Friday, May 21, 2010

Investment in Computers and Software


I'm a bit pressed for time this morning as I have to go to my son's third grade play, and then a busy day of work ahead.  So just a couple of quick graphs exploring the increasing automation of the US economy.  The graph above (from BEA table 1.5.5) shows all the components of non-residential investment in the economy as a percentage of GDP from 1950-2009.  So this is the same period that the working-age male employment/population ratio has fallen from about 95% to about 80%.

You can see that while non-residential investment has generally run in the range of 9-12%, the mix has shifted over time, with ever more going on computers, networking equipment, software, etc, while less is going on industrial and transportation equipment.

The next graph shows the same data, only broken out as separate lines:


A few points that strike me:

  • The structure line has two booms (associated with the early 80s S&L boom and crash, and the recent housing boom requiring associated non-residential investment).
  • The information equipment/software (IES) line increases to a peak in the 2000 dot-com boom, and then falls back a little, but stays at a high level.  The stuff goes obsolete so fast that it has to be constantly replaced.
  • Note that the IES investment here is non-residential; this is spending by businesses on computers, software, and network/telecommunications equipment.  It doesn't include consumer spending on personal computers, mobile devices, etc.
  • IES investment was not depressed by the 2005-2008 oil price boom; instead, it was increasing very slightly in those years.  This tends to argue against the idea that peak oil will cause de-automation.
  • Although IES investment fell in the 2009 recession, it was much less sensitive to the recession than the other investment components.  Look what happened to transportation equipment in 2009 by contrast!  Presumably, businesses don't see this spending as optional or deferrable.
  • Earlier recessions (1974, 1980-82, 1991) are also barely visible in the IES investment line.
If I had to guess, I would expect investment in IES to stay roughly flat as a fraction of the economy for the next few decades, neither growing massively, nor shrinking, but somewhere around the recent 4% level give or take a percentage point or so.  Most of it will be driven by the replacement cycle, and as long as Moore's law continues, the lifetime of computer equipment will be roughly similar.  A slowdown in Moore's law would lead to a lower level of IES investment as the stuff lasted longer.

5 comments:

Brian said...

"IES investment was not depressed by the 2005-2008 oil price boom; instead, it was increasing very slightly in those years. This tends to argue against the idea that peak oil will cause de-automation."

Maybe, but that depends on your estimation of the effects of peak oil and on your time frame. Personally, I think that, while price will go up, it will be volatility and scarcity over time that really cause the impacts. The combination of increasing volatility, price and scarcity may make future capital planning and financing very difficult.

Most of these technology purchases are the kinds of things that assume access to cheap energy (almost entirely in the form of cheap electricity). If underlying conditions force changes in the utilization of our energy mix, the situation could change. For example, if peak oil places price or availability constraints on transportation fuels, it is possible that the auto fleet might transition to natural gas, placing additional demand on gas and driving the price of gas higher for all uses, including electricity. The same effect might occur with a massive switch to electric vehicles, as that would probably require additional capacity and, by then, coal may or may not be politically viable.

Basically, I don't think you can draw too much from the last oil price run-up relative to the impact of peak oil on technology spending and adaptation. Might not have an impact, or it might.

Brian

KLR said...

Right. "Peak Oil" I think of as meaning real physical shortages and attendant impact on the economy's ability to function properly, not the problems brought on by a price run up, whatever its cause.

BEA is pretty vague with the term "nonresedential structures." What, like a storage shed? Here's their definition:

Nonresidential structures. .Investment in nonresidential structures consists of new construction (including own-account production), improvements to existing structures, expenditures on new mobile structures, brokers' commissions on sales of structures, and net purchases of used structures by private businesses and by nonprofit institutions from government agencies. New construction includes hotels and motels and mining exploration, shafts, and wells. Nonresidential structures also includes equipment considered to be an integral part of a structure, such as plumbing, heating, and electrical systems. Related terms: nonresidential fixed investment, fixed investment.

Moore's Law promises an exponential gain in computing power; what happens when the machine you bought yesterday is obsolete by the time you plug it in - and I mean seriously obsolete? It seems an absurdity, like the situation with currency in Weimar Germany or Zimbabwe, where inflation made it more sensible to use cash as fuel for stoves.

Stuart Staniford said...

KLR:

Nonresidential investment structures are offices, factories, shopping malls, that kind of stuff.

Moore's law is an exponential with fixed constant. So the replacement time stays constant too. Let's say you want to replace every time a new machine is 2x faster, then you will basically replace at the half life associated with the Moore's law exponent, and that is fixed.

Peak oil means maximum rate of oil supply! Some folks think it was in 2005 or 2008. I'm inclined to think it is a bit further out, but still I'd expect the main symptom to be price shocks. (Physical shortages would be more likely a result of price controls preventing the market from clearing, as in the 70s).

KLR said...

Even so we're building up a huge array of vintage tech that needs looking after; this article says the DOD dumps an estimated $10 billion a year on electronics obsolescence matters: IEEE Spectrum: Trapped on Technology's Trailing Edge. You could make some Tainter-esque statement about complexity and marginal returns here.

I like Rapier's term "Peak Lite." That fits the bill of the bumpy plateau we may be in at the moment. There's a definite distinction between an era of price surges hampering economies but not actually involving shortages, and undeniable declining supplies with nations having to outbid one another for resources, or threatening to go to war.

MisterMoose said...

Even if companies and individuals merely replace existing computers with new ones, total computing power will continue to increase. As computing power increases, and as software becomes more sophisticated, we'll see more and more labor-saving systems (think about all the secretary-receptionists who were replaced by "If you know your party's extension, press one."). Only now, we are beginning to see semi-intelligent systems that are beginning to compete with managers and professionals, not just low-level or blue-collar workers.)

This trend will only continue to reduce employment, which is already under pressure, as companies cut costs wherever they can. Normally, I'd look forward to more cost-saving tech, but as an under-employed engineer, I'm beginning to have some doubts.

When I got laid off my last job at age 62 I began collecting Social Security, so I'm not totally destitute. I am taking computer networking courses at the local community college, so theoretically I should be more employable in a high-tech economy. Unfortunately, there are about an order of magnitude more job seekers around here than there are jobs...