oregon law looks to the ordinary and general meaning of "evidences of debt" and "rights." See Waterman, 222 F.Supp. at 896 . Evidences of debt have been defined as "written instruments or securities for the payment of money, importing on their face the existence of a debt." Black's Law Dictionary 499 (5th ed. 1979). See Pacific Southwest Realty Co. v. Commissioner, 128 F.2d 815, 819 (9th Cir.) (Healy, J., concurring) (example of common usage of term "evidences of debt"), cert. denied, 317 U.S. 663 , 63 S.Ct. 64 , 87 L.Ed. 533 (1942); 13 C.F.R. § 101.2 (d)(2) (1988) (same); 12 U.S.C. § 91 (same); see generally Mortgage Guarantee Co. v. Welch, 38 F.2d 184, 186 (9th Cir.) (definitions of "evidences of debt"), cert. denied, 281 U.S. 759 , 50 S.Ct. 410 , 74 L.Ed. 1168 (1930). We observe that a bond is defined as "[a] certificate or evidence of a debt. . . ." Black's Law Dictionary at 161. "Rights," which we understand to mean "stock rights" (as opposed, for example, to constitutional rights), are defined as "[a] document (i.e. negotiable certificate) which gives an existing stockholder the privilege of buying additional stock of a corporation." Id. at 1272 . We see little difference for purposes of the question before us between these terms and others listed in the definition of "securities" such as "bonds," "debentures," "warrants," "coupons," and "certificates of stock." See Or.Rev.Stat. § 59.710(1) & (4) (1987) (defining "securities" and "bonds"). Indeed, the fidelity bond's definition of "securities" includes "all other instruments similar to or in the nature of the foregoing," so "evidences of debt" and "rights" may already be incorporated in the definition of property.

We see no basis for distinguishing cases interpreting similar language in a fidelity bonds and therefore we believe the Oregon courts would conclude that the Cumis fidelity bond does extend to cover "intangible" property. See Decatur, 424 F.2d at 315 ; Clarendon, 408 F.Supp. at 1169; see generally Or.Rev.Stat. §§ 307.020 & 98.302(9) (1987) (defining intangible property).

Portland’s building claim includes numerous types of loss the nature of which is not adequately developed in the record before us. Without a fuller understanding of each transaction and the way in which that transaction constituted a “loss” suffered by Portland, it would be premature and speculative for us to attempt to state definitely whether a particular transaction qualifies as a covered loss. On remand, the district court should develop the record more fully and determine whether, consistent with this opinion, each component of Portland’s building claim is covered. As with the loan claim, we leave it for the district court in the first instance to determine whether each component of the building claim qualifies as a direct loss rather than an indirect or consequential loss.

REVERSED AND REMANDED.