Diffusion Tracker

Two economists, Diego Comin and Martí Mestieri, noticed something odd about how technology travels. New inventions now arrive in lower-income countries almost as fast as in rich ones; how deeply they end up used keeps drifting apart. Arrival converged, depth diverged, and that quiet asymmetry explains a surprising share of why incomes haven't. I wanted to see it with my own eyes, so I timed seven technologies, from the fixed telephone to the electric car, across every country with data: how long each took to arrive, and how long to actually matter.

Median years until a technology reaches a country, by income group
Years from the global pioneer's first-milestone crossing to each country's own crossing of the same first milestone (extensive margin). Left-censored observations excluded. Technologies ordered by vintage.

Copy this snippet — the chart resizes with its container.

Worth noticing: the collapse is dramatic. A low-income country waited a median 41 years for the fixed telephone, 20 for the internet, 15 for broadband. For solar power the wait is 6 years, exactly the same as in high-income countries. Technologies that come in modular units you can buy by the container, rather than networks you must build, now arrive almost everywhere almost at once.

Median years from arrival to intensive use, by income group
Years between a country's first-milestone crossing and its third-milestone crossing (intensive margin), among countries that have reached both. Rich countries deepen adoption faster, and for recent technologies many poorer countries have not reached intensive use at all, which this survivor-only chart understates.

Copy this snippet — the chart resizes with its container.

Worth noticing: arrival is only half the story. High-income countries went from 5% to 50% internet use in a median 8 years; no low-income country has crossed 50% yet, so they don't appear in that bar at all; the divergence hides in what's missing. The exception is solar, where latecomers deepen faster than pioneers (3 to 4 years from 1% to 10% of generation, versus 9 in high-income countries): once a technology is cheap and modular, depth follows arrival quickly.

Arrival vs. deepening

Each point is one country × one technology. Horizontal: years the technology took to arrive (extensive lag). Vertical: years from arrival to intensive use (intensive lag). Colours are World Bank income groups. The cloud compresses left to right over successive technologies (arrival has converged) while the vertical spread persists: deepening has not.

Extensive vs. intensive margin lags
Only countries with both milestones observed (uncensored). Diagonal guides at equal lag.

Copy this snippet — the chart resizes with its container.

Worth noticing: squint and this is Comin and Mestieri's whole argument in one cloud. Newer technologies pile up against the left edge, meaning arrival has converged, while the vertical spread refuses to shrink: deepening has not. And the dots that are missing matter as much as the ones plotted; countries that never reached the intensive milestone don't appear here at all.

One technology, every country

Adoption trajectories for all countries with data, coloured by income group. Dashed horizontal lines mark the three milestones used in this tracker.

Copy this snippet — the chart resizes with its container.

Try this: select mobile: 203 of 214 countries have passed 50 subscriptions per 100 people, the fastest and most complete diffusion in the data, achieved by leapfrogging the fixed-line network rather than extending it. Then select electric cars: still a rich- and middle-income story, with no low-income country in the data past even the 1% mark.

One country, every technology

When each technology reached this country's three milestones. Open circles are left-censored: the country was already past the milestone when the data series begins, so the true crossing year is earlier.

Milestone 1 = arrival · Milestone 2 = intermediate · Milestone 3 = intensive use (thresholds per technology in the Methodology tab)
Milestone 1: arrivalMilestone 2: intermediateMilestone 3: intensive use○ Open circle = left-censored (true year is earlier than the data can show)

Copy this snippet — the chart resizes with its container.

Try this: policy can compress these timelines dramatically. Norway took electric cars from 1% to 25% of new sales in five years; Chile took solar from 1% to 10% of its electricity in six. When the technology is ready and the incentives line up, “intensive use” can arrive within a single government's term.

Methodology & sources

This page puts fresh, still-maintained open data under a known result: Comin and Hobijn's CHAT dataset tradition and Comin and Mestieri (2014), “If Technology Has Arrived Everywhere, Why Has Income Diverged?” (AEJ: Macro). It builds no index, ranks no one, and claims no causal effects. It just makes the pattern visible.

Technologies and milestones

For each technology × country, the tracker records the first year the adoption series crosses three thresholds. Thresholds are technology-specific, because a sensible “intensive use” level for solar's share of electricity is not the same as for mobile subscriptions:

    Extensive lag = country's first-milestone year minus the global pioneer's first-milestone year (the earliest crossing observed in the data). Intensive lag = country's third-milestone year minus its own first-milestone year.

    Censoring: read this before quoting numbers

    If a country is already past a milestone in the first year its series is observed, the crossing is left-censored: the true year is earlier than the data can show. Censored observations are marked with open circles in the country view and are excluded from all median-lag statistics rather than guessed. This mainly affects the fixed telephone (series start 1960) and electricity access (series start 1990) for early adopters. Survivor bias runs the other way for recent technologies: intensive-lag medians only include countries that have already reached intensive use, so they understate the true divergence.

    What this tracker explicitly does not do

    • It does not score or rank countries; the point is to make the diffusion pattern visible.
    • It does not claim causal effects of technology adoption on income.
    • It restricts to technologies with well-measured adoption paths, which excludes much “soft” innovation.
    • It does not include technologies whose adoption data could not be independently retrieved at build time: cloud computing (Cloudflare Radar requires an API token) and country-level generative-AI adoption (no comparable cross-country series is openly published yet). LED lighting and containerisation lack maintained country-level adoption series in open databases.
    • Historical technologies (steam, electricity generation, tractors…) from the original CHAT dataset (1800 to 2003) are not included; the tracker deliberately uses only series that are still maintained today.

    Sources