Special Series Part III: Quantifying the Impact of the SVB-Led Banking Crisis - Putting It All Together Through Scenarios
In last week’s special edition, we built factor mimicking portfolios to analyze the read-throughs to other sectors.
Our findings suggested a lopsided result, heavily favoring some sectors and negative on others.
This week, we combine the insights gained in Part I and Part II of our series to construct a scenario of continued stress on the US regional banking system and test its impact on major US benchmarks.
Though the circumstances of Silicon Valley Bank were extreme, many small banks still find themselves in a challenging position. These banks hold bonds at various durations on their balance sheets. Unfortunately, the value of those bonds has decreased dramatically as interest rates have risen steadily over the last year plus. While, for most, that drop in value represents unrealized losses, the downside risk is that they may not have the liquidity to cover redemptions should depositors look to pull their money.
Based on a recent article by American Banker,
“Across the entire industry, unrealized losses on bond portfolios totaled some $620 billion at the end of last year, according to the Federal Deposit Insurance Corp. In addition to the roughly 90 banks that American Banker identified, the analysis found that more than 200 other banks would see their capital drop to somewhat concerning levels in a scenario where they were forced to sell their underwater bonds.”
Based on the general stabilization in regional bank stock prices (S&P Regional Banking ETF price has been roughly neutral since March 14), the popular market scenario suggests less pain for regional banks going forward.
Most risk managers are paid to test the less popular scenarios.
Below, we construct a scenario under which we will see many US regional banks under stress over the next 12 months.
How To Construct A Scenario?
Building a quantitative scenario is done through three (3) steps:
- Predict the direction of the factors associated with the scenario
- Build a 'scenario mimicking portfolio'
- Compute the sensitivity (Beta) of each security in your universe to the 'scenario mimicking portfolio.'
Step 1 — Making a Prediction
As far as predictions are concerned, we’re far from pundits or armchair economists. Our views are rooted in our analysis in Part I of our series, where we used the Wolfe US Financials Risk Model to identify the style factors that best illustrated the market’s reaction to the SVB failure.
Our prediction is:
- Factors that saw positive moves during the SVB crisis will continue to see positive moves over the next 12 months.
- Factors that saw negative moves during the SVB crisis will continue to see negative moves over the next 12 months.
As a reminder, we list out some of the key factors below.
Step 2 — Building a Scenario Mimicking Portfolio
In Part II of our Series, we learned that we could construct a factor mimicking portfolio to emulate the movements of each Wolfe factor. Then, we used those factor mimicking portfolios to analyze the read-throughs to other sectors.
This week, we combine the factor mimicking portfolios from Part II into a Unified Scenario Mimicking Portfolio.
As shown below, we created a scenario mimicking portfolio as a market-neutral portfolio with long exposure to the positive factor mimicking portfolios and short exposure to the negative factor mimicking portfolios.
Step 3: Compute the Sensitivity of Stocks to the Scenario Mimicking Portfolio
We used the Predicted Beta calculation to measure each stock's sensitivity in the Russell 3000 index to the scenario mimicking portfolio. Our article, A Tale of Two Betas, explains how Predicted Beta differs from Historical Beta, the more general approach to computing Beta in our industry. The benefit of using Predicted Beta is that it considers regime changes and leverages the changing characteristics of companies. In contrast, Historical Beta is routed on past correlations purely based on prices.
Once we computed the Predicted Beta, we could then use that to calculate the implied return of each stock, assuming a two-standard deviation event over the next twelve months.
We used the using Axioma US4 Medium Horizon Risk Model for these calculations.
The table shows the aggregate predicted betas and implied returns of key sector and industry ETFs on an absolute basis and relative to the SPY ETF.
What is interesting to note is that this particular environment provides a generally favorable climate for the overall market. Based on this simulation, the SPY shows a 17% return, heavily driven by its large-cap technology weighting. We see a fair deal of dispersion at a sector and industry level. Communication Services, Information Technology, and Consumer Discretionary would be the biggest outperformers. Energy would be hardest hit, followed by Consumer Staples and Financials, driven mainly by Regional Banks.
We hope you found the last few weeks of articles helpful to understand better how to use factor models and quantitative techniques to:
- Identify the characteristics associated with a macro shock to the market
- Analyze those characteristics to develop read-throughs
- Build forward-looking scenarios and test your portfolio’s resilience to them.
Feel free to reach out if you'd like to go deeper on any of these topics or test your portfolio to the factors associated with the banking crisis.