A lack of resources over a short period of time cannot be said to define a situation of poverty, which needs to be assessed over a period of some years. Observation errors induce changes, which mistakenly suggest that the phenomenon is highly volatile. Statistical reconciliation methods can partially correct the observation errors, but the results are sensitive to the assumptions used. Reducing the factors that appear to induce spurious effects reduces the number of exits from poverty by approximately two-thirds. The correlation between the different forms of poverty remains low: the non-coincidence of the different forms of poverty is not an artefact. Other partial scales than monetary poverty, poverty in terms of living conditions and 'subjective' poverty could round out the multidimensional analysis of poverty, poor health resources and social capital, for example. The populations displaying these different elements do overlap in part and the different poverty symptoms do not occur at the same time. Subjective poverty appears to be a leading indicator, since material hardship occurs quite early on the impoverishment path. Health problems seems more a cause than a consequence of poverty. Foster has developed a poverty line defined as the geometric mean of an absolute threshold and a relative threshold, with an elasticity to be defined between 0, corresponding to the strictly absolute approach, and 1, corresponding to the relative approach. On the late 1990s the french value is close to 1, but significantly below: the representations are more in line with the relative concepts than with the absolute approach, but only give rise to partial indexing since the norms only gradually and selectively confirm the lifestyle changes.