Skip to content

Technology speedup graph

Dan Kahan sends along this awesome graph (click on the image to see the whole thing):

and writes:

I [Kahan] saw it at , which misidentified the source (not “visual economics”; visualizingeconomics.comwhich attributes it to Nicholas Felton, who apparently condensed this version, which I worry could cause a stroke). But it did have a good write-up that (I’m glad) caught my attention.

It made me [Kahan] start to wonder about what sorts of qualities of a technology will influence its dissemination & also about the availability of benchmarks for proliferation of various sorts of things (e.g, fads & trends, health-promoting behaviors, knowledge of a scientific discovery) that one could use to gauge how meaningful the apparent increase in rates of proliferation of these technologies has been over time. That in turn made me wonder whether — indeed, suspect that — some smart historian or economist has already addressed these points; I’ll have to poke around to see!

It’s not what anyone would or should use to report or illustrate data analysis, but a graph that puts you in the mood to wonder & conjecture — without negligently pointing you in a misleading direction — is a nice species of graph.

Imagining what you [i.e., me] might say, I’d reply that what I’m talking about— & what this is an example of —is different from Nightengale rose; that *was* supposed to be analysis but failed to convey the  data behind the intended inference — that deaths caused by poor medical treatment killed many more soldiers than combat — but didn’t in nearly as clear a way as it could have & also didn’t provoke conjecture — only perhaps a desire to make sense of something visually arresting.

My brief response: I think the graph could be cleaned up in some way but basically I like it. It has a directness that is to my taste, unlike the Nightingale graph, whose twists distract from the data message.


  1. Marc Levy says:

    I like the graph too, but it doesn’t seem to support the conclusion implied by the title. The internet and radio curves look almost identical, for example, as do the cell phone and color tv curves. Yeah, there’s nothing in the present as flat as the stove curve was during 1900-1915, but that’s surely a result of sample bias. There could be plenty of things moving as along the same lines as the stove was during 1900-1915, but we wouldn’t know it was interesting enough to bother graphing because its use is now so low (home generation of renewable energy might be something that ends up looking a lot like that stove graph, for example).

  2. Wayne says:

    The one misleading thing about the graph is that it seems to imply a cumulative effect, while my feeling would be that the demographics of who had and had not adopted something changed a lot over time. In particular, I’d guess that most who did not own a car in 1960 were poor, while many who do not own in 2010 are actually fairly well-to-do urbanites.

    The apparent 5% of holdouts on color TV also interests me, and I’d suspect that they don’t own a TV at all, which makes that category a bit misleading.

    • Andrew says:


      I think that income and car ownership are positively correlated in the U.S. as a whole and also in urban areas.

      • Wayne says:

        Andrew: Thanks! I was just looking at the county I live in, where there is a lot of high-tech employment within walking distance, where we have a subway to the nearest large city, where two different companies are competing to rent cars by the hour from parking spaces on most every block (with more set to enter the competition), and where another company rents bicycles (first half hour is free) from convenient racks all over the area. You really don’t need to own a car to live here, and the income and educational levels are above national and state averages.

        I think also of the DUMBO neighborhood in Brooklyn, NY, which has a subway, you can walk across the Brooklyn Bridge to your job, and owning a car is actually a liability. Price of admission is quite high there, too.

        Anecdotal, of course, so I trust your numbers/knowledge, but it did cause me to wonder if sometime in the 1960’s or early 1970’s such neighborhoods had disappeared, with our suburban love affair with the car and neighborhoods that required cars, leaving mainly the poor without cars. Now that you mention it, this probably isn’t true and there were probably always wealthy carless enclaves in places like New York, Chicago, Boston, etc.

  3. John says:

    I don’t have almost any of those technologies. I have electricity, a stove, a refrigerator, a computer w/ internet, and a cell phone. I’m glad I’m in the 1% by some metric.

  4. Phil says:

    I like the graphic, although the dashed lines have so much less visual weight than the solid lines that they are too easy to ignore; the thickly hashed lines are better. But with only three colors I guess it’s hard to find alternatives.

    We’re a much richer country now so it’s not surprising that just about anything useful gets adopted faster now, but I the title seems a too strong. Adoption of the car (from 10% to 60% in 15 years) and air conditioning (10 to 60 in 10 years) and especially the radio (10 to 60 in <10 years) were as fast as the more recent technologies, in spite of all of those being much more expensive in real terms than most of the recent stuff.

  5. kjetil halvorsen says:

    For us non-american readers it would be more interesting to see this graph for the whole world. But then it would have to include some more basic things such as rinning water.

  6. Marcin says:

    Pretty scary – since 1990 no major “technology” emerged (maybe with the exception of tablets).

    • Phinn says:

      Most of the major consumer technological advances since 1990 have been derivative of the Internet.

      The two most significant consumer items that I would include on the graph are: smartphones (2001-02, with the iPhone appearing in 2007), and Facebook (2004). Their market saturation has followed similar patterns, I believe.

  7. dmk38 says:

    there’s selection bias: technolgies that petered out are omitted. E.g., where’s the hoola-hoop?

  8. John Mashey says:

    I think this graph is accurate, but easily misleading, in the sense that “spreads faster today” may be true, but doesn’t get at underlying causes and misses multiple conflating factors. These consumer products/services have wildy-varying characteristics and dependencies.

    ‘It made me [Kahan] start to wonder about what sorts of qualities of a technology will influence its dissemination & also about the availability of benchmarks for proliferation of various sorts of things (e.g, fads & trends, health-promoting behaviors, knowledge of a scientific discovery) that one could use to gauge how meaningful the apparent increase in rates of proliferation of these technologies has been over time. That in turn made me wonder whether — indeed, suspect that — some smart historian or economist has already addressed these points; I’ll have to poke around to see!’

    First, nits:
    Telephone, to be more precise, should probably be landline.} so I ignroe that.
    Second, I’m not sure what “stove” means, because it could be {wood, coke, gas, electric … and they are different

    Relevant attributes:
    1) Dependency on external infrastructure and earlier capital investment to put it in place.
    Electricity: consumer needed somebody else to build big pervasive, infrastructure first.
    See Rural electrification for context on the electricity line.
    With the exception of those with off-grid supplies, consumers needed grid connections for
    {radio, refrigerator, clothes washer, clothes dryer (if electric), air conditioning, dishwasher, color TV, microwave, vcr, computer, Internet}, but not landline phones. It is slightly odd to see ~10 years where more households had radios than electricity.

    For landline phones, people needed wires to telco switches (likewise cable TV)
    For cellphones, people needed cellular radio infrastructure (a lot cheaper than running cables, why they get used so much in developing countries)

    For radio and TV, relatively modest broadcast infrastructure was needed.

    2) Network effects
    For some products or services to be useful, one needs them to be widely available.
    At one extreme, the only reason one cares if anyone else has one is that there be a market to induce suppliers.
    {Stove, fridge, washer, dryer, microwave, and sometimes computer} are like that.

    In the middle, {radio, TV, VCR} do not depend on any other specific users, but they do depend on having enough users that content is created for them.

    {Telephone, cellphone (voice}} are mostly useless unless people you want to communicate with have compatible devices/services.
    Likewise {Internet, computer} used for communication, not just computing or data access.
    Internet might usefully be expanded into {MySpace, Second Life, Friendster, Twitter, Facebook, etc} where network effects are very strong in making something successful or not.

    3) Physical
    Some of these are fairly easy to buy and add to one’s dwelling {radio, TV, computer, etc}.
    Bigger appliances are more space-constrained.

    4) Cost
    A new auto and its yearly upkeep dwarf any of the others in cost (and space) required. Of course, early on, there was a similar dependency issue on road quality, widespread gas stations, etc. The current analog is the chicken-egg dependency between EVs and availability of charging stations.

    In general, things spread faster if they are small, cheap, don’t depend on creating a new infrastructure.
    Cellphones were instantly useful without depending on other people having them, because there was already a phone system.
    Bell Labs Picturephones were not successful, because they were expensive … and useless unless your associates had them, also, and they didn’t.

    I’d suggest that the best analog of the more recent electronic products (which rise rapidly) was radio, and it rose pretty rapidly as well, until it ran into the electricity constraint. Electricity and landline telephones required pervasive capital-intensive buildout, and cars were expensive.

    People who do technology development worry all the time about chicken-egg and related dependencies, a key phrase being “technology insertion strategy,” found pervasively in the computer business at high velocity {hardware, software, netwworking and even connector standards.}
    A current hot topic is that of smart grids, EV and PHEVs that negotiate with the grid to either get charged or provide power, which requires both gird and cars to be smart. Smart grids want to be able to negotiate with fridges, dryers, etc to minimize peak loads, but that means there have to be enough of them to make it worth the bother, and few will add first cost if no smart grid is available to make it worthwhile.

    • jrkrideau says:

      I find something wrong with stoves as well. I think we need a better defintion of “stove”. I’m from Canada so pehaps our weather makes a stove more important but I would have thought that most rural — farm?– residences would have a stove for cooking and, in many parts of the country , heating. And if I recall correctly the US like Canada had a very high proportion of rural inhabitants in the 1900’s.

      Re telephones, while there is a requirement for the infrastructure, in the early days of the 20th century, not including the switchboard, much of it consisted of wooden poles, two glass insolators per pole, two strands of copper wire, a rather primative phone and two large dry-cell batteries per phone.

      On the other hand, as my father pointed out, while in the early 1900’s they could consider electricity in the house a luxury, a telephone had a lot of appeal as an emergency device. That, to him, was one of the major reasons for supporting a rural telephone system where we lived. My family got a telephone in 1915 but did not have electricity til 1948 or ’49.

      With a party line, a single ‘very long’ ring could alert neighbours,– those within 2-5 miles of the house in our case, to emergencies, especially to fires, and summon help in a minute or so, rather than, by my guess, 10 – 30 minutes required to notify people in person. Getting an alarm from a fire department would take 1-2 minutes as opposed to pehaps 30 minutes or more.

      I was really pleased to see the microwave curve as I had been wondering about microwave market penetration about two days ago and this helps anwser my question.


  9. […] Andrew Gelman wrote a post today that includes a great graph that shows how the innovation diffusion curve has played out for a number of major inventions over the past century. His post includes a discussion of how the graphic developed, an interesting story in itself. […]

  10. Tom says:

    I think the temporal relationships would be clearer if the y-axis were on a natural log scale.

  11. Jeff R says:

    Do the color categories mean sommething?

  12. Clark Andersen says:

    I’m bothered by the graph only showing the growth of current technologies, and not the decline of technologies which they replaced. Growth in cars should be balanced against decline in horse-drawn carriages/buggies (or simply horses), growth of electricity balanced by decline in gas-lights and wood-stove heating, etceteras. Neat graph, though.

  13. Donald A. Coffin says:

    There’s a version of this diagram in Gary Walton and Hugh Rockoff, History of the American Economy, 10th ed., p. 6. Their source is Michael W. Cox and Richard Alm, “Time Well Spent: The Declining Real Cost of Living in America,” Federal Reserve Bank of Dallas, 1997 ( They don’t provice a source for the data in the diagram.

  14. […] Technology speedup graph « Statistical Modeling, Causal Inference, and Social Science […]

  15. Phinn says:

    I find it somewhat odd that they would list “color TV” as an item but not simply television.

    I suppose color TV is what we have around today, but it was a relatively minor improvement over the basic black-and-white TV that preceded it by more than a decade.

  16. Jim Cobbs says:

    Be far more useful if it had been on a sem-log chart with trendlines.

Where can you find the best CBD products? CBD gummies made with vegan ingredients and CBD oils that are lab tested and 100% organic? Click here.