Chapter Three of Overselling the Web? looks at the impact of ICTs in wealthy countries, with a focus on the US. It opens with a discussion of the evidence regarding the total factor productivity (TFP) impact of ICTs. ICT using sectors have seen considerable investments in computing and communications --they accounted for as much as 25 percent of total investment by the end of the 1990s. This investment produced solid returns. At the same time, ICT investment did not produce especially high returns of the type that would show up in TFP statistics. The only sector to see particularly dramatic TFP growth was the ICT producing sector. And this was in large part connected to a new way of calculating output in the sector based on 'hedonic measures.' These measures adjusted industry output statistics to account for the rapid increase in quality for price in the sector. Such measures assume that if a computer had twice the power, it was worth twice as much even if it sells for the same price. That is a questionable assumption, and the chapter demonstrates why with the example of word processing:
In 1964, IBM brought out the MT/ST (Magnetic Tape/Selectric Typewriter) which allowed entered material to be edited without being typed onto paper, and provided electronic storage and replay, correction, and multiple re-printing… By the mid-1980s, the standard word processor already featured justification, alignment, insertion/overstrike, search and replace, copy/cutting, pagination, headers and footers, footnoting, table of contents and index generation, form letter merging and automatic spell check. Progress did not stop there, however… In 1995, Microsoft introduced what was already the eighth version of its world-processing software, Word 95. It had considerably higher requirements than its forebears. A 386 processor, six megabytes of RAM, 8 megabytes of hard drive and a 3.5 inch disk drive. But for this extra capacity, users were given a range of new features including an ‘answer wizard’ help system, auto-formatting and designer templates.
It should be noted that a study of 16 people using one version of this software found that of 642 commands available on the program, just 20 accounted for 90 percent of use. The average person used just 57 commands in six months, and all of the users together used only 152 commands in eighteen months –or less than one quarter of all those available.
Nonetheless, this did not stop Microsoft issuing new versions with more commands. Within two years, the company had issued Word 97 –it required a faster processor (a 486, with nearly 1,000,000 transistors on the chip), more memory (eight megabytes), a larger hard drive (20 megabytes) and a CD-ROM. In return, users now had access to a letter wizard with layouts and styles, 3-D effects and animated text, and the animated smiling paper-clip office assistant.
Word 2000 came next. Again, system requirements increased –a Pentium 75 processor, 20 megabytes of memory, 147 megabytes of hard drive. But users now had access to WYSIWYG (‘what you see is what you get’) font menus and improved hyper-link functions. This was also the version that replaced the animated paper clip office assistant with an animated Einstein office assistant.
The next version –Word 2002 (or Word 11.0) required a Pentium 133, 32 megabytes of memory and 375 megabytes on the hard drive. For which users were rewarded with content-sensitive smart keys, faster organization of mass-mailings (for which we can all be grateful), “a new auto-correct smart tag” that “lets you control how Word corrects text” –presumably allowing you to un-correct all of the mis-auto-corrected text from the Word 2000 version— and an auto save function that activates when the excessively baroque program crashes the computer.
Since then, we have had Word 2003, and Word 2007 is just around the corner. The latest version will use 42 times more RAM and 250 times more disk-space than Word 95. Assuming that the retail price of the computer on which Word 2007 runs is similar to Word 95, it might be arguable that the 'real' value of the computer is 250 times larger because it can run software with auto-correct smart tags and other unused features. Indeed, the excessively baroque nature of the programs that it runs might actually have reduced its value to the average user.
Worse, Word is in fact part of a 'baroque network.' The value of the program is in large part attached to the fact that everyone uses it. The utility of my word-processing software is considerably reduced if I cannot share files with collaborators. And if they have Word 2010 while I have Word Jurassic, I will not be able to read their files. Even if I am happy with the dancing paper clip mis-auto-correcting my text, if my collaborators will settle for nothing less than the dancing Einstein, I had better upgrade. This might be a particularly significant drain on the productivity of ICT investments in the developing world, where use of advanced applications requiring significant computing power is rare. But, along with the high failure rates associated with IT projects, the impact of baroque networks may help to account for no-more-than-normal returns to ICT investment in rich countries as well.
Looking at the impact of Internet use in particular, it appears that wholesale e-commerce enabled by the Internet is only a small improvement over the much older technology of Electronic Data Interchange, while retail e-commerce is not a significant improvement over off-line purchasing. This might help to explain why it is difficult to pick up any impact of e-commerce on macroeconomic indicators. Meanwhile, the Internet can be a powerful force for productivity at work, but it is also a powerful tool for wasting time. Internet use at home and at work looks very similar (eBay is the fifth most popular site at home and sixth at work, for example), and this suggests a lot about the major uses of the web at work (at one point in 2005, the top three search terms were "sex", "porn" and "Paris Hilton"). Perhaps 172 million hours a week are wasted in recreational surfing at work in the US.
What does this mean for developing countries? If the economic impact of the Internet and ICTs more broadly has been muted in the US, this is likely to be even more the case in countries with far lower usage rates. It appears that wealthy areas with skilled workforces and strong institutions are garnering most of the benefits from new ICT technologies to date. Overall, rich country evidence is not reassuring for those hoping for an ICT-enabled economic revolution in the developing World.