Monday, February 1, 2016

Monthly GDP data from Macro Advisers, or, Writing in the Dark

David Beckworth had me flying off the handle with his graph of "Total Dollar Spending":

Graph #1 via David Beckworth

1. Money times Velocity equals GDP, not total dollar spending.
2. GDP is final spending, not total spending
3. You don't start with money and velocity, and figure GDP. You start with money and GDP, and figure velocity.
4. Oh! His data is monthly. That's interesting.

In the comments, Beckworth says the data comes from Macro Advisers. Found it! And guess what? They say All MA MGDP data and historical files are available to the public at no charge.

Nice. That's what I live for. But I must have some setting wrong in my browser. I clicked the

MA’s Monthly GDP Historical Data (Through November 2015)

link at the MA site, and I got a pageful of garbage. Firefox opened their XLSX file (without a complaint) as a web page, not a spreadsheet. It was unreadable. What to do?

I just saved the page. It came out on my desktop as an Excel file, and opened easily then (in Excel) when I double-clicked it.

Well that was fun.

Anyway, the file contains monthly GDP data for 1992 through 2015. That's exciting because GDP data is always either annual or quarterly. Having access to monthly data means my graphs can display more detail. And that can be useful. I was grinning then.

And that's when I figured out what David Beckworth was up to. He was showing monthly data so I could see more detail. Not just to piss me off.

To see it myself, I duplicated his graph in the Macro Advisers download file:

My graph (Excel's first draft) shows the same pattern as Beckworth's. The range of cells from B186 to B209, outlined in blue, is the MA data I used to re-create the graph.


Beckworth's graph is evidence for what he describes as "the decline in ... nominal spending that starts in mid-2008." He's trying to show a decline in nominal spending. That's why he calls it "total dollar spending" instead of GDP.

By using monthly numbers, the graph shows a nice sharp peak there at the 15000.000 level. Using quarterly data, the peak would have been more rounded, and the moment of transition from increase to decrease would have been less well defined.

And Beckworth is not figuring his monthly GDP from money and velocity numbers. He's using monthly GDP from Macro Advisers. The "Money x Velocity" subtitle on his graph is only a reminder that his graph shows spending -- the quantity of money times the velocity of its circulation.

It's still not "total" spending. GDP is final spending. But that particular imprecision is common among economists and in the media. No biggie I guess, as long as you know about it.

The internal argument expressed above is typical for me. When there's something obviously wrong (like referring to GDP as "total spending") I start to see other things as wrong, too.
-> Why does Beckworth's graph show a sharp peak there at his dotted red line? GDP doesn't peak there. (Because he is showing monthly numbers, not quarterly numbers, dummy.)
-> Why is he telling me he figured GDP from Money and Velocity numbers? (He isn't.)
-> Why is the sky blue? (It isn't. The sun's not up yet.)

Since I had the monthly GDP numbers from Macro Advisers back to 1992, I figured I'd take a quick look at 'em. First, with a linear trend line:

Graph #3, Showing Linear Trend
This is the version that shows the economy running well above average in 2005 and 2006 and 2007.  You have to know that's not accurate. It also shows GDP nicely back on trend, by the end there.

That one's good for a laugh. Beware anyone who uses such a trend line.

The second one shows an exponential trend line. Because growth is exponential.

Graph #4, Showing Exponential Trend
The black trendline here is not based on the full length of the blue GDP line. It is based on GDP for the period from April 1992 to December 1999. That was the period when the economy turned out to be pretty good, catching Alan Greenspan quite by surprise.

I may have accidentally selected a period of time that starts out so-so and ends up better than average. Maybe. And maybe that particular selection gives me a trend line that goes up higher and faster than a less fortunately chosen time period would show. That would be a fair complaint.

However, it is fair to point out that the economy after 2001 fell below the standard set by the 1990s. And it is fair to point out that after 2008 the economy fell below both the standard of the 1990s and the lower standard of the 2001-2008 period. And it is fair to ask why these facts may be.

But these questions have nothing to do with the Macro Advisers data. Quarterly and annual data give rise to comparable trend lines, and bring forth the same questions.

No comments: