Os SHANE GREENSTEIN greenstein@kelloge.niraed Economic statistics usually involve something most cbvous, butt immecitely parent. Theitconstuction ‘often resembles that ald saying about ele- Dhants sitting in teas. The elephants are invisiole unti someone points them out After someone points them out, the ele- ents seem to be everywhere, and we onder how we missed them befor. Accordingly, recent federal government intiatves to measure electronic business ‘throughout the US get armixed resction. A skeptic would comtectly ask whether these intiotives measure anything other than the obvious. A supporter also correctly responds that society wants an accurate ‘count ofall the electranie elephants in ‘tees, then the US government is in a ‘900d position to measure the obvious. These initiatives raise one of the oldest questions about government agencies collecting statistics: What should these agencies collect and why? Many federal agencies aresdy collect data on different aspects of electronic commerce. These agencies include the Bureau of Labor Statistics, the National Telecommunications Industry Adminis- ‘wation the Bureau of Economic Analysis, and the Federal Communications Cor mission, just to name a few, To iustrate ‘the issues related to statistics collection, vill focus onthe US Census Bureau's E- Stats program to measure e-business. If Yyouwant to caserve E-Stats in action, 0 te hrtpiinww.census.gov/ and click on “eStats,” nN E-business infrastru cture ‘What do they dot itis important to understand thet the US Census does more than count the ‘entre US population every dacade. As 2 divsion ofthe Dapartment of Commerce, this agency lso spends considerabie time gathering economic deta about business establishments thoughout the county Thisis especialy usin the non. ecennia years, wien the staff hs time for special projects, suchas an econor c census of business establishments This data is of use to marketing exer, regional economist, and certain macro. economist. When the US Census fist began these economic censuses many decades 860, they primary surveyed manufacturing establishments As time has passed, how ever, the US economy has Become less evoted to manufacturing and. more focused on senices, thus, the US Census now surveys more than just manufactur The Etats program sone of several programs racking statisti formanutac- ‘uring and nonmanutaturing commerce. The motivation comes fom a variety of sources. Some agencies, such 3s the Federal Reserve, need more accurate sta tistics forthe part of the US economy behind recent produetivity gains and ‘more recent, vlatiltyin economic ct ity. Other agencies, suchas the Bureau of Economic Analysis, need @ better sense of what ration o gross domestic Broduct (GDP) is attibutabe 10 e-com- metee end simi phenomena Many publopotcy programs—such as the $2 billion E-Rate prograr—hinge on having unbiased data thet sorts between hyperbole, hypothesis, and mere specu lation. For example, debates about broad- band would beneft from more accurate ‘censuses of Internet use at homes and private business establishments. These initiatives a the US Census could ground Giscussion, orient debate toward the right set of questions, and, at @ minimum, ‘void common misconceptions. in this ‘sense, we can view these intatives in an overall postive light That seid, this type of measurement isnot easy to carty out. To illustrate, | row focus ona US Census proposal to measure e-business infrastructure, Why is this 50 interesting? To oversimplify ‘his abit, different regions of the coure tty need to know whether they are ahead or behind economically, Econo- ‘mists will use this date to explore ‘whether the mare informetionsintensive industries underwent more or less ‘growth compared to other industries, Economists are also curious about ‘whether investmentin e-business infe- stucture helped induce more revenue in some regional industias and not in others. We cannot make such detecmi- nations without date What is infrastructure? ‘To understand some ofthe difficulties with measuring elephants, let's aig below the surface. Measurement must take 0272-173201810 00 © 2001 IEEEplace in ways that recognize some ofthe Unique features of IT markets, Consider the measurement of software ‘and the problems that software's atibut- ‘8s pose in constructing an e-business infa- sStucture index. Whatever way We define it, e-business infrastructure comprises @ wide variety of disparate technological pieces witha wide set of uses, This makes ‘tGficu to identity impact, trace the tines of causality from infrastructure to out ‘comes, and s0.0n. One way to more east ly identify these linkages is to collect information about many circumstances in many diferent markets o industries. ‘Yet in practice this willbe challenging lnfrastructure is made up of routers, com- puters, optical communications, and soft ‘ware. But how about fax machines or CT ‘scanners? Are these part of e-business infrastructure? How about equipment used for radio and television broadcast: ing? What definition is appropriate, broad or narrow one? Here is another challenge. The general ‘ule of thumb within most corporate infor ‘mation groups is that peckaged software and hardware constitute a lage but not ‘verwheiming share of overall expense. Conversely, maintenance, administration, Support, and software programming are the largest expenses, These assets involve a lot of human capital, complex ‘investments, and the construction of ido- syncratic assets. Often, they have lie reso value butahigh service value tothe ‘owner Unfortunately, these patterns are Quite problematic for the US Census when carrying out measurements. What isthe proper way to estimate the value of sofware? | do not think anyone has 2 perfect answer. Whila packaged software, such as MS Office, has @ sat market value attached to it, the market value of much contract software from Consultants is ess lear. What is tha value fold, yet functioning software acquired 18 few yoars 290? Whats the best way to benchmark the value of @ program that took years to debug and cefine? These seemingly simple accounting questions villhave enormous consequences for any estimate of e-business infrastructure. This, is anartatbest and not yet a science. In adition, there is tremendous vari- ‘ance in ownership structures for IT ser Vices delivery. Normally, this would not cause deta collectors problems because there are ways to correct fr rentaliown- ership alfferences across fms andindus- ‘ies. However, many firms outsource ‘only parts of their business computing ‘and communications. Many firms do not have clean boundaries between assets ‘owned by the firm, rented from others but kept on the premises, and effective- ly tented from others whe provide end services. Data back up; hosting; and ‘many routing date functions in payroll, benefits, insurance claims, and other _2dministrative tasks fal into this catego '. Some firms own these tasks, and some outsource them. The same work ‘an occur in-house orofsite, butin each case, there isa diferent asset owner and, different work location ‘This trends aes significant, atleast {as measured by the many firms supply ing outsourcing, The most optimistic fore- ‘casters predict that outsourcing will ‘spread into many other facets of IT apal- Cation inthe form of application service providers, othe US Census has to figure ‘ut what part of corporate IT goss to a firm such as EDS or Perot Systems, and what part stays imhouse, In other words, 8 census cannot attibute a specie amount of e-business infrastructure to a spectic industry with- ‘ut also accounting for some fraction that is missing bocause the serview is rented from an upstream supplier. Because IT outsourcing is a huge industry, the com- panies init are constarty changing their organization, merging, disbanding, and recombining in new variations. In such an environment, nobody wants an e-busi ness infrastructure staistio that is sensi tive to arbitrary decisions about vertical integration or disintegration. ‘What should be left to markers? T markets differ from other infrastruc: ture in one other way: lot of commercial, ‘companies are already tracking consider. able information. For example, a trend ‘among commercial datacollecting firms is t0 learn about internet users, some- times by going directly tothe internet 12 track conduct. Commercially available data sets provide information about ‘whether households ara oniin, if bust nesses have access tointenet services, the time users spend online at work and athome, user surfing habits at adverts- ing-supported sites, online spending habits, and soon, Thase dets come from ‘such firms as Jupiter, Forrester, Media- Matrix, Plurimus, ComScore, Hare Hanks, IDG, DataPro, PNA, and many others. ‘These fms talor most of these data to ‘marketing departments neads, and some ofits expermental, but hese facts do not ‘make these data useless for every gov- lemment purpose. Said another way, if ‘many commercial providers supply infor ‘mation about electronic commerce, what {ype of information should a government agency spend ts time collecting? ‘The short answer goes something ike this: Governments shouldnt collact data that commercial firms ean track just as Well.On the other hand, there ara plenty of measures—such as GDP measure- mentand unemployment statistics—that matter to society st lage, but that com- ‘mercial fms do not comila.The US Cen- Sus has a comparstive advantage in ‘collecting this important data in tat itis ‘comprehensive, covering every estab- lishment in every location, oes this ebusiness infrastructure moa- ‘surement pay tothe comparative advan- tage ofthe census office, or could data from rate fms provide adequate information? Dowe needa government program fortis project? These are open questions that onomists ar stil debating, In the meantime, some wellmeaning ‘goverment empioyees are figuring out whether’ feasble tomeasure e-business infrastructure. The total wil probably reach into the hundeeds of bllons of dolar, so ‘the task is potentially quita important. At the sare time, the task is extremely chat lenging. These employees’ biggest hope is thatin the process, thoy'avaid creating an leven bigger white elephant.