VaR変化の要因分析
English Follows
もし、金利系商品を扱っているなら、まず、金利レベルの変化に注目。それは、方法は問わず、金利系の場合のvolatilityは、年率%換算のvolatility表示に、現在の金利レベルを掛けて計算されるからだ。
金利が1%近辺で5bp動いてもVaR変動は5%だが、金利が5bp近辺の場合、5bp動くとVaRの変動は100%(2倍)になってしまう。
VaRの計算方法にはいくつかある。また、対象商品によって計算方法が若干異なる。ざっくり言って、
- VCVの場合
VaR
= BPV X 2.33(C.I.99%) X Vola(%pa) / sqrt(240) X Current MarketLevel(bp)
Equityの場合
VaR = Position X 2.33(C.I.99%) X Vola(%pa) / sqrt(240)
- Historicalの場合
VaR =
Percentile ( BPV X HistoricalDailyChange(%) X Current MarketLevel(bp) )
Equityの場合
VaR = Percentile (Current Position amt X HistoricalDailyChange(%) )
- MonteCarloの場合
VaR = BPV X SimulationChange(%) X Current MarketLevel(bp)
Equityの場合
VaR = Position X SimulationChange(%)
VaR = Position X SimulationChange(%)
1)為替変動がベース通貨に対し再評価による変動を与えていないか?
2)Historicalの場合、シナリオの変化で説明できないか?
3)マーケットデータ採取は適切か?Bidしかない場合、Bidを使っていないか?
VaR Change Attribute Analysis
If you deal with fixed income products, check the level change of interest rate 1st. It is because, irrespective of methodologies, the volatility of fixed income products is calculated as a product of interest rate level and annualized % based volatility quote.
If interest rate of the market is around 1% and it increased 5bp, then VaR will increase 5% however, if interest rate is around 5bp and it increased 5bp, VaR will increase for 100%(doubles).
There are couple of approach to calculate VaR. Calculation method will differ a bit by products. In nutshell, roughly,
- VCV`s case
VaR
= BPV X 2.33(C.I.99%) X Vola(%pa) / sqrt(240) X Current MarketLevel(bp)
Equity's case
VaR = Position X 2.33(C.I.99%) X Vola(%pa) / sqrt(240)
- Historical's case
VaR =
Percentile ( BPV X HistoricalDailyChange(%) X Current MarketLevel(bp) )
Equity's case
VaR = Percentile (Current Position amt X HistoricalDailyChange(%) )
- MonteCarlo's case
VaR = BPV X SimulationChange(%) X Current MarketLevel(bp)
Equity's case
VaR = Position X SimulationChange(%)
VaR = Position X SimulationChange(%)
1) FX change might have impact against denominating base currency used
2) Scenario transition might have impact when historical simulation is used
3) Check if market data is taken properly such as taking only bid when only bid is available etc.