As I've explained before, to reduce random fluctuations in time series, you can do three things.
First, you can extreme-adjust the time series. I won't explain how this works, because it would take a couple of paragraphs, but essentially, the algorithm reduces large fluctuations up or down to nearer some "mean" by adjusting the error-term (or random fluctuation).
Second, you can add two (or more) time series together. If they are statistically independent (and they can be from the same sample population, as long as the sample is different), then the random fluctuations are reduced --- if one goes up, the other might well be going down. Again, I won't explain the maths behind it.
Third, you can fit some kind of moving average to the data. This is called "smoothing".
The line to watch is the chart below is the thick green line, which is an average of the extreme-adjusted US PMI and the extreme-adjusted US ISM surveys. However, it is not smoothed. It continues to decline, and only the Covid Crash has produced a lower average in the last 11 years. It's the second month in a row which has been below the 50% "recession line", where more than half the correspondents to the survey report actual declines in sales, employment, new orders, etc.
Economies respond to changes in interest rates with a lag. The recent increases in the Fed funds rate have not yet had their full impact. Moreover, it is likely that the Fed will go on raising interest rates, even if more slowly. Expect the slide to continue.
No comments:
Post a Comment