The Portfolio Composition Effect

85 Pages Posted: 28 Jul 2020 Last revised: 16 Aug 2020

See all articles by Jan Müller-Dethard

Jan Müller-Dethard

University of Mannheim

Martin Weber

University of Mannheim - Department of Banking and Finance

Multiple version iconThere are 2 versions of this paper

Date Written: July 2020


Does the evaluation of a portfolio of stocks depend on its composition of winner and loser stocks? To test this, we define a simple, counting-based measure of performance - the number of winner relative to the number of loser stocks in a portfolio - and examine how this composition measure affects individuals' willingness to invest in a portfolio. We derive testable predictions for the proposed composition measure from a framework which combines category-based thinking with mental accounting. Consistent with our predictions, we find across all experiments that individuals allocate larger investments to portfolios with more winner than loser stocks relative to alternative portfolios with more loser than winner stocks, although both portfolios (1) have realized identical overall portfolio returns and (2) show identical expected risk-return characteristics. Building on our experimental findings, we analyze fund flows of exchange-traded funds on leading equity market indices. We identify that the proposed portfolio composition measure is positively related to future net fund flows.

Keywords: investment behavior, Mental accounting, Portfolio composition, risk preferences

JEL Classification: D84, G11, G12, G40

Suggested Citation

Müller-Dethard, Jan and Weber, Martin, The Portfolio Composition Effect (July 2020). CEPR Discussion Paper No. DP15012, Available at SSRN:

Jan Müller-Dethard (Contact Author)

University of Mannheim ( email )

Department of Finance, L9, 1-2
Mannheim, 68131

Martin Weber

University of Mannheim - Department of Banking and Finance ( email )

D-68131 Mannheim
+49 621 181 1532 (Phone)
+49 621 181 1534 (Fax)

Here is the Coronavirus
related research on SSRN

Paper statistics

Abstract Views
PlumX Metrics