Investment arbitration is an expensive endeavor. In complex cases reaching full merits proceedings, combined party costs frequently reach into the millions of dollars, and sometimes even into the tens of millions. Yet, unlike most domestic jurisdictions, the World Bank's International Centre for Settlement of Investment Disputes (ICSID) and the United Nations Commission on International Trade Law (UNCITRAL) rules for the apportionment of arbitral expenses and attorneys' fees between parties leave significant room for argumentation and arbitral discretion. The possible effects of this uncertainty are large - ranging from reluctance to initiate arbitration on the part of claimants with relatively weak cases to effects on governmental regulation stemming from the prospect of being hit with damages, plus arbitral fees and legal costs in the event of an adverse cost award. Given the current popularity of investment arbitration, ensuring predictability of costs and fees is a worthy goal.

This Note builds on works of both quantitative and qualitative scholarship, in addition to my own analysis of the thirty-one public investment arbitration awards released in 2008 and 2009. Part I discusses previous scholarship on costs allocation - both normative and empirical - and presents new empirical results in the context of prior and current qualitative scholarship. Part II undertakes a more detailed analysis of the recent arbitral awards that comprise this Note's new data set. It contends that, in recent years, ICSID and UNCITRAL tribunals have reached widely divergent results that are inconsistent with the application of a steady costs-allocation regime. Rather, these outcomes result from the variable application of multiple factors. The result is a regime in which victorious claimants are substantially more likely to recover some measure of legal fees or arbitral costs than victorious respondents. However, this outcome is nothing more than a tendency: Depending on the other circumstances of the case - particularly excessive filings, wholly unmeritorious claims, or fraud - respondents can and do recover large amounts of expenses from losing claimants. Part III concludes by presenting a brief model for costs allocation that might unify, stabilize, and moderate the existing views of cost-shifting between arbitral parties.