What a Sri Lankan wage is really worth, and how far it still has to travel

The 2005–2018 run is the steadiest climb in the whole series — post-war reconstruction and a construction-and-services boom carried per capita income from $1,207 to a then-record $4,399 in 2017, better than a three-and-a-half-fold gain in twelve years, the kind of compounding that looks unremarkable year to year and only turns astonishing once you stand back from it. 2018 ticked down slightly on rupee depreciation, even while the underlying economy kept growing in its own currency — a reminder that this measure is denominated in a currency Sri Lanka doesn't print, and it will move with that currency's fortunes as much as with anything the country actually built in a given year.

Then came four years compressed into what felt like one long one: the Easter Sunday attacks, COVID-19, and finally the 2022 sovereign default and the currency float that came with it, which alone carried the rupee from roughly 200 to past 360 per dollar. Per capita income fell from the 2017 peak of $4,399 to $3,343 in 2022 — a 24% loss in dollar terms, on an economy that, measured in its own currency, shrank by "only" about 7–8% that year. The exchange rate did most of that damage; a currency can lose value a great deal faster than an economy can lose capacity, and 2022 is as clean an example of that as this series offers.

The recovery since has been sharp — up 14% in 2023, 19% in 2024, and 10% again in 2025, carrying the country past $5,000 per person for the first time, alongside real growth estimated near 5% in both years. 2026 opened with the government forecasting 6% real growth and the IMF closer to 3% — two parties looking at the same country and arriving at different numbers, which happens whenever the future is involved, and we won't pretend to know which of them is closer to right before the Central Bank and World Bank tell us in early 2027.

A currency, watched year to year, behaves like a voting machine — it registers a default, a rumor, a change of government, as fast as any trader can act on it. Watched across two decades, it behaves more like a weighing machine, and what it eventually weighs is how much of a government's promises turned out to be real. The two lines on this chart sit almost on top of each other before 2020 — US inflation was mild and steady through most of the 2010s, so the voting machine and the weighing machine were, for a while, reading the same number. They pull apart hard after that: US CPI rose about 29% from 2020 through mid-2026, and every nominal dollar since 2021 buys a smaller basket of goods than a 2020 dollar did. That divergence is simply the yardstick getting shorter — twenty-nine points of accumulated US inflation baked into every nominal dollar on the chart's second half.

The practical read is this: Sri Lanka's real recovery is genuine — from the 2022 low of roughly $2,957 (2020 $) to $4,021 in 2025 is a strong 36% gain, worth crediting on its own terms. In money that spends like 2020 money, though, 2025 still sits about 13% below the 2017 peak of $4,645 — the 2017 peak still stands. The "all-time high" in the headline is a nominal-dollar high, and in a world with inflation, that is a low bar to clear.

Two things stand out once the wage from Part One sits on the same chart as GDP per capita. The first is unsurprising: the wage line runs well below both GDP lines for the entire span, since GDP per capita is spread across the whole population — capital income, corporate profits, government spending, people who earn nothing at all — while the wage measures the pay of one specific slice of employees. The more interesting part is what happens to the size of that gap over time — whether it widens or narrows is the actual story here.

And the gap does move. With the reconstructed years folded in, the wage sat at 79% of GDP per capita in 2009 — just after the civil war ended — then settled into a steadier 56–67% band through 2021. It broke hard in the default year: 66% in 2021 down to 54% in 2022, the year the average formal worker's dollar income took its deepest cut relative to the national average. It kept falling in 2023, to 49%, since GDP per capita climbed faster than the wage did that year, and the partial 2024 recovery to $2,293 only carried the ratio back to 50%. Put plainly: the enterprise's per-head earnings have staged a real comeback since 2022, and the dividend that reaches the person doing the work has staged a smaller one. That gap opened in 2022, and four years on, it is still open.

Full-period correlation, 2009–2024: — — a strong positive relationship on its face, though a good deal of that strength is mechanical: any two series that both trend upward across two decades will show a strong correlation, whether or not they move for the same underlying reasons. The rolling window earns its keep here — it strips the shared trend down to five years at a time and shows the years where the wage and GDP per capita actually moved in step, and the years where each went its own way.

Two stretches earn a closer look. Through most of the 2010s the wage tracked GDP closely, several windows reading above 0.9. The 2015–2019 window stands apart: correlation against real GDP collapses to essentially zero, the one true stretch where GDP per capita and the average wage went their separate ways — GDP per capita rode exchange-rate and pre-crisis swings up, down, and partway back, while the wage kept its own steadier climb, indifferent to what the exchange rate was doing. The relationship reasserts itself hard afterward: every window touching the 2022 crisis and its recovery reads above 0.9 again, since 2021 through 2024 is the one stretch where both series fell and rose in the same rhythm, more tightly than anywhere else in the record.

Set against the full 2005–2026 run, the size of the ask is visible before you read a single number. The entire twenty-one-year climb from $1,207 to today — 2017 peak, 2022 crisis, and all — barely lifts off the bottom of the chart's axis, because that axis now has to stretch all the way to $96,200 to hold the projection. That compression is the whole point of drawing it this way: everything Sri Lanka has accomplished across two decades of hard, real work amounts to a small fraction of the distance this target still asks for.

The shape of the climb is the model's own choice, and it runs against the usual habit. A model tapering as it nears a ceiling is the shape economists tend to reach for — the cheap gains spent first, the hard ones saved for last. This model runs the other direction: growth opens at 5.6% in 2026 and climbs in every year that follows, closing near 18% by 2050, with the single fastest year sitting right at the end of the twenty-four-year run rather than anywhere near its middle. That is a real feature of the model behind the speech's own target, laid out here exactly as its author built it, and it's worth sitting with before the destination reads as more plausible than the path there — an economy asked to grow fastest in its final year is asked to defy the ordinary arithmetic of diminishing returns exactly when the base it's compounding on is largest. No curve accelerates forever, and this one is still speeding up on its last page. One year in the source workbook's own growth-rate schedule, 2037, also breaks its otherwise steady acceleration outright — the kind of thing a hand-built spreadsheet tends to pick up over enough rows, and we're naming it here while leaving the source workbook exactly as its author built it.

Is 9.9% a year for twenty-four years a reasonable thing to expect? Measured against this project's own record, this particular ask holds up well against precedent: real GDP per capita grew 10.7%/year across 2005–2015, the post-war reconstruction decade, an even quicker clip than the model's own twenty-four-year average calls for. What that decade doesn't answer is the question of duration. Sri Lanka has shown it can hold a pace like this for ten years. The 2050 target asks for well over twice that long, with every shock the last two decades handed it — an Easter Sunday, a pandemic, a default — presumed absent for the entire stretch. We would not underwrite that presumption ourselves, and we'd encourage the reader to hold it just as loosely, even with an encouraging precedent sitting right there on the record.

The nominal line remains the least useful number on this chart, and the one a headline will reach for regardless. At an assumed 2.5% a year of US inflation, "$40,000 in 2020 dollars" turns into a nominal figure near $96,200 by 2050 — the same lesson the corrected chart back in Part Two already taught, playing out again at a larger scale and over two more years besides: any nominal milestone announced in 2050 will look far bigger than the real target it is actually measuring, purely from a longer stretch of a shrinking dollar. A number that impresses a headline writer and a number that means something to the person living on it are not always the same number, and this one is a clean example of the difference.

The blue and orange GDP lines are here purely for scale — the same nominal and real (2020$) series from earlier in this letter, run alongside the estimated wage on one axis so the reader can see the gap directly rather than take our word for it. The wage sits well below both GDP lines across the entire 2009–2050 span, and the dollar gap between them keeps widening even as the wage-to-GDP ratio itself recovers — both things are true at once, because all three lines are growing, and the GDP lines are growing off a larger base to begin with.

Read as a share of GDP, this path looks almost ordinary — the ratio climbs from 2024's 50% back into the 56–67% range Sri Lanka held for most of 2010–2021, settles into that band around 2040, and drifts only a little higher from there, to 58% by 2050. Read as a dollar figure, the same path looks far more demanding: it calls for 10.5% real wage growth a year, every year, for twenty-four years, roughly three times the fastest decade this project's own wage data has actually recorded, which is 3.6% a year across 2009–2019. Both readings are correct, and both come from the same numbers; a ratio that stays comfortably inside a familiar band can still describe a dollar figure growing faster than anything on record, once the GDP base that ratio is measured against is itself assumed to grow at the pace Asanka de Mel's own model, rebased above, already asks for.

The regression underneath all of this is thinner than it looks. Refit it on the published-only 2014–2024 window, leaving out the years that lean on our own reconstructed estimate, and the slope climbs from β = 0.59 to roughly β = 0.81, which pushes the 2050 real wage figure from about $24,000 up to roughly $32,400. That is a wide spread to get from one modeling choice, and it's worth sitting with: this whole section rests on an eleven- to sixteen-point regression. We are extrapolating a slope, measured over sixteen noisy years, out across twenty-four more — the kind of exercise that deserves your skepticism as much as it deserves ours.