this paper evaluates to what extent the end-of-sample problem inherent to the hodrick-prescott (hp) filter may be ameliorated when estimating the output gap. for that purpose, the filter of st-amant and van norden (1997) is proposed using mexican data under both univariate and multivariate methods. as a reference, the filter of christiano and fitzgerald (2003) and a simple linear trend method are also used. on average, the results suggest that both the univariate method of st-amant and van norden and the linear trend method may significantly decrease the end-of-sample problem of the hp filter. these results are used to interpret the conduct of monetary policy in mexico during 2001-2002.