There are odd things hiding in the kinds of budget numbers collected by companies like IDC and Compustat. One of the oddest I saw recently was the claim (made in computerworld) that IT support costs per knowledge worker have fallen over the last two years - and that the average is now in the $8,685 per year range.
That was intriguing, but as it turns out explainable in terms of the interaction between budget management and Microsoft's development and release cycle. What seems to have happened was that some expenses surged in 2003/4 in response to XP/2003 uptake and fell in 2007/8 in response to Vista refusnikia, some expenses have been reclassified, and many continued their upward climb.
By themselves the IT expense numbers give us no good guidance on the overall effect on the return organizations get from their IT dollars because they don't cover user side costs or benefits. If, for example, users are spending more time waiting because server virtualization slows response, then the savings this technology offers the data center could be more than offset by user productivity change -but we'd never know from the published numbers.
Look at overall revenues and performance, however, and you do get a most intriguing hint about what's going on. On an across the board average basis the rate of change (not the amounts themselves, but the rate at which they change with each budget cycle) in IT expense appears to be positively correlated with both absolute spending and spending per employee.
Sounds like a "Duh?" moment doesn't it? but we're tripping the truly light fantastic here as we move from IT finance to economic theory - because what these numbers suggest is that IT expense may be dominated by one or more Giffen goods.
This bit from a Robert Jensen contribution to the Freakonomics blog about an attempt to find a Giffen good will explain, first what this is and secondly why it's interesting in an esoteric sort of way:
So, what's a Giffen good? It's a (theoretical) violation of one of the most sacred and holy laws of economics: the Law of Demand. It has excited and intrigued economists for over a century, though no verified example had ever been found.
The Law of Demand says that if the price of a good goes up, the quantity demanded decreases. A Giffen good is one where when the price goes up, the quantity demanded increases. It's named after Sir Robert Giffen, a 19th century British civil servant and economist who is believed to have first suggested the possibility.
For most companies the biggest upward change driver in IT costs over the last ten years has been associated with Microsoft's client-server architecture and expressed directly through Wintel upgrade, applications, security, and support processes and indirectly through business limits, actualized risk, and turn-over costs.
For something like the Wintel component in IT expense to qualify as a Giffen good it would need to meet three conditions: increased cost consistently precedes increased spending; buyers had, or have, rationally preferable (better, cheaper) alternatives during every price/demand surge; and, the increased spending isn't associated with increased value to the organization.
I don't know of a data source adequate to either prove or disprove the proposition that Wintel expenditures have generally fit all three criteria. Intuitively, however, it seems obvious: the direct cost history on implementing and maintaining the Microsoft architecture suggests consistent, closely coupled, price and demand increases; there have been better, cheaper, alternatives at every step; and, there's no evidence that companies upgrading from one Microsoft generation to the next improved their competitive positions or organizational returns by doing it.
There is a problem here: as anyone who has studied economic theory will know the utility argument against classifying Wintel expense as a Giffen good is along the lines of the claim that a million blondes can't be wrong - because the word "utility" in economic theory refers to perceived value received and so a million decisions to get along by going along with each Wintel cost surge establish its increasing market "utility" and thus that Wintel expense isn't a Giffen good. These same blondes, incidently, also form the collective basis of the argument advanced against classifying high fashion items - things like $2,300 shoes costing $8.00 to make and ship from China - as Giffen goods despite the reality that conspicuous consumption is always cost driven
Absent solid contrary data, however, I think the bottom line on Wintel expense is actually pretty clear: it seems to exemplify a class of business expense components exhibiting the characteristics of Giffen goods - cost increases lead to increased demand as people make double down" decisions and knowingly re-commit their companies to an increasingly dated, risky, and expensive technology.