Discussion

Our analysis reveals that, despite the theoretical double distortion mechanism, income-based fines generate higher welfare than flat fines under realistic US calibration. The deterrence equity gain from income-proportional penalties exceeds the labor distortion cost in the large majority of parameter configurations. This section discusses the sensitivity of our results to modeling choices, the role of institutional context, and implications for policy design.

Backward-looking income assessment¶

The most important institutional qualification concerns the timing of income measurement. Finland’s day-fine system calculates penalties from previous-year tax returns, not current income Kaila (2024). US systems vary: San Francisco’s pilot San Francisco Municipal Transportation Agency (2025) may use more contemporaneous income data, as would be feasible with real-time income verification through IRS data sharing. The Staten Island day-fine experiment Hillsman (1990) used recent pay stubs to assess income at the time of sentencing.

When fines are backward-looking, the contemporaneous link between current labor effort and fine liability is severed, and the effective marginal tax rate on current labor is $\text{MTR}_i$ regardless of speeding behavior. This design feature substantially weakens the double distortion mechanism. However, it does not eliminate it entirely. Forward-looking agents who anticipate future fines will internalize that higher future income implies higher future fines, generating a dynamic version of the labor supply distortion. The strength of this channel depends on the persistence of income shocks, the frequency of violations, and discount rates. Kaila (2024) finds limited evidence of labor supply responses in the Finnish system, consistent with the attenuation we would expect from backward-looking assessment.

A contemporaneous system---where fines are linked to current-year income---would exhibit the full force of the double distortion. As US jurisdictions design income-based fine programs, the choice between backward-looking and contemporaneous assessment is a critical design parameter that directly affects the magnitude of the labor distortion we identify.

Sensitivity to the labor supply elasticity¶

Our results are most sensitive to the labor supply elasticity, which governs the responsiveness of work effort to the effective marginal tax rate. The quadratic labor disutility specification in our model implies a Frisch elasticity of 1.0 (see Section 4), substantially higher than the meta-analytic consensus of approximately 0.25 from Chetty (2012) and Saez et al. (2012). Despite operating at this elevated elasticity---which amplifies the labor distortion channel by a factor of roughly four relative to the empirical estimate---income-based fines still dominate in 95% of draws. At the consensus elasticity of 0.25, the labor distortion would be correspondingly smaller, and the welfare advantage of income-based fines would be even more pronounced.

At the intensive margin, elasticities for prime-age men are typically 0.1--0.3 Keane (2011). At the extensive margin (participation), elasticities can be substantially larger, particularly for secondary earners and low-income workers. If income-based fines push some agents below their participation threshold---inducing them to exit the labor force entirely---the efficiency cost could be larger than our continuous model suggests, potentially narrowing or reversing the welfare advantage of income-based fines.

The elasticity also varies across institutional settings and income levels within the US. Workers in the EITC phase-out range may exhibit different behavioral responses than high-income earners, because the EITC phase-out already creates strong implicit taxes that interact with the fine-induced distortion. The US labor market, with weaker employment protections and more flexible hours than many European systems, may also permit larger intensive-margin responses, which would strengthen the double distortion and narrow the welfare gap.

The EITC phase-out and the double distortion¶

A distinctive feature of the US calibration is the interaction between income-based fines and the Earned Income Tax Credit phase-out. Workers earning roughly $20,000--$50,000 face effective marginal tax rates near 40% due to the combined effect of federal income tax, FICA payroll taxes, and the EITC phase-out Maag et al. (2012). Adding an income-based fine on top of these already-high rates creates particularly large deadweight loss because of the quadratic relationship between tax rates and efficiency costs Harberger (1964).

This interaction is especially concerning from an equity perspective. The EITC phase-out region contains many working-poor families---precisely the population that income-based fines are intended to help. While income-based fines reduce the level of the fine for these workers (relative to a flat fine), they increase the marginal tax rate on labor, potentially discouraging additional work effort. The net welfare effect depends on the relative magnitudes of the fine reduction and the labor distortion, which our Monte Carlo analysis quantifies.

Inequality and social preferences¶

The welfare ranking of fine systems depends on how society values equality, but with realistic US calibration the ranking favors income-based fines even under utilitarian preferences. Income-based fines dominate in 95% of Monte Carlo draws without any inequality aversion, and the advantage strengthens under more egalitarian social welfare functions.

Saez & Stantcheva (2016) provide a framework for incorporating diverse social preferences through generalized social marginal welfare weights. In their framework, the optimal degree of income-basedness depends on the weight society places on equity versus efficiency---a normative choice that our model can inform but not resolve. Our contribution is to show that, at empirically calibrated parameters, income-based fines are welfare-improving even before accounting for inequality aversion. The efficiency cost of the implicit labor tax exists but is quantitatively small relative to the deterrence equity gain.

This finding should not be interpreted as meaning the double distortion is irrelevant. At higher labor supply elasticities, higher fine rates, or with contemporaneous income assessment, the labor distortion channel would be larger. The relevant policy question is whether the efficiency cost is large enough to overcome the equity and deterrence benefits---and our calibration suggests it is not, at least under the baseline parameterization.

The value of targeted deterrence¶

Income-based fines are motivated by the observation that flat fines under-deter the wealthy and over-deter the poor. Our model captures this asymmetry: under flat fines, high-income agents speed more because the fine represents a smaller fraction of their consumption. Income-based fines equalize the deterrence margin across the income distribution, and our results confirm that this equalization generates a welfare gain of $\Delta W = 0.83$ (positive in 95% of draws).

The welfare value of this equalization depends on the curvature of the death probability function. Under the power model with $n = 4$, speeding carries sharply increasing risk, and under-deterrence of high-income agents imposes significant costs (through higher aggregate speeding and death probability). Our model focuses on private mortality risk; incorporating external harm to other road users would further strengthen the case for income-based fines by increasing the social value of deterrence.

Alternative policy instruments¶

Although our results favor income-based fines, alternative policy instruments may achieve similar objectives with different trade-offs.

One approach is payment flexibility: flat fines combined with income-contingent payment plans can address the regressivity concern without creating labor supply distortions. The fine amount remains flat, preserving the labor supply incentives, while the payment schedule accommodates liquidity constraints. Several US jurisdictions have adopted such systems, and the San Francisco pilot includes payment plan provisions alongside its income-scaling mechanism San Francisco Municipal Transportation Agency (2025).

Many US courts already offer community service as an alternative to monetary fines for defendants who cannot pay. This avoids the regressive burden of flat fines while not conditioning on income in a way that distorts labor supply. However, community service imposes time costs that may be more burdensome for low-income workers with less flexible schedules.

Non-monetary sanctions---point-based systems, license suspensions, and mandatory traffic safety courses---create penalties that are less directly tied to income. Bourgeon & Picard (2007) analyze point-record systems and show they can achieve effective deterrence through non-monetary channels. However, non-monetary sanctions may have their own distributional consequences if time costs or license dependency vary with income.

Finally, rather than linking fines to income, flat fines can be paired with enhanced transfers to low-income households. Kaplow & Shavell (2002) advocate separating the pricing function (deterrence through fines) from the redistributive function (transfers through the tax system). This approach avoids the implicit tax on labor while achieving distributional goals through a more efficient instrument. In the US context, expanding the EITC or other targeted transfers could offset the regressive impact of flat fines without distorting the fine-labor supply link.

Limitations¶

Our analysis rests on several simplifying assumptions that merit acknowledgment.

The model analyzes a single-period decision, abstracting from reputation effects, learning, and habit formation. Speeding behavior is likely persistent, and fines may have dynamic deterrent effects that our static model misses. Within each Monte Carlo draw, agents share common preference parameters ($\alpha$, $\beta$), differing only in wages and marginal tax rates. Heterogeneity in risk attitudes, time preferences, or driving needs could affect optimal fine design; for instance, agents who must drive for work face different trade-offs than recreational drivers.

The model bases income-based fines on gross employment income ($y_i = w_i h_i$), whereas real-world systems vary in their income concept. Finland’s day-fine system uses monthly net income minus a fixed deduction Kaila (2024); San Francisco’s pilot may use adjusted gross income or another tax-return-based measure. Using gross income in the model amplifies the double distortion because the fine compounds with existing taxes---the effective rate is $\text{MTR}_i + \phi s$ on gross income, rather than $\phi s$ applied to after-tax income. If a jurisdiction based fines on net (after-tax) income, the fine would not create an additional labor supply distortion beyond what the income tax already imposes, since earning more after-tax income would already reflect the agent’s marginal rate. The choice of income concept is therefore a consequential design parameter that affects both the magnitude of the labor distortion and administrative feasibility.

We assume uniform detection probability across income levels. If wealthy individuals can better avoid detection---through legal representation, choice of routes, or vehicle technology---the effective deterrence of income-based fines may differ from what our model predicts.

The budget constraint uses each agent’s marginal tax rate as a proportional rate, which is a linearization of the progressive tax schedule around the observed income level. This approximation overstates the average tax burden for agents in higher brackets (where the marginal rate exceeds the average rate) and understates it for agents on the EITC phase-in (where the marginal rate is negative). For the welfare comparison between fine systems, this bias is approximately symmetric: both fine regimes use the same linearized tax treatment, so the bias cancels to first order in the welfare difference $\Delta W = W_{\text{IB}} - W_{\text{flat}}$, which is our primary object of interest. The cancellation is not exact, because the behavioral responses to the two fine systems differ, and these responses interact differently with the linearization error. At empirically small fine rates ($\phi \approx 0.02$--0.05), the behavioral responses are small enough that this second-order interaction is negligible. The marginal rate is the correct object for analyzing agents’ behavioral responses at the margin, which drive the labor distortion channel.

While our use of PolicyEngine-computed marginal tax rates from CPS microdata is a substantial improvement over a single scalar tax rate, the CPS itself has known limitations. High incomes are top-coded at varying thresholds (roughly $200,000--$300,000 depending on year and state), compressing the upper tail of the income distribution where income-based fines have the largest bite. Some benefit variables (Supplemental Nutrition Assistance Program, Medicaid, housing assistance) are statistically imputed, and these imputations drive the benefit phase-out MTR spikes. The MTR estimates also assume current-law tax policy and do not capture informal economy participation or tax noncompliance.

The model treats labor supply as a continuous intensive-margin choice, with hours bounded below at $h \geq 1$. If income-based fines push some agents below their participation threshold---inducing them to exit the labor force entirely---the efficiency costs would be larger than we estimate. This is particularly relevant for low-income workers on the EITC phase-out, who face the highest baseline marginal rates and for whom the additional fine-induced tax wedge may be most consequential.

The model assumes that fines are proportional to speeding intensity rather than conditioned on discrete enforcement events. In practice, speeding fines are imposed only upon detection, with some probability $\pi < 1$. Introducing detection probability would rescale the effective fine: the optimal flat fine of $3,262 with $\pi = 0.05$ (a plausible estimate for speed cameras) implies an expected fine of approximately $163, close to the current US nominal average. This reinterpretation makes the optimal fine levels more plausible and connects to the Becker-Polinsky-Shavell framework Becker (1968)Polinsky & Shavell (1979), which emphasizes the probability-magnitude trade-off.

Broader implications¶

The double distortion mechanism applies whenever penalties are linked to economic productivity. Criminal day-fines, income-contingent environmental penalties, and means-tested regulatory sanctions all create the same implicit tax on earnings that we identify for traffic fines. The general principle is that linking any cost to income adds to the effective marginal tax rate, with efficiency consequences that coexist with equity and deterrence benefits.

This connects to a broader theme in public economics: the Tinbergen principle Tinbergen (1952), which holds that achieving $k$ policy objectives requires at least $k$ independent instruments. Income-based fines attempt to serve two objectives---deterrence and redistribution---with a single instrument. The Tinbergen principle would suggest that separating these functions---using flat fines for deterrence and the tax-transfer system for redistribution---would achieve both objectives more efficiently. Our calibration shows, however, that the efficiency cost of bundling deterrence and redistribution through income-based fines is quantitatively small enough that income-based fines still dominate flat fines. This represents a case where the single-instrument approach succeeds despite the theoretical prescription, because the implicit tax on labor is modest relative to the deterrence equity gain.

As the United States moves toward greater experimentation with income-based fines---following San Francisco’s 2025 pilot San Francisco Municipal Transportation Agency (2025) and building on earlier experiments like Staten Island Hillsman (1990)---the framework we develop here provides a principled basis for evaluating these programs. The key empirical inputs---the distribution of marginal tax rates across the income distribution, labor supply elasticities, and speeding behavior---are all measurable, making our welfare comparison empirically grounded rather than purely theoretical.