Paul Ehling (Norwegian School of Management (BI), Department of Financial Economics), Michael F. Gallmayer (Texas A&M University, Mays Business School, Department of Finance), Sanjay Srivastava (Georgia State University), Stathis Tompaidis (University of Texas, Red McCombs School of Business) & Chunyu Yang (University of Texas, Red McCombs School of Business) have posted Portfolio Choice with Capital Gain Taxation and the Limited Use of Losses on SSRN. Here is the abstract:

We study the consumption-portfolio problem with realized capital gain taxation and an important feature of real-world tax codes: that capital losses can only be used against capital gains. We find that this feature, which we call the limited use of losses (LUL), has striking implications for asset allocation and rebalancing of portfolios. In particular, when embedded capital gains are large, a capital lock-in effect dominates and makes it costly for the investor to trade out of a large equity position. As a result, investors in down markets hold significantly less equity than in up markets, and this creates a time-varying and path-dependent optimal equity holding that looks like increased risk aversion in down markets. These results contrast with intuition derived from existing work which assumes that use of losses is unrestricted, termed the full use of losses (FUL). Equity holdings are similar between FUL and LUL if investors have large embedded capital gains. Otherwise, even if embedded gains/losses are small, investors in an LUL world hold significantly less equity than in an FUL world, and the difference is most significant with large embedded losses. With FUL, rebates generated from capital losses artificially inflate the demand for equity, in fact even to levels above the no tax benchmark. We also show that the FUL case can lead to counterintuitive results; for example, an FUL investor can actually prefer paying capital gain taxes than being untaxed.