Predicting inflation

Here is an example of trying to predict inflation using the information transfer model (with the quantity theory of money). We posit a region of points near the last data point on the information transfer model surface. If we assume NGDP stays on trend (linear extrapolation of the log trend), that gives a reasonable estimate of the maximum radius of the region. Even if the monetary base stayed constant NGDP would only grow at best on trend. You could get more complicated by using an ellipse with a different radius for the monetary bound if you like. We'll just use a circle here. In this graph the points (blue) are shown near the last data point on the curve (also blue) -- they represent possible position states on the surface after some given amount of time (I extrapolated to 2020):

If we use these data points to show the price level, we obtain a prediction (assuming linear drift from the original point) which is essentially the vertical axis in the 3D surface plotted versus time:

We'll see if this comes true over the next seven years ...

Interestingly, the model says that we are near the top of a ridge so the surface is relatively flat, meaning little increase or decrease in the price level, i.e. a Japanese-style lost decade. In this model, it has little to do with a "liquidity trap", unless you define "liquidity trap" as being near the ridge of the surface. The reason is that we are near the ridge is that we are near where $Q^s \sim \exp \sqrt{\log Q^d}$: the monetary base are belong to us is too large relative to NGDP. Either NGDP growth or reductions in the growth rate monetary base relative to NGDP will move use away from the ridge.

Does this count as a policy prescription?