Tuesday, January 3, 2012

What's wrong with FRED

(Apart from revising the way they organize their data, I mean.)

FRED is fastidious about data integrity. Too fastidious. I think it's a problem.

At FRED, in the blue box at the bottom, under Monetary Aggregates we find

We often receive requests for monetary aggregates data that cover time periods prior to January 1959, the beginning date for the Board of Governors currently published monetary aggregates. Various data are available for these years, but not all data are consistent with the current definitions of the Board's monetary aggregates. This section discusses available data. The problem is not one of finding data; the problem is one of constructing monetary aggregates that are consistent with currently published definitions.

The problem is not to find the data. The problem is that the stuff they find doesn't match current definitions.

I wonder... If the economy changes, then perhaps it is appropriate that the definitions should change as well. If so, then the older definitions are not incorrect. They are correct for an earlier time. If this is true, then the older numbers are not "wrong" because they differ from newer numbers; they are merely "different". And hopefully, they are more relevant to the older period than current definitions would be.

Krugman ran into mismatched data definitions recently:

Source: Paul Krugman

I run into mismatched data definitions all the time.

Look at PK's graph. Suppose the graph showed only the "new series" numbers. It would be easy to think that the trend was always upward, and that it had never been higher than in the past few years.

Those would be very flawed assumptions.

It is far more important to know about the ups and downs than it is to have lines that match up perfectly. This is why FRED's fastidiousness is a problem. The fastidiousness is a cover-up of history. And I say that with love, because I do love FRED.

At FRED I click FRED Economic Data, then Monetary Aggregates, then M1 and Components to get a list of data series. Sorted by start date, the list offers two series beginning in 1947. Everything else begins in 1959 or after.

At ALFRED (Archival FRED) we read:

In general, economic data for past observation periods are revised as more accurate estimates become available. As a result, previous vintages of data can be superseded and may no longer be available...

So ALFRED offers older stuff.

But not the stuff I'm looking for. I click Monetary Aggregates: M1 and Components and sort the list of data series by start date. The oldest numbers that come up are for 1947. So it seems ALFRED offers not older data, but just older versions of current datasets.

One of the most pivotal books in all of economics was Friedman and Schwartz's A Monetary History of the United States, 1867-1960, a "monumental" work. Would you throw it away because its data definitions are not current?


But FRED won't use it. At the blue-box monetary-aggregates link, FRED includes that monumental history as one of five "common sources of historical monetary data". They even offer downloads for M1 back to 1929. But they have not integrated the data into their normal data series.

At FRED, they not only know about these sources; they recommend them. But they won't use those sources themselves. I think it's a mistake. They should set it up in separate data series, just like they do with everything else. And add a footnote about the mismatch.

I use FRED graphs a lot. I think one of the benefits of using FRED graphs is that it's not my graph. It's from a known, trustworthy source. I use FRED graphs because I think it increases the confidence people have in the information I present.

But I'm left on my own with the mismatched data. At the start, I thought maybe I was doing something wrong, to get the mismatches. Nope: Even Krugman gets mismatches. And FRED'll tell you why that is. But they won't touch the old data, themselves.

I think they should.


Jazzbumpa said...

I followed the link back to Krugman's post. He's saying a lot of things there you should really like.

He also notes that the up-slope of '29-'33 is due to declining GDP and not increasing debt.

The up-slope leading into 2006 cannot be so simply characterized.


The Arthurian said...

"I followed the link back to Krugman's post. He's saying a lot of things there you should really like."

I did like the post. I can't wait to see what he comes up with in his researches into debt.

"He also notes that the up-slope of '29-'33 is due to declining GDP and not increasing debt."

I know!

"The up-slope leading into 2006 cannot be so simply characterized."

The up-slope leading into 2006 is comparable to the up-slope leading to 1929. Not to the up-slope after 1929.

Jazzbumpa said...

I missed the linked post first time around.

You're right deflation is brutal.