Explanations

Sound Risk Management: Analysing a Portfolio using Risk Analytics and Stress Scenarios

For each of the themes I work out on this website, I have handpicked a few key risk statistics that give a better idea of the risks embedded in each of the portfolios that I work out.

1. Diversification. This is a rough idea of how well a given portfolio would diversify an imaginary 60% stocks / 40% bonds portfolio. So while the Robotics and automation theme essentially consists of only equities and would therefore be given a “low” diversification score, on the other hand the market crash theme scores “high” given that it consists of classic risk-off assets and defensive bond and stock holdings to a smaller extent

2. Volatility. Based on data from Bloomberg’s PORT risk system which is available on the terminal. Roughly around the time that an article is published I ran it through their ex-ante risk system and made an assessment of the predicted volatility given the market environment at the time. This is obviously subject to change, but if the conditions get too far out of whack I will assess new portfolios based on time horizons (for volatility and correlations) that are more consistent. Gives a general idea of how volatile the portfolio will be. As a general guide, 5% would be a relatively low fixed-income like volatility in normal markets, while anything above >10% is seen as equity territory.

3. 95% 1 year Value at Risk (VaR). Also sourced from Bloomberg’s PORT risk system. For the risk experts: this is based on a time-scaled 1 day parametric VaR at 95% confidence level. Gives an idea of what the maximum loss over a 1 year period could be with 95% certainty. Not a perfect risk statistic by any means, but gives a decent idea of the riskiness of a portfolio

4. Yield. The market value weighted sum of the yields paid out by the investments in the portfolio. Whether simply the distribution yield from the ETF, dividend yield from stocks or coupons received from the bonds

5. Stress scenarios. As VaR and volatility are not perfect measures of risk, I complement them for each theme with a set of relevant scenarios. In these scenarios, calculated in Bloomberg PORT, I model the suggested portfolio and recompute their value if the factor returns of a given scenario would play out (e.g. a 1.0% rise in interest rates or a 2008-like equity shock). Gives a better idea of what the sensitivities and worst-case scenarios could be

The stress scenarios tend to vary, but the ones I present most commonly are the following:

  • 2008 Lehman Brothers default period. A replay of historical factor returns between 14 September 2008 and 14 October 2008.
  • Interest rates +100bps. Shocks interest rates up by 1.0% and uses the correlations used within Bloomberg’s risk model to propagate that shock to other risk factors. From a very simplistic point of view, if a 1% shock creates a -3 standard deviation (-3x volatility) change in the prices of bonds, and their correlation with equities is 0.3, the scenario would also shock equities by 0.3*-3SD = -0.9SD. If the standard deviation of equities is 15%, stocks would lose -0.9*15% = -13.5% in the scenario
  • 2008-2009. A replay of historical factor returns between 31 December 2007 and 31 December 2009.
  • 2010 onwards. A replay of historical factor returns between 31 December 2009 and 12 March 2018 (an arbitrary cutoff that I however keep consistent between the portfolios)

Categories: Explanations

1 reply »

Leave a Reply