Pre-Tax and After-Tax Capital Bases

Why the Tax Basis Matters

Economic capital depends on the loss measure being modeled. A pre-tax loss records the gross amount of the loss. An after-tax loss reflects any tax relief that the loss actually generates. Because those amounts can differ, capital calculated on one basis should not be compared directly with capital calculated on the other without first identifying the tax assumptions.

The key distinction is between a tax rate and a usable tax benefit. A deduction reduces taxes only when it can be used under the assumed tax rules and timing. For a simple model with immediately usable deductions and a constant tax rate, after-tax losses are a constant fraction of pre-tax losses. In practice, limits on deductions, insufficient taxable income, timing differences, or different tax treatment of different loss types can make the relationship nonlinear.

Definition: Let $X$ be a nonnegative pre-tax loss. The after-tax loss is the loss remaining after the tax benefit attributable to that loss has been applied. Under the simplifying assumption that every dollar of loss is immediately deductible and the full deduction can be used at tax rate $\tau$, where $0\leq\tau<1$, the after-tax loss is $X_{\text{AT}}=(1-\tau)X$. If only part of a loss generates usable tax relief, $X_{\text{AT}}$ must instead be calculated from that specific tax-benefit rule.

For a loss variable $X$ and confidence level $p$, define $\operatorname{VaR}_p(X)=\inf\{x:F_X(x)\geq p\}$. In this tutorial, VaR-based economic capital is $\operatorname{VaR}_p(X)-E[X]$. TVaR uses the integrated-quantile convention, $\operatorname{TVaR}_p(X)=\frac{1}{1-p}\int_p^1\operatorname{VaR}_u(X)\,du$, and TVaR-based economic capital is $\operatorname{TVaR}_p(X)-E[X]$. These are stated capital measures; an organization may separately impose a floor or another capital policy.

When Fully Usable Tax Relief Scales Capital

Assume the after-tax loss is $cX$, where $c=1-\tau>0$ is constant. This assumption requires the same effective tax rate for all modeled outcomes, immediate recognition of the tax benefit, and no cap or other constraint on its use. Under these conditions, both the loss distribution and its expected value scale by $c$.

\operatorname{VaR}_p(cX)=c\operatorname{VaR}_p(X),\qquad E[cX]=cE[X],\qquad \operatorname{EC}^{\operatorname{VaR}}_{\text{AT}}=c\operatorname{EC}^{\operatorname{VaR}}_{\text{pre-tax}}.

The VaR identity follows from the quantile definition. Since $c>0$, $cX\leq y$ is equivalent to $X\leq y/c$. Therefore $F_{cX}(y)=F_X(y/c)$, and the smallest $y$ for which $F_{cX}(y)\geq p$ is $c$ times the smallest $x$ for which $F_X(x)\geq p$. Linearity of expectation gives $E[cX]=cE[X]$. Subtracting the scaled mean from the scaled VaR proves the capital identity.

The same scaling result applies to integrated-quantile TVaR: every quantile in the defining integral is multiplied by $c$, so the integral and the mean both scale by $c$. It does not follow that tax always reduces capital by precisely the tax rate. That conclusion depends on the proportional after-tax loss assumption, not merely on the existence of a stated tax rate.

Worked Example: VaR Capital With Fully Usable Deductions

A hypothetical insurer models a one-year pre-tax loss $X$ as follows: $X=0$ with probability $0.70$, $X=\$80$ with probability $0.20$, and $X=\$240$ with probability $0.10$. Every loss is immediately deductible at a tax rate of 25%, with no limit on use. Calculate pre-tax and after-tax 90% VaR-based economic capital.

Method. Calculate the left quantile from the cumulative distribution, then subtract the expected loss. Apply the tax transformation to each outcome before calculating after-tax capital. This avoids treating a probability exactly at the confidence level as if it required moving to the next outcome.

Pre-tax mean. $E[X]=0(0.70)+80(0.20)+240(0.10)=0+16+24=\$40$.

Pre-tax 90% VaR. The cumulative probabilities are $F_X(0)=0.70$, $F_X(80)=0.90$, and $F_X(240)=1.00$. Thus $\operatorname{VaR}_{0.90}(X)=\$80$, including because the cumulative probability reaches 0.90 at \$80. Pre-tax capital is $80-40=\$40$.

After-tax outcomes and mean. Here $c=1-0.25=0.75$, so the after-tax outcomes are $0$, $80(0.75)=\$60$, and $240(0.75)=\$180$, with the same probabilities. Their mean is $0(0.70)+60(0.20)+180(0.10)=12+18=\$30$.

After-tax VaR and capital. The cumulative probability at \$60 is still 0.90, so the after-tax 90% VaR is \$60. After-tax capital is $60-30=\$30$. As a check, $0.75(\$40)=\$30$, as the proportional-scaling result predicts. All amounts are exact to the cent.

TVaR and Discrete Losses

For a discrete distribution, TVaR at confidence level $p$ averages the worst $1-p$ fraction of the distribution. If the probability mass at the VaR outcome straddles the boundary of that tail, only the necessary fraction of that mass is included. This is why simply taking the conditional mean given $X\geq\operatorname{VaR}_p(X)$ can give the wrong answer: that conditioning may include more than the worst $1-p$ probability.

Worked Example: TVaR Capital With a Boundary Mass

A hypothetical equipment portfolio has one-year pre-tax losses of \$0 with probability 0.90, \$100 with probability 0.08, and \$500 with probability 0.02. All losses are immediately deductible at 20%, with no limit on use. Calculate pre-tax and after-tax 95% TVaR-based economic capital.

Method. Find the worst 5% of probability, including only the required portion of any outcome whose probability straddles the tail boundary. This avoids incorrectly averaging all outcomes at or above the VaR.

Pre-tax mean and VaR. The mean is $E[X]=0(0.90)+100(0.08)+500(0.02)=8+10=\$18$. The cumulative probability at \$0 is 0.90 and at \$100 is 0.98, so $\operatorname{VaR}_{0.95}(X)=\$100$.

Pre-tax TVaR. The worst 5% consists of all of the 2% probability at \$500 and 3% of the 8% probability at \$100. Thus $\operatorname{TVaR}_{0.95}(X)=\frac{0.03(100)+0.02(500)}{0.05}=\frac{3+10}{0.05}=\$260$. Pre-tax TVaR capital is $260-18=\$242$.

After-tax calculation. The multiplier is $1-0.20=0.80$, giving outcomes \$0, \$80, and \$400. The after-tax mean is $0(0.90)+80(0.08)+400(0.02)=6.40+8.00=\$14.40$. The worst 5% comprises 3% at \$80 and 2% at \$400, so after-tax TVaR is $\frac{0.03(80)+0.02(400)}{0.05}=\frac{2.40+8.00}{0.05}=\$208$. After-tax TVaR capital is $208-14.40=\$193.60$.

Check. $0.80(\$242)=\$193.60$, consistent with fully usable, proportional tax relief. All calculations are exact to the cent.

When Tax Relief Is Limited

A constant tax multiplier is not appropriate if only some of a loss can generate tax relief. One simple hypothetical rule is that a loss deduction is usable only up to a fixed amount $K$. At tax rate $\tau$, the tax benefit is then $\tau\min(X,K)$, and the after-tax loss is $g(X)=X-\tau\min(X,K)$. For $X\leq K$, this equals $(1-\tau)X$; for $X>K$, it equals $X-\tau K$. The marginal tax reduction therefore stops once the loss exceeds the cap.

This particular $g$ is increasing: its slope is $1-\tau>0$ below $K$, and 1 above $K$. Consequently, its quantiles can be found by applying $g$ to the corresponding loss outcomes. But the mean generally does not scale by a single multiplier, so the capital cannot be obtained by multiplying pre-tax capital by $1-\tau$. Other tax rules may require recalculating the full transformed distribution, including its quantiles.

Worked Example: Capital With a Capped Tax Benefit

A hypothetical firm models one-year pre-tax losses as \$0 with probability 0.60, \$100 with probability 0.30, and \$300 with probability 0.10. The tax rate is 30%, but the firm can use deductions against taxable income for at most \$150 of loss. Assume immediate use up to that cap. Calculate pre-tax and after-tax 80% VaR-based economic capital.

Method. Transform each loss using $g(X)=X-0.30\min(X,150)$, then calculate the after-tax mean and quantile directly. This avoids assuming that the tax rate proportionally scales every loss.

Pre-tax capital. The mean is $E[X]=0(0.60)+100(0.30)+300(0.10)=30+30=\$60$. The cumulative probability is 0.60 at \$0 and 0.90 at \$100, so the 80% VaR is \$100. Pre-tax capital is $100-60=\$40$.

After-tax outcomes. At $X=0$, the tax benefit is $0.30\min(0,150)=0$, leaving \$0. At $X=100$, it is $0.30(100)=\$30$, leaving \$70. At $X=300$, only \$150 is deductible, so the benefit is $0.30(150)=\$45$, leaving $300-45=\$255$.

After-tax capital. The mean is $0(0.60)+70(0.30)+255(0.10)=21+25.50=\$46.50$. Since the cumulative probability at \$70 is 0.90, the 80% VaR is \$70. After-tax capital is $70-46.50=\$23.50$. A proportional calculation would give $0.70(\$40)=\$28$, which is not the result because the deduction is capped. All amounts are exact to the cent.

Basis Checks and Common Traps

A capital figure is meaningful only when its loss basis, time horizon, risk measure, confidence level, and treatment of tax relief are clear. The pre-tax and after-tax calculations should use the same horizon and confidence level if the purpose is to isolate the effect of tax treatment. If tax relief is delayed, uncertain, or conditional on future taxable income, immediate full recognition is not justified unless the model explicitly assumes it.

Using the wrong discrete quantile. For the first example, the cumulative probability reaches exactly 90% at \$80. The 90% VaR is \$80, not \$240. Always apply the stated quantile definition before calculating capital.
Subtracting in the wrong order. Under the capital measure used here, capital is VaR minus the mean, not the mean minus VaR. If the result is negative, report it as negative unless a separately stated policy floors capital at zero.
Applying tax to capital without checking the loss transformation. Multiplying pre-tax capital by $1-\tau$ is justified when the entire after-tax loss distribution is $(1-\tau)X$. It is not justified merely because the tax rate is $\tau$.
Taking a conditional tail mean without checking the boundary. For discrete losses, TVaR includes exactly the worst $1-p$ probability under the integrated-quantile convention; it may use only part of the probability mass at the VaR outcome.

Key takeaway: First specify how tax converts each pre-tax loss into an after-tax loss. Then calculate the chosen risk measure and capital on that transformed distribution. Proportional capital scaling is a useful result under fully usable, immediate deductions at a constant rate, not a universal tax rule.

Check Your Understanding

For each problem, use VaR-based capital as VaR minus the mean unless TVaR is specified. Assume all amounts are in dollars and report capital to the cent.

A loss is \$0 with probability 0.50, \$50 with probability 0.40, and \$200 with probability 0.10. All losses are immediately deductible at 30%, with no cap. Find pre-tax and after-tax 90% VaR-based capital. (Answer: Pre-tax \$10.00; after-tax \$7.00)
A loss is \$0 with probability 0.92, \$120 with probability 0.06, and \$400 with probability 0.02. All losses are immediately deductible at 25%, with no cap. Find pre-tax and after-tax 95% VaR-based capital. (Answer: Pre-tax \$104.80; after-tax \$78.60)
A loss is \$0 with probability 0.80, \$60 with probability 0.15, and \$260 with probability 0.05. All losses are immediately deductible at 20%, with no cap. Find pre-tax and after-tax 90% TVaR-based capital using the integrated-quantile convention. (Answer: Pre-tax \$138.00; after-tax \$110.40)
A loss is \$0 with probability 0.50, \$80 with probability 0.30, and \$250 with probability 0.20. The tax rate is 25%, but only the first \$100 of loss can be deducted, with immediate use up to the cap. Find pre-tax and after-tax 80% VaR-based capital. (Answer: Pre-tax \$6.00; after-tax -\$3.00)
For a loss $X$, deductions are fully usable immediately at a constant 18% rate. If the pre-tax VaR-based capital is \$75, calculate after-tax capital and state the assumption that permits the calculation without re-estimating the distribution. (Answer: \$61.50; the after-tax loss must equal $0.82X$ for every outcome.)

What You'll Learn