Friday, April 13, 2012

Looking at the same graph -- seeing an entirely different picture

There's plenty to talk about with this chart (which comes to us via Andrew Gelman), but the thing that struck me (and this happens a lot) is that, when you look at it closely, the graph actually undercuts the point it's supposed to be making.

The idea that technology is coming at us faster and faster is one of the most ingrained ideas in our society. Given the title of this chart, its creator would certainly seem to share this notion, but does the chart really show what its name indicates or are we seeing something more like a distant cousin of Simpson's paradox where shifts in make-up and other factors create the illusion of a trend?

(Before we get into these factors, though, take a minute to look at the graph. It looks like the steepest curve is associated with the VCR. That's over thirty years ago, a long time for a record to hold if we really are seeing an upward trend.)

As you're looking at the graph, think about these three things:



Big ticket items

Going last to first, big ticket items requiring heavier manufacturing always had slow adoption curves regardless of when they were introduced (check out the dishwasher). Much of the appearance of acceleration can be attributed to a shift in composition to smaller items.

Then there's the period of 1930 to 1945, pretty much the ultimate in anomalous data. During the Depression most people couldn't afford new products. During the war, new technology was rapidly developed and adopted, but by the military, not consumers.

Finally there's infrastructure. The adoption curves of electricity and telephones are almost entirely governed by the hard work of developing infrastructure (something we have arguably gotten slower at) and infrastructure at least indirectly constrains every line on this graph (check out the Reivers for an account of what cross country driving was like at the beginning of the last century).

The infrastructure constraint points to perhaps the most impressive case of a technology exploding. Take a look at the curve for radio. Now look at the curve for electricity. Even allowing for a few crystal radios and battery operated sets, you have to conclude that radios achieved almost 100% penetration of houses with electricity in about a decade. By that standard, the record for fastest adoption is over eighty years old.

Update: There's a good string of comments on this over at Andrew Gelman's site, some of which make some of the same points I've made here..


  1. Cellphones and internet are also infrastructure items, though, right? So at the turn of the century we were able to complete two large projects (roads; electric and phone lines), and in the early 90s we were again able to complete two large projects (cellular phone/satellite coverage; high speed cable penetration).

  2. "Cellphones and internet are also infrastructure items..."
    Yes, but with a distinctive difference. Cell and Internet initially worked over existing infrastructure - the electric and telephone grids. Furthermore, other products were dependent upon existing infrastructure (think color TV and number of broadcasts that were in color). Then there's the personal modification requirement - your apartment may not have room for a clothes/dish washer or dryer. And finally rural electrification was a government effort, where the urban infrastructure was built by private entities.