Dynamic asset allocation bootcamp, using OECD data to select ETFs
By Allan Lane (pictured), Algo-Chain – At Algo-Chain when re-balancing our Model Portfolios we have always been big advocates of using macro-economic data to help guide the asset allocation decision process, and central to that is the data provided by the Organisation for Economic Co-operation and Development, OECD.
In a world of data overload, we all know a picture is worth a thousand words, and very occasionally the data throws up such a simple chart that captures the moment like never before. And so it was with the most recent print of the OECD’s composite leading indicators, CLIs, that are published every month for many countries and regions. As an amalgam of a country’s economic statistics, such as a country’s GDP, GDP growth or measures of consumer and business confidence, the CLI as a single metric provides a very handy snapshot of the health of the world’s economies.
Source: Algo-Chain.com, OECD.org
The baseline number for the CLI is 100. A reading of 100 or higher generally indicates that the economy is expanding while a value below 100 signals contraction. On a secondary level, one can distinguish whether the value of the CLI is trending upwards or downwards. From this one defines the four states of an economy. One talks of the expansion phase when that value is above 100 and still trending upwards, meaning the economy is doing well but at risk of overheating.
The downturn phase is characterised by an economy still expanding (value is above 100) but starting to slow down (value is trending downwards), whereas in a slowdown phase it has further contracted to a value below 100 and is still trending downwards which also explains why the gloomy outlook usually means negative returns for risk-on assets like equities. Finally, the economy turns again in a recovery phase with a value still below 100 but green shoots arising.
In the world map shown above, it is quite remarkable to see that South Korea appears to be the only country that is now in recovery phase. Over the recent weeks much has been said about Korea being ahead of most countries in the Covid-19 cycle, which coupled with the fact that it is well known for having an economy that is very tech-centric, does go some way to explain why it is the only area on that map that is showing green.
The basic idea of using macro-economic data as a tool in a dynamic asset-allocation process is driven by the belief that certain asset classes do better than others in any particular regime. For example, if one compares the year-to-date performance of the Franklin FTSE Korea ETF (ticker: FLXK, fee 0.09 per cent per year), with the iShares Core FTSE 100 ETF (ticker: ISF, fee 0.07 per cent per year), as of 10th June 2020, there is an outperformance of nearly 16 per cent delivered by investing in large cap Korean equities in GBP.
Many critics would argue that when used for investment purposes the OECD data is too slow, but our own experience suggests that when used within a low-turnover framework, it has proven to be a very useful tool. Using a variation of that idea to drive country-by-country allocation decisions, as highlighted above, has its merits, but its real strength comes to the fore when used to develop trading signals on a cross-sectional asset allocation basis.
Using the OECD data to first determine where in the cycle the US economy is, one naturally wants to next figure out which asset or sub-asset classes are best suited in the prevailing regime? At this stage, several statistical models are at one’s disposal. In the end though, if one believes in using empirical data, it boils down to a counting exercise to determine the Bayesian probabilities of which investments show the most promising returns. Knowing how small cap stocks behaved before and after the Global Financial Crisis of 2008, and any other crisis we’ve seen recently, in essence, allows one to compare the odds of choosing between large and small cap stocks.
This line of thinking is most effective when one applies the model to a wider list of US assets, such as Treasuries, Corporate Bonds, TIPs, tech stocks, utility stocks etc. To see this model in action, let’s take the example for the period mid-May to mid-June 2020, where in the figure shown above, the top three ETFs with the highest signals score track short-term US Treasuries, US Aggregate Bonds and the S&P 500 tech sector respectively. All scores are standardized to lie between -2 and +2, where 2 constitutes a strong buy signal and -2 a strong sell signal. The remarkable thing is that these recommendations simply arise as the output of a Bayesian estimation process and fit quite well with what we now know to be very sensible suggestions.
How well this framework would have held up during the Great Depression of the 1920s is anyone’s guess. The big worry is that we might be about to find out!