Purpose – The purpose of this paper is to empirically investigate the effects of farm income variability, farm size, and other socio-demographic characteristics on the precautionary saving behavior of farm households and to estimate the influences of the identified factors on the amount of savings by self-employed farm households. Design/methodology/approach – Using 2003 Agricultural Resource Management Survey (ARMS) data and a Double-Hurdle procedure, the likelihood and the amount of savings by farm households are estimated. Findings – An important empirical finding of this study is that variability in income plays an important role in explaining precautionary savings of US farm households. Findings suggest that farm households facing higher income risk save more and accumulate more wealth. It is indicated that several farm, operator, household, and demographic attributes contribute to the precautionary savings of farm households. In particular, results show that educational attainment by operator and spouses have positive impact on the decision to save. In addition, results from this study show that farms that specialize in cash grain are likely to have precautionary savings. Practical implications – Farm households today are virtually indistinguishable from non-farm households in their levels of income and diversity of employment. As a result, government policies that influence general economic conditions have much more profound impacts on farm families. Federal support of farm income warrants continued scrutiny. This paper shows that greater income uncertainty increases savings and wealth of farm households. Therefore, farm policies that reduce income variability or uncertainty will have an impact on precautionary savings and wealth of farm households. Originality/value – Several studies have investigated savings of households; however, these studies are limited to entire US population, older Americans, or non-self-employed individuals in the USA. Little is known about the savings behavior of self-employed US farm households owing to a lack of household survey data and because of the complex relationship between the farm household and farm business in terms of resource allocation (both capital and labor).