Voluntary Disclosure

This is the second of 2 blog posts on PZ Cussons.  I thought I might as well publish it before H1 Results are out on Tues 28th January.

The first post looked at how PZ Cussons had been a 1000 bagger from the 1950s, but had recently (in the last 10 years) lost its way – I used Markov Chains to analyse how the share price had reacted to announcements from the last two decades.  The first decade of this century the share price rose by 10x, whereas in the most recent 10 years the share price has halved (a minus 7% CAGR not including dividends). Along with Burford, which I also own, I’m intrigued by PZ Cussons as a multi bagger that has gone wrong.
This blog post I was going to use text analysis to try and work out what has gone wrong. We will get to that later.

Equity markets closed when announcements are released

Before I do, one difference between equity markets and currency markets that I thought is worth pointing out, is that equity markets are closed overnight. A company should make a price sensitive announcement at 7AM, 1 hour before markets open. The investors and market makers then have an hour to read and digest the news – at which point the price can open at a very different level to the previous nights close.  This has implications for any strategy based on using Natural Language Processing to read announcements and trade on the information. Computers can do a lot of things faster than humans, including filtering through hundreds of pages of announcements. However, that advantage disappears if the computer then has to wait 1 hour for markets to open.

Subtle changes

Hence, I’m not particularly interested in filtering company announcements and using something like Naïve Bayes to separate announcements into immediate positive and negative share price reactions. It just wouldn’t be much use, because generally the market makers react to announcements and adjust prices before even the fastest computer could trade.
Instead, I think it is worth using computers to flag subtle changes, or hidden states, that the human reader finds harder to spot. For instance, looking at how commentary evolves, and whether outlook statements would have helped a long term buy-and-hold investor identify key turning points.  To outperform with the same information, you need to spot strong signals that others, not analysing the text as closely, miss. Like a cryptographer with a private key, who is receiving the same coded signal as everyone else, but only he has the private key to convert the information into meaningful information.  Or maybe someone using Hidden Markov Processes to crack the code, and understand messages for which he isn’t the intended recipient. 

Lazy Prices

One approach that I’ve come across, that might help with “when to sell” decisions for a buy-and-hold investors, (that is: identifying signals that highlight a reversal in a long term trend) is the absence of information. That is making changes to their own “voluntary disclosure” in the management commentary.  There’s an interesting paper called “Lazy Prices” which looks at the impact of companies withdrawing a phrase that they have tended to repeat year after year. Management discussion is voluntary disclosure, the people running the company have a lot of freedom to decide what to tell (and what not to tell) their shareholders. So it makes sense that actively changing disclosure or withdrawing information should be a strong signal that something is up. The Lazy Prices * paper suggests that firms that change their reports in a significant fashion are associated with lower returns, and that these returns accrue over 18 months and do not reverse – implying they are fundamental.

Goal posts unmoved

I spent considerable time with the Quanteda package, ** tracking Key Words In Context around “Outlook” and “Nigeria”, discovering only that management were very open and early to warn about the problems in Nigeria.  For instance warning early in 2012 about the State of Emergency in the country and it’s effect on their business. Later, in July 2013, when the share price was at 400p (close to it’s all time high), the Outlook statement was warning about that the company faced “difficult trading environments in most of the geographies in which we operate”. I bought the shares 2.5 years later in December 2015, thinking that the bad news was already in the price.  But by 2016, the company was warning that:

The liquidity squeeze and restrictions in foreign exchange availability in Nigeria, caused by the fall in the oil price, have created some of the most difficult trading conditions we have seen for some time and I am proud to see our 130 year experience in Nigeria carry us through this challenging period with our brands holding or growing share.

Performance has continued to worsen.  The trouble is if management warn openly about problems then everyone can interpret the same information.  My mistake was to think that performance wouldn’t continue to deteriorate, since management had been making cautious statements for at least 2.5 years.  In fact, we have now seen almost 7 years of worsening performance.  It could have been worse, Atlas Mara – Bob Diamond’s African bank he founded after he left Barclays is down 90% over 10 years.

On the positive side, management’s open communication gives me confidence to keep faith with the company, hoping a revival will eventually come in its key Nigeria market. I’m hoping that the Outlook statement will begin to sound more upbeat next week when the company reports H1 Results. 

Multi baggers gone wrong, can also return to their good old multi bagging ways.  Bear in mind that even Games Workshop saw declining revenues for 6 years, until it started growing dramatically again in 2016.  Impax another multi-bagger whose purchase I timed much more fortuitously declined went from 70p at the start of 2011, to 30p 2 years later, and only hit 70p in 2017.  The share price is now just below to 400p.   

Burford’s cash NAV

I’m still hopeful that text analysis of management commentary can work.  As an example, I did spot with my eyes (old fashioned reading old Annual Reports, rather than programming) that another one of my investments – a multi bagger gone wrong-  had withdrawn information.  Burford used to give cash NAV, saying in 2011:

…We still believe strongly that litigation and arbitration returns are inherently speculative and are most appropriately accounted for by holding investments at cost until a cash realisation has occurred, as opposed to taking unrealised gains into income before a litigation resolution has occurred. Moreover, the appropriate metric, in our view, for Burford Capital’s share price measurement is its relationship to net asset value based solely on actual cash realisations. Thus, for the guidance of investors, we publish a cash NAV figure alongside the requisite IFRS-based NAV, and we encourage investors to consider the cash NAV as the appropriate valuation metric.

But then one year later, withdrawing their disclosure of cash NAV. And blaming “lack of interest from the analyst community” !?!? 

We have also historically published a “cash NAV”, but that measure has not been embraced by analysts and investors, and given the increased complexity in its computation following the First assist acquisition and the relatively modest and clearly identifiable unrealized gain in the portfolio, we do not intend to continue to use or publish it after this set of accounts. We will, however, continue to deduct unrealized litigation gains from NAV for purposes of the Adviser’s performance fee computation, so no change of any substance is occurring in that regard, nor do we propose any change in the dividend policy.

In hindsight perhaps this signalled trouble for Burford shareholders?

Wait for it

Except….wait for it.
The announcement where Burford withdrew its cash NAV metric was April 2012, when the share price was 125p.
Perhaps an investor should have sold then? Well, if you’d sold then you’d have missed out on a 16x bagger, because the shares rose to over £20 in August 2018, before investors began selling and Muddy Waters launched their shorting attack.  

Conclusion: trying to build a portfolio of high performance 10 baggers using computers to analyse text discussion of qualitative factors is hard. That’s my own voluntary disclosure.

Photo by Mpumelelo Macu on Unsplash

*Lazy Prices:

Lauren Cohen
Harvard University – Business School (HBS); National Bureau of Economic Research (NBER)

Christopher J. Malloy
Harvard Business School; National Bureau of Economic Research (NBER)

Quoc Nguyen

DePaul University

** Quanteda Benoit, Kenneth, Kohei Watanabe, Haiyan Wang, Paul Nulty, Adam Obeng, Stefan Müller, and Akitaka Matsuo. (2018) “quanteda: An R package for the quantitative analysis of textual data”. Journal of Open Source Software. 3(30), 774. https://doi.org/10.21105/joss.00774.