Next to smvival, food, clothing and shelter, education is children's greatest need after the agernof six or seven. Without education, chi ldren can do little when they become adults, to improverntheir living conditions and to participate in development activities.rnPrimary education eqUIps people with the ski lls of literacy and numaracy and thereforerncontributes to poverty reduction by increasing labour productivity. Besides, primary educationrnserves as a bridge towards slower population growth and plays a critical role in demographicrntransition. Past and current evidences in Ethiopia however, reveal that many primary schoolrnpupils are leaving the system at early stages. For instance out of 1,107,751 pupils enrolled inrngrade one in 1994/95, only 39.1 % of them completed grade four in 1998/99. This wastage isrnestimated to be more than Eth. Birr 186,850,423 .00. The 1998 Welfare Monitoring Surveyrn(WMS) also shows that 19.1 percent of the primary school pupils registered in the academicrnyear 1996/97 have dropped out. This situation therefore, calls for a study to understandrncovariates of primary school pupils attrition and capture policy makers' and planners' attentionrnto reduce this prevailing dropout problem.rnThi s study uses the 1998 WMS data collected by the Ethiopian Central Statistical Authority.rnThe data comprises a sample of 22,787 pupils regi stered in primary schools during thernacademic year 1996/97.rnThe study applies the multivariate logistic regresslOn model to identify some mall1rndemographic and socio-economic covariates of attrition. It also investigates differentials inrnxirndropping out of primary schools in Ethiopia, by age, sex, pupil's relationship to the householdrnhead, household size, pupil's marital status, di stance to the nearest primary school, place ofrnres idence, household head' s educational attainment, pupil's working status, and householdrnincome. Besides, it tries to identify the covariates which most affect dropping out andrnestablishes the magnitude and direction of the effects. Logistic regression analysis isrnappropriate for the prediction of a binary dependent variable as used in the present study.rnPreliminary analyses are also made using descriptive statistics.rnSix hypotheses regarding primary school pupils attrition were tested in the study. As a result,rnage-grade mismatch of the pupil, sex of the pupil, household size, urban rural place ofrnresidence, household income, and pupil 's working status are found to be significant 111rnpredicting pnmary school pupils attrition 111 Ethiopia. In this study, household Size andrndropping out of primary schools were found to have inverse relationship, contrary to therngeneral literature.rnAlthough not stated as hypotheses, other covariates included in the study such as, pupil'srnrelation to the household head, pupil 's marital status, distance from the household to thernnearest primary school, living in some of the regions, and household head's educationalrnattainment were also found significant. Moreover, marital status of the household head, sex ofrnthe household head and working status of the household head were found to be statisticallyrninsignificant. Pupil 's working status, pupil's marital status and Afar region in this order werernfound to be the tlu'ee covariates with the strongest effect on primary school pupils attrition