Friday, April 21, 2023

An inflation prediction

Graph #1: Price Increase from Previous Month

The last data shown is for March 2023. 

Inflation became newsworthy after the covid recession of 2020, and reached a month-on-month peak in June 2022. The following month, July, the number dropped to zero. According to the stats at least, prices didn't go up in July. And since July, the monthly increases have been pretty tame, similar to what we had for 10 years before 2020.

The news doesn't report the month-on-month increases that are shown on the graph above. They report the 12-month change: the price level for the month, relative to the price level 12 months before. For example, the "percent change from a year ago" inflation rate for the high point, June 2022, included all the price increases since June 2021.  This makes the inflation number much higher than it is on the first graph. The second graph shows the inflation data in this more newsworthy form:

Graph #2: Price Increase from a Year Ago

This graph shows the same source data, but it figures 12-month increases rather than 1-month increases. Notice that June 2022 is a high point on both graphs. But inflation is almost 9 percent on the second graph, and not much more than 1 percent on the first. One is newsworthy, and one is not.

Next, a prediction.

On the first graph, from June to July 2022 inflation drops all the way to zero. But on the second graph inflation is still high in July. That's because the second graph figures 12 months, not just one.

For the June 2022 number, they take June 2021 as a starting point and then count all the increases from July 2021 to June 2022.

Then for the next month, the July 2021 number is used as a new starting point. This time it doesn't count as an increase. The July 2022 number drops the July 2021 increase, counts the other 11 increases used the month before, and adds in the July 2022 increase. This is like why people don't like math I guess.

Anyway, the last data point shown on both graphs is for March 2023. March of 2022 is the starting point, and April 2022 is the first of a dozen monthly numbers used to figure the March 2023 number on the second graph. That's the process. It's kind of an incremental shift, one month at a time.

The high point on the first graph was the June 2022 increase. For June 2023, that increase no longer counts when they figure "from year ago" inflation. Instead, it serves as the starting point for the calculation. 

The last time the June 2022 increase (the big one) figures into the inflation rate is for May 2023. As of June 2023, all of the increasing inflation, from 2020 to June 2022, will be out of the calculation. The first monthly increase that comes into the June 2023 "from year ago" number is the July 2022 number -- the zero inflation shown on graph #1. 

And the inflation we got in the months after July 2022 was about as low as the inflation we had for the first ten years shown on graph #1. So my prediction is that the inflation for July 2023 and after will be "back to normal" or close to it. It will hardly be newsworthy anymore. 

That doesn't mean prices will come back down, of course. But at least we won't have to listen to them talk about it on the news!

1 comment:

The Arthurian said...

In my old "Growth in the 1970s" post, Graph #2 shows in gray the annual growth rate of Real GDP, and in blue a nine-year moving average of the same data. The paragraph above Graph #2 offers this observation:

"To smooth the jiggy growth data (gray) on the graph below, I show a 9-year moving average in blue. It's a centered moving average: the last point shown for example, 2014, is the average for the years 2010 thru 2018. (And by the way, the two rightmost data points of the blue line show sharp uptrend because 2008 and 2009 drop out of the calculation.)"

In the moving average, the set of 9 values changes as you move from year to year. In the second inflation graph shown above, the set of 12 values changes as you move from month to month. It is the same technique, applied to different data.

When you add the newest data into the average of 9 (or the accumulation of 12) values, you drop out the oldest value. Therefore the resulting value can change substantially if the *earliest* value changes substantially.