We analyse the effects of wage uncertainty on the labour supply of self-employed workers, using PSID data on self-employed American males. The standard deviation of past wages, as a measure of wage uncertainty, is the key determinant of male self-employed labour supply, with a significant positive effect. In contrast there is no effect from the (instrumented) wage or other explanatory variables. Our findings are consistent with the self-employed 'self-insuring' in response to greater uncertainty by working longer hours, and they can also help explain why self-employed Americans work longer average hours for lower average wages than their employee counterparts.The self-employed account for around 10% of the workforce in most developed countries, employ a similar number of additional workers, and run most of these nations' firms (Parker, 2004). Despite this, relatively little is known about the labour supply of self-employed workers -partly, no doubt, because of data limitations pertaining to self-employed individuals. In contrast, there is an extensive and growing literature on the labour supply of employees (Blundell and MaCurdy, 1999). If for no other reason, the burgeoning policy interest centring on the promotion of entrepreneurship and labour market flexibility (with which the selfemployed are often associated) makes it timely to explore more closely what factors bear on the labour input of this important group of workers.Although pronounced, the literature's neglect of self-employed labour supply has not been complete. Early work by Wales (1973) estimated a simple static model of work hours using a sample of American self-employed proprietors. More recently, researchers have explored the labour supply behaviour of specific selfemployed occupations, including dentists (Boulier, 1979;Scheffler and Rossiter, 1983), physicians (Thornton, 1998), and taxi-drivers (Camerer et al., 1997. We depart from these studies by focusing on a factor that we argue is a particularly salient one for the self-employed: income risk. There is now clear evidence that self-employed workers have more variable and unequal incomes than employees do (Carrington et al., 1996;Parker, 1997). A separate body of evidence also shows that self-employed workers face a greater hazard of involuntary exit from their occupation (Evans and Leighton, 1989;Phillips and Kirchhoff, 1989;Taylor, 1999). That is a distinct form of risk that we do not focus on in this paper; nor shall we explore the decision to participate in self-employment instead of paid employment. Instead, we focus specifically on uncertainty of returns within selfemployment, and ask how this affects the labour supply of self-employed workers.