Saturday, March 22, 2014

It works about as well as everything else the Fed does

I think the new FRED goes astray as soon as you get an error while entering the formula. Unfortunately, if you want to type a-b you have to type the "-b" really fast. Otherwise, in the fraction of a second between typing the "-" and the "b" FRED will give you an error. That's not something new; the old FRED was a pain in the ass that way, too. But with the new FRED, after you get the error message everything falls apart.

I managed to "add data series" for total debt, and the federal portion, and the state-and-local government portion, all without error. The graph all the while showed the first series correctly.

Then by typing "-b" really fast, taking time to position my fingers, and typing "-c" really fast, I got the formula to be "a-b-c" without getting an error message. And the graph actually showed something that matched the formula.

I was trying to duplicate the U.S. debt portion of Steve Keen's "Change in Private Debt" graph that I showed yesterday -- change in private debt as a percent of GDP. So next I added GDP as series d. (FRED identifies the series by letter; the first one you select is "a", the next is "b" and like that.)

But now I had to change the formula to read "(a-b-c)/d". Here's where I made my big mistake. I used the backspace to delete the formula, and before I could type the "(" FRED gave me an error message. It was all downhill from there.

I ignored the error message and entered the correct formula. Amazingly, the graph looked okay. But Keen's graph shows change in debt relative to GDP. So I went back to the a, b, and c data (my three debt measures) and changed the units from "Billions of dollars" to "Change, Billions of Dollars". I just changed the three debt series, not the GDP series.

But when I looked back at the graph, it had figured "Change, Billions of dollars" for all four series, including GDP. The graph was all spikey and obviously not like Keen's graph.

From there, things only got worse. I removed the "-d" from the formula and clicked "apply". The graph changed correctly. Then I added "-d" to the formula and clicked "apply" again. But the graph didn't change. Even though the formula shows "(a-b-c)/d", the graph and the graph borders show only "(a-b-c)".

What else did I do? Oh yeah, I clicked "max" up at the top, above the graph, to set the date range. And FRED set the date range, all right. It set the start date equal to the end date. But the graph didn't change!

But, hey: That was on the 20th. Maybe it's all fixed by now.


jim said...

Hi Art,

I also looked at the private debt to GDP relation at FRED and got tangled up in errors similar to what you describe. I did eventually get some answers.

The thing is that private debt is much larger than GDP. Since private debt is more than twice as large as GDP if private debt is growing at 1% and GDP is growing at 2% then private debt will increase in dollars faster than GDP will increase in dollars.

Lately, GDP has a rate of growth that is greater than the rate of growth of private debt, but the dollar increase in private debt is greater than the dollar increase in GDP.

I notice you exclude State and local govt debt from the private sector. I would include it as part of private debt.


The Arthurian said...

Hey Jim. If FRED had been cooperative, I would have included or excluded S&L government debt as needed to duplicate Keen's graph.

I understand a lot of people would do it your way -- I think because Federal is the money provider and stands apart from all the rest, which I surely agree with.

But a lot of people would lump S&L in with Federal and call it "government" and think of it as "the enemy". I don't, but I do like to look at the numbers their way, too. Covering all bases.

Glad I'm not the only one having trouble with the new FRED.