This paper discusses how the forecast accuracy of a Bayesian vector autoregression(BVAR) is affected by introducing the zero lower bound on the federal funds rate. As abenchmark I adopt a common BVAR specification, including 18 variables, estimatedshrinkage, and no nonlinearity. Then I entertain alternative specifications of the zerolower bound: replace the federal funds rate by its shadow rate, consider a logarithmictransformation, feed in monetary policy shocks, or utilize a rejection sampler. The lattertwo are also coupled with interest rate expectations from future contracts. The comparisonis based on the accuracy of point and density forecasts of major U.S. macroeconomicseries during the period 2009:1 to 2014:4. The results show that the performance of thespecifications is greatly different, suggesting that this modeling choice is not innocuous.The introduction of the zero lower bound is not beneficial per se, but it depends on howit is done and which series is forecasted. With caution, I recommend the shadow ratespecification and the rejection sampler combined with interest rate expectations to dealwith the nonlinearity in the policy rate. Since the policy rate will remain low for sometime, these findings could prove useful for practical forecasters.