Previous studies on cash flow prediction mainly focus onthe information provided in financial statements. Studies suchas Barth et al. (2001) discussed the power of accrual terms andhow disaggregating accruals into its major components enhancescash flow prediction. This study applies a number of models tocash flow data for South African firms. Three main cash flowmodels are investigated in this study, mainly, linear regression,which has been widely adopted in similar and relevant studies,moving average model, which is mostly applied in time-seriesanalysis, and vector autoregressive model, that has been widelyapplied in macroeconomics and finance. The latter two types ofmodels are applied for the first time to cash flow prediction.The results reported in this study contrast that reported else-where. Disaggregating cash flows into its major componentsdoes not appear to enhance cash flow prediction for the aver-age South African firms compared to results reported by Barthet al. (2001) for USA and Farshadfar and Monem (2013) forAustralian firms. The results suggest that implications of stud-ies conducted elsewhere cannot be extrapolated across othercountries without taking into account country context and dif-ferences. The reported results in this study show that modelsincorporating income statement information seem to result inworse out-of-sample prediction. However, when two lags ofindependent variables are included in prediction models, resultsdiffer in several ways which suggests that variable inclusionin cash flow modeling in South Africa should be treated withcaution. In addition, prediction accuracy, as measured by R2for South African firms are high compared to extant studieselsewhere. Studies on cash flow prediction pool all firms’ datatogether and ignore heterogeneity that exists among firms andindustry. There is therefore the need for studies on cash flowprediction that focus on industry and individual firms.