What is Coincident Index？

[hardmann]

Dear Tom:

In Section 6.1 of the RATS Handbook for State Space Models, the Coincident Indicator is the cumulative value of the estimated factors. There is a coincident index on published by the Federal Reserve Bank of Philadelphia (USPHCI on FRED). In the Python stastmodel example, these code has gotten the same filtered factor. Furthermore, it reconstructs the mean of the (differenced) factor. Both coincident indexes is different from that of Winrats. Should the Coincident Index obtained by Winrats be similar to its four indicators?

This confuses me, what exactly is Coincident Index or Indicator?

https://www.statsmodels.org/dev/example ... dent-Index

Best Regard
Hardmann

[TomDoan]

"Coincident" (and "Leading" and "Lagging") indicators are observable series which are thought to provide information about the (in order) current, future and past states of the economy. In the U.S., the Conference Board (a non-profit NGO) has produced its "Index of Leading Economic Indicators" for many years which seeks to produce a single number which summarizes the information in the various data that are seen as Leading Indicators. The idea behind Stock and Watson and others is to produce a mechanical procedure which can produce something like that for Leading and for Coincident Indicators. (I'm not sure to what extent the Conference Board numbers are mechanical vs judgmental). There is no single way to combine the information from multiple series into a single index---it depends upon all sorts of assumptions and, of course, it depends upon what set of series that you choose. Note that, as an index, its actual values don't really mean anything---it's only the changes from period to period that matter.

For obvious reasons, there isn't much study of the "lagging" indicators---it's just that it's known that certain series like unemployment tend to rise several months after a business cycle peaks and fall several months after the trough, thus the phrase "unemployment is a lagging indicator".

[hardmann]

Dear Tom:

Thanks. I have modified the code of ss_6_1.rpf, Indicator Models, RATS Handbook for State-Space Models(2ed). It read directly the data from online FRED and remove the mean term. However, the results with and without Mu=beta is significently different.

The result without Mu=beta seems to be similar to the its indicator.

I would like to inquire about the reasons for the results.

Best regard
Hardmann

[TomDoan]

One is (in effect) linearly detrending the indicators; the other is not. If you are trying to gauge where you are in the business cycle, it would seem the first of the two would make more sense. And it's the changes that matter, not the absolute level.