Arnold Kling  

Hal Varian on Modern Capital

PRINT
More on Trade and Empire... Moore on Climate Preferences...

He writes,


The only thing one could do with those railroad tracks was carry trains. It would have been fantastic if the miles of excess railroad tracks could have been transformed into highways to service the new growth industry coming on in the next decade: the automobile.

But today, if you have fiber you can carry voice, data, audio, video, transactions, whatever...

That's a prime capability of computers. It means you're less likely to end up with an oversupply in the classic sense, which we saw in these other historical periods. The technology boom and bust cycle isn't eliminated, but it will have less of an effect than in prior decades.


The Austrian theory of the business cycle is that businesses over-invest in capital during booms. So you build too many railroads in the 1880's, and you have a bust later.

Varian's claim is that the capital we are investing in now is more fungible, so that if you believe the Austrian model then recessions ought to be milder.

I think that in general the economy is more elastic than it used to be. Years ago, a decline in housing construction of the magnitude that we have observed this year would have already caused unemployment to rise by a couple of percentage points.



COMMENTS (17 to date)
Chuck writes:

I think so far the unemployment from the construction slow down may be largely illegal immigrants, whose lost employment won't show up on the official stats. Plausible, I think?

John V writes:

Arnold,

I'm no economist by any means, but even I know the ATOTBC is a little more involved than just that.


Come on.

Peter G. Klein writes:

I think Hal is right that fiber-optic cable tends to be less industry-specific, firm-specific, and transaction-specific than railroad tracks. But there is little evidence that asset specificity, more generally, is lower today than in the past -- particularly when one considers the increasingly important role of intangible capital (knowledge, capabilities, reputation, etc.).

There is a huge (microeconomic) literature on asset specificity, mostly in the context of vertical integration, and no consensus about if and how asset specificity can be measured consistently across firms and industries. So, unfortunately, we can do little more than speculate about these questions in the aggregate.

8 writes:

Fiber doesn't stay dark for long because we always want more and we quickly find uses for excess capacity. It's not that the railroad changes to a highway, it's that the amount of goods being shipped is increasing exponentially.

david foster writes:

I'm not very impressed by this argument. The ability of fiber to carry different kinds of data is not different, in economic terms, than the ability of railroad tracks to carry different kinds of freight. If there is general overcapacity in fiber (or data center space, or whatever) the ability to change the mix of data will not prevent prices from plunging.

"Exponential growth" is no guarantee against overcapacity unless there is some reason why it is impossible to expand capacity as fast as traffic is growing. This is certainly not the case with fiber.

Indeed, it is much harder to build new RR tracks than to expand fiber bandwidth, implying that railroads have a broader "moat" than long-haul bandwidth providers.

spencer writes:

Except maybe for fiber-optic cable the durability of most new technology equipment is very short. A rail road engine or coal car or blast furnace would last for decades but the useful life of a computer is only two or three year--especially in a corporate environment. In the 1980s and 1990s it was about two years but it has improved in recent years.

Consequently gross capital expenditures now have to be much larger then they use to be to get the same net addition to the capital stock. To see this look at the depreciation data in the national account. In other words we have to run faster to stay even.

But I expect this is a significant reason why the capital spending cycles are not as big as they use to be.

Ajay writes:

Yeah, not a very strong argument by Hal. Yes, fiber can be put to many more uses than railroad tracks but as David says, it is sort of like railroads carrying new types of freight. The point does hold though that fiber is less likely to be superceded by some new telecommunications channel. Also, I was going to make spencer's point that computers are replaced much faster so Hal is being a bit slippery when he says fiber is long-lived and then starts talking about computers. However, spencer is looking at this the wrong way when he says "we have to run faster to stay even." The reason people replace their computers so fast is not because the computers wear out but because the pace of Moore's law is so fast. So the computers you bought 2-3 years ago are now 25-50% as powerful as the latest ones, a pace of technological change unprecedented in human history, and most companies upgrade just to stay competitive within the market (home cycles can be much slower however as the old computers are perfectly functional for everything you used to do, if not for the latest and greatest software that utilize the added capacity). So it's anything but running faster to stay even, it's more like paying the same price every couple of years to escalate your capacity 2-4X, a trend that will probably last another couple of decades.

It is interesting how both sides mislead to make their point. Hal, presumably a technology optimist, makes an implicit and misleading comparison between the long-lasting utility of fiber and computing capacity. Spencer, the doom and gloom liberal, frames the steady replacement of computers every 2-3 years in terms of durability and capital stock replenishment, when in fact it's because we're currently on the escalator of Moore's law and computers that are significantly better come out every year or two.

Bill Stepp writes:

Re: a more elastic economy, one of the factors in this is the deregulation movement that began in the late 1970s, especially in banking and transportation. Lower trade barriers have also helped mitigate the business cycle.
Rothbard said the Austrian theory posited malinvestment, not overinvestment.

david foster writes:

"So it's anything but running faster to stay even, it's more like paying the same price every couple of years to escalate your capacity 2-4X, a trend that will probably last another couple of decades"..how much of this capacity is gobbled up by the operating system (and various applications) due either to inefficiency or to features that are not really required for someone's tasks? I suspect it's a lot.

For a typical office environment, with applications like e-mail, simple spreadsheets, and word processing, it's questionable whether a 5X improvement in CPU speed has much meaningful impact on productivity. For more structured environments, like a bank teller system that exists mainly to communicate with a mainframe CICS system, this is probably even more true.

spencer writes:

The issue is the impact of capital spending on business cycles and/or depressions. It does not matter if the short lives of computer is due to durability or Moore's law, the results or the same.


We made the same point and reached the same conclusion, yet when I said it was liberal gloom and doom.
I love it when I run into the belief that facts have a liberal or conservative bias.

Ajay writes:

David, the OS doesn't grab much resources as long as you don't use all the features in the install, which most people don't. The additional capacity is used to multi-task like crazy and to allow webapps with their highly inefficient javascript to be viable, allowing new and cheaper competitors like google apps to compete. Your argument is self-defeating because those who don't need the capacity don't upgrade (spencer noted the recently lengthening upgrade cycles) and those who do obviously need it. Unless you're arguing that people are upgrading when they don't need to?

spencer, where do I begin? As you note, there are two separate points here. The main point I addressed is your pessimistic framing of computer upgrade cycles, a point which Hal did not make so I don't see how you made the same point. As for your conclusion, where you do agree with Hal and which I did not address, I wasn't sure what to make of it. How do larger gross capital expenditures imply smaller capital spending cycles? I'm not an economist but I don't see a connection there and it seemed to me that you just pulled that conclusion out of thin air. Anyway, my point wasn't regarding your conclusion but that your pessimistic framing of faster upgrade cycles was wrong.

david foster writes:

Ajay...people eventually have to upgrade if they don't want to run unsupported software. Old software is eventually going to lead to unfixable problems.

All kinds of things gobble up cycles. Web pages with idiotic animated ads often bring this machine (the one I am writing on) almost to a halt. Usually this is a CPU-hogging problem, not a bandwidth problem, since it goes away when a faster computer is run on the same connection.

Resources which are very cheap tend to be used inefficiently, and this negates part of the benefit of their cheapness.

spencer writes:

Ajay -- you do not have to be an economist to see my point. All you have to do is look at it as a businessman.

if you invest in a $2,000 in piece of capital that has a 20 year life you do not have to make that expenditure again for another 20 years. So you can continue to benefit from having that piece of equipment and all you do is include $100 each year in your depreciation expenses.

But if spend $2,000 of a computer with a two year life span after two years you have to spend another $2,000 to replace the first computer so your annual expenses are going to be a lot higher.
That is what I meant by having to run faster to stay even.

I'm sorry, I assumed that someone who knew enough to write your first comment would know enough economics or business accounting that this line of thinking would not have to be spelled out.

Back in the 1980s and 1990s the pace of change in computers and software was such that they were obsolete in two years. But now for standard business uses the pace of change is much slower
and one hardly notices a change in capacity when updating to a new computer after a couple of years.

Ajay writes:

Aah, Spencer, it's just funny to read your obtuse responses. Either you do not understand what I'm saying, perhaps because you read it too fast, or you're intentionally mixing up what I say so that you can compose nonsensical responses. I'm leaning toward the latter interpretation right now. I was very clear about which question might require an economist's understanding. Let me repeat it however because your answer implies that you didn't understand. You said that the differences in historical technological spending explained the differences in business cycles. How does one imply the other? Rather than answer this question, you took my sentence fragment about not being an economist, which was in reference to this business cycle question, and combined it with a different point I made about how running faster to stay even is not an accurate description (what's funny is that even your answer to this made-up question is nonsensical: a one-time buy of long-life capital equipment costs much more than or is roughly equivalent to the cumulative recurring costs of buying computers every couple of years). You do not need to bother trying to answer this question anymore, as anyone who either doesn't understand what I wrote or childishly composes made-up questions out of sentence fragments from my original comment clearly wouldn't know enough about economics to answer it.

Axel writes:

[Comment removed for supplying false email address. Email the webmaster@econlib.org to request restoring this comment. A valid email address is required to post comments on EconLog.--Econlib Ed.]

spencer writes:

Ajay -- all I can tell you is to go look at the data.

Ajay writes:

I suppose that would make sense if our discussion were about the data, rather it's about your purported explanation of the data.

Comments for this entry have been closed
Return to top