Michael Orwin

Long only, growth at reasonable price
Michael Orwin
Long only, growth at reasonable price
Contributor since: 2015
I think I'm better at digging things out of the accounts than assessing management, and I can't remember much about anyone except the CEO. I don't have a strong opinion but on balance I favor keeping Witty as CEO. I can see the argument that's he's had enough years to turn the company around, during which results have been disappointing and the stock has underperformed. That's a powerful argument. I think he's done the right things, though it's taken a long time for the effect to show up.
After years of trying to be nimble like biotech, I think this quote from GSK's transcript is some evidence of efficient R&D -
"I also thought you might be interested to know that, when we look at things like the CMR database, GSK is now running clinical studies 20% faster, so our execution times are 20% faster than our major Pharma competitors. Our enrolment time for studies is 20% faster than our major competitors, and we have the lowest clinical trial drop-out rate compared to major Pharma. Over the period since 2010, we have reduced R&D headcount by a third, and we have reduced our R&D footprint down to two from five global centres, all of which has taken huge amounts of fixed costs out of the organisation. At the same time, we have massively increased the number of programmes which are in the clinic and in advanced clinical development. Those are really some of the things that are driving that rate of return calculation."
That kind of efficiency isn't enough if the products aren't commercial, but with £2 billion of new product sales in 2015, and £682 million in Q4, the target of £6 billion of new product sales has been brought forward from 2020 to 2018. I find that very encouraging, but in business there's nearly always uncertainty about prospects, and here the outcome depends on factors like what the competition brings out, pricing pressure, if any safety issues emerge, etc.
Also, if the industry does worse than expected due to pricing pressure or any other reason, the expansion of vaccines and the consumer business could look very good.
"are they getting ready to make an offer on that ~20% ?"
I don't know, so I can't answer your question, but what GSK are doing seems odd to me and is worth commenting on. The simple version is they've given up the right to refuse "put" options, just so they could put the liability for them on the balance sheet, and got nothing in return from Pfizer or Shionogi. There are complications to that, such as under the original arrangement, to refuse a put option they couldn't make a "subsequent distribution", meaning they couldn't pay a dividend but I don't know how long that would have lasted for each refusal. No-one wants the dividend stopped but the way I see it, there's some finite chance of the put being made at a time when the company needs the cash so much that they'd rather pass the dividend than pay for the "put". It isn't just bad news that could make GSK need the cash, I'm not saying this is likely but there could be an acquisition opportunity that GSK can't take because they'll need to reserve funds for the ViiV "put" liability. I don't mean buying out Novartis, because their "put" is already reserved for, although GSK might want to buy out Novartis at a bit more than the market value.
You've probably read this already in the transcript but for anyone else, it looks like the liability will hit the balance sheet in Q1 -
"so that is how you come up with the £2 billion number, you have got to, kind of, strip out the preference shares before you land on the number that will eventually go into the Q1 balance sheet"
That will probably be excluded from core earnings but hit free cash flow, meaning worse cash conversion between the two in Q1 and 2016. I should have mentioned in my long previous comment where I used my "underlying" leakage figure of 4.73%, that I have not included the effect of having to reserve more for the "put" options, because I find it too hard to predict. The better ViiV and the consumer business perform, the more the estimated market value will be and the more GSK will have to reserve for the puts.
For the details about the current and previous exit arrangements for ViiV, find "Exit rights" in the 2015 Q4 results announcement. There's something about the IPO alternative which I haven't covered here. For the terms about Novartis's option find "acquire its 36.5% shareholding" in the announcement.
Thanks for bringing that up, sorry I couldn't answer your question or keep it simple.
This is about 2020 core EPS and how projections based on GSK's statements are affected by the latest results. I estimate that given "worst case" interpretations of GSK's statements on a "constant exchange rates" (CER) basis, in 2020 the dividend cover would be 1.15 (or 115%). That's for cover from free cash flow (by GSK's definition) plus proceeds from sale of intangibles. GSK have made a positive but uncertain statement about the impact of exchange rates in 2016. Including that, the dividend cover rises to 1.20 times (or 120%).
Readers need to decide how reliable GSK's statements are. Also it's the actual core EPS that gives rise to actual free cash flow that dividends can be paid from, not the CER figure. In other words exchange rates could have a significant effect on the dividend cover.
On the plus side, GSK's original statements are looking conservative. From GSK's 2015 results transcript, about the effect of new products doing well -
"That therefore, it is not unreasonable for you to think about how you lift your 2020 numbers potentially to reflect that and the way in which that might affect your assumptions on pharma margin and all of those good things."
Given the room for error and the variation in cash conversion, rather than make a precise and confident prediciton for 2020, I prefer to say that GSK's statements and my analysis of cash conversion implies they can very likely cover an 80p dividend from free cash flow, on average, in the early 2020s.
The rest of this is about the details.
Core EPS fell by 15% at CER in 2015. That's at the high end of expectations (i.e. good) because they'd expected a “decline at a percentage rate in the high teens”. In the link in my previous comment I'd assumed a "worst case" of -19%, and -15% is obviously better. However there's bad news when you look at the actual growth, not at CER, where I've calculated a change of -20.65% (that's down 20.65%), which is worse than my CER "worst case" of -19%.
GSK give the figure 75.7p for 2015 core EPS, which is (naturally) the result of the actual fall, not the smaller fall at CER. While the statement about growth of core EPS (CER) in 2016 is unchanged, it's now growth based on 75.7p for 2015.
GSK's statement for 2016 is "we will return to growth in core EPS in constant currency terms at a rate that reaches double digits". I assume 9% growth for the 2016 "worst case", which on 75.7p gives 82.513p.
GSK's statement for 2016 to 2020 is that they expect “CAGR of mid-to-high single digits” (CER). I assume a "worst case" of 4% growth. Four years of 4% growth is like multiplying by 1.16985856, and doing that to the 2016 figure gets a 2020 figure of approx 96.53p.
In my "systematic guesswork" piece I estimated an adjusted leakage rate of -4.73% based on the 2012 to 2014 figures (and 2015 is not likely to be a typical year), which implies about 95% cash conversion (from core EPS to free cash flow + proceeds from sales of intanbible assets). Applying 95% conversion to the estimated 96.53p core EPS gives 91.70p (of free cash flow + proceeds from sales of intanbible assets). That covers an 80p dividend 1.15 times (or 115%).
There's also good news on the currency front, but it's not certain - "The currency impact for 2016 is clearly difficult to predict given recent volatility but on the basis of January average rates, we are presently expecting currency will be a tailwind this year of approximately 2% to the top line and 5% to core EPS growth due mostly to the impact of the stronger US dollar. (GSK's transcript for 2015 results)
Incorporating the good-but-uncertain news means more calculations. Applying the 5% currency benefit to 2016 core EPS -
82.513p * 1.05 = approx 86.64p
Then applying the 4 years of "worst case" 4% growth -
86.64p * 1.16985856 = approx 101.35p
Applying 95% conversion to the 101.35p estimate of 2020 core EPS gives 96.29p (of free cash flow + proceeds from sales of intanbible assets). That covers the dividend 1.20 times (or 120%).
The second (higher) estimate of core EPS combines -
2015 core EPS of 75.7p (actual, not CER)
2016 growth of 9% CER, and +5% from currency effects
2016 to 2020 growth of 4% CER
A problem with GSK's statements and any method based on them, is that they (reasonably) don't predict the impact of exchange rates with much confidence for 2016, or at all from then to 2020.
The leakage rate of -4.73% based on 2012 to 2014 figures was given in my "systematic guesswork" piece. It's based partly on excluding categories that I think are either avoidable or not likely to repeat, for example I excluded tax because over time the cash payments for tax are likely to be about the same as the tax charged in the income statement.
I've written "Estimating GSK’s free cash flow per share in 2020 given “worst case” core EPS" http://bit.ly/1nE3vvY . I conclude that there’s reason to believe that on average the dividend will have cover from free cash flow even in the event of the “worst case” core EPS implied by GSK’s statements. The conclusion depends on my own judgment and on the plausability of the statements by GSK.
"For us, as GSK current or future small shareholders, why should we care about conversion rate?"
For a given amount of earnings, the higher the conversion rate, the more cash flow there is.
The lower the conversion rate from core EPS to free cash flow, the higher core EPS has to be for the dividend to be covered by free cash flow.
Management have made statements about the growth of core EPS up to 2020. If you think the statements are plausible and you can estimate core EPS in 2020 from them, it won't give you any idea of what free cash flow per share will be unless you have some idea of what the conversion rate will be.
If you think core EPS is totally unpredictable, then the conversion rate from core EPS is no use.
I hope that helps, but if not, you could try the link at the top, find "cash conversion ratio".
The market already knows the bottom lines in the cash flow statements, so I don't think the bottom line in this piece has any big implication for the stock. What this piece does is to allocate the leakage (between earnings and cash flow) to the various accounting areas (major restructuring, intangible assets etc.). That lays the foundation for further analysis. The further analysis requires judgment. I applied my own judgment in the "guesswork" piece and the result was much less leakage. The "less leakage" doesn't change the past, and might not accurately predict the future, but I believe that at the very least, GSK is not doomed to repeat the poor cash conversion since 2012 (when they first disclosed core results and gave the reconciliations I used). For any given core result (such as core EPS), better cash conversion means more cash cover for the dividend.
Also the site didn't pick up the aggressive accounting by Tesco, with good M-Scores for Feb 2014 and Feb 2015 (at the bottom, here http://bit.ly/1nYCeoh ).
This confirms the poor M-Scores for Vodafone http://bit.ly/1nYCgfU for Mar12 and Mar14 (at the bottom).
BTW I should have said "I'm sure you know" (about low M-Scores being good) not "probably", given your use of them.
I've put the calculations here http://bit.ly/1KRVkl6 in case anyone wants to check them.
That's a good point. I looked up AstraZeneca and they have a better M-Score http://bit.ly/1nYwHy8 but I take your point that sound UK companies have had bad M-Scores "at some moment in time".
This is just a technicality which you probably know, but a "low M-Score" is actually good.
Thanks for the comments.
The "bottom line" is in the comment where I gave corrected figures for the "leakage" between earnings and free cash flow.
2012 to Q3 2015 ~ 44.67%
2012 to 2014 ~ 30.20%
The figures are for the percentage of Profit attributable to shareholders that didn't get to free cash flow (by GSK's definitions) plus the proceeds from selling intangible assets.
The detail matters as well though. for example the tables at the top show how much leakage was due to the change in working capital. The changes in working capital are not likely to repeat (assuming the accounts are prepared honestly), and it's fairly common to exclude the change in working capital from cash flow. In my "guesswork" piece (which I've linked to) I apply judgments such as the change in working capital likely to be zero (or as likely to go down as up), and end up with a much lower "ongoing" leakage (or cash conversion rate closer to 100%). I kept that separate from this piece because it depends so much on my own opinion. Using the data I've shown, anyone can apply the same technique using their own judgment.
I found a mistake. Where I said -
"Using those figures to adjust the leakages in the table at the top of the article, the total leakage (including “other”) adjusted for non-controlling interests and the sale of intangibles is –
2012 to Q3 2015 ~ 42.12%
2012 to 2014 ~ 28.65%"
I'd subtracted the adjustment, which only gives a rough approximation. I've recalculated and got:
2012 to Q3 2015 ~ 44.67%
2012 to 2014 ~ 30.20%
"Very few people who comment on SA understand non recourse project debt"
I used to comment about the non-recourse debt fairly frequently. Back in August 2014 it was stated that the project debt was not amortized:
"Unlike the corresponding asset, this debt is non-amortizing and the full balance is extinguished upon the last lease payment typically 20 years after the project is sold, at which time a large GAAP gain will be realized. Accordingly, the non-recourse debt will always exceed the corresponding asset balance." http://tinyurl.com/zut... from "SunEdison's (SUNE) CEO Ahmad Chatila on Q2 2014 Results - Earnings Call Transcript", Aug. 7, 2014 (seekingalpha.com).
I have not found that repeated in the two most recent transcripts, but it doesn't seem like the kind of situation that's likely to change suddenly. I found this about it in the 10-K for 2014:
"As a result, most of our lease payments are reported as interest expense, the principal of the financing debt does not amortize, and we expect to recognize a gain upon the final lease payment at the end of the lease term equal to the unamortized balance of the financing debt less the write-off of the system assets net book value."
Although it's not likely to change I suggest investors check for a similar statement in the next 10-K.
Back in November 2014 the CEO said that non-recourse debt was not included in covenant calculations. "Note, for the purposes of our debt covenant calculations, non-recourse debt as well as our convertible bonds are not included. Overall, we remain comfortable with our balance sheet and our liquidity position." http://tinyurl.com/zut... from "SunEdison's (SUNE) CEO Ahmad Chatila on Q3 2014 Results - Earnings Call Transcript", Nov. 6, 2014 (seekingalpha.com).
I can't find "covenant" at all in the latest two transcripts. I haven't checked the transcripts between Nov. 2014 and Aug 2015. I haven't found anything about it in a 10-K or 10-Q. "Not found" does not necessarily mean it's not there, but equally it could be because non-recourse debt is now included in covenant calculations. I stopped commenting about SUNE's project debt being non recourse when I noticed it had not been excluded from covenant calculations.
There's also "SunEdison - Now with $739 Million in Extra Recourse Debt?" http://tinyurl.com/hok...- November 16, 2015 (bloomberg.com).
While the non-amortization of the project debt means that debt is overstated, there's some doubt about non recourse debt staying non recourse. Note also that the company did not clearly announce the change from non recourse to recourse, and stopped returning emails and phone calls from CreditSights who broke the story. At the time, CreditSights weren't sure if they'd spotted a real story or just a typo. IMO if you aren't confident about knowing SUNE's finances better than CreditSights, you can't be confident about having a good understanding of SUNE's finances and an investment in the stock is speculative.
Morningsidepark - I don't mean to imply that your remarks about non recourse debt are wrong, just that there's other information that needs to be considered.
I'm long SUNE but only have a quarter of my original holding. I wasn't sure if the expansion into wind was good, I didn't like the Vivint acquisition, and I can't figure out how good the projects are and how well they'll perform in the long term.
Thanks canidam. I don't think I can summarize it better than the second and third points in the summary at the top, unless there's a reader who doesn't understand "mean" in the context. It's the usual average, for example the mean of 6, 6 and 9 is 7, because that's what you get if you add the numbers and divide by 3 (because I added 3 numbers). I said about core EPS "The projections for 2020 range from 102.38p to 128.20p, with a mean of 114.85p.".
The other point is about the chance of getting 2020 core EPS at the low end or at the high end of the range, compared to a value somewhere in the middle. If you roll a dice the smallest possible number is 1 and the biggest is six, and a low number like 1 or 2 is just as likely as a middle number like 3 or four or a high number like 5 or 6. But if you roll the dice six times and add the numbers you get, the smallest possible sum is 6 and the biggest possible number is 36, and both of those are very unlikely. You are much more likely to get a number around the average of 21 than at the high or low end. When I say that the outcomes for 2020 core EPS are evenly spread, I could translate that to the outcomes being more like rolling a 36 sided dice once, than rolling a normal dice six times and adding up the numbers. That's not a perfect analogy, for example the dice would have to be numbered from 102.38p to 128.20p rather than 1 to 36.
In finance it's common to just assume a "normal distribution", which is like the sum of rolling a dice many times, with a vanishingly small chance of getting the smallest or biggest possible values. If you've heard the expression "fat tails" when it all goes wrong, it means the extreme cases were more likely than was thought, as a result of wrongly assuming a normal distribution (which has thin-tails either side of a peak).
I recommend reading an old PDF: "Buffett Lecture at the University of Florida School of Business October 15, 1998" http://tinyurl.com/j5l... . Find "You were rumored to be one of the rescue buyers of Long Term Capital". It's about a hedge fund run by Nobel-prize winning statisticians, that was too big to fail, and Buffett was involved in the bail out. When it mentions a "six Sigma event", it means an event that's stupendously unlikely, IF the true distribution is normal, as was (wrongly) assumed.
Getting back to GSK, I believe that the even spread of outcomes means that investors should probably be wary of betting a lot on GSK's 2020 core EPS not being at the low end of the range I've given. I'm sorry but I can't avoid that sentence being hedged and complicated, because investors vary in the risk they should tolerate, no-one pretends that GSK's statements are 100% reliable, there are reasons to believe that GSK's statements are conservative, and there are various factors which could interact in complex ways.
I hope I'm helping to clarify and not just adding to confusion. There's also a risk that I've over simplified. I'll repeat that the work depends on GSK's statements and the way I translated them into growth figures.
All I have done is analyze the implications of GSK's statements. I have not put a probability on those statements proving accurate, and I don't recall GSK doing that either.
For anyone who knows what a probability density function is, I've estimated one for 2020 core EPS on http://bit.ly/1mgJcUt
Truebluealpha, thank you for your comment. I expect most readers will agree with you. With hindsight maybe I should have just stated my refined conclusions as a comment under my previous article and not explained how I derived them. However I prefer showing my work to asking readers to just trust me.
I have a problem now. I want to write about GSK's cash conversion from core EPS to free cash flow. I don't have a test audience and I won't know if it's too complicated before publishing it.
"one new blockbuster drug approval and the projections you made will completely change."
That's a fair point, but I also think it was worth looking into the implications of GSK’s statements about growth as objectively as I could manage.
I had the pipeline in mind when I wrote -
"If you believe GSK’s statements are conservative rather than accurate without positive or negative bias, then it’s reasonable to believe that bottom-of-the-range outcomes are unlikely."
I prefered to focus on the implications of GSK's statements about growth, rather than elaborate.
Earlier this month I wrote a piece with "Upside In The Pipeline" in the title.
I'm reluctant to risk spoiling work I hope is fairly objective with an opinion that could prove wrong. I don't claim my judgment is better than other investors' judgment. I believe I can figure out some bits and pieces but I'd rather let readers put them together themselves. I hope that isn't too aggravating.
I'll guess that everyone understands a range of projections, and the problem is in understanding the relevance of the distribution within the range.
Suppose I'd projected a range between 20 and 30 with a mean of 25. That does not tell you if the outcome is more likely to be in the middle or not. Now suppose I could break that result down into five cases. If they were like this:
20, 24 ,25, 26, 30
then the outcome is more likely to be around the middle. If they were like this:
20, 21 ,25, 29, 30
then the outcome is not likely to be around the middle. If they were like this:
20, 22.5 ,25, 27.5, 30
then the outcomes are evenly spread.
For GSK's core 2020 EPS, the outcomes are evenly spread.
Does that help? Does anyone want me to explain why it matters?
Thanks matratra! I still regard the article as a failure if most readers don't get it, but it's nice to know someone likes it.
I guess this one's a communication failure. What I've done is to take GSK's comments, like "CAGR of mid-to-high single digits", and put them into figures, like low case 5% CAGR, mid case 7% CAGR, and high case 9% CAGR. From the figures I derived a range for 2020 core EPS. I also showed that results at one end of the range or the other are about as likely as a result in the middle, and tried to explain why that matters.
I believe in showing my work so it can be checked, but in this case it added a lot to the length and complication.
I'll try to answer specific questions, but faced with general bemusement, I don't know how to explain it any simpler.
Thanks for the feedback.
"No discussion of mega downside insurance factor. The worse things get the more the possibility of a profitable buyout/split up/reorganization."
I'm not sure if you mean in the article or in the comments. I touched on something like that, without the brevity and without covering your point exactly, in the paragraph with "one possibility is a further reorganization".
Thanks for the comment Janedoeseeker, but I don't think you should rely on me for your stock selection. I'm happy to take credit for my research but no-one should confuse research effort with judgment. I only aim to inform so that some investors have a bit more information to base their own judgment on.
Thanks for the encouragement! I haven't nailed exactly where the problem is in converting profit to cash. I've got the various reconciliations from core earnings to GSK's own free cash flow, lined up in a spreadsheet, and you'd think most of the problem was between "Net cash inflow from operating activities" and "Free cash flow", but the items in that reconciliation are mostly costs that are in core EPS, for example "Interest paid" is deducted, but it isn't very different to "Finance expense" in the income statement and in core earnings. I'm still digging, but if anyone knows where the cash conversion has been failing, please say.
"1. Insurance pricing would adjust to higher interest rates. Premiums would go down, but probably not as much as the economic benefit of higher rates."
Here's a quote about how low interest rates have been some restraint on the competition. From "Markel's (MKL) Q3 2015 Results - Earnings Call Transcript" Nov. 6, 2015 http://seekingalpha.co... (seekingalpha.com), Mike Crowley, President and Co-Chief Operating Officer, said -
"Obviously, the investment side was the toughest part of the house for us. But we breathe the same air as everybody else. The investment operation – environment we are operating in is the same that every insurance company is. So while everything is competitive, I don’t think anybody is under the delusion that they can operate on a sloppy fashion on the underwriting side and make it up on investments. So the nature and tone of competition and where that bar is set in terms of underwriting profitability is lower this time around than what it would have been in previous cycles, so that’s somewhat good news. It’s hard to get the volumes and we have to compete for every piece of business that we have. But I think everybody is competing in a more disciplined fashion than they were in previous cycles. And that’s a function of both lower level of investment returns and better underwriting data and technology that gives people better insights into what the costs or what they are writing are."
The "better underwriting data and technology" are good to the extent that they stop irrational pricing, but they might also level the playing field to some extent. I realize that technology is not the same as prudence, but IMO it's easier to be imprudent when the situation is opaque. An over-optimistic insurer would not buy a security for $100 if the only benefit was $95 in five years time, even if that was a better deal than the insurance it was writing. There'll be a limit to what software can do, particularly in assessing catastrophe risk.
About acquisitions, from the 2014 10-K -
"The Company does not discount reserves for losses and loss adjustment expenses to reflect estimated present value, except for reserves assumed in connection with an acquisition, which are recorded at fair value at the acquisition date."
If acquired reserves reflect estimated present value, I don't see how you can regard them as part of the float without double-counting the benefit from the delay between acquiring the liability and having to pay out. I don't see how the statement is consistent with -
"Our A&E reserves are not discounted to present value and are forecasted to pay out over the next 40 to 50 years."
where there's nothing about acquisitions. On the plus side -
"We do not purchase or sell finite reinsurance products or use other structures that would have the effect of discounting loss reserves."
though I don't think that's as common as it used to be.
I'm long MKL
That's interesting. I came across Element Six http://bit.ly/1Pb7Lip a while ago. They make artificial diamonds, with some hi-tech applications like all diamond optics for high power lasers. Element Six are just a tiny part of the massive miner Anglo American.
Apparently, 45% of Americans aged 18 to 35 prefer synthetic to natural diamonds, possibly for ethical reasons ("blood diamonds"/"conflict diamonds"). That's from "Want to Make a Diamond in Just 10 Weeks? Use a Microwave" http://bloom.bg/1Pb7Lis . Unfortunately you can't really make diamonds in the kitchen!
There's a piece about BDCs on ft.com, "Shadow banks push for easing of loan caps" by Ben McLannahan http://on.ft.com/1NQ6WWa . There's disagreement about whether or not easing restrictions would increase the risk of failure, and I suppose most people's views depend on what they already think about deregulation. My view is that allowing more debt for the same equity would probably increase the risk, which supports investing in the more prudently run BDCs, although there are other factors.
About gold, and this is probably unlikely anytime soon due to practical difficulties - "Senate Votes To Legalize Space Mining" http://bit.ly/1lL7iqH . There's gold in them thar asteroids - Wikipedia's "Asteroid mining" http://bit.ly/1lL7iqI . This on ycombinator.com claims there's more gold in one asteroid than ever extracted on earth, but I don't know where their fact is from, http://bit.ly/1lL7iqK .
If gold from space flooded the market, it would count as a "black swan" event to gold bugs who didn't know about the possibility or discounted it. I'd say "black swans" ought to be included in risk assessments and Kelly bets, although it's not easy to quantify them (and for non-survivable events there's no point anyway). That might be a point in favor of half-Kelly.
I think valuation and volatility are related. If you trade on valuation, you want volatility. On Wikipedia's page about Ben Graham's description of "Mr. Market", http://bit.ly/1QHkfyj , point 6 is "Will offer you a chance to buy low, and sell high.". High volatility can mean maybe you should wait or be underweight, but to take advantage of it you have to buy at some point.
There are people who don't seem to pay attention to valuation. I've nothing against reinvesting dividends, but some people who do it seem determined to keep reinvesting no matter what, at least that's my impression from some comments I've read. It means the gains they make by reinvesting when their stocks are down are probably lost when the stocks are overvalued. Then there are pundits, who's jobs depend on having enough tips each week, usually buys.
I don't mean to knock theory, but I'd guess maybe you'd write a little less of it if you were finding great bargains in the market?
I've also been thinking about rebalancing between different asset types, though all I've done about it is a little selling as valuations moved higher, and a little buying when Asian stocks had a bad day.
A while back I noticed that since inception, NASDAQ outperformed the S&P 500, but with more volatility, and I'll look into buying a NASDAQ tracker if it takes a big dive, then sell when it looks too expensive relative to the S&P 500. I've forgotten the details so I can't absolutely promise that NASDAQ had a higher return including dividends.
I've done some research using data from multpl.com http://www.multpl.com . After calculating CAPE (cyclically adjusted PE) against its 60 year rolling average, and charting that against the future ten year total return, I made a "bubble avoider" spreadsheet table (with no optimization, just getting thresholds from looking at the chart). It works by selling progressively as valuations go into bubble territory, and buying on the way down. By using thresholds and trading just once a year, it made a profit instead of just reversing previous trades as the bubble collapsed. It didn't make very big profits, and didn't trade much outside of the Dotcom bubble, but it avoided the risk of a large holding of stocks in a bubble. Obviously a backtest doesn't prove a strategy will work in future, even if the data wasn't tortured. I think if a real full-blown bubble developed I could probably beat my simple automatic system, and so could most investors who didn't go crazy.
There's also data for Treasury rates on multpl.com http://bit.ly/1dsUTns . The link is for ten year rates but there are links on the page for other periods. I'm thinking that with the S&P 500 data that might be enough data to backtest rebalancing between equities and bonds. I'm never sure of such things unless I actually try and it works.
One problem I have with multpl's data is they credit as sources Standard and Poor's and a book by Robert Shiller, and not being a lawyer I don't want to publish charts and have S&P, Shiller or multpl suing for breach of copyright.
Ben, it looks like your "Cash flow from operations less changes in working capital" is before "Taxation paid" for 2012 to 2014.
I like the pharmacy automation acquired from CareFusion. See Rowa http://bit.ly/1PAjIhx and Pyxis http://bit.ly/1PAjFCo . Automatic storage and dispensing eliminates human error. In a manual system, drugs are filed alphabetically, and very similar names can be for very different drugs. Packaging for generics often features branding rather than the condition the medicine is for, which also makes mistakes easy for busy staff. I suppose there's some risk of competition from cheaper models, but buyers will want to be sure a system is reliable. Now there's a Californian hospital where robots are distributing drugs - http://bit.ly/1PAjIhB (YouTube). I hope BD keep up with that because it fits with pharmacy automation.