본 연구는 연령효과, 코호트효과, 시간효과에 초점을 맞추어 최근 25년 동안 우리나라 가계저축률을 분석하였다. 이를 위하여 가계동향조사의 미시자료를 이용하였고 기본방법론으로 코호트분석을 적용하였다. 주요 결과는 다음과 같다. 첫째 소득의 연령별 프로파일은 낙타등모양이었고, 저축률의 연령별 프로파일은 M자형이었다. 저축률의 M자형 프로파일은 부양비, 특히 40대에 집중되었던 교육비 때문이었던 것으로 보인다. 둘째 1955~1975 년생의 소득은 이전 세대보다 많은 편이었지만 1976년생 이후 세대의 소득은 비슷하였다. 한편 1940~1964년생의 저축률은 높은 것으로 나타났지만 1965년생 이후 세대의 저축률은 높지 않았다. 셋째 시간효과를 보면, 1991~1994년의 저축률은 높았지만 1997년 이후의 저축률은 낮았다. 따라서 최근 25년간의 저축률 하락은 외환위기 이후 모든 코호트에서 나타난저축률 하락과 1965년생 이후 세대의 저축률 하락이라는 두 가지 현상이 작용한 결과로 판단된다. This research is to review basic facts on Korean saving rate in the recent twenty five years, and analyze the determining factors behind its trend. We use the micro-data(MDSS) from Household Income and Expenditure Survey from 1990 to 2014 and apply the cohort analysis of Deaton(1985). We tries to characterize the life-cycle pattern of household income, consumption, and saving on a cohort-by-cohort basis. To test the existence of cohort effect and time effect, we test the cohort effect and regress saving/consumption on several variables. Evidences say some results. First, age-profiles of income and consumption had similarly hump-shaped. and income and consumption reached their peak at an earlier path of their lifetime, compared with those of the seniors. Second, age-profiles of saving rates of all cohorts were M-shaped and shifted to the left during the sample period. The dependency ratio was an important factor in determining saving, but the M-shaped age-profiles of saving rates can be explained mainly by large education expenditures during 40s. Third, cohort effect and time effect were all important. Saving rates of cohorts born in 1940~1974 were higher than those of cohorts born before 1940, but saving rates of cohorts born after 1975 were not. And saving rates of all cohorts in 1991~1997 were higher than those of all cohorts after 1998. So we conclude that decreases in saving rates in the recent twenty five years are a mixed phenomenon of a universal downward change across all cohorts and downward changes of younger generations.