A Sober Look post from a couple of years ago discusses the issues associated with using rolling historical period Value at Risk (VAR). The approach has a highly pro-cyclical impact on bank capital. During periods of low volatility and generally high profitability for financial institutions, the historical VAR method allows firms to lower their capital usage. During periods of crisis and therefore high volatility, VAR forces banks to allocate higher capital for the same positions. As an example, for the same dollar equivalent position in a diversified portfolio of US equities, a bank would have to allocate 3 times as much capital in 2009 as it did in 2007. Yet in 2009 these equities were no "riskier" than they were in 2007 - in fact one might argue that the inverse is true.
Basel III attempts to correct this approach. The new paradigm for market risk capital is expected to more than double the current capital requirements (that's partially why EU banks will need to raise so much more capital). A large portion of this regulatory capital increase comes from SVAR - the "stressed value at risk". SVAR still uses the standard 99% confidence interval (one-tailed), 10-day holding period, and at least one year worth of data, but requires it to be "calibrated" to a stress period (such as 2008).
This is certainly an improvement on the existing methodology and should reduce the pro-cyclical nature of the approach. But since the Basel Committee did not specify the stress period, and in fact requires that banks consider multiple stress periods, the measurement is still open to interpretation. One for example could see a case where regional supervisors would choose different stress periods for the same asset classes. That would mean that the same asset held in two different jurisdictions could potentially require different capital requirements.
The broader issue here is that VAR, even if it's now SVAR, is still the preferred regulatory approach to capital requirements. Risk practitioners in the financial services industry (both on the "buy" and the "sell" side) have long preferred to apply stress testing/scenarios when analyzing risk for internal purposes. SVAR would not have captured the risks at many financial institutions going into 2008. Since regulatory capital requirements drove performance, in many instances internal stress scenarios were ignored. The gap between what should be a superior approach to measuring risk and what the BIS bureaucrats are using to determine capital requirements has not been closed. Aaron Brown at AQR put it well when he said:
"People have tried to change VAR in lots of ways to explore the tail, but VAR is really for the center risk. In the tails you don't have enough data, so you don't know if something is one in a million, one in ten thousand or one in a hundred. But auditors and regulators want precise tools for exploring the regions that you can't explore with precise tools. They want to take a protractor into the jungle – it just doesn't work."In addition to SVAR, the market risk framework under Basel III will also add capital charges from the incremental-risk charge (counterparty credit), a new asset securitization charge, and the so-called comprehensive-risk measure (more on these new rules later.) Risk "management" departments at banks will now grow more bloated (and probably even less effective), just to keep up with all this.