The association between academic self-concept and achievement is assumed to be reciprocal. Typically, the association is analyzed by variants of the classical cross-lagged panel model. Results with more recently developed methodological approaches, for example, the random intercept cross-lagged panel model, its continuous-time implementation, and the latent change score model, are not available so far. The present study aimed to test the association between reading self-concept and reading achievement with different models to evaluate whether positive cross-lagged effects can be demonstrated with the alternative models. Drawing on a sample of 2,014 elementary students from Grade 1 to Grade 4, results of alternative models yielded noticeable differences. Support for reciprocal effects between reading self-concept and reading achievement was found using the classical and the full-forward cross-lagged panel models. No such effects were found in the other models. Thus, the reciprocal effect model can be called into question for elementary schoolchildren. The results are discussed with regard to the theoretical assumptions and appropriate methodological approaches.