Turkey's Informality Tax

A general-equilibrium model of Turkey's dual labour market says the cost of taxing formal work doesn't show up where we usually look for it.

The bottomline:

• Making Turkey's transfer to the unemployed a third more generous raises unemployment from 8.6% to 9.2%. That is the honest cost, and it is not large.
• How you pay for it matters more than whether you pay for it. Funded by payroll taxes, the poor end up 0.6% worse off than before the transfer was raised — the policy defeats itself. Funded by VAT, they are 0.5% better off.
• The reason is not unemployment, which is nearly identical under both. It is informality. A higher payroll tax pushes formal jobs into the unregistered sector, where the wage is 43% lower and nothing is taxed.
• Turkey is not on the wrong side of the payroll-tax Laffer curve — revenue peaks around 53%, well above today's ~37.5% wedge. But the marginal cost of those funds is high enough that a broader base wins anyway.
• VAT is not free either. It leaks into informality too, just less steeply per lira raised.

1. The thing that makes Turkey different

Most macro models of unemployment insurance ask a single question: how much does a more generous benefit weaken the incentive to take a job? In an economy with one labour market, that is the whole story.

Turkey has two. Roughly a quarter of Turkish workers — about 27% by TurkStat's unregistered-employment measure — work outside the formal system entirely. They pay no payroll tax, generate no social security record, and are invisible to most of the policy levers a finance ministry has. That sector is not a rounding error or a measurement problem. It is a live outside option that every formal worker and every formal firm is bargaining against.

This changes the question. When Turkey raises the tax on formal labour, workers don't simply choose between a formal job and idleness. They choose between a formal job, an informal job, and searching. The cost of the tax gets paid, but it can be paid in three different currencies — unemployment, informality, or lost output — and the headline unemployment rate only shows you one of them.

The model below has all three margins. The main finding is that Turkey pays mostly in the two you can't see in the unemployment statistics.

2. What's in the model

It is a two-agent New Keynesian structure with search-and-matching in the formal sector, solved for its steady state. The pieces that matter:

• A coherent labour force. Every worker is formally employed, informally employed, or unemployed and searching: . So u is a rate you can compare to TurkStat's.
• Informality as a choice. A worker not in a formal job is indifferent between informal work and searching. That indifference — the informal wage must compensate for giving up the option value of search — is what pins down how big the informal sector is:

where rr·w_f is the transfer to the unemployed, h the flow value of home production, p_find the quarterly job-finding rate and S_W the worker's share of the match surplus.

• Wage bargaining with a tax wedge. Firms post vacancies until the expected profit covers the cost; the wage splits the match surplus. Because the worker's income is taxed and the outside option isn't, the tax drives a wedge:

where η is worker bargaining power and mpl_f the marginal product of formal labour. Raising τ_w raises the pre-tax wage a firm must pay, which is the channel through which the payroll tax destroys formal jobs.

• VAT that gets evaded. Households buy a CES bundle of formal and informal goods, and VAT applies only to the formal one. This is the single most important modelling choice in the post. If you tax all consumption uniformly, VAT cannot touch the labour market at all, and any finding that "VAT saves jobs" is an artefact of the assumption rather than a result.
• Two household types. Savers own the capital stock and absorb the fiscal residual; spenders are hand-to-mouth. Half and half.

One structural fact worth stating up front. The formal sector has constant returns and the interest rate is pinned by the discount factor, so the formal marginal product of labour is constant. That means VAT moves the allocation of labour without moving wages or market tightness at all — it shifts workers between formal work, informal work and unemployment while leaving what each is paid untouched. The payroll tax moves both. Every VAT number below should be read as a quantity effect.

3. Does it look like Turkey?

Five parameters — matching efficiency, vacancy cost, informal productivity, the consumption weight on formal goods, and the value of home production — are solved for jointly with the model so that it reproduces five Turkish moments. Everything else is set from the literature and left alone.

Moment	Target	Model	Status
Unemployment rate	8.6%	8.60%	targeted
Informal share of employment	27%	27.00%	targeted
Formal / informal wage gap	1.75×	1.75×	targeted
Market tightness (v/u)	1.00	1.00	normalisation
Relative price, informal good	1.00	1.00	normalisation
Tax revenue / GDP	~31%	31.3%	not targeted
Quarterly job-finding rate	—	54.3%	not targeted
Vacancy costs / formal output	—	3.9%	not targeted

The revenue-to-GDP ratio is the useful one: nothing in the calibration aims at it, and it lands at 31.3% against a Turkish figure in the low thirties. The job-finding rate and the recruiting-cost share are also in the range the search literature reports.

A number that will look wrong, and isn't. The model puts informal activity at about 10% of GDP while 27% of workers are informal. That is not an inconsistency: informal work is low-productivity and uses no capital, so its value-added share is far below its employment share. It is also not comparable to the "shadow economy is 30% of Turkish GDP" estimates you may have seen, which measure a different object — those include under-reporting by registered firms, which this model has no channel for.

4. The experiment: a third more generous

Turkey's unemployment insurance replaces about 40% of prior earnings, but reaches only a small minority of the unemployed — contribution-history requirements exclude most. Combining the two, the model's baseline transfer is 15% of the formal wage, averaged over everyone out of work. The experiment raises it to 20%: a third more generous, and roughly what it would take to move Turkey toward the coverage its OECD peers have.

Left unfunded — with savers absorbing the cost through a lump-sum residual — the effect is exactly what the textbook predicts, and it is modest:

• Unemployment rises from 8.60% to 9.20%, six-tenths of a point.
• The poor gain +0.98% in real consumption.
• GDP falls 0.23%.

A rich country would take that trade without much argument. But Turkey is running a consolidation, so the interesting question is the next one.

5. Who pays?

Figure 1. The same policy, paid for three ways. The unemployment panel barely moves — which is exactly the point. The cost of the financing choice shows up in output and in the consumption of the households the policy was meant to help.

Closing the budget with the payroll tax requires raising the wedge from 37.5% to 38.6%. Closing it with VAT requires going from 20% to 20.6%. Both are small changes. They do not produce small differences.

Scenario	Unemployment	Informal share	Real GDP	Consumption of the poor
Baseline	8.60%	27.00%	—	—
Unfunded (savers absorb)	9.20%	26.64%	−0.23%	+0.98%
Payroll tax 37.5% → 38.6%	9.19%	27.25%	−0.91%	−0.60%
VAT 20.0% → 20.6%	9.18%	26.82%	−0.35%	+0.52%

Read the unemployment column first. Three very different fiscal policies; three indistinguishable unemployment rates. If you were evaluating this reform by watching the unemployment print, you would conclude the financing choice didn't matter.

It matters enormously. Under payroll-tax financing the poor end up worse off than before the transfer was raised — the extra benefit is more than eaten by the higher tax on the formal wages that most of them earn, and by the shift of jobs into the low-paying informal sector. The policy defeats its own purpose. Under VAT financing, most of the gain survives.

GDP tells the same story with a factor of nearly three between the two instruments. The payroll tax costs 0.91% of output to raise the same revenue that VAT raises for 0.35%.

6. Two Laffer curves

The standard version of this argument is that Turkey is at or past the peak of its payroll-tax Laffer curve. In this model it isn't — and the more careful finding is more useful than the dramatic one.

Figure 2. Revenue (top) and informality (bottom) for each instrument, on shared scales. Turkey sits well short of the payroll-tax revenue peak — but look at the bottom-left panel to see what getting there would cost.

Total revenue peaks at a payroll wedge of about 53%. Turkey's ~37.5% is comfortably on the right side. So the claim that Turkey cannot raise more from payroll taxes is false: it can.

The bottom-left panel is the reason it shouldn't. Walking the payroll tax from 5% to 70% takes informal employment from 15% to 56% of the workforce. The tax base doesn't disappear because people stop working. It disappears because they stop working formally. Every point of extra revenue past here is bought by moving another slice of the economy beyond the reach of the state — including beyond the reach of the social insurance the revenue is meant to fund.

The right-hand column is the honest case for VAT, and it is a qualified one. VAT revenue never turns over in the plausible range, so there is no fiscal cliff. But the bottom-right panel shows VAT leaking into informality too, from 21% to 38% across the sweep. Taxing formal consumption when informal consumption escapes is still a tax on being formal. It is simply a flatter one.

7. How much of this survives a different model?

The result rests on two parameters chosen without much of a Turkish empirical anchor: the elasticity of substitution between formal and informal goods (σ), and worker bargaining power (η). Varying both:

Specification	Payroll tax needed	VAT needed	Poor, payroll	Poor, VAT
Baseline (σ=2, η=0.5)	38.63%	20.60%	−0.60%	+0.52%
σ = 1.5	38.59%	20.63%	−0.53%	+0.49%
σ = 3	38.76%	20.53%	−0.78%	+0.58%
σ = 4	38.99%	20.46%	−1.11%	+0.64%
η = 0.35	37.99%	20.27%	+0.73%	+1.19%
η = 0.65	39.04%	20.79%	−1.44%	+0.14%

The ranking never flips. VAT financing leaves the poor better off than payroll-tax financing in every specification, and costs less output in every specification. The magnitude moves a lot, and the sign of the payroll-tax effect is not robust: at low bargaining power (η=0.35) workers absorb less of the tax and the payroll-financed policy still leaves them slightly ahead. So "payroll financing actively hurts the poor" is a parameter-dependent claim. "Payroll financing is the worse of the two" is not.

8. What this doesn't show

Four limits worth being explicit about, because each one could change a policy conclusion:

• VAT regressivity is understated. Both household types consume the same CES bundle, so the model misses the fact that poorer households spend a larger share of income on necessities. A real VAT increase would hit the poor harder than this says. The ranking would narrow; whether it flips depends on how the increase is structured.
• Enforcement is fixed. The informal sector's size responds to taxes but not to audit intensity, penalties, or formalisation incentives. Since those are the actual policy levers Turkey has been pulling, the model can't evaluate the most interesting alternative — raising the payroll tax and tightening enforcement.
• It is a steady-state comparison. There is no transition path, so nothing here speaks to timing, and the short-run effects of a VAT increase on inflation — which in Turkey's current situation is not a side issue — are entirely absent.
• No participation margin. The labour force is fixed at one. Turkey's low female participation rate is arguably the largest single distortion in its labour market, and the model cannot see it. Taxes and benefits that would move people in or out of the labour force show up here only as movements between the three states of those already in it.

9. What follows:

1. Fund social protection from the broad base, not the payroll. The difference is not marginal: it is the difference between a transfer that reaches the poor and one that is cancelled by the tax raised to pay for it.
2. Stop treating the unemployment rate as the scoreboard. The most expensive consequence of Turkish labour taxation is invisible in it. Informality and formal output are where the cost lands, and both should be reported alongside any reform's employment effects.
3. Don't oversell the fiscal cliff. Turkey has room on the payroll-tax Laffer curve. The argument against using it is the marginal cost of those funds, not insolvency — and that is a stronger argument, because it holds regardless of how close to the peak you think the country is.
4. Pair any VAT increase with a targeted offset. VAT wins on efficiency in this model, and the model understates its distributional cost. Those two facts are compatible with a VAT increase being right and with it needing a compensating transfer.

10. Reproducing this

The model, calibration and every number above are in the repository. It solves in a couple of seconds with scipy; there is no solver licence to buy. calibrate_to_turkey.py reproduces the fit table, financing_comparison.py Figure 1, and laffer_curves.py Figure 2. Each writes a CSV alongside its chart, so the tables here can be regenerated rather than trusted.

On the data. Calibration targets are approximate 2024–25 figures: the unemployment rate and unregistered-employment share from TurkStat's labour force survey, the tax wedge from OECD Taxing Wages, the 20% standard VAT rate in force since July 2023, and the formal–informal wage gap from survey estimates. Check them against current releases before quoting the model's results.

Citation. Copy verbatim:

Nxumalo, Mpumelelo. 2026. Turkey's Informality Tax: Social Protection and the Marginal Cost of Public Funds.