A multinomial approach to predicting the probability of inflation/deflation

Due to the Great Recession and “global financial meltdown” central banks reduced key interest rates dramatically while also significantly increasing central bank credit to private banks and markets. Meanwhile, fiscal policy in many governments injected trillions of dollars through stimulus packages. Some analysts, because of the above mentioned factors, worry about inflation. Alternatively, some analysts have argued that the private economy has changed and this has created the possibility of deflation. First, as a result of lower growth rates (GDP is less than its potential level for many developed countries), and therefore a large output gap, downward price pressures are expected. Second, higher unemployment rates along with huge jobs losses during the Great Recession have reduced wage pressures. Third, higher expected taxes to cover large budget deficits in many countries with serious debt-management issues have lead to expectations of restricted fiscal policy and to uncertainty about the pace of future growth and the risk of deflation. In sum, at present, a high level of uncertainty exists about future path of prices toward inflation, deflation or price stability.

This paper provides a multinomial approach that estimates the 12-months ahead probability of three distinct scenarios for prices: inflation, deflation or price stability. We use CPI as proxy for general price level. A dummy-variables (Y=0, 1, 2) is created: (1) Y equals to one if CPI (yoy) is greater than 2.5% (assuming that higher than 2.5% growth rate of CPI will alter the inflation outlook), (2) Y equals to two if CPI (yoy) is less than 1.5%( again assuming that a less than 1.5% growth rate will bring to light a deflation-averted strategy) and (3) Y equals to zero if CPI between 1.5% and 2.5%. The model estimates the probability of being in inflation or deflation relative to the probability of being in the price stability.

The traditional way of forecasting inflation is to predict a single level or/and growth rate of CPI; however, this approach suffers two problems. First, it is not useful for the option/risk facing decision-makers. Second, point estimates of inflation convey a sense of overconfidence. Our method is completely different and more practical for decision-makers who must hedge their benefits for alternative states of the world. The approach is also more useful for policymakers, investors and consumers if we attach a probability with the each more-likely scenarios of future price levels: inflation, deflation or price stability.