The purpose of this paper is to quantify the relationship between poverty and education using a multinomial logit measurement. Starting with a subjective measurement of households’ poverty in the region of Tlemcen, we then move to an application of a multinomial regression approach to non school factors as the main determinants of the relationship between poverty and education. Education plays a vital role in economic and social development, and due to its complexity and multidimensionality, should be comprehended within a general approach, in order, first to pinpoint the relevant factors for its efficiency, and second, to assess its outcome (individual and collective). It follows therefore, that higher levels of education and enlarged access will lead to productivity gains and income, hence reduced inequality and poverty. The current practice chosen for this analysis consists of two steps: in the first, we attempt to identify a subjective poverty measure. In the second step; beside using a theoretical analysis of the linkage between poverty and education on the basis of domestic data , we want to better understand the linkage between poverty and education through the use of a multinomial regression model drawn on a survey of 500 households in the region of Tlemcen . We consider that the main variables that may determine this linkage are non school factors such as the level of education of the head of the households, gender, education expenditure, and any additional courses for children. The outcome shows that for the first model (very poor relative to intermediate) the individual housing, the collective housing, the gender (male), the age of the head (20-25) and the level of instruction of the head of the household have lesser probability for the very poor to improve his well off level to a higher i-e intermediate situation. For the second model (poor relative to intermediate) only the individual housing, the university level of male, the level of instruction and expenditures for education have a lesser probability for the poor to improve his well off level to a higher i-e intermediate situation. The last model (rich to intermediate) shows that only the age category under 31 years for the head of the household, the primary and secondary level of instruction of the male head of household, the instruction level of female head of household have a negative impact on the rich level, i-e that the subjective probability of feeling rich is questioned through these variables leading to a transfer from a rich level to intermediate real level. Our approach can help Algerian policy makers to identify the actual missing variables that are important to the education sector, particularly if the state maintains his actual policy.