• Macroeconomics holds significance for asset managers
• Equity, interest rates, credit, commodities are key variables
• Actual or realised observations are frequently different from expected levels
As we conclude the third quarter of 2014 today, a question that I hear being asked with increasing frequency when it comes to strategic asset allocation is:
“Does global macro matter? Isn’t it all just random noise?”
In essence, this asks whether or not there are significant historical relationships between the relative performance of portfolio components, i.e. the risk associated with investment factors.
These factors include equity, interest rates, credit, commodities, alternatives and so forth.
And in terms of shifts in underlying global macroeconomic conditions?
Since the financial crisis of 2007/8, there is plenty of evidence to support the view that risk factors are critical as key drivers of portfolio returns. In short, macroeconomic shifts are highly significant when one looks to optimise asset allocation.
The objective is to produce superior returns and hence, access to insightful macroeconomic analysis and strategic planning can add value, leading to an outperformance of a given benchmark or peer group (also known as “alpha”).
It is therefore helpful to have a well-reasoned central scenario or reference model which can be overlain and/or underlain by a more optimistic and/or pessimistic alternative. One can use these three scenarios to evaluate the potential impact on factor returns.
It should be assumed that simple regressive models cannot completely replace a forward-looking investment process built on reason and due diligence. Rather, scenario modelling can be regarded as a complementary process that can allow the asset manager/investor to consider macroeconomic shifts that are enhanced by wisdom, judgement, risk-factor specific variables and an informed opinion on current and future economic and political events. This is not the same as running a rear-view mirror optimiser.
Incorporation of market expectations
Asset pricing theory generally contends that market-based risk factors will change in anticipation of forthcoming macroeconomic data. Indeed, current asset valuations should be seen as the best reflection of market agent’s interpretation of the available information and future expectations.
However, do not believe that this is a linear process, or one in which a positive data point will automatically make assets rise. The market is a swirling mix of cross currents and undertow; how often we can point to the old adage:
"Buy on the rumour and sell on the fact…”
One has to be able to see through the mist and identify anticipated changes in macroeconomic variables as well as the unexpected.
Flexibility in defining a scenario
It is too simplistic to live in a world of comparative statics where we describe return estimates for scenarios in wide ranging categories of significant variables such as Gross Domestic Product growth or consumer price inflation as “low... stable... high", nor can we do the same for specific past phases in the economic cycle.
Instead, it is critical to try and accommodate various combinations of GDP and inflation views to extrapolate estimates of GDP growth and inflation into risk factor returns.
Creating the concept
It is best to commence with a relatively straightforward conceptual model that can zero in on the following core risk factors:
- Equity
- Interest Rates
- Credit
- Commodities
We can leave “alternatives” aside for the time being as art, wine, vintage watches and cars are characterised by fashion fads and emotion.
In most diversified portfolios, these risk factors will be seen as accounting for the majority of the return variation.
However, how can we stress test the portfolio so as to derive an estimate for the portfolio performance under different macroeconomic scenarios.
We do so by looking for connections between factor returns and the macro surprise that can catch the investor unaware.
Specific Factor Return is given by the following formula:
SFR = α + β GDP x SGDP + β Inflation x S Inflation + β Interest Rate x SGDP x SGDP + E {1}
Where:
a = Economic base line
β = Expected data point outcome
S = Unexpected or surprise date point outcome, >/< β
E = Error term
In statistics, "E" (or: errors-in-variables models, or measurement error models) are regression models that account for measurement errors in the independent variables. In contrast, standard regression models assume that those regressors were measured exactly, or observed without error. That would imply all values of S are zero… which is usually not the case.
The term or interaction labelled {1} measures the effect between growth and inflation and accommodates an impact that arises from surprises.
For example, if we are to consider the credit or fixed income element of a portfolio, then the impact of a surprise in actual inflation on yields will be more pronounced if accompanied by a positive shock to GDP growth. Why so? It is highly likely that it will signal a higher probability of a permanent increase in the path for future inflation (and in turn the path of central bank interest rates).
The simple model illustrated here can be updated on a quarterly basis and can be regressed against risk factor returns. If we look to the medium term, we can plan our portfolio on a rolling one-year horizon given that GDP data is released on a quarterly basis.
For more active or high frequency traders that trade on specific events and agency news flows, trading based on technical analysis is a more recognised approach.
Through back-testing, the model can reveal that if one takes the US as the observed economy then over the period 1980 to 2013 we can see that on a “one year ahead” basis the variations in expected and realised GDP growth and inflation are as follows:
• GDP growth percent: Expected 0.0 to 6.0, Realised -5.0 to 9.0
• Inflation percent: Expected 2.0 to 9.0, Realised -2.0 to 14.0
What are the β coefficients for various factors? Below they are listed as listed as: Factor, β coefficient SGDP, β coefficient, S Inflation.
- USD LIBOR Money market, 0.5, 0.7
- Treasury 10 year yield, 0.2, 0.4
- Treasury 10/30 spread, -0.1, 0.0
- “A” Credit spread, -0.2, 0.0
- S&P 500 Index, 3.9, -2.2
- Commodities, 2.7, 6.3
This means, taking the S&P 500 as an example, that if GDP is 1.0% higher than was expected, (i.e. the surprise is 1.0%) then as a factor, equities as measured by that n=benchmark will rise by 3.9%
Commodities appreciate with upward surprises in growth and inflation whereas, just as one might expect, the 10-year Treasury will see yields rise (and prices fall) if growth or inflation delivers an upside surprise.
This simple model can be created to accommodate multiple combinations of GDP and inflation views. Therefore it can escape a binary “low/high" choice or even a triple “low… stable… high” scenario.
Macroeconomic variables are important drivers of the returns on investable risk factors. These are in turn the underlying components of asset class returns. I firmly believe that establishing an econometric framework to assess the impact of unforeseen surprises (good or bad) to GDP growth and inflation provides a critical link between macroeconomic forecasts and portfolio returns.
What really is critical in any specified scenario is how the investable factor will pan out relative to these expectations.