Monday, 27 February 2017

CPI Inflation Model - USA CPI-U inflation forecasts for Feb & Mar 2017. Spike over 3% likely.

I have compiled my first inflation forecast report forecasting 'Headline' year-over-year all items CPI inflation for all urban consumers (CPI-U) in the USA. This includes forecasts of February and March 2017 forecasts. The actual inflation figures will be announced by the Bureau of Labor Statistics (BLS) on 
  • 15 March 2017 (CPI for Feb 2017) and
  • 14 April 2017 (CPI for Mar 2017) respectively.
The inflation model forecasts are now given at the bottom of this post. Most applications of the model give 3%+ for both months.

In addition to the forecasts based upon my proprietary model, the report shows correlations between different components of the CPI statistics and demonstrates clearly the significance of energy prices in determination of CPI figures for the past three years, with many charts included. All of the data used are year-over year (y-o-y) figures.

I am also preparing a second report, which will refine the model after the Feb 2017 CPI inflation announcement on 15 March and will also provide potential inflation forecast scenarios covering the whole of the remainder of 2017. This second report should be available in late March.

The importance of energy prices to the level of CPI is crucial at this time because most other components of the CPI (services, food, commodities / apparel / durable goods categories have shown low volatility over this time period. There is therefore an opportunity to use energy prices to forecast future CPI inflation in this environment. For example, services inflation is relatively constant at around +3% and has contributed very little to the changes in overall inflation in the past couple of years.

I believe that February 2017 CPI inflation figure announced on 15 March is going to be a crucial announcement that could have a considerable effect on markets if my forecast is correct. Although I have used data from the past three years in my model, there is still some uncertainty in the upside to the CPI energy component since year over year energy prices increases have been significant only in the January 2017 inflation report so far, despite the bottoming of the oil price in February of 2016.

[I have posted a rehearsal of a presentation on this subject on YouTube at: - please forgive the shaky hand held camera!]

Once the Feb 2017 inflation figures are announced further refining of the model will be possible and I intend to publish a second report later in March to be summarized on this blog.

The basis of the two reports is that CPI is critically linked to energy prices and energy prices are strongly correlated to changes in the oil contract price. The oil price action alone is giving a strong indication of the magnitude and direction of changes in the inflation rate. This has not always been true historically but in his period where the oil market has turned quite sharply in direction there is a strong relationship.

Here are a couple of charts from the first report:

Figure 1: Correlation of CPI All Items index with energy sector price changes. Note that Core CPI and Services prices have contributed very little to variations in the CPI in the past 3 years and the vast majority of CPI changes are due to energy prices:

Figure 2: The turnaround in oil prices in 2014-2017 showing year over year changes:


UPDATE 2017-03-02

Inflation forecasts from my first two models are as follows. These are obtained purely from year over year moves in the WTIC oil price based on the correlation between CPI and energy and oil prices over the past 3 years.

Model 1 gives forecasts for inflation up to 2 months in advance.

Model 1 gives forecasts for inflation up to 1 month in advance.

Model 1:
Feb 2017
(BLS release 15 Mar 2017)
Mar 2017
(BLS release 14 Apr 2017)
CPI year over year High estimate
CPI year over year  Low estimate

Model 2:
Feb 2017
(BLS release 15 Mar 2017)
Mar 2017
(BLS release 14 Apr 2017)
CPI year over year High estimate
CPI year over year Low estimate

Two more models are currently being tested. These further forecasts will be posted soon but the results are not too far different from Model 1. Then comes the task of modelling inflation scenarios to the end of 2017.

Thanks very much for reading!

Please consider sponsoring my website on Patreon at

or by using the PayPal button below:

No comments: