Most of the debate about AI and labor markets is a debate about workers: who gets displaced, who gets augmented, who gets retrained. Almost none of it is about the entity standing quietly behind every payroll: the treasury. Governments in most middle- and high-income countries raise roughly half of their revenue from labor — personal income tax plus social security contributions. If AI-driven capital deepening shifts factor income from labor toward capital, that base erodes at precisely the moment demand for social spending rises. Displacement, retraining, and safety nets all cost money, and the money comes disproportionately from taxing the very jobs under threat. Call it the fiscal scissors: the blades close from both sides.
That much is familiar as a macro proposition. What is missing from the discussion is microdata: which occupations actually carry the labor tax base, and how do they line up against AI exposure? In the companion analysis we mapped every two-digit occupation in Türkiye's 2024 labor force survey into the four transition archetypes of the Richmond framework — Automation Risk, Reorganize, Grow with AI, and Less Change. Here we push every private-sector formal wage employee through Türkiye's full 2024 labor tax schedule: progressive income tax, employee and employer social contributions, stamp duty, the minimum-wage exemption. The result is a fiscal map of AI exposure, covering about 14.3 million workers and a labor tax base of roughly 2.5 trillion lira a year.
The tax base lives where the displacement risk lives
The headline finding is a concentration result. Automation Risk occupations — clerks, administrative and business associate professionals, the white-collar middle — account for 19 percent of private formal employment but 25 percent of labor tax revenue and 29 percent of personal income tax. These are exactly the occupations the framework flags as most substitutable.
Figure 1. Employment, wage-bill and labor-tax-revenue shares by archetype, private-sector formal wage employees, Türkiye 2024. The gap between the light bar (employment) and the dark bar (tax revenue) is the fiscal story: Automation Risk widens, Reorganize narrows.
The cleanest summary is the ratio of the two. Call it the Fiscal Exposure Index — an archetype's share of labor tax revenue divided by its share of employment. Above 1, the archetype pays in more than its headcount implies.
Figure 2. Fiscal Exposure Index by archetype, private-sector formal wage employees, Türkiye 2024. An index above 1 means the archetype contributes more to labor tax revenue than to employment.
| Archetype | Employment | Tax revenue | Income tax | Index |
|---|---|---|---|---|
| Automation Risk | 19.1% | 24.7% | 28.7% | 1.29 |
| Grow with AI | 9.1% | 10.9% | 12.3% | 1.20 |
| Less Change | 47.3% | 47.2% | 48.2% | 1.00 |
| Reorganize | 24.5% | 17.2% | 10.9% | 0.70 |
Automation Risk's index is 1.29: every job in this archetype carries about thirty percent more tax revenue than the average job.
The mirror image is just as telling. Reorganize occupations — dominated by sales workers and cleaners — are 24 percent of formal employment but only 17 percent of revenue and just 11 percent of income tax, an index of 0.70. The reason is a policy choice: since 2022, the minimum-wage portion of earnings is exempt from income tax and stamp duty, and around thirty percent of the formal private workforce sits at or near the minimum wage. For those workers the effective tax wedge falls to a floor of 27.7 percent, essentially all of it social contributions.
The exemption has already hollowed out revenue collection from the bottom of the distribution. Whatever one thinks of that policy on equity grounds — and there is a good case for it — its fiscal consequence is that Türkiye's labor tax base has no cushion. The revenue is concentrated at the top of the wage distribution, in occupations where the displacement risk is also concentrated. Erosion starts where the money is.
Informality is usually a buffer. Not this time.
In an economy with a large informal sector, labor-market shocks normally hit the treasury less than the headlines suggest, because many of the affected workers were not remitting much to begin with. AI displacement inverts that logic. Automation Risk workers are the most formalized segment of the Turkish labor market — 84 percent are in registered wage employment, against 44 percent for the least-exposed archetype. The occupations most exposed to displacement are the ones most fully inside the tax net. There is no pre-absorbed loss; displacement there converts into revenue loss one-for-one.
Our scenario work makes the arithmetic concrete. In a central scenario — 30 percent of Automation Risk workers displaced over roughly a decade, 40 percent of those transitioning to informality and the rest re-employed formally at a 20 percent wage penalty, mild wage compression in Reorganize, a 10 percent wage gain in Grow with AI — about 3.7 percent of the formal labor tax base is at risk, roughly 95 billion lira a year at 2024 values.
Figure 3. Decomposition of revenue at risk, central scenario, as a percentage of baseline labor tax revenue. Informalization dominates: a worker who exits the formal sector leaves the tax base entirely, while a re-employed worker at a lower wage still remits most social contributions.
The decomposition is the most policy-relevant part. Informalization alone accounts for about three-fifths of gross losses — nearly three times the loss from wage cuts among the re-employed. The mechanism is structural, not parametric: a displaced worker who lands in informal work exits the base entirely, whereas one re-employed at a lower wage keeps paying social contributions, which are flat and floor-insensitive.
For fiscal purposes, where displaced workers land matters far more than how far their wages fall. That single result reorders the policy priority list: formal re-employment services, wage-subsidy bridges into registered jobs, and enforcement against informal absorption do more for the treasury than wage insurance.
The result is not an artifact of the central parameters. Across the sensitivity grid — varying the displacement rate and the share of displaced workers absorbed informally — revenue at risk runs from about 1 percent to over 8 percent, and the informal-transition share moves the answer at every displacement rate.
Figure 4. Labor tax revenue at risk as a percentage of baseline, across displacement rates for Automation Risk workers (rows) and the share of displaced workers absorbed into informality (columns). The central scenario is the middle cell, 3.7 percent.
The wedge is financing its own erosion
There is a second, more uncomfortable layer. Türkiye's aggregate labor tax wedge in this sample is about 36 percent — among the higher wedges in the OECD orbit. A high wedge does not just sit atop the labor market; it changes relative prices. Every lira of tax on labor makes the capital substitute — the software license, the model subscription, the automated workflow — relatively cheaper at the margin.
Acemoglu, Manera and Restrepo (2020) made this argument for the US tax code's favoring of equipment over labor, and the logic carries: labor taxation is not merely a vulnerable revenue source in the AI transition, it is an accelerant of the very substitution that erodes it. The treasury is, in effect, subsidizing the technology that shrinks its base.
What we are not claiming
Three honest caveats. First, these are comparative statics on a fixed 2024 cross-section — no general equilibrium, no productivity spillovers, no new occupations. History counsels humility: labor-share collapse has been predicted before, and the reinstatement of labor through new tasks is real.
Second, the offset assumption matters. The 10 percent wage gain we credit to Grow with AI workers absorbs about a quarter of gross losses in the central scenario, and it is an assumption, not an estimate; at zero offset, revenue at risk rises toward 5 percent.
Third, our baseline understates the true statutory base — a nontrivial share of registered workers report earnings consistent with wage under-declaration, a well-documented Turkish phenomenon — which makes the percentages here conservative.
If not labor, then what?
The menu is familiar and every item has a catch, which is precisely why the transition should start before it is forced:
- Consumption taxes are robust to factor-share shifts but regressive, and Türkiye already leans on VAT heavily.
- Capital income taxation is the theoretically clean answer and the practically hardest one — mobile, deferrable, and planned around. Though it is worth noting that AI rents may prove less mobile than textbook capital: they are embedded in data, market access, and compute located somewhere.
- Reducing the wedge itself, financed by base-broadening elsewhere, attacks the accelerant problem directly — making formal labor cheaper relative to its automated substitute while the substitution margin is still contestable.
- Sovereign participation in AI rents — equity stakes, windfall structures, spectrum-style auctions for compute or data rights — remains exotic but is the only item on the list whose base grows with the disruption rather than despite it.
None of this is an argument that the sky is falling. It is an argument that a fiscal system built in the age of the payroll should be stress-tested against the age of the model — and that the stress test should use microdata, not vibes. On Türkiye's numbers, the vulnerable fifth of the workforce carries a quarter of the labor tax base. That is a fact worth planning around.
Comments
Post a Comment
Constructive feedback is always welcome. Thank you