The price elasticity of labor-saving devices

Andrew Gelman passes along a great chart (via here and originally here):

In his post, Gelman praises the cleanliness of the graph, and I think that's fair--it's pretty. However, it's not obviously a great data visualization, since neither the color nor the type of the lines used convey any information beyond making the graph readable, which seems a waste (couldn't the colors had been chosen based on time-to-80% adoption or something?).

Like others, however, I find this a useful spur to thinking. In my case, looking at the S-shaped curves of adoption for most (but not all!) technologies, I thought of diffusion theorizing. In this case, what seems to be driving differential adoption rates is first the fact that relative prices for all of these innovations have fallen really far as per-capita GDP has risen and prices for manufactured goods have fallen. (Consider the "computer" in this context, which has surely changed the most of all of the innovations over time.) 

Second, what is most obvious (to me, anyway) is the profound labor-price elasticity here. 

Note that the technologies that diffuse fastest are those that do things that human beings can't substitute their labor for. (No matter how hard you try or how much elbow grease you apply, you can't make things colder, which is why the refrigerator diffuses so quickly. On the other hand, you can substitute time and labor for washers and dryers.

Third, general economic conditions play a huge role in distorting what seem to be, essentially, two waves of product innovation. Had there been no Great Depression, then a lot of things would have reached their post-1945 peak a lot faster, and the period 1915-1945 would  look a lot more like the period 1970-2000. In this case, though, the telephone stands out as a real puzzle: why did it take such a useful gadget so long to diffuse? It must have been much more expensive in relative terms circa 1925-1930; look at the profound elasticity it displays at 1929, which is almost as pronounced as that for the automobile.

Fourth, the presentist bias in this chart is extreme in two ways. First, we forget things that we don't count as "technology" anymore (e.g., toilets, coal furnaces, sewing machines), and so they are left off. Second, we don't know what innovations are at low levels of adoption right now--imagine someone in 1960 trying to predict the adoption arc for personal computers!--and so our current rates of adoption are vastly overestimated compared to what the same chart will look like in 50 years.