The accuracy of inflation forecasting has been a challenge in most developing countries which do not have high frequency data.This study evaluates the performance of leading bivariate and trivariate models of inflation in forecasting inflation for Zimbabwe. Root Mean Square Forecast Errors (RMSFE) of the benchmark autoregressive model are compared with those of bivariate and trivariate vector autoregressions. The pseudo-out-of-sample forecasting results show that (1) the trivariate model using lags of inflation, crop production index and the parallel market exchange rate forecast inflation better (2) the term structure of interest rates also performs well and (3) monetary aggregates fare badly. Further, the study analyses the institutional set-up of the Reserve Bank of Zimbabwe (RBZ) to determine its independence. Results show that political interests have been driving the monetary policy and this has undermined the credibility of the central bank. Even if independence is not the complete answer, it is at least suggestive that a more autonomous central bank would be effective even though it is neither necessary nor sufficient by itself for achieving and maintaining low inflation.