OBJECTIVE Data from the National Household Survey 2004 was analyzed to compare differences in prevalence among moderate or severe food insecurity. Also, it was compared food security or mild food insecurity households in relation to the assets and other socioeconomic and demographic conditions of the household. METHODS Private permanent households, with per capita monthly income of up to one minimum wage and with the Brazilian Food Insecurity Scale answered by a household resident were studied (n=51,357). Association of variables with the dependent variable (food security) was verified using χ² test, with 5% significance level. Crude prevalence ratio, respective 95% confidence interval and adjusted analyses were carried out using Poisson multiple regression Stata 8.0. It considers the weights of the complex sampling design of the survey. RESULTS The per capita monthly household income was the variable with strongest association to food security. Both in urban and rural areas, there were higher risk of moderate or severe food insecurity prevalence ratio when the head of the household was a female, black color, presence of six or more members in the household, metropolitan area and with absence of some specific assets (stove, water filter, refrigerator, freezer, washing machine and cellular phone). In a model that, among assets, included just the refrigerator, it was observed the highest prevalence ratio for household income of up to ¼ of a minimum wage and after this, the absence of refrigerator among households headed by white and black males and white or black female. Although female and black headed households have greater food restriction, internal differences among these groups were higher for households headed by white males and lower for those headed by black females. CONCLUSIONS At national level and households with monthly income of up to one minimum age, poor socioeconomic conditions are associated to household food insecurity. This situation is worse among those headed by women and black people. Among poor people, the absence of assets identifies the most vulnerable population to food insecurity and may be used as complementary indicator, mainly in local studies with poor technical resources for data collection and more sophisticated analyzes.