Analyze inflation impacts on specific portfolios and factors

In a regime of higher inflation, geographical, sectorial, and idiosyncratic specificities are going to be radically differentiated among companies. Some will benefit or suffer less from the effects of inflation than others.

Using Bloomberg’s BQuant Enterprise platform, we can analyze the effects of inflation on a specific portfolio and select the factors to watch in the context of inflationary time.

Summary

Bloomberg has developed a project in BQuant Enterprise, Bloomberg’s quantitative investment platform, that will allow you to build a more in-depth analysis of the dependencies of your portfolio on interest and inflation metrics (consumer price indices, interest rates, etc.) or factors linked to inflation (as either components of inflation or dependent on inflation).

We use datasets and compute libraries available in BQuant to perform the following pre-analysis before using ML libraries in BQuant to predict portfolio returns under different inflation regimes. These are:

 1. Sensitivity analysis of a portfolio of interest against inflation metrics and factors linked to inflation:

The quant project first considers a set of basic parametric linear analysis on normalized, market beta adjusted daily factors:

 

 

• Correlation analysis: building a correlation matrix across all factors of interest to select among correlated ones the most relevant to avoid collinearities
• Univariate analysis: computing betas across the factors preselected from correlation analysis and over a predefined rolling window, with a predefined lag, adding significance statistics
• Multivariate analysis with elastic net penalties to estimate more robust multivariate betas on the set of factors we can preselect from the above two analyses

 

 

 

 

3. Fundamental durations and convexity:

Now the analysis can focus on more specific security-level inflation-based indicators by considering two types of signals: duration and convexity.

We compute a second order polynomial fit per each security in the portfolio of interest on inflation metrics. The security duration is approximated by the first order parameter of the fit and the convexity by its second order parameter. On Bloomberg’s BQuant platform, we build a double-sorted top and bottom quintile portfolio based on duration and convexity metrics, and then plot the performance of this portfolio across time to evaluate the exposure of the portfolio of interest to duration and convexity. Duration approximates linear dependencies, while convexity introduces nonlinear dependencies with inflation.
 

2. Event study of factor returns linked to inflation around changes and surprises of inflation metrics (often lower frequency economic data):

In a second analysis, the project runs an empirical study of changes in inflation metrics and surprises on factor returns. The idea is to empirically observe the impact of a change in an inflation component’s new released value, and (if forecasts/consensus are available) the effect of a surprise between this newly released value and the consensus on specific factors.
 

 

4. Prediction model:

Finally, BQuant Enterprise enables us to use these three pre-analyses to select all the relevant factors that correlate to inflation. The analysis leverages a gradient-boosted random forest model to predict portfolio returns under inflation regimes.

Using Bloomberg’s BQuant Enterprise platform, we can then analyze the effects of inflation on a specific portfolio and select the factors to watch in the context of inflationary time.

 

Using Bloomberg’s BQuant Enterprise platform, we can analyze the effects of inflation on a specific portfolios and select the factors to watch in the context of inflationary time.