Background:
Over the span of two weeks earlier this February the S&P index lost approximately 10% of its value. The correction might have been triggered by fears of inflation as investors suddenly became worried that tax cuts and an already overvalued stock market would require a faster than expected Fed intervention.
When we see market corrections (and we have seen plenty over the decades) we like to go back and analyze what our models were forecasting in terms of risk propagation. For example: if we travelled in time to a few weeks before the correction, what was our best estimate of contagion for a -10% drop in the S&P at that point? Common sense would have anticipated volatility going up, credit spreads widening, Yen appreciating and so forth. But by how much?
Our models use sophisticated statistical methods to retain only relevant information about these contagion effects and learn from them. We can’t nail exactly what will happened in the future (surprise!) but we can provide a robust projection that improves its estimates as more events are observed.
Before the correction:
On 01/10/18 (approximately 2 weeks before the correction) we were predicting the following propagated moves from a 10% up and down S&P shock to a few selected macro indices/currencies:
|
SP500 |
VIX |
BB spreads |
JPYUSD |
|
+10% |
-36.14% |
-38.61 bps |
-2.19% |
|
-10% |
+119.15% |
+43.60 bps |
+2.61% |
Our calculation engine was expecting volatility to jump roughly 120% if the markets dropped 10%, i.e. going from 9.82 on that date to 21.60 (lognormal move). It also expected BB spreads to widen by 43.60 basis points from a 10% equity correction, i.e. going from 2.03% on that date to 2.46% (normal move).
Our models expected a flight to quality that would result on a +2.61% appreciation of the yen versus the dollar.
Because we generate full distributions of contagion, we were also providing a range around the expectations above on a -10% SP shock:
|
|
VIX |
BB Spreads |
JPYUSD |
|
Extreme positive(*) |
8.2% |
0 bps |
+0.13% |
|
Expectation |
+119.15% |
+43.60 bps |
+2.61% |
|
Extreme negative(*) |
395.23% |
+108.12 bps |
+5.58% |
(*) average of the best and worst 5-percentile PLs.
How did these forecasted propagations fare with what actually was observed in the markets a couple weeks later? The table below shows side-by-side:
|
01/26/18 |
02/08/18 |
Realized PL from 01/26/18 to 02/08/18 |
Forecasted PL on 01/10/18 |
|
|
SPY |
284.16 |
255.45 |
-10% |
-10% |
|
VIX |
11.08 |
33.45 |
201.89% |
119.15% |
|
BB spreads |
1.92% |
2.20% |
+28 bps |
+43.6 bps |
|
JPYUSD |
0.009118 |
0.009161 |
0.47% |
2.61% |
We were expecting a larger reaction from flight to quality on the yen, a less pronounced effect on volatility and more BB spread widening, but our calculation engine largely captured the magnitude and direction of contagion, conditional on a 10% S&P drop. The observed contagion to other macro indices were all within our extreme positive and negative ranges from the previous table.
After the correction:
Each new event in the market, such as what percolated during those 2 weeks in early February, provides additional information for our calculation engine. Thus, right after the correction on 02/10/18, our forecasts reflecting the fresh information would have been:
|
SP500 |
VIX |
BB spreads |
JPYUSD |
|
+10% |
-39.44% |
-21.55 bps |
-0.66% |
|
-10% |
+151.46% |
+24.6 bps |
+0.96% |
As per table above, the observed correction and resulting propagation of risks are incorporated into the model to produce a forecast with more pronounced increase in volatility, slightly less widening in BB spreads and slightly less JPY strengthening from a flight to quality.
Summary:
We demonstrated in this short report how Everysk can reliably propagate macro shocks. We compared the forecasted risk propagation with an observed one, by measuring closing prices of a few macro indices/currencies during a 10% S&P decline earlier this year.
We also showed forecast improvements after our model incorporates new event data, by performing a forecast just after the event.
The example above shows how a exogenous shock propagates to other generic macro indices and currencies. The same principle applies when the exogenous shock is being propagated to all the risk factors within the portfolio. We can produce global stress tests that differ from a localized approach that stress tests securities via their pricing formulas, without propagation. We content our methodology is more relevant to buy side investors interested on big picture effects.
Additionally our methodology generalizes to shocks on any exogenous index, even when the shock does not affect directly the securities via their pricing formula.