This　paper　proposes　a　new　approach　to　measuring　Malmquist　productivity　index　(MPI)　by　using　both　optimistic　and　pessimistic　data　envelopment　analyses　(DEA)　simultaneously,　which　we　refer　to　as　the　double　frontiers　data　envelopment　analysis　(DFDEA).　The　MPIs　measured　from　the　two　different　DEA　points　of　view　are　geometrically　averaged　to　generate　an　integrated　MPI,　which　we　refer　to　as　the　DFDEA-based　MPI.　This　DFDEA-based　MPI　reflects　the　productivity　changes　of　decision　making　units　(DMUs)　over　time　more　truly　and　more　comprehensively　than　the　traditional　optimistic　DEA-based　MPI.　The　proposed　new　approach　is　tested　with　a　numerical　example　and　applied　to　the　productivity　analysis　of　the　industrial　economy　of　China.　©　2011　Elsevier　Ltd.