A Mysterious Force Took Over Investing. I Know What It Is. (value vs growth, ML)

[Liberty]

https://www.institutionalinvestor.com/article/b1j2j2r6ljl8pf/A-Mysterious-Force-Took-Over-Investing-I-Know-What-It-Is#.XdLLbaRtp48.twitter

 

 

Interesting read.

 

h/t @kylerhasson

[Jurgis]

I agree with the author that it will become harder and harder for humans to outperform computers.

 

I also agree with the author (although he might not be saying this much) that outperforming indexes is hard.

 

However, I think that author's arguments why AI can do better in exploiting and removing value advantage vs. growth advantage are weak. The author tries to provide some basis for their arguments, but I don't think the private equity comparison strongly confirms their hypothesis.

 

I'm not saying there is an easy way to validate author's claim. Maybe the right thing to do would be to see how value returns in less-information-available area vs more-information-available area compare to growth returns in less-information-available area vs more-information-available area. But there are many factors that might skew this too.

 

Personally, I'd hypothesize that growth outperformance might be due to (a) zero-marginal-cost growth in software (b) cheap capital. Though (b) applies to value companies as well. So maybe it is just (a).

[jschembs]

I agree with the author that it will become harder and harder for humans to outperform computers.

 

I also agree with the author (although he might not be saying this much) that outperforming indexes is hard.

 

However, I think that author's arguments why AI can do better in exploiting and removing value advantage vs. growth advantage are weak. The author tries to provide some basis for their arguments, but I don't think the private equity comparison strongly confirms their hypothesis.

 

I'm not saying there is an easy way to validate author's claim. Maybe the right thing to do would be to see how value returns in less-information-available area vs more-information-available area compare to growth returns in less-information-available area vs more-information-available area. But there are many factors that might skew this too.

 

Personally, I'd hypothesize that growth outperformance might be due to (a) zero-marginal-cost growth in software (b) cheap capital. Though (b) applies to value companies as well. So maybe it is just (a).

 

Agree with this. The story started with growth vs value, then spent the rest of the article talking about how increasingly difficult it is for value to generate excess returns, which doesn't have anything to do with absolute returns of growth vs value.

 

Just spitballing here, but it'd be interesting to see how growth vs value under/outperformance overlaid with rate cycles. As rates have declined since the early 80s, future growth becomes increasingly valuable (and of course more sensitive to changes in assumptions).

[Sharad]

As algorithms start taking in all the information and interpreting the data, will COBF add an algorithm section so we can discuss which algo's we are investing in?

[bci23]

How about zero interest rates driving the out performance of growth vs. value? When interest rates are zero, the terminal values of DCFs become much more valuable. When interest rates are high, the terminal values are worth much less. Growth company typically implies low current earnings but high hopes for big future earnings far in the distance. Zero interest rates keep those future high hopes alive and well in investor's minds. Thoughts?

[Spekulatius]

The article states that growth has outperformed for at least 25 years. So in a way, one could say that the timeframe from 2002 to 2007 was the aberration, not the time since 2008. Maybe investors systematically underestimate the propensity of growth to persist and business disruption and overestimate the ability of business to reverse to the former mean?

[RuleNumberOne]

I think it is proportional to interest rates. A higher interest rate will result in low return-on-capital companies getting kicked to the sidelines.

 

Look at today's addition to the S&P 500: ServiceNow (NOW)

- Never made a GAAP profit

- $53 B fully-diluted market cap.

- CY 2019 revenue guidance $3.24 billion giving it a P/S > 16.

 

Do you think NOW will ever be able to pay a dividend? It started as just a website to file IT tickets. We just need to add more such companies to the S&P 500 to lift its P/E. Low-tech, no profits, no dividends.

[oddballstocks]

I guess the question then is will we ever have high rates again?

 

 

[RuleNumberOne]

The only cloud stock I found in the Renaissance Technologies 13-F is TEAM and I think it was around position #20.

 

If something breaks, either too much inflation (e.g. we run out of workers) or too much speculation (e.g. speculation pushes every unprofitable cloud stock into the S&P 500 with a $50 billion market cap) or too much debt, Renaissance will outperform the market in a big way again I guess.

 

 

 

I guess the question then is will we ever have high rates again?

[RuleNumberOne]

The bizarre thing is the low interest rates that have inflated market caps of unprofitable tech companies have also made huge amounts of venture capital available to startups that compete with those same unprofitable tech companies.

 

For example, I think Gusto ($200 million Series D) is a better product than Workday. I can also think of at least one startup that targets ServiceNow ($75 million Series B). Such startups are able to raise large amounts of venture capital.

 

 

[LC]

I guess the question then is will we ever have high rates again?

 

Exactly.

 

Then the question becomes: what enables high interest rates. And the answer there is, scarcity of capital. Now, I would argue that human capital is cheap. Financial capital is cheap. Material capital is cheap. So, what will cause a contraction in capital availability? That is the question.

[Ice77]

I think the traditional measures of value, those that rely more on numbers such as multiples, earnings and cashflow will be disrupted (are being disrupted) by algos and AI, whether in the garb of momentum or relative value, call it whatever. If its easy to calculate for you, it will be easy to calculate for the machines as well and they will get better at it with time, iteratively. The value investments that rest on deep qualitative insights - e.g. that Amazon's 1 millionth customer will be more valuable than the 10th customer, will continue to flourish. That tells me that early stage investing lends itself to less disruption as there just isn't enough reliable data or "value" data. While failure rate is much higher, the rewards for deep insights there represent an edge that isn't going to go away easily IMO.

 

What is obvious to you, will be increasingly obvious to machines as well. What is less obvious to you, will also likely be less obvious to machines (though not necessarily). I'm not advocating people go down the early stage route - we are in the adverse part of that cycle. But if you are looking for areas where disruption is less likely it is those areas which have greater impact of power law (VC, Biotech/Healthcare, Early stage exploration etc). Either that is because of asymmetric information or lack of it or deep non qualitative insights.

[DooDiligence]

I guess the question then is will we ever have high rates again?

 

Exactly.

 

Then the question becomes: what enables high interest rates. And the answer there is, scarcity of capital. Now, I would argue that human capital is cheap. Financial capital is cheap. Material capital is cheap. So, what will cause a contraction in capital availability? That is the question.

 

Incineration of capital?

[SharperDingaan]

I guess the question then is will we ever have high rates again?

 

Exactly.

 

Then the question becomes: what enables high interest rates. And the answer there is, scarcity of capital. Now, I would argue that human capital is cheap. Financial capital is cheap. Material capital is cheap. So, what will cause a contraction in capital availability? That is the question.

 

The excess capital (human. financial, material) is not what it seems; and assumes ongoing continuous external intervention, to maintain 'near perfect' conditions. Continuously keep dosing with a drug, and after a while - the intervention becomes toxic. Ultimately, the patient collapses.

 

We are in the land of negative interest rates - how long do you really think that can be maintained?

We have the addiction to QE - even a mild reduction in the doses now causes economic fits.

And widespread automation, with nowhere for the displaced to go.

 

Value investing relies on stable environments.

Per the metrics, today's ABC stock is cheap/expensive relative to its history - which isn't much different from today.

But if today is a very different place ..... are those comps still relevant?

 

Hence Taleb's bar-bell keep coming to mind.

Find time-tested safe places to keep the majority of your wealth in - and bet the rest in asymmetric bets on market failure.

The what's in your head, the gold bars/bitcoin, bolt-holes in various places, etc.

 

SD 

 

[Jurgis]

I think the traditional measures of value, those that rely more on numbers such as multiples, earnings and cashflow will be disrupted (are being disrupted) by algos and AI, whether in the garb of momentum or relative value, call it whatever. If its easy to calculate for you, it will be easy to calculate for the machines as well and they will get better at it with time, iteratively. The value investments that rest on deep qualitative insights - e.g. that Amazon's 1 millionth customer will be more valuable than the 10th customer, will continue to flourish. That tells me that early stage investing lends itself to less disruption as there just isn't enough reliable data or "value" data. While failure rate is much higher, the rewards for deep insights there represent an edge that isn't going to go away easily IMO.

 

What is obvious to you, will be increasingly obvious to machines as well. What is less obvious to you, will also likely be less obvious to machines (though not necessarily). I'm not advocating people go down the early stage route - we are in the adverse part of that cycle. But if you are looking for areas where disruption is less likely it is those areas which have greater impact of power law (VC, Biotech/Healthcare, Early stage exploration etc). Either that is because of asymmetric information or lack of it or deep non qualitative insights.

 

I agree with you that going early stage might be one way to go.

 

Just a note on biotech/healthcare - I think this is an area that is ripe for AI to outperform humans. Basically you have probability based outcomes for a wide spectrum of prospective drugs/devices. I'd think that slurping in research papers + FDA results/documents + any additional info available ( chemical/bio similarity and related info? ) would provide machine way more info than single person has about the field. And that results in better probability prediction IMO. If I was doing investment AI ML, I'd possibly do biotech/healthcare. ( Klarman might be doing a similar thing by having human biotech team. )

[LC]

Even more so I think the benefit of AI comes in medical diagnostics. Huge structured training sets, if people were allowed to anonymize their health records and sell them, en masse to developers, that is decades of medical history over millions of people. Perfect conditions to train models.

[Jurgis]

Even more so I think the benefit of AI comes in medical diagnostics. Huge structured training sets, if people were allowed to anonymize their health records and sell them, en masse to developers, that is decades of medical history over millions of people. Perfect conditions to train models.

 

Yeah, definitely. This is being done by a lot of companies in a lot of medical areas.

[stahleyp]

I totally agree that interest rates are what's driving the divergence. I also think that applies to real estate.

 

As far as thing that will drive inflation, I think debt forgiveness, free college and free healthcare could drive that.

[cameronfen]

I don’t think it’s AI driving down value investor returns.  A majority of AI investing is trend following.  That’s basically momentum investing and swing trading on a computer ie mostly short term growth investing.  It could be interest rates and it could be as the market has gotten more liquid people are better at providing correct valuations on value stocks than growth stocks. 

[RuleNumberOne]

It has to be interest rates. That is the only way you can explain the difference between CSCO (P/E  14) and ServiceNow (P/S  17).

 

Renaissance doesn't own any cloud stocks in its top 20 positions (except a small position in Atlassian at #20).

 

I think the "fundamentals-based" investors are the ones who buy momentum stocks like ServiceNow.

 

I don’t think it’s AI driving down value investor returns.  A majority of AI investing is trend following.  That’s basically momentum investing and swing trading on a computer ie mostly short term growth investing.  It could be interest rates and it could be as the market has gotten more liquid people are better at providing correct valuations on value stocks than growth stocks.

[cameronfen]

It has to be interest rates. That is the only way you can explain the difference between CSCO (P/E  14) and ServiceNow (P/S  17).

 

Renaissance doesn't own any cloud stocks in its top 20 positions (except a small position in Atlassian at #20).

 

I think the "fundamentals-based" investors are the ones who buy momentum stocks like ServiceNow.

 

I don’t think it’s AI driving down value investor returns.  A majority of AI investing is trend following.  That’s basically momentum investing and swing trading on a computer ie mostly short term growth investing.  It could be interest rates and it could be as the market has gotten more liquid people are better at providing correct valuations on value stocks than growth stocks.

 

Using Renessainces 13-f to explain what they are doing has limited value.  They hold probably are running their algorithm on every liquid stock in the market and hold for a couple days before selling.  To some extent it just so happened those stocks had higher positioning due to favorable momentum factors. 

[RuleNumberOne]

Half of what you say is what works against the other half. It is much easier to rollout a global service, which is why even after many years the return on capital has stayed at zero.

 

WDAY and NOW IPOed in 2012, but the return on capital (as in Profits divided by Capital) have been zero.

 

Compare this to the high-tech companies such as GOOG, MSFT, CSCO, FB, and so on. They had very wide GAAP profit margins right at their IPO because they had wide-moats.

 

What specifically are the companies that deserve the high multiples?

 

It has to be interest rates. That is the only way you can explain the difference between CSCO (P/E  14) and ServiceNow (P/S  17).

 

Renaissance doesn't own any cloud stocks in its top 20 positions (except a small position in Atlassian at #20).

 

I think the "fundamentals-based" investors are the ones who buy momentum stocks like ServiceNow.

 

I don’t think it’s AI driving down value investor returns.  A majority of AI investing is trend following.  That’s basically momentum investing and swing trading on a computer ie mostly short term growth investing.  It could be interest rates and it could be as the market has gotten more liquid people are better at providing correct valuations on value stocks than growth stocks.

 

I think it's more than just interest rates. As Bezos mentioned in his talk in Germany, try building a business around space exploration. It is very very difficult - there is little to no infrastructure that is available for a private company to use. This is unlike the internet oriented businesses where the "railroads" of fiber, towers, routers, servers have been laid out for anybody to use. Moreover, many of these businesses have minimal/zero cost of acquiring a marginal customer, unlike the more traditional brick and mortar businesses.

 

This zero marginal cost, together with universality - its so much easier to rollout a global service now at limited cost than before given the global internet "railroads" - are critical factors underlying this. Most of these businesses are not asset heavy, they are not labor heavy, they have high incremental margins, so high returns on incremental capital and so deservedly command higher multiples in many cases. Sure there are ridiculous valuations out there, there are bubbles, but I increasingly see the traditional value investor not noticing the melting ice cube and the changing

business landscape where Zero Marginal Cost and Universality are becoming common.

 

One could argue that these interenet/SAAS oriented businesses also have to pay exorbitantly for Adwords/FB/AWS et al (by some estimates 50% of all the VC $$ go to that) but the overall margins/returns profile for many of them are still superior to classical businesses. Sure the biz life cycles are materially shorter now but the growths during their peak phases is also materially higher.

[RuleNumberOne]

Let me put it another way to show that I wasn't cherry-picking. Why can't companies that build an app on AWS show a profit? GOOG, FB have their own clouds - servers, switches, databases. They are in a different class altogether.

 

But WDAY and NOW are apps on AWS. Can anyone explain why companies that are an app on AWS never seem to show any profits despite strong revenue growth?

 

On the other hand, Vertex Pharmaceuticals (VRTX) shows a $1 billion profit on $3.7 billion in revenue (similar revenue as the profit-less WDAY and NOW). VRTX is one of the top Renaissance Holdings whereas "cloud" stocks are nowhere to be found. @cameronfen i looked at the last 5-10 13-F filings of Renaissance.

 

GOOG, FB are apt examples of the same. Minimal cost of acquiring marginal customers, universality, not asset heavy etc. If you are just going to cherry pick the companies that didn't work for whatever company specific reasons, that doesn't dispute the point. This is not an argument about the importance of moats but simply that the new age moats are different from older ones as they scale faster, at lower incremental costs with quicker and easier global footprints, with often a winner take all outcome. They may have less longevity but have longer periods of high growth. That's it.

[Spekulatius]

I don’t think the comparison of WDAY and NOW with CSCO is fair as they are totally different business. a better comparison would be with another software company like ORCL and MSFT. the latter trades at ~8.5x sales now.

 

I bought NOW a while ago (for a bounce after a CEO change) and it was trading at ~13 sales, but given the fact that the year is almost over and they have good visibility, the relevant number was ~10x forward sales. While it is true that NOW isn’t GAAP profitable (due to stock option expense), when you exclude this expense, it is quite nicely profitable. My own take is that it can’t be ignored, but when I solely look at revenue growth rates/share and hence incorporate the roughly 5% dilution from stock based comp and their convertible raises, that it compounds intrinsic value quite nicely (~30% annually) and at that point was a better deal than MSFT.

 

Just a different point of view. I missed the run up in SAAS stocks totally and yet, I think there is a lot of folly in this sector, however, some products are quite sticky and the stocks deserve a high valuation. I think NOW and to a lesser extend WDAY belong in this group, while TEAM doesn’t.

[Liberty]

But WDAY and NOW are apps on AWS. Can anyone explain why companies that are an app on AWS never seem to show any profits despite strong revenue growth?

 

Because they reinvest all their margins in SG&A and R&D, because at this stage of their maturity, it's the best way to create long-term value.