The trouble with models

From Paul Krugman's INCIDENTS FROM MY CAREER:

There are several different ways of doing good economics. ... But what has always appealed to me, ever since I saw Nordhaus practice it on energy, is the MIT style: small models applied to real problems, blending real-world observation and a little mathematics to cut through to the core of an issue. 

The first summer I worked for him, Nordhaus began with only a vague sense of how to think about the problem of appropriate pricing of energy. I was able to watch the process by which he crystallized that vague sense into a model, and then was able to see the way in which that model transformed everyone's perception of the issue.

In one summer -- three months, or some portion thereof -- Nordhaus went from having no idea about the appropriate pricing of energy, to transforming everyone's perception of the issue.

My first computer was a Radio Shack PC-1 "pocket computer" with one K of memory and the BASIC language built in, a gift from my mother-in-law.

I read the manual and tried to figure out how to use the BASIC language. But I could make no headway. So I got a little book of computer games written in BASIC, and typed one in. When I ran it, there was an error. So I compared what the book showed to what I had typed. There was a discrepancy: As I remember, I used a semicolon where I should have used a colon. Something like that, anyway. 

I fixed the problem, and ran the program again. But there was still an error. I checked my work, found the discrepancy, fixed it, and tried again. But the error was still there.

About that time I noticed that it wasn't the same error every time. Each error was a little further down in the code. Suddenly I realized that the computer was helping me fix my errors, one at a time, and that we would eventually run out of errors.

Fortunately it was a short program. We ran out of errors, the computer and I, before I ran out of patience. And suddenly, instead of getting an error, the little game worked. After that, computers got a lot easier. And I was hooked.

That all probably took me three months, or some portion thereof.

When I felt like I could do more than my little computer could do, I signed up for a course in BASIC programming. Loved it. After that, I got my first desktop computer, a Commodore 64. In addition to BASIC, I taught myself assembly language and tried to learn the C language.

Later on, I graduated to an IBM-PC compatible, learned assembler all over again, and finally succeeded in learning C. Later on again, I found the "Visual Basic" language that comes with Excel, and picked that up. By this time, Krugman was probably a full professor.

I wasn't into collecting recipes, and I hadn't yet started using the computer to keep notes and quotes from my reading. But I was always interested in using the computer to solve problems that came up at work -- mostly problems like "this is tedious and boring". Things are very often tedious and boring because they are repetitive, and computers are very good at doing repetitive things. So I always had little puzzles to work out on the computer, even if I didn't get to use them at work, and I learned a lot.

During that time, Krugman's beard would have been turning gray.

The last place I worked, I was there 15 years before retiring. We had a metal shop where we fabricated steel components. We had a wood shop where we built large assemblies, always similar in design but differing in dimensions and details. Repetitive work.

My job was to make drawings for the wood and metal shops, based on customer requirements. Everything was made of lumber held together with steel components. Because the stuff we made was almost always similar to something we made before, for a new job we'd start with some old drawings and start making changes. Again, repetitive work.

You know how it is: Work is work. It's good most days, but once in a while it gets to ya. The longer I was there, the more some particular tasks seemed particularly tiresome. I would take one of those tasks, take it home, and see if I could use the computer to automate it. Sometimes I could.

Because we used computers to do our drawings, sometimes I would bring my automation stuff to work and start using it, and actually make my job less boring. But after you make the most boring task less boring, you soon find that some other task is now the most boring. So I always had something to work on.

Long story short, after several years I had enough stuff automated that my computer was doing a good portion of my drawings for me, while I just coached it. My drawings got more standardized. And because AutoCad, like Excel, came with Visual Basic, I was able to automate the process of making parts lists in AutoCad, transferring them to Excel sheets, and generating the information the production shops required. Talk about relieving tedium!

I retired five or six years ago now, and they still use the system that I came up with on my own time.

The trouble with models

The point of this story is that, when you're dealing with a complex system like manufacturing or the economy, you don't sit down for a couple days and have the whole thing worked out.

You don't start by knowing nothing about it, and end up two months later knowing so much that you transform the world. Not in manufacturing, and certainly not when you're dealing with the economy. (Not that Krugman is saying otherwise.)

You don't start with a fragment of the production process or a fragment of the economy, work out some nifty little thing, and call it done.

The world of computer programming didn't go my way. They made standardized accounting software and standardized  manufacturing software, and you're expected to pay hundreds of thousands of dollars to buy it. And then you've got so much money invested in it that you refuse to listen when your people tell you that the new software doesn't work the way your company works.

I didn't automate anything for a long time. My first task was to get familiar with what we do and the way we do it, familiar with it to the point where the boredom was becoming painful. And then the mind would wander, and I would think of a way to automate some little piece of the boring process.

 

Or the mind would wander, and I would realize that no one addresses the cost problem behind cost-push inflation. They just quash the inflation.

Or I would realize that the cost problem, unaddressed, causes the fall of profits and living standards.

Or I would realize that the fall of profits and living standards is exactly the problem we face every day.

Or I would realize that the cost problem might not go away on its own. It might survive for decades. It might cause profits and living standards to fall for decades, maybe longer.

Or I would realize that the growing cost of finance is exactly the sort of problem that could cause this long-term decline.

 

The only thing I haven't realized yet is how to convince people of the importance of these realizations.