Network margins with growth

I got talking to a friend about Metcalfe’s Law: that the value of a network grows at the square of the number of connected users. That is, as users grow in a linear fashion 1,2,3,4,5 the value grows exponentially 1,4,9,16,25. My friend is a hipster barman, but he used to work for an Australian telco, and wanted to know if he should sell his shares. I have no idea. But we did start discussing a puzzle:

Why had telco network operators around the world been such poor investments for the last 20 years?

Surely Metcalfe’s Law should mean these networks should have been increasingly valuable – like google or Facebook?
I’m not an expert and I don’t know the answer. But I did have a hypothesis that I wanted to test. I got the idea from reading Nabokov’s Favourite Word Is Mauve by Ben Blatt.

Rather than all this super clever Natural Language Processing / Machine Learning  I thought that I would try a simple word frequency test.
My hypothesis is that telcos have been too focused on “growth” – and forgot about profitability and “margin(s)”.

So I picked a couple of large telcos and compared with other companies. I didn’t choose the others at random – I figured I would pick a low growth power utility network like National Grid, and a couple of boringly profitable consumer companies like Unilever (washing powder etc) and Reckitt & Benckiser (headache tablets etc). With the value of hindsight, I knew that the NG had done OK.  And that it was a network that supplied power to households, rather than information.  I also knew that the last two companies had done well. My hypothesis was that they’d done well because their growth had been more profitable, and they would talk a lot more about “margins”.

TMT Madness

If you’d suggested to anyone in the TMT bubble that Unilever and R&B had much better prospects than Telcos Media or Technology companies – they would have thought you were mad. But although TMT stocks were all about growth, many of them failed to convert growth into much of a return. Vodafone’s Return on Capital Employed has averaged less than 1% in the last 5 years (according to Stockopedia numbers). Reckitt & Benckiser’s by comparison has averaged 20% over the same time period.

Anyhow – I just did a simple word frequency test on the 2000 Annual Reports. This didn’t involve any programming.  I just went to the documents and did a “CRTL + F” search, counting the number of times “growth” appeared and comparing it to “margin(s)”. It’s pretty interesting:


Comparing frequency of “growth” versus “margin(s)” in 1999 Full Year Results
“growth” “margin(s)” Ratio Share price
Reckitt & Benckiser 29 13 2.2 +858%
Unilever  34  26 1.3 +234%
National Grid 6  1  6 +86%
Vodafone 38 0 n/a -30%
BT 10 1 10 -60%
Share price performance from 1 Jun 2000 (after all companies had released their 1999 Full Year Results) to 01/01/2018

The growth focused Telcos barely mentioned margin – and their subsequent share price performance has been dire. In fact I was wondering whether to give BT minus points, because although the company did use the word “margin” once, (one more time that Vodafone, that didn’t use the word at all in their 2000 report) the context was negative: “increased lower margin wholesale business”.
So my experiment worked.  It’s a positive correlation (above 50%) between the ratio and share price performance… but not perfect.   For instance Unilever talked more about “margin” than Reckitt, but Reckitt has the better numbers (and superior share price performance).
The sample size is tiny.  2 words. 5 companies.  But it’s interesting how the word frequency changes (or doesn’t change) over time.


Comparing frequency of “growth” versus “margin(s)” in most recent Full Year Results
“growth” “margin(s)” Ratio Ratio FY 1999
Reckitt & Benckiser 82 37 2.2 2.2
Unilever  64 41 1.6 1.3
National Grid 38 1  38 6
Vodafone 123 37  3.3 38
BT 6 1 6 10

R&B and Unilever haven’t changed much.  But National Grid has become more focused on “growth”, and Vodafone is now talking about “margin” a lot more.  I’d be tempted to buy some Vodafone, and sell National Grid.  But that’s not advice – that’s just me thinking if I had to back my own research with cash, that’s what I would do.

You might think this approach would have been incredibly valuable for professional fund managers. Any information that persuaded a fund manager to sell telcos shares and bought Unilever and Reckitt & Benckiser would have outperformed hugely.  But sadly I don’t think there is a way to get paid for this simple heuristic. Fund managers like complexity – you can’t just phone up Fidelity or Blackrock and leave a voicemail suggesting they buy / sell billions of dollars of Vodafone shares because you’ve counted two words. Or you can – but they won’t reward you for it. I used to build simple models that worked, but never got very far. Whereas I worked with an analyst who would get things hopelessly wrong, but because he got things hopelessly wrong with a complicated excel spreadsheet model, fund managers gave him lots of votes. 

I guess if you can get paid lots of money by clever people for being inept, it means you are probably not so inept after all.  That statement also goes for the management of Telcos companies, where fund managers voted through pay packages worth millions, despite their shares being poor investments.

My heuristic still doesn’t really explain why Telecom companies have not been more profitable and seen higher margins, when the amount of data and connections has increased massively.  I don’t have a full answer, but I am suspicious of the assumption that more data, more complexity is valuable in of itself. I suppose that the difference between Facebook and Vodafone is that Facebook get users to generate their own content, and it doesn’t cost them much to maintain and grow their network.  Like a railway, Vodafone’s physical network is expensive to maintain, and they find it expensive to acquire new customers, who have come to expect faster speeds and more data.  That’s a simple model – I don’t think spending hours building a complicated excel spreadsheet DCF model with forecasts out 10 years would have given me that insight.

In fact, I have rarely found that more inputs and more years of forecasting in a model reveals an insight that a simpler version doesn’t.  What seems valuable to me, is the ability to use a computer and a simple “rule of thumb” together, to make a better decision …don’t invest in companies that use the word “growth” frequently, while rarely using the word “margin”.

Anti Rival Good

Metcalfe’s Law is an example of an “anti rival good” …the more people share, the greater the benefit each sharer receives. Programming languages are an example of an an “anti rival good” – programmers like it when more people learn python or R or whatever. They don’t see other programmers as rivals who will drive down the hourly wage for programmers, but instead will create more code and techniques that can be shared.  Creating more value. 
So that’s why I’m sharing this idea. Maybe a simple “CRTL + F” word frequency test might be useful for other private investors to try? 

Picture Credit: Photo by Slava Bowman on Unsplash

2 Responses

  1. Hi

    I always enjoy reading your blogs when I come here. A few points on this one

    1. I think the telecom companies were hobbled by massive over-investment in fibre optic cable in the early 2000s. 95%+ of new capacity turned out to be excess capacity. Pages 127-151 of “Capital Account” by Edward Chancellor explains this, you can probably read bits on Amazon or Google Books.

    2. In the UK the telecom companies also made vast overpayment in competitive auctions for mobile spectrum (from the govt’s perspective very well designed by UK academics). http://www.nuff.ox.ac.uk/users/klemperer/biggestpaper.pdf

    3. Metcalf’s law is wrong for large networks anyway: n log (n) is a better rule for the value. The growth is between linear and quadratic. https://spectrum.ieee.org/computing/networks/metcalfes-law-is-wrong

    4. This looks like quibbling, but it’s important: Metcalf’s law is quadratic, not exponential! It’s n^2. Exponential would be e^n, which is much faster. Admittedly you are in good company: almost everyone misuses “exponential”.

    Exponential is when growth of the “thing” from n to n+1 is proportional to the “thing’s” current size. So when the “thing” is small, the growth rate small! Think of the lily on the pond, doubling in diameter every day: when the lily is small, the growth from day n to n+1 is small. Too small to notice. This is the point about exponential growth: it stays very small for a long time, then it becomes large very quickly. When you notice it, you’re probably already too late.
    Aaron Brown’s book “Red Blooded Risk” is good on this.

    5. While I’m here: I liked your explanation of why London prime property has done well, and Russian companies listed in London have done badly.

    1. Bruce Packard

      Thanks – yes I had previously thought that anything n squared was exponential (linear 1,2,3,4,5 v quadratic 1,4,9,16,25) accelerates away, but not as fast as your example of exponential. So I meant quadratic. Thanks for the feedback! I also agree on your other points. The telco’s were too focused on investment and growth, without asking if either of those would lead to higher returns. And there were clues in the text of their Annual Reports that they were too focused on these wrong metrics.