Controls
Presets
Parameters
Quality
Camera
Model checks
The science
The model. Each of 200 workers picks taxable income z to maximise (1-t) z - z0-1/e z1+1/e/(1+1/e) (consumption minus the effort cost of earning z). The first-order condition gives z = z0(1-t)e, so e is the elasticity of taxable income with respect to the net-of-tax rate 1-t (Feldstein 1995; empirical estimates for broad taxable income are typically in the range 0.2 to 0.7, with much lower values for labour hours alone, so the sliders go well beyond them for illustration).
The Laffer curve. Revenue per unit of untaxed income is R(t) = t (1-t)e. Its slope is (1-t)e-1(1 - t - e t), zero at t* = 1/(1+e). Above t* a lower rate raises revenue. The gap between mechanical revenue t and R(t) is the behavioural response. Any spending target G below R* can be financed by two rates, one on each side of the peak; the lower one is always better for welfare.
Welfare. With quasi-linear utility the excess burden (the welfare lost beyond the revenue itself) is DWL = [1 - (1-t)1+e]/(1+e) - t (1-t)e, equal to the integral of τ(-dz/dτ) (the envelope theorem), and approximately e t²/2 for small t (the Harberger triangle). The marginal cost of public funds is (1-t)/(1-t-e t), which diverges at the peak.
Simplifications. One flat tax rate, no income effects, no tax avoidance channels separate from e, no externalities, no general-equilibrium or wage effects, no optimal-tax redistribution (Mirrlees, Saez), and workers differ only by a lognormal productivity (sigma 0.5, mean 1). Real top-rate estimates depend on the base, the country and the horizon; the presets are illustrative. Checks: golden-section peak versus 1/(1+e); revenue versus the integral of marginal revenue; excess burden from indirect utility versus the integral of the lost-income formula; 60 workers maximising utility numerically versus the closed form; the elasticity recovered by finite differences; and both rates that raise G.