It's About Time: An Examination of Loss Reserve Development Time Horizons. Issue 2 (3rd April 2019)

• Record Type: Journal Article
• Authors: Barth, Michael M.; Eastman, Evan M.; Eckles, David L.
• Is Part Of: North American actuarial journal. Volume 23:Issue 2(2019)
• Page Start: 143
• Page End: 168
• Publication Date: 2019-04-03
• Subjects: Life insurance -- Research -- North America -- Periodicals; Actuarial science -- North America -- Periodicals; Web sites; Electronic journals; 368.010973
• DOI: 10.1080/10920277.2018.1538804
• Languages: English
• ISSNs: 2325-0453

Abstract

Abstract : A rich body of academic research has addressed the question of earnings management in the property-casualty insurance industry via manipulation of loss reserve estimates. This study analyzes the variability of reserve estimates at different development horizons to determine whether the predominant practice of relying on five years of development is appropriate. We examine two common measures of reserve estimation error, calendar year development and accident year development, and compare and contrast the two approaches. We also consider the appropriateness of the common practice of aggregating lines of business. After examining reserve development patterns for each of the major lines of business, we conclude that the appropriate development horizon to adequately establish ultimate liability may be longer than the current maximum reported horizon of 10 years found in Schedule P for most lines of business, including the aggregate reserves. Although longer-term development horizons are necessary to establish insurers' ultimate liability, relatively short-term development horizons may be more appropriate when attempting to identify deliberate manipulations or to assess solvency risk, where the short-term variations are the primary object of interest. Ultimately, this article investigates the degree to which methodology originally developed for estimating loss reserve errors is appropriate today, in particular, relative to current data availability.