According to Schweser: The historical method and Monte Carlo Simulation both suffer from modeling risk. How does historical method suffer from model risk. You are just basically collection historical data as opposed to MC where you are inputting various assumptions to create a model.
I agree, no modeling risk is inherent in historical VAR. Perhaps they were referring to historical simulation where you use parameters from a portfolio simulated to mimick current situation, hence eliminate historical VAR’s disadvantage where past returns distribution does not necessarily reflect future returns distribution. Long short…I think it’s misleading
isn’t historical VAR are based on input? The output of those inputs are as good as the input so if you are taking incorrect inputs, your output will be incorrect too and hence modeling is gone wrong.
Well, the inputs are historical data, so there really isnt assumption as such being inputed. You can’t change the output much, because there is only one set of inputs. You look at the data and select the VAR from it. For me, that is a big difference from constructing a model from scratch with various assumptions.
May be because you can take different time periods to calculate historical VAR.
Hmm…that could be a point. Oh well…
because you are basing your calculations on historical data, the inherent assumption is that there is some basis in the past to explain or predict the future. HUGE risk as we saw in the subprime crisis, where people failed to stress their models adequately because they only considered the past and assumed it would continue to the future
Yea…makes sense…thanks a lot guys…appreciate your comments.
Theres such a thing as “historical simulation”? If it’s historical, it’s no longer a simulation right? Did I miss this in the readings?