This paper examines the problems of estimating risk measures and their stability in thin markets. It shows analytically that conventional approaches used in previous studies can lead to serious overestimates of the stability of risk measures when shares are subject to thin trading. It then demonstrates, using UK data, that this is, in fact, a serious practical problem, and that the resultant biases are of precisely the form predicted. Finally, the paper presents reliable evidence on the stability of UK risk measures by using an estimation method designed to avoid thin trading bias. Using this approach, risk measures are found to be as stable in the UK as they are in the USA.