This paper investigates the possibility of Granger causality between the saving ratio (the proportion of gross domestic saving in GDP) and the growth rate (annual percentage change of real per capita GDP) in eighty-four countries of the world from 1961 to 2000. For technical reason, on the basis of their per capita GDP in 1995, the countries have been classified as high-income (at least 10000 $US), medium-income (between 1000 and 10000 $US) and low- income (less than 1000 $US) countries, and the three panels of twenty-six, thirty and twenty- eight countries, respectively, are considered separately. Granger causality is tested for with a new panel-data approach based on SUR systems and Wald tests with country specific bootstrap critical values. The results indicate two-way Granger causality between the saving ratio and the growth rate in Austria, one-way causality from saving to growth in Ireland, Trinidad & Tobago and the Central African Republic, and one-way causality from growth to saving in Finland, France, Japan, Sweden, Switzerland and Niger. There is also some support to causality from saving to growth in Mauritania and from growth to saving in Saudi Arabia, but in all other cases there is no empirical evidence of Granger causality in either direction.