Pro-cyclical effects of VaR-based capital measures

[brian.field]

Can anyone help me understand what BIS means when they say "pro-cyclical effects?"

I see it referenced numerous times in the Messages from the Academic Literature on Risk Measuring for the Trading Books reading.

Thanks

Brian

 

[desaiha]

In business cycle terms; Any quantity that is positively correlated to the economy. For eg gdp is procyclical.

In reference to economic policy, it refers to any aspect of economic policy that magnifies economic or financial fluctuations.

 

[brian.field]

Thank you @desaiha.

If I am understanding you correctly, you are stating that Pro-Cyclical VaR measures would be positively correlated with economic indicators. So, if GDP rises, for instance, which would suggest an improving economy, then a Pro-Cyclical VaR would also rise, thus indicating a corresponding increase in risk? This is a bit unclear to me still....

A strong or improving economy might induce lower risk than a struggling economy no?

Yet, higher returns would correspond to a stronger economy and higher risk....and thus, higher VaRs. Hmm....I am confused.

 

[ShaktiRathore]

Hi Brian,
Yes it does seems counteintuitive. But yes procycle Var could increase during high growth period. We know Var=return*PortflioValue. During high Gdp period Portfolio value increases which increases the Var assuming return constant. So Var($) do seem to increases with Gdp.
However i am not sure of return Var(%).it could go either way,shall can improve too.
Thanks

 

[Thomas Obitz]

A procyclical measure is one which makes things worse when they are bad - and gives you a false sense of security when they are good.

In a nutshell - your VaR will be low under calm market conditions, because your volas are low, which go into the calculation. Now when the markets become more volatile, the VaR goes up - without you having done anything.

If you now react by reducing position sizes, you are creating additional volatility, which make other market participants close positions etc. Have a look at the VIX during the 2007/ 2008 period...

 

[ami44]

As I understood it, the VaR is countercyclical because rising asset prices will normally result in a lower VaR (assuming a portfolio long the assets). If you use historical simulation you will see less days with a loss and normally also less severe losses. Of course you can find counterexamples but in general that should hold.

In the last part of the article is the sentence "Since VaR per value of assets is countercyclical, it directly follows that leverage is procyclical ..."
here the VaR per asset value also decreases when asset prices rise, because the asset prices are in the denominator.

 

[Thomas Obitz]

Hi Ami,

Brian spoke of "pro-cyclical effects" - its use amplifies existing trends in both directions. This is quoted in other papers as well (e.g. below).

I agree on the effects of positive market returns.

Regards
Thomas

"Specifically, the 10-day VaR calculation did not adequately capture credit risk or market liquidity risks; incentivised banks to take on tail risk; inadequately captured basis risk and proved procyclical due to its reliance on relatively recent historical data. " BIS Fundamental Review of the Trading Book

 

[ami44]

Hallo Thomas,
you are right of course, the question was about procyclical VaR. I did not read carefully enough.

 

[ShaktiRathore]

I also got confused by Brians second post above didnt read his first post though,Thomas is right yes procyclical effects seems to make worse even more worse while good even more good. Due to this effect during worse economics when potential loss Var is already high it gets even worse when investors short positions creating more volatility in panic while during good economics with already low potential losses Var the low activity ad traders do not want to trade but keep long positions results in low volatility which further reduces Var. Due to these procyclical effects banks do keep buffer capital to face additional lossed due to these effects.
Thanks

 

[daweinzettl]

all those who are still confused about the term "pro-cyclical" the attached paper by T. Adrian & H.S. Shin (both are by far the most well known academics in this field) should help to alleviate any concerns.
Yes, Basel II was indeed heavily criticised for it's pro-cyclical nature, but in Basel III the focus is on "counter-cyclical" regulation.

 

[afterworkguinness]

@Thomas Obitz and @ShaktiRathore great answers. Can either of you explain the later bit of the section in question?

"The effects of time-varying volatility on the accuracy of simple VaR measures diminish as the time horizon lengthens. In contrast, volatility generated by stochastic jumps will diminish the accuracy of long-horizon VaR measures unless the VaR measures properly account for the jump features of the data.. Distinguishing between time-varying volatility and volatility changes that owe to stochastic jump process realizations can be important for VaR measurement."

As I understand it, a stochastic jump process is an immediate increase in a variable's value rather than a slow continuous one. If volatility in a model is increased due to a rapid increase in modeled prices, why does this have an effect on long-horizon VaRs ?

 

[ShaktiRathore]

Hi
Long term Var is less affected by time varying volatility as time lengthens because the time varying effects of volatility avrerages out so that the time varying volatility has its effect diminishes as time lenthgens, but spikes/jumps do not averages out over time these jumps can have sizeable effect due to their sheer magnitude,1 or more such jumps can certainly affect overall volatilty or Var.jumps can certainly effect volatility to high levels and certainly Var also as Var is itself dependef on volatility. Var(T)=z*volatility(T), thetefore as volatility is affected so slao is Var due to these jumps.
These jumps need to be taken into a/c otherwise they shall distort the overall volatilty to unreasonable/abnormal level which can distort the overall Var we forecasted over time horizon. As time horizon increases the probability and number of such jumps increases therefore increasing the inaccuracy of Var further.

Thanks

 

[afterworkguinness]

Thanks for the detailed reply. Is the reason the jumps don't average out because they are stochastic and not pro-cyclical?

 

[ShaktiRathore]

Yes jumps are stochastic are random shocks they shock prices to high levels contributing to higher vilatility,hence affects Var.

Thanks