March 2013 National and Local Inflation Forecasts
Assumptions & Methodology
On March13th, 2013 the Chief Economist of King County presented inflation forecasts to the Forecast Council. They were adopted by the Forecast Council and will be used to construct a balanced budget for 2014. The purpose of this document is to provide transparency regarding the assumptions and methodology to users of the forecasts, both inside and outside of County government.
Mandated Assumption by the Forecast Council
The forecasts are constructed with a 65% level of confidence. Since inflation is a cost driver, this means that forecasts are set so there is a 65% chance that actual inflation will be less than forecasted, and only a 35% chance that actual will exceed forecasted. This is more conservative than an “expected value” forecast, which would be at the 50% confidence level.
The County budget requires forecasts of several price indices, each targeted for specific budget items and uses. They are:
- Consumer Price Index for all urban consumers (CPI-U) for the US. This is the most closely watched measure of “background” inflation in the country. We forecast the12-month average rate of inflation.
- Consumer Price Index for office workers and wage earners (CPI-W) for the US. This tracks the cost of a basket of goods matched to the consumption pattern of this group. We forecast the September-to-September rate of change as that is the basis for cost of living adjustments in some County labor contracts currently in force.
- Consumer Price Index for all urban consumers in Seattle (CPI-U-Sea). Actually, this index covers Seattle, Tacoma, and Bremerton. We forecast the 12-month average rate of inflation.
- Transportation component of CPI-U for all of the US (CPI-Transportation). We forecast the12-month average rate of transportation inflation.
- Pharmaceuticals component of the Producers Price Index for all of the US. We forecast the12-month average inflation rate of prescription drug and other medicines.
CPI-U-US Inflation Forecasts
We combine six forecasts from four professional services. The combination gives us an expert consensus view as well as revealing the range of disagreement between the experts. The sources are:
- Global Insight: baseline, pessimistic and optimistic forecasts.
- Puget Sound Economic Forecaster (PSEF)
- Washington State Economic & Revenue Forecast Council (ERFC)
- Blue Chip Consensus Forecasts
Chart 1 shows the range of forecasts. The average of the six sets of forecasts represents an “expected value” or 50% confidence forecast. The Forecast Council requires a more conservative forecast, one set at a 65% confidence, meaning there is a 65% probability that actual inflation will be below forecasted. For each out-year we have n = 6 forecasts that form a distribution, which is assumed to be a Student’s t-distribution with n-1 = 5 degrees of freedom. The 0.65 percentile of the distribution produces the 65% confidence forecast.
CPI-W-US Inflation Forecasts
This forecast is used to estimate cost-of-living-adjustments (COLA) that are embedded in some labor contracts currently in force with King County. The COLA is based on September-to-September rate of change of the seasonally unadjusted CPI-W index. That is what we use for historic data and that is what is forecasted.
First, we regress the CPI-W index on the CPI-U 12 month average. The historical correlation between the two is 0.89. Then we use the 65% confidence level forecast of the CPI-U through this regression equation to generate 65% confidence forecasts of the September-to-September CPI-W. The results are in Table 2.
CPI-U-Seattle Inflation Forecasts
For the Seattle CPI we use three sets of forecasts to obtain a 65% confidence level forecast:
- Forecast of the Seattle CPI by the ERFC
- Forecast of the Seattle CPI by the PSEF
- Regression of the Seattle CPI on the CPI-U for all of the US. Forecasts of the Seattle CPI are then generated using the mean of the six forecasts of the CPI-U for all of the US.
Chart 2 shows the three sets of forecasts and the 65% confidence level forecast calculated from their distribution. Table 3 displays the forecasts.
Transportation Component of the US CPI-U
We regress transportation inflation on the price of gasoline and the CPI-U. Three sets of forecasts are generated by using the Global Insight’s baseline, optimistic, and pessimistic forecasts of the price of gasoline along with the Blue Chip Consensus forecasts of the CPI-U. The forecasts are in Table 4.
Pharmaceutical Component of the PPI
We regress the Pharmaceutical Producers Price Index on personal consumption expenditures on drugs and pharmacies. Three sets of forecasts are generated from the regression model: one each from Global Insight’s baseline, pessimistic, and optimistic forecast of personal consumption expenditures on drugs and pharmacies. Finally, we also generate forecasts using an exponential smoothing method called Holt-Winters. This method relies solely on the past momentum of pharmaceutical prices to forecast its future path. These forecasts and the resulting 65% forecast are displayed in Chart 3. Table 5 has the forecasts.
 Johnson, Montgomery and Gardiner (1990), Forecasting and Time Series Analysis, McGraw-Hill, 2nd Ed.