Modelinglog-mortalityratesonO-UtypeprocessesandforecastinglifeexpectanciesareexploredusingU.S.data.IntheclassicLee-Cartermodelofmortality,thetimetrendandtheage-specificpatternofmortalityoveragegrouparelinear,thisisnotthefeatureofmortalitmodel.Toavoidthisdisadvantage,O-Utypeprocesseswillbeusedtomodelthelog-mortalityinthispaper.Infact,thismodelisanAR(1)process,butwithanonlineartimedriftterm.BasedonthemortalitydataofAmericafromHumanMortalitydatabase(HMD),mortalityprojectionconsistentlyindicatesapreferenceformortalitywithO-UtypeprocessesoverthosewiththeclassicalLee-Cartermodel.Bymeansofthismodel,thelowboundsofmortalityratesateveryagearegiven.Therefore,lengtheningofmaximumlifeexpectanciesspanisestimatedinthispaper.
