Thursday, July 5, 2012

White-Ray Electrical, or ...

... how I learnt about data --  a morality tale.

My sister used to work at a company in Cape Town called White-Ray Electrical.   I'd just started working as an economist and I was intrigued about data.  How to collect them, how to analyse them, whether they showed what was really happening in the economy.  I was the most junior person in the department so I was given the task of maintaining the times series in the Red Book, our record of what we considered to be relevant data series.  My sister told me that she used to fill in the Department of Statistics survey about employment, and, years before, the actual number of employees had deviated from the number they would put in the dept of stats survey.  So she would look at the change in their payrolls from the month before and adjust the number she put on the dept of stats survey by the this change.

I was horrified.  You're telling me that the numbers bear only the most tenuous relationship to reality?  How am I supposed to make decisions involving millions of dollars when I can't trust the data?  Eeep!

So I started a major project.  I graphed (by hand!  this was long before PCs) 20 major time series and put the charts up on the wall all round our office.  And guess what?  They all tended to show booms and recessions.  Oh, the amplitude of the waves would vary, the exact timing was different, but broadly, they tended to give the same picture.

And that was one of the best lessons about economic data I ever learnt.  Don't rely on just one series, look at different series from across the economy.  Allow for the White-Ray Electrical effect, but also assume that if you look at enough time series, they will give you a real picture of what's happening.

Things are seldom what they seem
Skim milk masquerades as cream

Anyway, the purpose of that little discursion across my personal history as an investment person is to provide an intro to the chart below.  Now I should have shown this before the ISM was published, but I had to persuade a friend to give me back data for the Chicago PMI, which I didn't have.  Notice how the average for the Philadelphia, Dallas and Richmond Fed surveys and the Chicago PMI closely follows the national ISM manufacturing survey, even though they from regions around the country, not for the country as a whole, and they're also taken at different dates during the month.

Of course, it still doesn't tell us whether this is "payback" for earlier mild weather.  But ... if it is, we should start to see upturns in the regional numbers when they come out over the next couple of weeks, and if we don't, then the risk that this is a renewed (albeit mild, perhaps) downturn rises sharply.  And the nice little stock-market rally will come to an abrupt end.

There's method in my madness.




No comments:

Post a Comment