The offshoring wage floor has stopped falling.
For three decades, labor arbitrage rested on a single assumption: somewhere, there is always cheaper labor. Across eleven delivery geographies and 4.2 million employment records, that assumption no longer survives contact with the data.
The assumption was never stated that plainly, of course. It arrived dressed as a cost model, with a discount rate and a five-year horizon and a sensitivity table that moved two variables and held the rest still. But strip the model back and the load-bearing beam is always the same: that the wage differential which justified the move will persist long enough to pay for the move.
It has not. Median real compensation in tier-2 delivery cities rose 6.1% annually across the study window, against a modelled 2.3%. That gap is not noise. Compounded across a decade it inverts the business case entirely, and it is present in every geography we examined, which is to say it is structural rather than local.
What makes this more than an accounting correction is the second finding. Fully loaded replacement cost, once attrition is honestly priced, now exceeds the nominal wage gap in four of the eleven geographies. The arbitrage in those markets is booked but not realized. It exists in the model and not in the accounts.
Why nobody caught it
Because almost nobody was looking without a position in the answer. Vendors published research showing vendors were right. Advisors published research showing advisors were needed. The wage series everyone cited traced back, through three or four layers of citation, to survey instruments administered by parties who benefited from a particular reading.
We built the series from national employment microdata and firm-level payroll returns. It took four years and it is tedious and it is the only reason the finding can be defended under questioning. The full method, including every transformation and every series vintage, is in Appendix A. Where a figure could not be traced to a primary source, it does not appear.
Four claims, each traceable to a named series.
Real delivery wages converged faster than modelled
6.1% annual growth against a modelled 2.3%, consistent across all eleven geographies. The direction of the error is the same everywhere, which points to the model rather than the market.
Attrition is the true cost line, not salary
Fully loaded replacement cost exceeds the nominal wage gap in four of eleven geographies. The arbitrage is booked but not realized.
Capability moved before cost did
Judgment work migrated on a different curve than transaction work, and the receiving markets repriced accordingly. This was foreseeable and was not foreseen.
The next decade is a supply problem
Demographic arithmetic, not policy preference, sets the ceiling. Fourteen of the twenty largest economies face a binding constraint before 2040.