I am running an OLS model with a continuous asset index variable as the DV. My data is aggregated from three similar communities in close geographic proximity to one another. Despite this, I thought it important to use community as a controlling variable. As it turns out, community is significant at the 1% level (t-score of -4.52). Community is a nominal/categorical variable coded as 1,2,3 for 1 of 3 different communities.

My question is if this high degree of significance means I should be doing regressions on the communities individually rather than as an aggregation. Otherwise, is using community as a controlling variable essentially doing that?