29 February 2012

Manufacturing Stats

I caught up today with last week's EconTalk with Adam Davidson. The topic was American manufacturing, and it was based on some reporting he did at an autoparts factory in South Carolina.

One of the main points of discussion was the skills needed by the machinists who operate multi-axis CNC machine tools, and the extreme precision with which they turn out fuel injectors.

This reminded me of something I wanted to post about years ago, but I don't think I ever got around to it. We've got good statistics for the number of people employed in manufacturing both through the years and across countries. And decent statistics of the dollar value of manufacturing outputs. But do we have good numbers for the tolerance of manufacturing processes across time or borders? Or the number of components being assembled? Ceteris paribus, I'd be impressed by more pieces being assembled at higher precisions.

Obviously computing the averages for these things if difficult, if it's not undefined entirely. But I'd like to see someone take a try at estimating them.


  1. I'm not entirely sure how you would collect the data, or even what it would mean if you could. The tolerance for a given manufacturing process is a totally different animal than the tolerance of an assembled fit. And how would we go about quantifying "more pieces at higher precision"?

    The cost of tighter tolerancing drives a lot of manufacturing decisions, also. A good engineer will create a design where all the tolerances are just barely good enough - which means that a better design may incorporate more low-precision processes than a worse design. This implies that large numbers of low-precision parts may reflect less in the way of machining capabilities and more in the way of economic efficiency.

    I think maybe a good way of getting at what you want may be on a cost per part basis for a given standardized part in a given job size with a given tolerance. If you're interested in the precision of turned parts, you create some standard part that uses only that process, and commission 100-part lots of them from various machine shops, changing only the tolerancing on the drawing (e.g. 25mm +/- 0.1, 0.01, 0.001, etc). Then you could get an average cost for the part with a given tolerance from a given process.

    That's as good a way as I can determine to roll productivity and precision across borders into some sort of standardized measure.


  2. Yeah, the more I think about it, the more I'm concluding that the relevant statistics aren't just hard to collect, they're undefined.

    Maybe something more like a case study is needed. I remember reading about some researchers that tracked down where all the components of an iPhone come from, what the value-added was at each step, what the effect was on import & export numbers for each country involved, etc. Perhaps something like that is needed, where you could chart for each step along the supply chain how many pieces were being assembled and at what tolerance they were being made or fitted together.