Featured Image

The Impact of Sentiment on the JSE Small Cap

Technology and Tips

By Adam Locher  |  March 11, 2019

The FTSE JSE Small Cap (J202) has often been ignored by the average South African investor due to the risk inherent in investing in the constituents of this index. More interest has been in favor of the JSE Top 40 listings and the safety they provide due to their size when considering market capitalization. While value exists in the small cap space in South Africa, identifying factors that can lead to superior stock selection has been challenging over the last few years.

As traditional approaches to seeking alpha look at fundamentals and pricing factors, investors are looking for new ways to generate returns in a bear market. In the quant space, seeking new content sets that can drive the investment process has led to the niche development of news sentiment scores. 

Through data providers such as Alexandria Technology, investors have the opportunity in leveraging pure machine learning, to uncover hidden sentiment, critical facts and key relationships to investment workflows.

Trends in the Small Cap Space in South Africa

As many asset managers in South Africa benchmark portfolio performance against the JSE Capped Shareholder Weighted All Share Index (J433), from the below chart and comparable returns analysis, taking a three-year horizon, the cumulative differential between this index and the JSE Small Cap translates to -.72% (with cumulative total return sitting at 13.9%).

At the peak of the Small Cap Index total return relative to the Capped SWIX, March 2017 saw the index outperform more than 17 %. Following the total returns crossing in November 2017, no clear outperformer can be discerned, adding to the complexity and uncertainly faced by the investment community.

Broker consensus revisions for the number of companies with upward sales estimates are 6 for Dec 2019E and 6 for 2020E. Contrast this with the number of downward sales estimates, the figures are more bearish with 8 and 7 respectively. Please note that only 31 companies were given sales estimate figures by brokers out of an index universe of over 60.

In essence, this tells us that broker coverage is limited for the Small Cap index (50% coverage), leading investors to consider expanding their data ingestion process to cater to additional content providers over and above consideration for the standard price and fundamental metrics.

Secondly, the industry is anticipating a bear market in 2019 and 2020. Thus, in light of the anticipated struggles the investment community faces in South Africa, the search to seek alpha requires out of the box thinking, shifting the paradigm to machine learning and artificial intelligence.

 

 Chart 1: 3-yr Return Analysis J202 Vs. J433

FTSE Small Cap 

Applying Sentiment Analysis

Through the use of a backtesting engine (FactSet’s Alpha Testing in this review), we can isolate sentiment factors with Alexandria Contextual Text Analytics (ACTA) which can be considered for use in a South African quant model, illustrating the use case behind this artificial intelligence content set.

 

The primary goal is to flag factors to assess the relationship between one or more variables and subsequent returns over time. To do this we looked at standard investment factors (i.e. non-Alexandria scores) relative to ACTA factors having an average market impact.

The standard investment factors in the model will cater to:

Size                                        Multifactor rank (MFR) of security market value and total assets

Momentum                        MFR of the relative strength and 90-day security alpha

Growth                                MFR of the return on common equity and the return on average total assets

FCF                                         Free cash flow per share

 

The ACTA scores will cater to the average market impact related to:

Equity news corporate actions, corporate governance, dividends, management, earnings and operations.

When validating a coverage check for the JSE Small Cap, we have (as of this writing) a constituent count of 63 securities. Alexandria cover 57 of these counters in the index, leaving only six securities uncovered. Certainly, representing a greater coverage when looking purely at broker estimates.

Hence, the model has 90% coverage and will be run on a monthly basis over the last 10 years to ensure sufficient historical validity.

When running the model, the first statistical measure we address is the Alpha score, which is a measure of active performance. In the below summary statistics table, we see that Alexandria provides a market beating factor return when focusing on the first fractile.

Table 1: Summary Stats

Summary Stats: Fractile 1

Alpha

Beta

Residual Risk

Information Ratio

Sharpe Ratio

Alexandria Eq Earnings

2.66

0.67

9.35

0.76

0.32

Alexandria Eq Dividends

2.11

2.47

6.94

1.37

0.34

Momentum

0.96

0.69

3.02

0.74

0.42

Growth

0.76

0.90

2.60

0.87

0.41

Alexandria Eq All News

0.53

1.29

9.55

0.12

0.12

Size

0.36

1.14

2.59

0.61

0.31

Free Cash Flow

0.35

1.23

2.26

0.80

0.33

 

As Table 1 isolates earnings as a factor, Alexandria caters to a parameter named “total positive news to total negative news.” This parameter represents the ratio of total positive to total negative news within the Earnings news topic on the date of aggregation.

When aggregating this factor over the last year, we see seven companies with the highest positive to negative news ratio. They are the below counters featuring in the J202:

Table 2: Highest Positive to Negative News Ratio

Name

Sum 1 Year Earnings

Adcock Ingram Holdings Limited

0.6

Mpact

0.6

Spur Corporation Limited

0.6

MiX Telematics Limited

0.5

AfroCentric Investments Corporation Limited

0.3

Clover Industries Limited

0.3

Tradehold Limited

0.3

 

By creating a portfolio with these securities (equally weighted) and subsequently reviewing the portfolio performance from end 2017 to end January 2019 relative to the J433, one can see the predictive power of this data set visually in the below chart.

Chart 2: Performance over time

performance over time

Conclusion

The South African market has seen a significant downturn in 2018 with the headline JSE losing investors over 11 % over the course of last year. The broker consensus points toward 2019 and 2020 as predicting a rough ride ahead as investors face declining returns on their portfolios.

The need to look for small cap securities as a source of alpha in the portfolio construction process and the impact of news sentiment in the investment process creates an opportunity to leverage artificial intelligence to generate market beating returns through backtesting engines.

Niche data sets, add a modern machine learning/artificial intelligence perspective to the task of flagging sentiment factors and further isolating which of these factors are correlated to subsequent returns. This helped us concretely isolate seven securities from the small cap space that outperformed the broader index, adding alpha to a portfolio in a turbulent year ahead faced by the investment community in South Africa. 

New call-to-action

Adam Locher

Senior Consultant

Adam Locher joined FactSet in 2014 and is based in Johannesburg.

Comments

Info

We use cookies to personalize content and ads and to analyze our traffic.

We also share information about your use of our site with our advertising and analytics partners. See details.