The goal of this article is to estimate the New Keynesian Phillips Curve for Brazilian economy. Due to some specifications problems in regressions estimated by IV method, the GMM-HAC methodology was used in order to address them. We noted the robustness of the results performing a detailed investigation that was based not only on different proxies for the dependent variable and the regressors but also on samples with distinct temporal dimensions. Among the main achievements of this study, the following are more remarkable. Firstly, the inflationary inertia and expectation of inflation are important variables for the dynamic of inflation although the relevance of expectation seems to be more relevant in a more recent period from 2002 onwards. When one estimates the NKPC using data from 1995, the effect of expectation get down and becomes close to the inertia. In second place, the majority of regressions did not reject the hypothesis derived from structural form that the sum of coefficients of lagged inflation and expectation of inflation is equal to unity. Thirdly, the effect of unemployment on inflation seems to present in short term. In the long run, the estimations do not able to detect any impact of unemployment on the dynamic of inflation. Finally, the structural break marks the relationship between the exchange rate and inflation. The regressions estimated with data from 2002, the effect of exchange rate shock is negative. But, when one uses data from 1995, this shock has a positive impact on inflation.