The following essay won the Letter of the Week award on 2004-02-24 at Mark Steyn's web site, http://www.marksteyn.com .
My Fellow Canadian ~
I once read an excellent Isaac Asimov non-fiction essay on really big numbers. Humans are in general really bad at understanding big numbers. Because of my math / science / engineering background, I'm maybe a bit better than average, but I'm no Asimov. I have though learned a few ways to help me better understand big numbers, so that I can better deal with them when I need to. This essay shows how some of those methods work.
The initial Government of Canada estimate for the gun registry database system was $1 million. Technically, I think that's probably a bit low. Based on my on three decades of work in the field of distributed multi-user database transaction processing systems like the registry, and on some systems I'm currently working on which are of that type, I think $3 million would have been a better estimate.
If someone from the Government of Canada can provide me with a simple accounting showing some component of the system that I've missed, I'd be more than happy to adjust my analysis of the situation to take that data into account. My current analysis is based on the numbers I have collected from the public media over the last few years.
Given how important it is for state monopolies to serve citizens to the highest possible ethical standard, let's throw in a factor of three-ish over my base estimate and call it $10 million, to be as careful as possible.
Now, say you had such a $10 million contract with some customers. And then, say you spent three times that: $30 million. Does it occur to you that your customers (in this case, we citizens) might be, oh, shall we say, somewhat angry? Ok, let's say it's another factor of three: $90 million. How are your customers doing now? Fine. Let's throw in another factor of three, so we're now up to $270 million. How angry are your customers now? In more primitive times than we live in, would you still be alive? But wait, there's more. How much would we pay for another factor of three? Oh, about $810 million. Say, that's interesting, the gun registry database system has, according to the CBC, cost $750 million.
It didn't cost 3 times as much. Or 3 times 3 times as much. Or 3 times 3 times 3 times as much. It cost 3 times 3 times 3 times 3 times as much.
That's like planning to have two children, and ending up with 162 (two times three to the power of four). Now stop. Think about that number, 162 children. It's unimaginable that you could legitimately have 162 children, in any way whatsoever. Likewise, there is no way I can come up with to imagine how the registry database project could legitimately cost $750 million, whatsoever.
Never mind that it is to me unfathomable that it could take eight years to develop the registry database, and never mind that according to the CBC it doesn't actually work; $750 million divided by eight years is about $250,000 per day. That's right, they spent what should have been, at its most extreme, a $10 million budget for the entire project, they spent that much every 40 days, for eight years.
Here's another way to look at it. The database system has cost about 750 / 8 = $94 million per year, for eight years. Loaded full-time staff costs in this field are about $100,000 per year. That means the development of this system employed 940 full-time staff per year for eight years. How the hell can a database fundamentally designed to store and retrieve 7 million gun records distributed across 3 million person records take 7,500 man-years to develop? What is this, the Pyramid of Cheops?
(Actually, the CBC's total cost figure of $2 billion for the entire gun registry "file" amounts to about $685,000 per day, which means they spent the entire initially estimated budget of $2 million, again according to the CBC, every three days, for eight years straight. That's 20,000 man-years, to register 7 million guns. But what do I know, I'm just a software guy, I'll stick to the database system.)
Corporations get sued for cost overruns on the order of tens of percent. Assuming my careful $10 million estimate is reasonable, the registry database is not 10 % over. It's not 100 % over. It's not 1,000 % over. It's 7,500 % over.
It's not 10 times less than the standard we citizens are held to, under threat of criminal prosecution. It's not 100 times less. The state's standard of performance for itself is 750 times less than the standard it holds us to. Who the hell, exactly, do these people think they are, and why the hell shouldn't we smite them?
Ok, I'll tell you what. Section 380 of the Criminal Code of Canada calls for jail for up to 10 years for fraud over $5,000. Applying the 750 factor apparently used by State Canada, let's just say that any politician or civil servant that is guilty of fraud over $5,000 should go to jail for 7,500 years.
That's starting to sound like a big enough number for me.
My Fellow Canadian ~
I once read an excellent Isaac Asimov non-fiction essay on really big numbers. Humans are in general really bad at understanding big numbers. Because of my math / science / engineering background, I'm maybe a bit better than average, but I'm no Asimov. I have though learned a few ways to help me better understand big numbers, so that I can better deal with them when I need to. This essay shows how some of those methods work.
The initial Government of Canada estimate for the gun registry database system was $1 million. Technically, I think that's probably a bit low. Based on my on three decades of work in the field of distributed multi-user database transaction processing systems like the registry, and on some systems I'm currently working on which are of that type, I think $3 million would have been a better estimate.
If someone from the Government of Canada can provide me with a simple accounting showing some component of the system that I've missed, I'd be more than happy to adjust my analysis of the situation to take that data into account. My current analysis is based on the numbers I have collected from the public media over the last few years.
Given how important it is for state monopolies to serve citizens to the highest possible ethical standard, let's throw in a factor of three-ish over my base estimate and call it $10 million, to be as careful as possible.
Now, say you had such a $10 million contract with some customers. And then, say you spent three times that: $30 million. Does it occur to you that your customers (in this case, we citizens) might be, oh, shall we say, somewhat angry? Ok, let's say it's another factor of three: $90 million. How are your customers doing now? Fine. Let's throw in another factor of three, so we're now up to $270 million. How angry are your customers now? In more primitive times than we live in, would you still be alive? But wait, there's more. How much would we pay for another factor of three? Oh, about $810 million. Say, that's interesting, the gun registry database system has, according to the CBC, cost $750 million.
It didn't cost 3 times as much. Or 3 times 3 times as much. Or 3 times 3 times 3 times as much. It cost 3 times 3 times 3 times 3 times as much.
That's like planning to have two children, and ending up with 162 (two times three to the power of four). Now stop. Think about that number, 162 children. It's unimaginable that you could legitimately have 162 children, in any way whatsoever. Likewise, there is no way I can come up with to imagine how the registry database project could legitimately cost $750 million, whatsoever.
Never mind that it is to me unfathomable that it could take eight years to develop the registry database, and never mind that according to the CBC it doesn't actually work; $750 million divided by eight years is about $250,000 per day. That's right, they spent what should have been, at its most extreme, a $10 million budget for the entire project, they spent that much every 40 days, for eight years.
Here's another way to look at it. The database system has cost about 750 / 8 = $94 million per year, for eight years. Loaded full-time staff costs in this field are about $100,000 per year. That means the development of this system employed 940 full-time staff per year for eight years. How the hell can a database fundamentally designed to store and retrieve 7 million gun records distributed across 3 million person records take 7,500 man-years to develop? What is this, the Pyramid of Cheops?
(Actually, the CBC's total cost figure of $2 billion for the entire gun registry "file" amounts to about $685,000 per day, which means they spent the entire initially estimated budget of $2 million, again according to the CBC, every three days, for eight years straight. That's 20,000 man-years, to register 7 million guns. But what do I know, I'm just a software guy, I'll stick to the database system.)
Corporations get sued for cost overruns on the order of tens of percent. Assuming my careful $10 million estimate is reasonable, the registry database is not 10 % over. It's not 100 % over. It's not 1,000 % over. It's 7,500 % over.
It's not 10 times less than the standard we citizens are held to, under threat of criminal prosecution. It's not 100 times less. The state's standard of performance for itself is 750 times less than the standard it holds us to. Who the hell, exactly, do these people think they are, and why the hell shouldn't we smite them?
Ok, I'll tell you what. Section 380 of the Criminal Code of Canada calls for jail for up to 10 years for fraud over $5,000. Applying the 750 factor apparently used by State Canada, let's just say that any politician or civil servant that is guilty of fraud over $5,000 should go to jail for 7,500 years.
That's starting to sound like a big enough number for me.
Comment