A tale of spin, not spin-offs

The Canadian International Development Agency was just caught in a bit of a fib. It claimed to know that, by distributing vitamin A in the Third World, it had saved 1.5 million lives. But auditors say the agency didn't count, it just ran a computer model. Bow your heads, O people: You stand in the presence of social science. It bestrides the modern world like a clay-footed colossus. When we were in Cleveland recently, an AFL-CIO agency bombarded us with statistics on manufacturing job losses by quarter and county. The per-cent bone's connected to the graph bone, the graph bone's connected to the decimal bone, now hear the word of the Lord: "45,734 Ohio jobs lost between 1995 and October 2003 can be directly traced to international trade." Not 45,733 or 45,735; 45,734. The Pope should be so infallible.

Look, if you like playing SimCity, go ahead. But in those terrible middle ages no one thought real knights jumped over things in an L or castles zoomed about in straight lines, the way they did in chess (many a king may privately have felt his queen was more powerful, but never mind). And no peasant could be so superstitious as to believe all human experience could be reduced to a series of linear equations, then solved to yield orderly, scientific happiness. Especially not after watching Robert McNamara run the Vietnam War that way. But like Ptolemaic circles, if you're committed to the concept, you patch the model here and fudge a variable there and get so focused on it you forget the original point lay outside your study.

Remember the headlines in late 2000 that man-made global warming was even worse than the alarmists had thought? Well, maybe it is and maybe it's not. But all that happened then is programmers punched even more pessimistic assumptions into their models and got even more pessimistic projections.

It's like increasing the hypothetical interest rate on your mortgage-payment spreadsheet, seeing your "payments" rise and thinking you're poorer, when all you really just learned is 12 times six is more than 12 times five. If that; increasing one variable and having the result rise doesn't give any reason to think you got the other variables or the formulas right.

Or take government job creation ... please. When something like Canada's infamous Infrastructure Program is projected to create so many jobs and then later hailed for doing it, did you ever wonder how they knew? Going out and counting would be useless because they had to know what was going to have happened in order to compare it with what did happen and take credit for the difference.

So they create a computer model of the economy and tweak it until it "works," that is, converts 10 things they think were measured right in 1995 into six things they think matter in 2003. But the model says nothing about reality because lots of different sets of equations can turn 10 numbers into six other numbers and they can't do test runs of the economy to see which really applies. Despite which, they run their model before implementing the program and call the output a projection. Then they do it again afterwards and call it a result, when all it really proves is that unless you spill coffee into a computer, it always thinks two plus two is four.

So fling econometrics out the window, including any notion of "spin-off" (or "multiplier") effects. Sure, if a government program pays a construction worker a dollar, he spends it on lunch, the waiter then spends it on braces for his kid and so on. But since that dollar came from a taxpayer who now doesn't spend it at the cafe, it's all a giant wash, leaving only the boring question of whether the government is likely to spend the dollar more efficiently than the private sector.

No, it's not a loaded question. It obviously won't in many areas. But government must provide real infrastructure (though pouring the asphalt can be contracted out). It all comes down to a common-sense judgment: Do we really need a road, canal or airport here?

Of course, you need to count your sheep to know if wolves are eating them. I was horrified to learn that during the First World War the British admiralty didn't keep track of whether U-boats were sinking ships faster than they were being built. But you should collect numbers to give common sense something to work with, not to replace it. So forget social science and study hard sciences or humanities.

If vitamin A observably improves immune system responses in individuals (it does), and CIDA distributes lots of it in poor countries (it does), it clearly helps many people live longer. Maybe about 1.5 million. I don't know. Or need to. Doing good is not a branch of mathematics.

So what are these computers doing in our temples?

[First published in the Ottawa Citizen]

ColumnsJohn Robson