Subscribe
Featured Image

The Value of Local Factors When Evaluating Portfolios

Risk, Performance, and Reporting

By Ravinder Dosanjh  |  February 25, 2021

Risk models have become commonplace in the financial industry as firms seek to gain a deeper understanding of what is driving the risk and return of their portfolios and ensure that their products are following their preferred investment styles. Risk models, however, can come in a variety of different flavors, one of which being the universe of securities used to construct the model and which the model will cover.

Global risk models can be used for portfolios that hold securities from across the world, but given their wide coverage, they can also be used for single-country/region portfolios that have a narrower investment universe. The question is, does one-size-fit-all, or can single-country/regional models provide advantages for more targeted mandates?

Global vs. Country/Regional Models

Single-country models are designed to capture the sources of equity returns within a local market—the correlations of the factors are estimated using a universe of stocks belonging to that specific country. Naturally, single-country models should provide accurate risk forecasts for portfolios that are built from stocks within that same country.

Global models are calculated using a much broader, global universe of stocks and therefore can also be used to estimate risk for single-country portfolios due to their broader coverage universe. But doing this could potentially lead to a trade-off between coverage and accuracy, possibly reducing the granularity of any analysis.

Because of the significant market events over the last 12 months due to the COVID-19 pandemic, this question is potentially more relevant now than in previous years. Using FactSet’s GeoRev database, we can see how globalized markets really are and whether the events of 2020 have caused any changes.

As countries entered lockdowns across Europe, there was a significant shift in the market and in how businesses operated and reached their customer base. We are potentially beginning to see the impact of this shift across Europe—since the end of Q1 2020, the trend towards countries/regions becoming more globalized has paused, if not slightly reversed.

We can see in the chart below that across various markets in Europe the proportion of revenues generated by that market’s local country was generally lower by the end of 2019 than at the end of 2015, and in some cases like the Netherlands, by quite a significant margin. However, from the end of Q1 2020, when many of the lockdowns across Europe came into effect, we see signs of a pause or slight reversal in this pattern, as the proportion of revenues generated by each market’s local country began to rise slightly. With the UK completing its exit from the EU, this will be an interesting trend to monitor over the coming year.

Trend Toward Market Globalization

Comparing Value Factors Across Countries and Regions

In risk model terms, a common understanding is that all style factors work equally across different countries and regions. While it may be true that these style factors exist and are constructed in similar ways across different countries and regions, it is not true that they behave in the same way. The chart below compares the value factors across different countries and regions from different factor models.

Value Factors Differ by Country and Region

In the chart above, the variation in the performance of the value factor across different regions is clear and particularly pronounced when comparing the UK to the U.S. The UK value factor also differs significantly from the global value factor, bringing us back to the original question of whether a global model can be suitable for portfolios with narrower mandates.

Another way to analyze the divergent behavior of the value factor across regions is to plot the correlations between the UK, U.S., and global value factors. From the chart below, it’s clear that while there are periods where they are more highly correlated, there are also periods where the correlations are close to zero or in fact negative.

The chart also shows that more recently the correlations between the factors converge, as is often seen with return correlations with different asset classes during periods of market turmoil. However, in general, the correlations between these factors are not consistent, the average correlations over the period are lower than might be expected, and the factors behave differently.

UK US Global Value Correlations

Incorporating Value Factors into a Risk Model

In general, the correlations between U.S. value and global value are higher than the UK value correlations—this is intuitive as risk models tend to be market cap weighted. This means the U.S. is heavily weighted within global models and skews the behavior of these factors to be closer to U.S. factors than may be the case in all regions. This is potentially an issue for managers investing predominantly in ex-U.S. securities but using a global model, where the behaviors of those factors in the regions their portfolio is invested may be significantly different from the behavior of the factors in the model, as illustrated above. This discrepancy could also lead to stock selection issues or inaccuracies when measuring factor exposures.

So, what impact does this really have when you are analyzing a portfolio? We can check this by comparing a sample of UK value-tilted indices relative to their non-value UK counterparts and running the analysis using both global and UK-focused risk models.

Contribution to Tracking Error

Across all three indices, there is clearly a higher contribution to tracking error coming from the value factor in the UK model as opposed to the global model. Given that the main difference between the indices is the fact that one is value tilted and the other is not, we would expect there to be a large contribution to tracking error coming from the value factor. We see that the UK model is better at picking this up with the UK indices as compared to the global model.

Risk Factors Effect

This trend persists in the return attribution with a significantly different contribution coming from the value factor when analyzing using the UK model versus the global model. This corresponds to what we saw earlier when analyzing the factor returns for the value factor across different regions The UK value has performed much better than global value and we are seeing the positive impact of that when using the UK risk model; however, the global model tells a different and perhaps misleading story.

Conclusion

In short, due to the differences in the behavior of style factors in different countries and regions, and the low correlations among these, local factors have an advantage when evaluating single-country/region portfolios. Despite the global coverage of a global risk model, using a country/regional risk model may lead to better evaluation of a portfolio as the factors are more representative of the local market and will pick up the nuances specific to those markets.

Risk Live Panel

Ravinder Dosanjh

Specialist in Risk and Quantitative Analytics

Mr. Ravinder Dosanjh is a Specialist in Risk and Quantitative Analytics at FactSet. In this role, he is one of FactSet’s experts for Portfolio Risk and Quantitative Analytics and has spent the last four years specializing in workflows and solutions for portfolio and quantitative analytics including, but not limited to, factor research, portfolio construction, optimization, performance, and risk and factor attribution. Starting as a Consultant in 2012, he spent three years working with some of FactSet’s largest buy-side clients across the UK. He then joined the Analytics team in 2015 covering the same region, before working with clients across EMEA from 2019. Mr. Dosanjh is a CFA charterholder since 2017, and earned a Bachelor of Science in Economics from the University of Birmingham.

Comments