Saturday, January 4, 2014


Pure logic won the day here. But pure logic should be taken with a grain of empirical salt, because what logicians ignore can matter more than logical deductions made from what they consider. If the data contradicts the logic, then the logicians have ignored something crucial in their otherwise convincing argument.

You see it all the time -- in analysis after analysis that evaluates everything in terms of government debt. As if debt other than government debt was a ceteris paribus constant without cost or consequence.

(click here to see how this data was compiled – as well as a longer term estimate for US debt that extends back to 1834: the data is downloadable from here)

The first of these two links brings you to Business Spectator, where Steve Keen reviews his data sources and his method of fitting various data series together. To me, it was fascinating.

Keen's second link gets you an Excel file. One sheet, six columns: household debt, business debt, private debt, government debt, GDP, and the date.

Most interesting? Date. Keen provides monthly data, but the month-and-year value is expressed as a decimal number. Same values as feet-and-inches, if you know inch values divided by 12.

No, huh?

Anyway, we end up with over 2,600 rows of data. Well, since 1790... 200 years to 1990... 210 years to 2000... 222 years to 2012... 222 times 12 is 2664. Plus a few months of 2013.

201 years from 1790 to 1990, if we include 'em both. 223 years total. Times 12 is 2676. Plus 7 entries for 2013 is 2683. Plus the row of category titles, 2684. Yup.

Keen's spreadsheet starts out as shown at right. As you can see, row 2 contains the first value for 1790. Row 14 contains the first value for 1791. And Row 13 contains the last value for 1790. The "917" there, that's 11 divided by 12. If it was 6 feet 11 inches it would be 6.917 feet (though I use two decimal places when I use such numbers at work: 6.92 feet. It's much easier to think in two decimals than three.)

The odd thing with Keen's date numbering is that his 11 value, his "917" represents the 12th month of the year. Not the 11th month.

The first month of the year 1790 is given in Row 2: 1790 point zero. The second month is in Row 3, point oh eight three, which is ONE divided by 12.

It doesn't matter much, I guess. His numbering system works. It's just odd, that's all. To me, it looks like he ended up doing it like that because of the way Excel works. It's just odd.

Personally, I'd rather use quarterly values or (if we're going back so many years) annual values. It's easier to work with and easier to understand.

No comments: