Next step in style and size

Historically, Fed tightening, combined with a range of other key factors evident in current market conditions, presages a 12 month period in which US and Asian value styles tend to outperform growth styles and Asian small caps tend to outper-form large caps.

Next steps in style and size

Firth Investment Management is a firm with two strategies, both (to varying degrees) value-biased and one small cap-focused. Consequently, considerable thought goes into future size and style bias in market performance. Given the recent back-up in US bond yields (as at mid November 2016), rising US dollar, increased expectation of imminent Fed rate hikes and the election of Donald Trump, questions of future performance bias are particularly relevant.

The challenge with such questions is the enormous range of potential factors to consider. At the simplest level, for example, it is not clear whether a steepening US yield curve is positive or negative for future style bias. Intuitively it depends on which end of the yield curve is driving the change, the starting point for change, the magnitude of change and innumerable externalities.

In such situations we put significant stock in the utility of decision tree models.¹ These models employ quantitative techniques to identify the most critical factors to explain/predict economic regimes. Essentially, once fed a multitude of factors, the model selects the one critical question to ask regarding regime identification. Having identified that first factor (and thus creating two “branches”) the model then identifies the next best question to further differentiate between regimes, resulting in additional branches. This process continues until the model is unable to improve upon the solution without introducing additional error. The result is a decision tree – effectively a flow diagram – for identification and prediction of regimes.

The strength of this approach is its ability to identify the most useful decision factors in complex multi-dimensional problems. The key weakness is, if misused, this approach can be no more than useless data mining. It is therefore important that we are able to identify a sensible economic rationale for each branch/node of the tree. In addition, by evaluating different, but related decisions, across a range of time periods, we can look for common themes which would signal greater model robustness.

In this brief paper we present the results of decision tree analysis on four regime tests:

1. Asia style bias
2. US style bias
3. Asia size bias
4. Asia style bias within small caps

For each of these tests the model was provided with a wide range of common economic factors (including interest rate data, yield curve data, credit spreads, US business confidence, US market valuation and recession indicators). These factors were selected on the basis of a considerable body of academic research highlighting their importance for economic regime identification. All bond-related data are in yield terms. A description of each variable is provided in the appendix.

The dependent variable in each test was the relevant market bias over the period 12 months forward from the explanatory factors. So for style bias the measure was whether or not value had outperformed growth over the period 12 months after each factor data point. For size bias the question was whether or not small firms outperformed large firms over the following 12 month period.

The maximum available dataset was employed in each test, with the key limitation being the length of MSCI index series for the performance bias measures. MSCI US value and growth indices from December 1975 were employed in this analysis along with MSCI Asia small cap and small style indices from May 1994 and MSCI Asia style indices from December 2000 (all in USD total returns with net dividend reinvestment).

Our focus is on Asia data, but the longer time series for US data can provide indications regarding the robustness of Asia results.

Beginning with US style bias, figure 1 illustrates the resulting tree. All decisions are presented with a true response to the left and a false response to the right.

To interpret this diagram, begin at the top with the deciding factor being the US corporate credit spread (US BAA corporate bond yields less AAA bond yields). Historically, when this has been very low (<0.7%) growth has outperformed value in the US over the subsequent 12 months. By way of comparison, this represents a credit spread level roughly one standard deviation below the long run average over the period analysed. This intuitively makes sense, implying a relative boost to growth stock valuations from the impact of a reduced discount rate on valuations (i.e. lower perceived risk).

The numbers below the credit spread decision node indicate 273 observations in which growth outperformed value over the subsequent 12 months and 206 observations of value outperforming growth. In terminal node 1 we see that there were a total of 76 observations (65+11) for which the credit spread was below 0.7%, with 65 of those coinciding with growth outperforming value over the following 12 months. Hence, a success ratio for this node of approximately 84%.

Figure 1. 12 month forward US equities style bias (December 1975+)

Splitting criteria are applied to each monthly observation to divide the sample into periods in which growth outperformed value or value outperformed growth, e.g. credit spread<0.7%. If the splitting criteria is true, then move to the left on the tree, if false then move right. Numbers below each node reflect the number of monthly observations meeting the previous splitting criteria, split into periods in which growth outperformed value (left) or value outperformed growth (right). E.g. For terminal node 1 there were a total of 76 observations for which credit spreads were less than 0.7%, with 65 of these coinciding with growth stocks outperforming value stocks for the subsequent rolling 12 month period. 11 observations subsequently saw value outperform growth. The ratio of 65:11 in favour of growth therefore identifies this as a growth-biased node.

In the more common situation that the credit spread is above 0.7% (as is the case currently), the model identifies Fed interest rate activity as the next critical question. If the Fed has been raising rates over the last 6 months the model signals a subsequent value bias in the US equity market at terminal node 2 (value outperformed growth by a ratio of roughly 2:1 in frequency terms). This is significant in the current context of a market increasingly anticipating an imminent Fed rate hike.

If the Fed has not been raising rates the model identifies US market valuation as important. When the S&P 500 PE has been below 22x and the long end of the US yield curve has not been falling substantially over the last 12 months (12 month change in US 10 year yields), growth has tended to outperform value. When the long end has been falling aggressively, value tends to be favoured. The former coincides with Fed rate cutting and a weak growth environment favouring growth stocks. The latter is consistent with periods in which high bond yields negatively impact growth stocks (one has to go back to the early 1980’s to see such a mix of characteristics).

Currently the S&P 500 PE is significantly above 22x, and hence the next appropriate question is the size of the yield spread (defined here as US 10 year government yields less the Fed target rate). When the yield spread is at roughly long run average levels or lower, growth stocks are subsequently favoured. Hence, growth stocks perform relatively better even when valuations are high as long as the yield curve is not particularly steep. If the yield spread is wider than average and increasing significantly, growth is favoured (e.g. early stage recovery post recession), otherwise value. For much of this year the market has been at terminal node 5, favouring growth.

Importantly, this analysis suggests favouring value over growth if the Fed is raising rates, and favouring value over growth even if the Fed is not raising rates, but bond yields rise and the yield curve steepens significantly on the back of stronger growth and/or inflation expectations – behaviour evident in recent market activity.

Applying this approach to Asian data (over a shorter time horizon) we obtain similar results. If US 10 year yields are very low by historic standards, growth is favoured over value—terminal node 1 in figure 2 (i.e. low growth expectations for the market overall benefits growth stocks, and they receive a positive impact from a low discount rate). This has been the case over the last couple of years. However, if bond yields continue to edge higher and the ISM (US manufacturing business confidence) has been increasing, value is overwhelmingly favoured. The exception to this is when the Fed has been very aggressively tightening (during a strong acceleration phase of the cycle). If the ISM is very high (>58), value is favoured and when the ISM is increasing rapidly, value is favoured.

Effectively, growth benefits from very low bond yields and/or weak growth expectations. If bond yields increase and growth expectations improve, value is strongly favoured over the following 12 months.

Two more decision trees are provided for Asia size (small caps versus the wider market) – figure 3 – and style within Asia small caps (small value versus small growth) – figure 4.

Historically, small firms were favoured relative to large firms after periods of increasing credit spreads (terminal nodes 2 and 5). This was particularly so if US market valuations were very high (>25x) or lower than a threshold of 21x. Only when the market was trading in a sweet spot between 21x and 25x, when credit spreads were widening, did large firms tend to outperform small firms (terminal node 4). Narrowing or steady credit spreads and Fed rate hike increases have tended to favour small firms (terminal node 1). If credit spreads are not widening, the Fed is not raising rates and the US economy is not in recession then large firms tended to outperform small firms over the following 12 months (terminal node 6).

Critically, if credit spreads are widening and/ or the Fed is raising rates this analysis points to future outperformance by Asian small firms versus Asian large firms.

Figure 2. 12 month forward Asian equities style bias (December 2000+)

Splitting criteria are applied to each monthly observation to divide the sample into periods in which growth outperformed value or value outperformed growth, e.g. US 10 year Government yield < 2.6%. If the splitting criteria is true, then move to the left on the tree, if false then move right. Numbers below each node reflect the number of monthly observations meeting the previous splitting criteria, split into periods in which growth outperformed value (left) or value outperformed growth (right). E.g. For terminal node 1 there were a total of 45 observations for which 10 year yields were less than 2.6%, with 41 of these coinciding with growth stocks outperforming value stocks for the subsequent rolling 12 month period. 4 observations subsequently saw value outperform growth. The ratio of 41:4 in favour of growth therefore identifies this as a growth-biased node.

Finally, in figure 4 we see a strong tendency for small value to outperform small growth after periods of “normal” bond yields – yields less than 6% or greater than 1.9%. This was particularly so when the US yield spread was wider an 2.3% – terminal node 5 (77 small value observations, 0 small growth observations). However, even when the yield spread was narrower (as is the case currently), small value still outperformed small growth by more than 3:1 in frequency terms (terminal node 4).

I noted earlier the importance of common themes across these related investment decisions, which would imply model robustness. Whether evaluating US style, Asia style, Asia size or Asia small style there are numerous common factors. It is evident across all of these that Fed action, credit spreads and US long rates are important decision factors.

Applying these results to current data and extrapolating forward, these models imply any Fed tightening would favour US value over US growth, Fed tightening combined with higher US 10 year yields would favour Asia value over Asia growth, Fed tightening and current US valuations would favour Asia small caps over Asia large caps and this would in turn be an environment consistent with future outperformance of small Asian value over small Asian growth.

Firth Investment Management offers two strategies – a systematic mid-large cap Asian equity strategy with a value bias and a small cap valuebiased Asian equity strategy. We believe both are well positioned for what increasingly appears to be a transition phase for global markets.

Appendix – Data definitions

Credit spread
US BAA rated corporate bond yields less AAA corporate bond yields. Changes are in basis point terms over the period referred to in the node description.

Fed target rate
US Federal Reserve target rate (and upper bound). Changes are in basis point terms over the period referred to in the node description.
US 10 year
US Government 10 year bond yield. Changes are in basis point terms over the period referred to in the node description.

Yield spread
US Government 10 year bond yield less the US Federal Reserve target rate.

S&P 500 PE
Price to earnings ratio for S&P 500 stocks, based on trailing twelve month reported earnings.

ISM
Institute for Supply Management Purchasing Managers’ Index. Changes are in simple difference terms over the period referred to in the node description.

NBER
National Bureau of Economic Research dating of US business cycle expansions and contractions. A value of 1 refers to a contraction phase. This variable is only available with a substantial lag. However, it represents a single proxy indicator for a range of other factors indicating economic expansion or contraction.

¹ Technically, we employ recursive partitioning using the RPART library and R Core Team (2015) – R. A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/.

Written by: Dr. Hamish Macalister