Mention ‘foreign trade' and most of us, even in this wired age, are still apt to think of ships and planes bearing goods produced in one country to customers in another.
Such images are one reason we place such great stock in merchandise trade statistics. Businesses use them to identify potential new markets. Government officials factor them into export promotion strategies. Wall Street analysts study them for clues that might herald shifts in interest rates or currency values. And, of course, whenever the numbers indicate that one nation is running a sizeable trade deficit with another, an international dispute is likely to ensue.
But what if this way of keeping score can also mislead us? What if the whole system of collecting information on international transactions is based on concepts that are hopelessly out of date?
Consider the case of the computer software industry. Arguably the most dynamic and internationally competitive segment of the US economy, it is also the industry most likely to confound efforts to measure international commerce and gauge its importance to the local economy.
According to the Business Software Alliance, this industry's growth in recent years has been so explosive that its contribution to the nation's economic output is poised to exceed that of the automotive industry. By most estimates, as much as one-quarter of this hard-charging industry is based here in California, where it employs well over 200,000 workers. Yet, almost paradoxically, a growing number of California software firms have positioned themselves to earn very substantial revenues from overseas sales without contributing appreciably to the state's merchandise export trade.
Outside the U.S., sales of packaged software reached over $135.4 billion in 1997, according to estimates by IDC, a market research company. By all accounts, American companies dominate the world market for software products, both the customized variety written to meet the needs of individual clients and the packaged software consumers buy in the familiar shrink-wrapped boxes. In Germany, for example, six of the top ten software providers are American-owned firms. U.S. Commercial Service reports indicate that American companies control as much as 70 percent of software sales throughout Europe.
One might expect, therefore, that the US software industry would have a substantial, beneficial impact on the nation's merchandise trade figures. Surprisingly, no one really knows how much this industry exports. According to Anne Griffith, the research director for the Software & Information Industry Association, there are "no hard data on this question," adding that anecdotal evidence suggests that conventional exports represent "a very small share" of the total value of US software products sold abroad.
The principal reason for the relatively meager volume of software exports is that US software firms have aggressively embraced a business strategy that American manufacturers have pursued ever since an enterprising New York company founded by Isaac Singer opened a plant in Scotland to assemble its newfangled sewing machines. Today, U.S. companies do more business overseas through their majority-owned foreign affiliates (MOFAs) than they do through direct exports.
In the latest Bureau of Economic Analysis benchmark survey of U.S. direct investment abroad (1994), total overseas sales of goods and services by these MOFAs amounted to $1,282.3 billion. By comparison, US merchandise exports totaled $502.4 billion that year, which also saw an additional $199.7 billion in exports of services.
Manufacturing overseas is especially attractive for the software industry because of the very nature of its product. Essentially, software is more digital than tangible. Taking it to market, therefore, does not involve the often cumbersome logistics of merchandise exporting. As the SIIA's Griffith points out: it is "much more expensive to ship a box with all the documentation than to reproduce and distribute regionally overseas." So while a software program may be developed in the U.S., master copies are shipped or, more likely, transmitted electronically to facilities overseas where the CD- ROMs or diskettes are duplicated, manuals and other documentation are printed, and the familiar shrink-wrapped boxes assembled for distribution.
Odd as it may seem, many of these foreign sales are not classified as international transactions. When a German citizen buys a copy of a software program developed in the United States, there is a fair chance the software had actually been duplicated at a plant in Germany. Because this sale involves two German parties, it does not register as a U.S. export. Nor for the same reason do the Germans regard it as an import.
Ironically, the largest European exporter of packaged software appears to be Ireland which, according to a recent U.S. Commercial Service report, ships some 60 percent of the packaged software sold throughout the rest of Europe. The reason for Ireland's apparent dominance in the European software market? Nearly all of the largest U.S. software firms use Ireland as their primary base of operations for servicing markets in Europe, Africa, and the Middle East.
Keeping an accurate tally of software exports is made even more difficult by the fact that software is routinely shipped pre-installed on computers, printers, and other items of hardware. Yet knowing where the hardware is eventually sold offers little help in following the trail of embedded software exports. Apple computers distributed in Europe are made in Ireland just as Asian Apples are manufactured in Singapore. Likewise, Hewlett-Packard printers sold in the Far East carry a Made in Singapore label. How the software installed in these foreign-made creations of California ingenuity got overseas is not a question the current export data collection system is designed to answer with any degree of accuracy.
Further complicating the task of government trade statisticians is the question of whether software is a good or a service. Unfortunately, the Solomon-like bureaucratic solution -- that customized software is a service, whereas the mass-produced packaged software programs are goods -- has only led to further confusion, in part because different agencies using different data collection methodologies are involved.
U.S. merchandise exports are tracked by the Customs Service and the Census Bureau, whereas trade in services is monitored by the Bureau of Economic Analysis. Merchandise trade data are derived from information provided by exporters on the Shippers Export Declarations (SEDs) that must accompany every export shipment valued at more than $2,500. Data on services are gleaned from periodic mail-out/mail-in surveys of service providers, which invariably target the larger and more established ones and often miss newer, smaller firms.
If keeping track of the volume of software exports is problematic, placing a dollar value on software shipments has proved downright exasperating. The Foreign Trade Division of the Census Bureau has ordained that, for export documentation purposes, packaged software "is to be treated as a commodity and the value reported on the SED should represent the full selling price of the software." The value of customized software, meanwhile, is to "based on the value of the media."
The Census Bureau's rationale for not recording the true value of customized software exports on export documents is to avoid double counting. (It is presumed that information on these service exports will be obtained through a BEA survey at a later date.) The problem, however, is that shippers of packaged software have evidently taken to valuing their merchandise in the manner prescribed for customized software exports.
As one industry researcher observes: "exports are valued at the price of the CD-ROM itself, not the intellectual property contained on it. [A CD-ROM is worth about $.70 and a diskette about $.10.] The software publishers are perfectly happy with this arrangement since it reduces the likelihood that software will be subject to various duties upon arrival in other countries."
Apart from raising questions as to if and where taxes are being paid, what this means, of course, is that the officially reported value of U.S. software exported to foreign consumers is apt to be sharply discounted from its market value. For example, Mexico, which currently lacks adequate facilities for duplicating and distributing packaged software, is one of the few countries to which US software firms export their products the old-fashioned way. Yet, as industry insiders concede, a shipment of one thousand shrink-wrapped boxes containing, say, a $100 personal accounting program to a chain of computer stores in Mexico would probably not be valued at $100,000 but more typically at something closer to $700 (assuming each box contained the program on one CD-ROM).
The distorting effect of the data will only grow worse as the use of the Internet (on which there are no customs posts) spreads worldwide. Today, American consumers no longer have to leave their homes or offices to buy new software. Instead, they may simply download the program from a Website. In much of the rest of the world, this is not currently an attractive distribution mode because Internet users typically pay a per-minute online charge to the local phone company. However, once hook-up charges become as marginal as they now are for most U.S. Internet users, it is expected that software will be sold almost entirely over the Internet, downloaded from servers that could be located nearly anywhere in the world. When a consumer in Stockholm downloads a program written by technicians in Sunnyvale from a server located, presumably for tax purposes, in St. Kitts, the relevance of conventional trade data will have become nil.
The Internet is also likely to spawn different kinds of relationships between software users and vendors. Rather than buying a program, businesses and individual consumers worldwide are even now going online to "rent" programs on an as-needed basis. In one recently reported case, Intuit Inc. says 388,000 people filled out their 1998 tax returns on its Web site rather than by buying its TurboTax software. By dispensing with the physical product, Intuit and other software vendors can dramatically slash manufacturing and distribution costs and, one hopes, pass the savings on to the online customer. The exact impact of such marketing and distribution innovations utilizing the Internet are generally hard to foresee. One thing is certain, though. Novel ways of bringing products to market will befuddle statisticians trying to make sense of the nation's international business activities.
All of this vastly complicates the already tricky problem of calibrating just how internationalized the Golden State's economy has become. It also raises some prickly issues about what types of economic and trade development programs may be appropriate for the state. Historically, state and local officials have been eager to boost exporting companies, the rationale being that such firms are likely to be prime sources of new jobs and tax revenue. But what of businesses that, while highly competitive internationally, export very little? Are they similarly worthy of government assistance if most of the employment and tax benefits of their overseas sales accrue to foreigners? Similarly, how can state policymakers be expected to make sound decisions about where to establish new foreign trade offices when the information available to them is often inaccurate, incomplete and misleading?
With the encouragement of a Legislature long skeptical of the virtues of many of the state's export promotion programs, the Davis administration is now conducting a comprehensive trade policy review. If this exercise achieves nothing else, it should be considered a success if it manages to rid public debate of some popular but outmoded notions about world trade. It will be an outright triumph if state leaders also decide to equip themselves with new analytical tools designed to yield information that more precisely depicts the real world of international business.
Revised August 2, 1999