This paper presents a detailed analysis of the composition of household portfolios, using both aggregate and micro-data. Among the key findings are that: Most household wealth is held in the form of housing and pensions. Over time, there has been a shift away from housing towards financial assets, driven largely by the growth in life and pension funds. Liquid financial wealth (excluding life and pension funds) is not predominantly held in risky form. By far the most commonly held asset is an interest-bearing account at a bank or building society account. Of people with positive (liquid) financial wealth, more than half is held in savings accounts. The importance of risky assets in an individual's portfolio varies according to their characteristics. The unconditional portfolio share held in risky assets (i.e. averaged across those with and without any risky assets) rises with both age and total financial wealth. However, most of the variation in unconditional portfolio shares is due to differences in ownership rates as opposed to the proportion of the portfolio held in risky assets. Looking only at the people within each wealth decile who have risky assets, the conditional portfolio share is relatively constant across wealth, suggesting a possible role for entry costs or other fixed costs in explaining portfolio holdings. Multivariate analysis shows that the conditional portfolio share in risky assets actually falls with age as classical portfolio theory would predict. Finally, the tax treatment of savings products has an effect on portfolio choice. Separate probit regressions for the ownership of tax-favoured assets and similar assets without the tax exemption, show that, controlling for other factors, marginal tax rates are important in determining asset ownership. These results are in accordance with those found by Poterba in the US.