This paper analyzes retirement timing decisions of DC pension plan members. In a first step, a model of optimal retirement timing decisions is proposed which incorporates the optimal time to annuitize the DC pension wealth. An individual obtains utility from leisure, labor income before retirement and pension benefits after retirement. These benefits include the income from an annuity which is bought at an optimal time. The optimal annuitization time depends on current and future expected financial market performance. Based on the model, a forward looking retirement likelihood measure is derived, which describes the probability that an individual retires within the next few years. In a second step, the retirement likelihood measure is used to predict transitions into retirement which took place between the first and second wave of the English Longitudinal Study of Ageing (ELSA). It turns out that the retirement likelihood measure has strong predictive power for actual retirement timing decisions. More precisely, for various groups of individuals the correlation between model predictions and actual retirement outcomes reaches 94% and the root mean square error can be as low as 5%.