Investors wishing to analyse the risk of their portfolios, and to rebalance them in the light of the results, often do not have detailed stock-level data. Typically, however, asset class data is available: for example, the manager of a real estate portfolio may not have each individual stock's risk characteristics, but will usually have access to the risk characteristics of the relevant sector indices. In these circumstances, the conventional approach is to assume that each sector index as it stands is a good representation of the corresponding sub-portfolio, ignoring the sector information. Sometimes this will be a reasonable assumption, but it is unlikely to be so if the sub-portfolio is not an index fund, closet index fund, or commingled fund with similar characteristics, that is if (a) it is genuinely actively managed, and/or (b) the portfolio is too small to warrant holding an adequate number of stocks to diversify away residual and other risk. // This paper suggests a general methodology by which the investor can use the sector-level information to improve on the conventional approach, and * obtain a better estimate of total portfolio risk, whether in absolute terms or relative to a benchmark * more accurately estimate the risk contributions of the different parts of the portfolio * allocate assets more realistically * more specifically, decide whether commingled funds are a better choice than direct holdings for all or part of the portfolio. // The methodology has now been implemented in practical software and, with the assistance and encouragement of IPD, applied to their database with revealing results.