It was great to have Stephen Penman participate in our round table discussion this past week. It is worthwhile to listen to someone else’s views on standard setting, particularly when those views have been somewhat critical of the current state of standard setting. Although I still have many questions about the completeness of Stephen’s framework for financial reporting (a framework he described as transaction-based), I feel like I have a better understanding of that framework and places where it still needs further development.

One place that seems to need further development was highlighted toward the end of our round table discussion. Stephen had provided a high-level description of a transaction-based framework for financial reporting. My understanding of that framework is that it seeks to recognize amounts in the financial statements that derive directly from transactions, not hypothetical events. We were talking about how a product sold with a warranty usually results in the recognition of a warranty expense and an accrual for anticipated repair costs. Stephen explained that the warranty accrual would be appropriate in a transaction-based framework because you could base that accrual on a history of many previous repair transactions. Because these transactions exist, the warranty accrual can be estimated with some objectivity at the point of sale.

As I was thinking about this example, I started to wonder why the same logic couldn’t be applied to a separately sold extended warranty. In that case, an entity may have just as much history on which to base its expected repair costs as an entity that sales a product with a warranty. If so, why couldn’t the entity recognize the full amount of revenue for the extended warranty at the point of sale and accrue an expense based on the anticipated repair costs for the warranty? Of course, the obvious rebuttal to this idea is that the entity providing the extended warranty has not yet provided any warranty coverage. But the same could be said of the entity that sales a product with a warranty. At the point of sale, no warranty coverage has yet been provided in either case, and yet all of the revenue associated with the sale of a product with a warranty is recognized up front.

My question here is how the transaction-based framework would distinguish between these two types of warranties. In both cases, a history of repair transactions with the customer suggests a reasonable estimate of the expected repair costs. So what in a transaction-based framework tells me to recognize all revenue at the point of sale for a product sold with a warranty but none of the revenue at the point of sale for an extended warranty sold on its own?

Again, let me express my thanks to Stephen for his willingness to share his views in our round table forum and for thinking on the fly about how a transaction-based framework would handle this and other situations. I hope we can have more discussions like this one in future round tables.