Two approaches based on regression models are proposed to estimate competition from census data. The "static" approach is based on censuses of population sizes among species at one point in time over many sites. The "dynamic" approach relies on a time series of species abundance data to examine whether per capita changes in one species are associated with the abundance of other species. We estimated competition interactions in a Negev rodent community consisting of 10 species using both approaches, basing on 8 years (16 half-year periods) of observations. The static approach revealed significant competitive interactions in four of 45 pairs of species, whereas the dynamic approach did so in the same four plus two more pairs. For each species pair, both approaches revealed significant negative interactions in only 1-4 of 16 seasons. The static approach provided nearly symmetric estimations of competition, whereas estimations of dynamic approach were asymmetric. Moreover, estimations of the two approaches did not coincide in time. Cases of negative interactions recorded by the static approach were more frequent at peak and increase phases of population density dynamics, whereas those recorded by the dynamic approach were more frequent at peak and decline phases. Results of field removal experiments with Mus musculus and Gerbillus dasyurus supported predictions of dynamic but not static approaches. We hypothesized that in harsh and fluctuating desert environments that disrupt equilibrium, the dynamic approach indicates true (exploitation) competition, whereas the static approach reflects negative interspecific spatial association (interference).