28 February 2018

wHo NkeW

One of the most interesting uses of R popped up on "R-bloggers" today. It's a test of reading. In R? Yes. Moreover, the earliest cite in the article is 1981, and text makes reference to research done in the 70s. Who Knew?

An earlier missive here asked what sense it makes to build computers that are more "brain like". Text recognition has been of interest for some time, and highlights the (nearly?) unique capabilities of the human brain. Not so much for processing Big Data, alas.

24 February 2018

The Asymptote of Progress - part the fifth

If there are any stragglers among the readership who deny the reality of the Asymptote of Progress©, here are a couple of links from comments to a piece on Anandtech.

The first is from 2017, and is generalist. And is, on the whole, optimistic.

The second from this year, on the other hand, is decidedly not.
In fact, depending on what aspect of Moore's Law you look at, it ended a long time ago (power scaling), or recently (cost per transistor scaling) or still has several nodes to go (technical barriers).


And, in conclusion, it gets worse:
Moore's Law is, indeed, ending. But that is only one part of the problem:

MOSFETs have run out of steam, with newest nodes requiring more money for very limited improvements
FETs of all the weird types are not going to save us
Von Neumann compute is reaching its limit

Read that last point. Let it sink in. Offer up a superior alternative. This is his:
On a longer term still, we will end up moving away from von Neumann architectures to something more "brain like."

Of course, we don't even have a definition of how the brain actually does what we call computing. Show me a brain that computes Big Data? It's one thing to conjure up the words to "War and Peace" in a "brain like" brain, and quite another to run multiple regression on a million row data frame in a "brain like" brain. Why would a neuron based compute engine be successful at such a task? All that nonsense about only having a hammer seeing the world as nails. Any answer welcome.

22 February 2018

I Still Hate Neil Irwin - part the tenth

It's the nature of quants to be lazy, despite all the brow sweat they brag about. This laziness leads them to seek out easy data. Which, in turn, leads them corporations, i.e. the micro sector, who collect as much data as they want. And, since corporations are mostly rolling in dough, fat salaries to be had. So long as the quant doesn't find the requirement to pander undesirable. Hog heaven for the quant. The problem is that, as any macro quant will tell you: 99.44% of micro performance is the result of the macro environment. Do the CxO class, and their quant servants, really think Mr. Market's spectacular rise over the last years is due to their brilliance? The CxO class has had, at most, an itsy bitsy teeny weeny yellow polka-dot effect; almost entirely M&A to destroy competition.

So, today is another reason to hate Neil Irwin!! As stated here more than once, Say is crap. If it were true, there'd never be recessions or depressions because business would ramp production (and employment and wage incomes) in the face of falling demand. I dare anyone to cite any company that did so. You won't find one. They follow the Lemming Law: each attempts to conserve cash and profits by cutting production and employment. Thus constructing a downward spiral of global contraction. What the individual sees as rational through his own eyes, leads to a mess when everyone does it.
The latest wrinkle is that the researchers now believe that productivity growth depends not just on the supply side of the economy — what companies produce and what technologies they use to do it — but also significantly on the demand side.
[C]ompanies need to increase production to match demand for their goods, and a shortage, either of workers or of materials, forces them to think creatively about how to do so.

As Homer says, "D'oh!!". And, it's worth noting, McKinsey has an extreme rightwing reputation in the consulting world.
McDonald comments that apart from slashing cost by slashing headcount (increasing employee churn in order to keep costs low), there have been many instances where the recommendations that have come from McKinsey seem to focus on cold blooded cost cutting at the expense of long term staff loyalty, creating a culture of fear and short-termism in the organisational culture.

It is kind of amazing that McKinsey and Irwin seem to ask, "What would Jesus do?". Again, the repeat: go read Baumol, as well. FIRE has eaten GDP, and measuring productivity using tools created for a manufacturing economy will, by definition, show lousy productivity as GDP skews more toward services. Which is not to assert that some other measure of productivity would both be honest and at much higher levels than we've actually seen. After all, productivity is supposed to be a measure of labor output. A few years back, GDP's definition was fiddled to add IP related activity. No one seems to ask the fundamental question: is there any productivity benefit to a mobile phone that runs "Candy Crush" 10% faster???

Here's a long-ish article making these points.
A problem with GDP even when it is being asked to do nothing more than measure production is that it is a relic of a period dominated by manufacturing.

Just like I said. Of course.

21 February 2018

A Brother From Another Mother

We all now know, beyond doubt that Dear Leader is an illiterate moron. With the brain of chicken, as Connie used to say when she did something silly. Orange Julius Caesar does nothing else. So, today brings us some stark contrast.
"How did you end up here, Marilynne? What happened? Was it libraries?"

In 2015, President Obama sought out the novelist Marilynne Robinson, the author of the acclaimed Gilead trilogy, for a wide-ranging conversation. They spoke with the ease and deft mutual flattery of friends, touching on democracy and education, their Midwestern roots and love of literature. Obama's question lay at the crux of their discussion.

Is there any way to imagine Carrot Top even considering books???

Here's the quote that made me stop.
As a child, her teachers told her, "You have to live with your mind your whole life"; they taught her the value of building it, making it worthy. It was, she has said, the most important lesson she ever received.

Since at least high school, I've always answered various existential questions with, "you live between your ears". A point of view that doesn't drive one to keep up with the Joneses, naturally. All these years later, it's more than a tad comforting to find it in print.

An ironic way to close the piece
"Slander" is the story of Robinson's strained relationship with her mother. "With a little difficulty we finally reached an accommodation, an adult friendship," she writes. "Then she started watching Fox News." Her mother and her fellow retirees began to share "salacious dread over coffee cake," fretting over the rumored "war against Christmas." "My mother lived out the end of her fortunate life in a state of bitterness and panic, never having had the slightest brush with any experience that would confirm her in these emotions, except, of course, Fox News," Robinson writes. The essay brings all the abstractions home, makes real and painful the cost of ceding an independence of mind.

19 February 2018

Thought For the Day - 19 February 2018

17 of my classmates are gone. That's 17 futures, 17 children, and 17 friends stolen. But you're right, it always has to be about you. How silly of me to forget. #neveragain
Very sad that the FBI missed all of the many signals sent out by the Florida school shooter. This is not acceptable. They are spending too much time trying to prove Russian collusion with the Trump campaign - there is no collusion. Get back to the basics and make us all proud!
-- Trump/2018

18 February 2018

Shooting Fish in a Barrel

N. Gregory Mankiw is a Marie Antoinette Class quant type, with a very public platform. Recently he expounded on free trade. Much of it is baloney, but this bit
To be sure, expanding trade hurts some people in the short run, especially those in import-competing sectors who have to find new jobs. That fact may call for a robust safety net and effective retraining. But it does not undermine the conclusion that free trade raises average living standards.
[my bold]
deserves shooting. Like a fish in a barrel.

I wonder how he feels about all those coal miners? Not to mention that, being a card carrying economist he knows better, citing "average living standards". Given the continuing income skew towards the 1%, and documented destruction of the remaining 99%, that cite rather then median income (or, as even some Trumpians cite, vis-a-vis 1970s income) which has barely budged since the OPEC oil embargo sent the economy into the ditch. IOW, any data analysis of known skewed distribution, income or wealth or widget weight using mean rather than median (or some other skew reducing/removing measure) is politics, not analysis.

14 February 2018

Two For the Seesaw

Regular reader likely realizes that I don't often just say, "go read this". Mostly because the mainstream pundocracy hasn't caught up yet. Eduardo Porter, on the other hand, does so more than the others. Well, not counting Krugman, naturally.

Today Porter takes on Mr. Market. As has been the case with the rest of the pundit class, he avoids asking the basic question, "where have all the dollars gone?" But he does deal with the obvious question of overall stock prices.
What I contend is that if the American economy behaved in the way that most economists say market economies should, stocks would in all likelihood be cheaper.

The Standard & Poor's 500-stock index increased 8 percent per year from 1970 to 2015, on average.
What makes this particularly puzzling for scholars reared on the classical models of competitive economies is that all this happened despite a persistent decline in real interest rates.

The money quote, buried in the middle
The ratio of the capital stock — the value of factories, machines and such — to the nation's economic output has actually declined a little since the 1970s.

So where have all those dollars gone? Well, to Treasuries and other fiduciaries. Capitalists were, are, and always will be risk averse. Little to no rise in physical investment in a decade of low rates, yet all that idle moolah keeps buying Treasuries, and knocking down the opportunity cost of real investment. And they still won't bite. The obvious reason is that, intuitively, they understand that only massive capital (e.g. Fab 8) can eke out a teeny increment in progress. So they toss in the towel and buy Treasuries. In due time Orange Julius Caesar will demand that Treasury cease the auction and sell at named interest rate. Chaos follows.

12 February 2018

Bad Vibrations

The recent fibrillations of Mr. Market have been the grist for countless pundits. Rather than attempt to precis all of them, I'll just note this one from Jeff Sommer
"What we have seen in the last week or two is minuscule compared with the amount of real risk that is coming in the months and years ahead," said Salil Mehta, an independent statistician with deep experience in troubled markets and their consequences. He was the director of research and analytics for the federal Pension Benefit Guaranty Corporation and for the Treasury's Troubled Asset Relief Program, which was set up to help stabilize the financial system in the 2008 crisis.

Is that a grand oops? Perhaps so. What Mr. Sommer, and his correspondents, miss (and such media pundits do so frequently) is that Mr. Market can only go on a Jabba The Hut binge if he's fed lots of moolah, increasingly. Where does the moolah come from? As well, so far as stocks are concerned, it's the opportunity cost (and imputed risk aversion) of Treasuries. But opportunity cost is a two-say street. Many/most pundits take as an article of faith that "government borrowing steals from private investment". That's never been true, of course. There are, fundamentally, three things to be done with idle moolah. You can stuff your mattress. You can buy Treasuries. You can buy corporate instruments. From left to right, increasing (universally assumed) risk. If you want to make more moolah, you go corporate. If you want safety, you go gummint. And so it went after the Great Recession. Why? If you read up the recent piece on Fab 8, you can see that returns on corporate instruments is limited by science/engineering progress. Ever more Bongo Bucks to spit out chips. The only way that can generate financial returns is if there's an expanding market for those chips. Remember that the 300mm wafer was dead meat more than a decade ago? Still around.
If the foundry can keep its wafers-per-hour production rate on 450mm wafers close to its 300mm wafer production rate, it can produce vastly more chips per hour. This helps reduce costs, provided that the semiconductor economy is healthy and the foundry utilization rate is high.

Hasn't happened. So, Treasuries and private physical investment are competitors, but the former is driven by psychology (mostly) while the the latter is driven by physics (mostly). The added risk, as the Extreme Tech piece tells us, is that cheaper unit fixed cost leads to capital returns if, and only if, output can be sold at high enough price. As we've seen in the Treasuries market, all that moolah chasing a, more or less, fixed supply drove up prices; and down imputed return. Now that more and more corporates find themselves in the tech cul-de-sac of the asymptote of progress, returns there still have risk, but lower intrinsic return. Unless, naturally, floor sweeping wage earners do earn enough to want and pay for the iPhone X. Good luck with that.

And, by the way, here's another quote from the Fab 8 story:
Growing up is realizing that fabs are sophisticated sinkholes of ever-increasing capital expenditure in the eternal quest to outsmart physics, so much so that where there used to be dozens of foundries and IDMs jostling at the cutting edge, there are now just four: TSMC, Intel, Samsung, and GlobalFoundries. Of them, only TSMC and GlobalFoundries are pure-play foundries, only taking orders from fabless chip designers rather than fabbing their own designs.

Don't screw with Mother Nature.

09 February 2018

Where Have All The Dollars Gone?

The recent fibrillations in the market, at least on a post hoc analysis basis, have been driven by a smidgen of increase in wages for the 99%. In retaliation, the Marie Antoinette Class has burned down the asset markets. Spite. The thing is: where have all the dollars gone? The inflation in the asset markets could/can only happen if there's excess moolah chasing the instruments. Right? Treasuries represent the lowest risk opportunity cost to the MAC, and the MAC has shown over the last decade or so that they'd rather hold Treasuries than expand the real economy. Spite. One might even harbor the notion that the MAC would prefer the "return" from catastrophic deflation (held moolah buying a bit more each day as prices plunge) over building out new physical capital.

But, you say, what if there's a sharply diminishing venue for new physical capital? Fab 8 is instructive (see recent missives). What if the notion that new knowledge of the scientific/engineering kind is slowing? What if Treasuries aren't the opportunity cost, but it's really the other way round? What if the MAC look at their world from the other direction? What if they say, "I'm not going to dump Treasuries and make real investment until the real return difference is way higher than it is today"?

If so, then the turbulence of recent days is just pique at the fact of a bit more in the pockets of the 99%. It won't last just because the Era of Low Returns is here to stay. All that idle moolah will, soon enough, go back to chasing Treasuries. As the punch line of the adage of the frog and scorpion says, "It's just my nature".

08 February 2018

Brave New World - part the second

Despite the Pollyanna spewing from some folks, here's the future:
The operations engineer would tell us that this was the most automated fab in the world, where no one moves or touches the wafers at all. GlobalFoundries had built Fab 8 from the ground-up to be highly automated, and the 'vehicles' were the key part.

As one version of these escapades puts it, "It's the distribution, stupid". If capital continues to displace labor, who's going to have any moolah to buy the neato products???? Riddle me that, Batman.

05 February 2018

Dee Feat is in Dee Flation - Part the Thirty Third [update]

So, what's going on with the 1,000 point (or so) drop in the DOW? All together now: "it's the 1%!" Recall all the braying about TARP and QE and such leading to ruinous inflation? Well, as reported here many times, ruinous inflation in stocks and bonds has been going on since 2009; March of that year.

The 1%, being utterly risk averse, mostly have plowed their idle moolah into Treasuries, and some into private instruments. Enough to drive DOW over 26,000. But now that there's just a whif of wage growth in the 99%, they bail. Figures. So, yes, that's really all there is to it. They've stuffed their mattresses, again. But not for long. Economies are caught on the leveling part of the asymptote of progress; there is no steam engine to invest in. You want to live on 10% vig? Hop in your time machine to 1900.

Well, well, well. Today the estimable Andrew Ross Sorkin weighs in:
Sometimes you can have too much of a good thing. Don't forget what set off the plunge on Friday: better-than-expected job growth numbers.

A day late, a dollar short!

01 February 2018


Tomorrow's (in a few hours) the day for the monthly labor report. It's been discussed here a few times, mostly around the reliability of the sampling. 99.44% of damn gummint macro-data is from sampling surveys, not population data. And, if the Right Wingnuts get their way, the 2020 census will be as polluted as a Superfund Site.

Anyway, a recently minted NYT reporter offers some additional thoughts. Worth your time, if you care about macro-data.
[I]t's really two reports stitched together. One, based on a survey of employers, provides information on jobs: how many were created (or eliminated) the previous month and how much those jobs pay. The other, based on a survey of households, focuses on individuals: how many are working or not working, along with information on their age, race, education and other characteristics.

For me, the second most interesting aspect (beyond the difficulty in tracking the SDs of the surveys) of the report is that it is released on the first Friday of the following month. So, BLS/Census have to crunch the numbers in a very short window for this report. As Casselman mentions, reports are subsequently updated, and short window reports being the most volatile.

Thought for the day - 1 February 2018

Well, this didn't take long. Now more reporting on the thin ice bitcoin, et al, skate on.
Tether offered a preliminary report last year from Friedman, the accounting firm, suggesting that it had bank accounts with dollars corresponding to all the Tether that had been issued. But the report was far from conclusive and Tether has never produced a real audit, leading to suspicions that Bitfinex may be printing virtual money backed by nothing.

Off to the Yukon!!! We don't need no damn gummint!!!
"Greater assurances are needed that the trades taking place are in fact legitimate and reflect buying and selling by independent actors," said Mr. Moore, an assistant professor at the University of Tulsa. "Unless and until such oversight is implemented, we cannot trust the exchange rate to reflect only legitimate sources of supply and demand."

I Cudda Bin a Contenda

Not for nothing, but the continuing quest for the "best" intro to R book finds a new contenda. It's name is "The Book of R" by Tilman Davies. Yes, it likely, on purpose, sounds like "The R Book" by Michael Crawley. Moreover, Davies is a Kiwi. A bit of island rivalry? And, yes, while Pluto is no longer a planet, Australia never was a continent.

So, what's the difference between them? Let me count the ways.
Crawley: hardcover, best binding I've encountered in ages, ~1,000 pages, only base graphics, liberal use of color, 2012, expensive
Davies: paperback but RepKover so a beneficial binding, ~800 pages (the page size is the smaller, so easier to tote around), relies on ggplot2, monochrome with some color plates tacked on at the end, 2016, quite inexpensive

If you look at the ToC of the two, you'll see the differing emphasis. Both offer some preliminaries, and then diverge. Crawley offers the remaining chapters by stat/quant procedure, including GAM, Bayes, and survival analysis. Davies, on the other hand, views R as a programming language, so devotes more of the text to such concerns. The furthest he gets is multiple regression.

Davies is a contenda if your interest lies in programming R. If your interest is doing quant studies, you'll definitely need other sources.

It's about time for Crawley to make a 3rd, with a substantial chapter on ggplot2. That would make it a no brainer. Wiley shows it as out of stock, so may be.