Colin's Journal

Colin's Journal: A place for thoughts about politics, software, and daily life.

November 21st, 2002

The economics of computing power

There’s been a running theme through Brad DeLong’s Semi-Daily Journal regarding the tech boom, and the question of the economic growth due to advancements in technology. An aspect of this question is put nicely in this short post regarding the investment in computers and peripherals.

The question being asked is how can there have been an excess in capital spending growth on computers and surrounding technologies, if even today, we are seeing steep growths in “real” expenditure. The problem with this questions is of course in the definition of “real” expenditure. The real expenditure is calculated based on the amount that would have had to have been spent in the baseline year (in this case 1996) to purchase the same thing. The problem with this measure is that the extra computing power purchased today versus 1996 is not really used for anything. The fact that it’s present might add slight benefits to the user experience, but it’s economic impact is actually very low. Whether you are using your 2002 computer or your 1996 computer it will take the same length of time to write that memo or send that email.

The point raised is, however, correct: there isn’t a huge overhang in capital spending on computers. There is some overhang of course, mostly in telecoms, but also present elsewhere, it’s just not that large. One reason for the size of the gap being small is simply that the life cycle of a computer hasn’t really changed much. After 3-4 years you need a new one. It’s not simply a matter of the latest software requiring it, nor the peripherals not being compatible with it, but that it simply wears out. Hard drives start failing, keyboards need replacing, even mice can only do so many miles. Laptops are a particular problem of course – they travel around the world and have more moving parts. The result is that after a while it costs more to keep an old computer going than it does to replace it. When the replacement is made it’s with a machine of a comparable initial cost to the first one but it will, on an economists scale, be considerably better. In practise their was no option to choose a machine of the same spec as the original for a lower price, and so the newer machine will be a better specification regardless of whether or not this was a purchasing criteria.

The other major reason why so much of what was invested in the tech boom has not created a large overhang is nothing to do with hardware, but is instead about software. For any given system or project the hardware costs are always lower than the software costs, and yet in an asset sale the hardware retains it’s value far better than software does. Most software is developed or tailored for a particular situation, and so it has no value outside of that situation. During the tech boom lots of software was developed that is now simply irrelevant, has no value, and so can not contribute to an investment overhang. Unlike the hardware it’s not possible for someone else to come along and pick it up and use it for a new venture – in fact it would cost more to do so than to start from scratch.

In summary: most of the capital expenditure made through the tech boom has to be written off. The hardware that can be sold off does create a modest overhang (except in telecoms where the hardware has a very long shelf life), particularly for servers and other data centre infrastructure, but nothing in comparison to the total expenditures made during the boom.

Comments are closed.

November 21st, 2002

Messing around with sound

I’ve been messing around with sound – trying to record the output from a music player into an mp3 file. I’ve managed to do it, but it took some messing around. I can make the player use esd to play the sound, and then use esdmon to get the sound stream. The stream however is not handled by lame, so I have pipe it through sox first to convert it. The reason it took so long to figure out however was that esdmon would fail if the DISPLAY variable was set – yuck.

In better news I now have a link assistant in my weblog program for quickly generating links to sites. Nothing special, but it should be a time saver!

Comments are closed.

Copyright 2015 Colin Stewart

Email: colin at