Market Risk: Value at Risk and Stop Loss Policies

Prof. Giulio Tagliavini

University of Parma - Istituto di Credito, Finanza ed Assicurazioni
Via J.F.Kennedy, 6 - 43100 Parma (Italy)
[email protected]

Introduction
According to a basic approach, the principal aim of the bank management is to create shareholder value, which also
means the bank capital growth. By this expression we try to look at the ability to create a present income or the
appreciation of the share value, by having at the same time the ability to arrange investments as regards its risk and
profit. It appears obvious that there is a need to chose a favourite habitat along the risk - profit line, that is to decide
the kind of risks to face, with which steps and in which combination to reach the best performance.
The financial management of a bank implies the "opening of new positions on the market". The modern banking
permits and obliges to chose differentiated risk patterns so that financial strategies can assume divergent aspects. In
this kind of situation it is worthwhile carrying out risks control modalities, based on appropriate survey tools.
Therefore it is important to make the best use of the capital that can be risked in order to support the open positions.
The growth of a bank's equity, thanks to external capital inflows and retained earning, depends on the success of the
financial strategy. The possibility to create proper competitive strategies depends on the growth of the equity. The
result is that a proper financial strategy is an unavoidable element for having reliable growth options. The value of
the economic capital expresses the level of solvency; the capital gains are the results of the transformation and
development projects, but they are its results also as regards the next market cycles.
The supervision rules impose specific capital endowments linked to the current activity. This point as well obliges
financial intermediaries to an efficient monitoring of the level of exposure at risks of the economic capital's market
value fluctuation.
The capital is by definition a short resource, that is why it is necessary to plan its use in different operation areas
(loans, bonds, currency) on the base of open positions with the desired features of prospective risk - profit. To carry
on this working project we need good profit indicators - and this seems not to be difficult - and reasonable risks
indicators - and this, we know, is particularly hard. It is particularly difficult to have reasonable risks indicators, that
can be used transversally in the mentioned operation areas. Problems arise when we consider tools with no linear
performance comparing to market dynamics (especially options).
We have to affirm that in contemporary economics, with strong financial development, the price for risk became less
advantageous. In the past there were some fundamental elements which made a more interesting economic outcome
possible thanks to big discrepancies between provision costs and investment profits in a medium steady situation.
Market globalisation and the refining of hedging, arbitrage and trading techniques of financial intermediaries and
industrial and commercial enterprises made financial activities less profitable and this means a need for sophisticated
monitoring tools and risk pricing.
Regarding market risks we have to admit that there were intense and very interesting methodological developments to
measure risks. However the market risk increased. More often we noted the reversal of market tendencies which
seemed to have a good start. We add to the inborn difficulties of following markets the possibility of fraudulent
behaviour of the management which is not controlled, as well as the possibility that control system appears improper
(model risk).
It is difficult to design the market risk because it is linked to the possibility to do worse, but also to do better than the
medium result, which is usually more likely. We should also consider the links between tendencies of different
markets. This situation is very different if compared with the situation that regards credit risk; for this one it is only
possible to worsen the performance rather than to improve it. Again, for this one, operations are less immediate and
harder to perceive.
Even if in a context of unsolved methodological complexities, the Committee of Basle adopted moral season
measures in order to spread the use of internal models to calculate property requirements for market risk. The
Committee gave, as basic condition, the respect of qualitative and quantitative parameters, which characterise
efficient and probably concrete useful models.

The value at risk techniques

Value at risk (VaR) is an important improvement compared to past models of risk management. More precisely VaR
considers different types of investments and gives a "transversal" solution for their management.
VaR has an emphasised probability nature. It consists in estimating the potential loss by considering the volatility of
the invested object and the relationships of market prices belonging to different, but linked sections. By using statistic
techniques, even if not really satisfactory, we have to evaluate the biggest loss that a financial intermediaries could
undergo during the holding period (which is the time of reference) with a specific level of probability, in the event
that markets develop unfavourably. It is also a very interesting conventional approach. The standardisation of
different elements which are useful to find a solution permits, even if in a little indefinite way, to compare different
banks and to manage market risk rate of a single bank in a long period. This method requires a complex information
management and calculation algorithm.
Now I wish to make a few remarks on the main conventional elements.
Holding period means the period of time in which we want to measure the potential loss. If for instance the VaR of a
position is of 100.000 $, we mean with this expression that in the worst case, the position, during x days will reach a
loss of maximum 100.000 $. The kind of holding period is chosen by the investor and has to be coherent with the
level of possible liquidation of the opened position.
Choosing a long holding period for a position that may be easily liquidated is incongruous. As a matter of fact even if
things get worse, the position can be closed avoiding to reach the VaR; the previously calculated VaR had
overestimated the risk of position. Choosing a short holding period for an object which may be hardly liquidated is
incongruous as well. As a matter of fact if things get worse it is not possible to close the position within the holding
period and this generates a loss which is higher than the previously calculated VaR, which in this case was
underestimated.
The choice is therefore not obvious, but crucial to give a meaning to the VaR on which we work. The choice depends
on the dimensions of the analysed position (compared to the usually negotiated volumes) and on markets' conditions
in which it is opened. In any case we should consider that market liquidity, when things get worse, can be quickly
spoiled, thus making it impossible to close the position according to the original evaluations. As we know the
proposal of the Committee of Basle is a 10-day period.
Besides the holding period we need to establish a reference to describe the worst situation which is needed to estimate
the maximum possible loss. Of course it has not to be the absolutely worst loss in the most disastrous situation of
history, because in this case it would only be sufficient to abandon a high VaR and the capacity to support risky
situations would be excessively compromised. Actually as "negative situation" on which we estimate the effect on the
economic capital of the bank, we choose the real case which is better for one, three or five per cent if compared with
the worst cases. By so doing, we do not consider the absolutely worst VaR, but 5% of the worst cases. That is why
actually we loose more than the VaR, then this describes the best of the worst cases. Is this kind of thinking
reasonable? Certainly it is not perfect, but it is useful. It is based on a conventional method, therefore it requires a
certain level of homogeneity between the VaR users' behaviour. The question of the identification of the worst case
as a reference point is obviously linked to that of the holding period. If this one is too short and we choose a 5%
reference the result are underestimated risks; if the holding period is long and we choose a 3% reference the result is
a bigger risk of the opened position. The absolutely best solution does not exist, but it is not difficult to find a
reasonable solution to compare and think on the proceeding during the risky period. As we know the Committee of
Basle proposes a 1% level.
In any case we should consider the collapse of Wall Street on the 19 October 1987 and the Italian rate curve shift in
August 1994 presented a difference of respectively 20 and 10 times the daily average deviation. These are market
events located in the worst part of distribution and the VaR accumulated according to the current methodologies was
not able to cover the loss. Consider also that many analysis showed that the profit distributions are more than those
indicated by normal distribution. This depends on the possibility for markets to follow the trends and to persist in a
chosen direction. The VaR estimating system should consider this aspect and consequently the VaR's estimate should
be raised.
Another aspect is that of correlation. The VaR system openly considers links between markets and tries to strengthen
risky situations by considering the diversification effect, which tends to mitigate the entire risk that is less than the
sum of its parts. Considering correlation - which are never perfect - means being able to resist risks that are more
accentuated than those resulting by monitoring the sum of risks linked to single positions. This means a major range
of risks because there will not be negative events on different markets all together. At this point it must be said that
the exemplification of models are important; this is the basis of a simplified conventional methodology compared
with a complicated problem.
Every single VaR model is based on statistic parameters, of historical nature, about the risk of every market and
about correlation between markets. These parameters are estimated on long periods (to be more meaningful) and at
the same time on recent periods (to show present situations). As it is not possible to reach both things simultaneously
we will have to decide to which one we give our priority. When different choices regarding reference data produce
different results in the parameters, the consequence is an indefinable risk of the model (the estimating risk).

VaR and stop loss

VaR is an ex-ante system of limitation of the assumed risk and when managed dynamically, it is an insurance system
of a minimum performance, as it shows the maximum loss that we are ready to risk. The maximum loss corresponds
to the value at risk, identified in the different units that manage the positions. It is a system that tends to close
positions when the loss takes a value next to VaR. The VaR is very similar to the stop loss, both derive from the same
psychological feature of trader who wants to limit the assumed risk; they are two risk management techniques with
the same basic assumptions that have to coexist in the best way.
If the risk managing model is unique for the financial intermediaries, they can close the same positions when markets
assume a definite position over the limit value. VaR is a mechanism that tends to destabilise markets, which means
that it convinces groups of operators to abandon their positions in some market circumstances, working in this way to
reinforce market tendencies. In this situation it is profitable for operators to risk moderately; the spread of VaR
control systems leads operator to abandon loosing operations after reaching a certain level.
In this case, is the VaR really a stop loss? If not identical, which are the differences between them?
First of all, we can say that the VaR is a kind of stop loss at a bank associated level. When VaR is eroded the
financial institution knows that it cannot run any risk on more capital, at least not without damaging the basic
minimum risk capital needed for the planned activity (without the financial activity concerning open positions on the
markets). The VaR puts a general limit over which we should close positions (without the possibility to open other
positions, because in this case you would need another VaR which is not available any longer).
Which are the results for an institution using the VaR logic, on the activity of single traders used to think on stop loss
based on single open position? How should the stop loss game be modulated (self-imposed or imposed by a
hierarchical organisational structure) with VaR limits for every organised structure and every trader?

Positioning principles for stop loss

Obviously a unified theory, sufficiently shared, on the stop loss positioning does not exist. A stop loss is a position's
sale order predetermined in advance; in this way the trader's psychological elements do not delay the order if
necessary. Trading without positioning a stop loss generates the risk that the - unmotivated - hope to reduce loss or to
enlarge actual profits creates very poor performances spurred by the failed ability to declare the preceding trade move
as closed. Preparing in advance the circumstances to leave the positions gives you the possibility to manage the
trading better. It is in any case better to cut losses in every trading system used; the stop loss is exactly oriented in
this direction. In this we can see a similarity with the VaR, which even if ex ante, avoids the transformation of little
losses into big losses. They are both protection systems.
The first type of stop loss helps identifying a wrong position immediately after its launch (INITIAL STOP). If the
trader opens a position on the rise (falling) and the price is falling (rising), he has to know on which price level the
action has been inappropriate. For a bulling (bearish) position the stop loss is on lower (higher) price compared with
the initial price. A fixed rule to position the stop loss does not exist, but there are different references and ways of
thinking. Usually we refer to quite near support and resistance levels, we take position on a bit lower (higher) prices
to be sure of the meaning of the signal. Sometimes it reasonable to risk 1-2% of the opened position before making
the mistake. A 1% initial stop of the initial risked amount is reasonable when there is not the risk to participate too
often in the market.
A second type of stop loss is linked with successful positions, even if moderately for the moment. If the price follows
the expected tendency it is better to shift the stop loss as soon as possible to a level that gives you the possibility to
obtain a minimum acceptable result (BREAKEVEN STOP). If the price retraces, the exit would be with a price that
generates a little but satisfying profit (as well as trading's operational costs recovering).
A third type of stop loss is used for clear successful positions. If the price shifts to the expected part, we wish to stop
a part of the theoretic profit, shifting the stop loss towards the profitable direction (TRAILING STOP). This stop
should not be fixed near to the current price to avoid to give rise to a brief correction which is followed the original
tendency, but not far away from the current price, in order to avoid the risks to the profits already obtained. In this
case too, it is useful to observe supports and resistance and to put the stop loss on suitable technical levels.
A fourth type of stop loss refers to the maximum duration time of a position that is not giving important results
(TIME STOP). In this case it is not profitable to pledge capitals that do not yield as we like, when there are other
trade possibilities of which we want to take advantage.

Remarks on a combined use of VaR and stop losses

The VaR and the stop losses are kinds of "safety web" that should be made compatible.
First of all, we should remember that the budget period assigned to a trader does not coincide with the holding period.
If the trader uses the whole VaR, there is the possibility, even if just in 1% of the cases, to loose the VaR only with 10
days of the opening position. If they are the first 10 days, the trader has to stop for the whole remaining year, unless
he receives another VaR, but this is unlikely.
If the budget period includes one year, there are still enormous possibilities to loose the assigned VaR (approximately
every 5 years the loss of the capital would be sure), but this has been avoided by a reasonable use of stop loss.
However people who co-ordinate financial investments policies do not like to be on the loser's side of the market,
they want to be on the winner's side and they will judge traders in this sense. The VaR allows to assemble the risks of
the single trader, the stop loss permits to limit conveniently the losses of single traders, even if it does not fit to
"consolidation" of the single trader's positions in a general view.
With the VaR, financial speculations are seen as equal bets in a casino where the classic rules of betting are valid.
Obviously traders act with a different logic and, minute after minute, they try to understand market's actions.
This line of reasoning helps you understand that the "safety web" of the VaR is a useful general government's tool of
the bank, but it does not come out of a credible market's situation for traders. They believe in the speeches which
convinced them to create a stop loss logic even if not unified.
If we want to carry on with our analysis, we have to ask the following questions: is the technique of stop losses
modified by the introduction of a government's logic of trading structures based on VaR or is it unaltered? In other
words, in the definition of stop loss, is it important to know how much VaR has been used? And how much VaR has
been destroyed with unlucky speculations? Or how much extra VaR has been created by successful speculations?
These are very practical topics; we can answer on a logical base or with regard to the past experiences. I try to
summarise some basic remarks:
1. It is not logical to think that the use of VaR could influence the positioning of stop loss. In the context of technical
analysis approach the stop loss has a reference to the logic of single speculations, not in the general context of
portfolio risks. If it was not like this it would be possible for the trader to use little of the assigned VaR risking on
temporarily unsuccessful speculations. In short, it would be possible to dissimulate unsuccessful situations with
limited upsetting possibilities.
2. It is reasonable to think that the initial stop is related to the VaR "used" by the security object chosen for the
trading. If this security needs more capital, this means that it is more volatile and this means more tolerance in
defining the initial stop.
3. It is reasonable to think that the break-even stop is defined in relation with the VaR used in the operation, more
precisely we have to consider the cost of the capital needed for the position that tries to recover that cost.
4. Obviously the trailing stop is defined in relation with the volatility of the position and therefore with its VaR.
5. The success in previous market operations should assign more free spaces to the trader, meaning more VaR or less
restrictive stop losses.
6. The time stop should be fixed by considering the financial cost of the VaR used.
In conclusion, I wish to remind that the stop loss does not mean to stop the trader's activity which shows cumulated
losses corresponding to an alarm level (this kind of tool is called “trigger point”). Nonetheless if with this expression
we mean this, it is reasonable to think that the stop loss acts on a certain part of the assigned VaR. In this case the
stop loss is completely different from the VaR, because it is simply an ex-post tool that allows to be sure on the
application of the policies previously defined by the VaR.

Bibliography
Arnold, Curtis (1993) Placing Stop-Loss Orders, Technical Analysis of Stock and Commodities, vol. 11, n.4, pp. 178-
81.
Bulkowsly Thomas (1997), Beta-Adjusted Trailing Stops, Technical Analysis of Stock and Commodities, January,
pp.30-6.
Chande Tushar S. (1992) Smart Stops, Technical Analysis of Stock and Commodities, vol. 10, n. 12, pp. 507-10.
Cifaldi Patrick (1993), The Importance of Stop Orders, Technical Analysis of Stock and Commodities, vol. 11, n. 10,
pp. 440-2.
D'Acunto R. - Poglia F. - Lanni V. (1996) Il Value at Risk e l’organizzazione dell'area finanza, Spot, n. X3,
December, pp. 27-9.
D'Acunto R. - Foglia F. - Lanni V. - Nicolucci T. (1997) I limiti di VaR e la scelta della politica di stop loss, Spot,
84, March 1997, pp. 44-7.
Ehlers, John (1989) Using stops - a new approach, Technical Analysis of Stock and Commodities, vol. 7, n. I1,
pp.408-10.
Greepan William I. (l 994), Using Stops To Your Advantage, Technical Analysis of Stock and Commodities, vol. 12,
n. 3, pp. 125-8.
Hutson Jack K. (1987) Maximising profits with stop orders, Technical Analysis of Stock and Commodities, vol. 5, n.
6, pp. 192-4.
Luisi Joe (1996) On Using Stops, Technical Analysis of Stock and Commodities, January, pp. 66-71.
Lusignani Giuseppe (1996), La gestione dei rischi finanziari nella banca, Bologna, Il Mulino.
J.P. Morgan (1997), RiskMetrics: Technical Document, New York.
Sironi Andrea (1996), Gestione del rischio e allocazione del capitale nelle banche, Milano, Egea.
Sweeney John (l9BS), Where to Put Your Stop, Technical Analysis of Stock and Commodities, vol. 3, n. 5, January,
pp.15-8.