Growth at Risk (GaR) methodology developed by Adrian et al. (2019) has been of special interest by policymakers since it provides a measure of the relationship among macrofinancial variables. GaR requires estimating a set of predictive quantile regressions (QR) where future economic activity (GDP growth) is linked to current financial conditions, measured through a set of alternative market or bank related indicators. As GaR methodology increased in popularity among policymakers, recent literature has stressed the need of model evaluation of GaR results. For instance, Reichlin et al. (2020) evaluate the out-of-sample performance of a GaR model and find little evidence of predictability beyond what can be achieved using timely indicators of the real economy. Moreover, Brownlees and Souza (2020) use a Garch-type model to forecast the distribution of future economic growth, and compare their forecasting power against GaR model, finding that a Garch-type model outperforms a GaR model. Taking into consideration the need for a proper evaluation of GaR results, our work implements several model evaluation techniques to increase the accuracy of a Growth at Risk model for the Peruvian Economy. Considering a broad sample of parametric and nonparametric distributions to fit the GaR results, we use log scoring, probability integral transform and entropy tests as model evaluation tools to select the best density forecast that fits Peruvian data. Once we obtain a more reliable GaR results, we use this model to implement a counterfactual analysis to evaluate the impact of Reactiva Peru, a government program that support the credit to firms during the lockdown due the Covid-19 crisis. Our results show that Reactiva Peru had a sizable impact in macroeconomic and financial stability, since it avoided a much deeper decrease in economy activity during the covid-19 crisis.