Defining Portfolio Loss for VaR with Contributions and Rebalancing
Summary
The document addresses how to estimate Value at Risk for a portfolio that receives regular contributions and is rebalanced between a risky asset and a risk-free bond. Its central distinction is between external cash flows and investment performance: deposits and withdrawals are not trading profit or loss, so they should not be included in the loss measure.
For a historical scenario, apply that period’s asset returns to the portfolio value at the start of the period using the chosen allocation, then define profit or loss as the resulting value minus the starting value. Rank these scenario outcomes to select the tail corresponding to the desired confidence level. The answer assumes rebalancing costs are immaterial and describes a one-period historical replay, rather than giving a complete procedure for estimating the distribution of terminal wealth over the full investment horizon. The loss sign convention and scaling should be kept consistent with the VaR convention used in the analysis.
Key ideas
- Portfolio contributions and withdrawals are external cash flows, not trading profit or loss.
- Measure each scenario’s performance by comparing the resulting portfolio value with its starting value.
- Apply historical returns to the chosen allocation to construct scenario outcomes.
- Rank the resulting profits or losses to identify the desired VaR tail.
- The suggested calculation assumes rebalancing costs are immaterial.
Tags
Full text
# Expected loss of investment over 30 years
# Expected loss of investment over 30 years
Let $V_{30}$ denote the value of my portfolio after 30 years. Each year, I add 1000\$ and rebalance my portfolio such that invest $c_1$ in a risky asset and $c_2$ in a risk-free bond. Denote the return of the risky asset over the $k$ year as $R_{k+1}$ and that the risk-free interest is $r$. The value process looks like this:
\begin{equation} \begin{split} V_0 &= 0 \\ V_{k+1} &= (V_k + 1000)(c_1R_{k+1} + c_2 e^r) \end{split} \end{equation}
Now I want to calculate the empirical Value at Risk over 30 years. How do I define the "loss" $X$?
In my textbook, over 1 period, we set $X = V_1 - V_0 R_0$ and $L = -X/L_0$. Given a sample of losses $\{L_1, \dots, L_n \}$, the empirical estimate of $\text{VaR}_p(X)$ is given by $\hat{\text{VaR}_p} = L_{[np]+1,n}$, where $L_{1,n} \geq L{2,n} \geq \dots \geq L_{n,n}$ is the ordered sample.
My question is, from a sample $\{V_{30}^1, \dots V_{30}^n \}$, define the loss $X$?
regards
## Answer by Dimitri Vulis (score 1)
https://quant.stackexchange.com/a/77431
The amounts that you add or withdraw are not part of your trading profit and loss, and shouldn't affect the calculation.
Assuming that the cost of rebalancing is immaterial,
If you have some amount $V_l$, and invest $c_{l,1}$ in the risky asset and $c_{l,2}$ in the riskless asset, and the market behaved from time $l$ until time $l+1$ like it did in the historical period from time $k$ until time $k+1$, then the new portfolio value would be $V_{k,l+1}=V_l \left(c_{l,1}\left(1+R_{k+1}\right) + c_{k,2} \left(1+e^{r_k}\right)\right)$, and the profit or loss would be $X_k=V_{k,l+1}-V_l$, you can rank those $X_k$'s, and look for the desired VaR's confidence level.Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.