Is the US Bureau of Economic Analysis Undermeasuring the Tech Economy?


When I do my US tech market sizing and forecasting, I start with the data on business investment in computer equipment, communications equipment, and software in the quarterly National Accounts of the US Bureau of Economic Analysis (BEA).  As Forrester’s recently published report on the US tech market noted (see September 28, 2012, "US Tech Market Outlook Dims For 2012 To 2013 -- US Tech Market Spending Will Maintain A Mediocre 4% to 5% Pace"), the BEA in July revised the historic data on these categories of business investment going back to 2009, significantly reducing the size of tech investment in this period and lowering the growth rate of business tech investment to a pace not appreciably faster than the growth rate in the US economy.

While I adjusted my tech market sizings and forecasts to these lower numbers from the BEA, I have been wondering whether the BEA in their data collection is missing key segments of new technology, and thus understating the level of tech buying that is actually going on.  We have no good way of answering this question since the BEA has not publicly indicated that there have been any changes in data sources and aggregation methods that would signs of undermeasurement.  Still, here are the questions I would ask BEA if I had the chance.

  • Did the BEA's numbers on server investment miss sales by hardware vendors to their services’ arms?  Vendor data on sales of server and storage  equipment showed moderate and slowing growth in the US (and globally) in 2010 and 2011.  The revised BEA numbers, on the other hand, shows business investment in servers dropped in 2010, but growing at double-digit growth rates in 2011.  But 2010 was the year of the rush to infrastructure-as-a-service offerings, with hardware vendors like IBM, HP, and Dell joining in the rush.  While sales of servers to telcos and telecommunication services who were busy creating their own IAAS infrastructure would certainly be reflected in the BEA numbers on server investment, it is possible that it missed the sales by IBM, Dell, HP and other server vendors to their own sister units offering IaaS services, as a left hand to right hand internal transctions.  If BEA's data missed that pattern, that would explain why its data shows declines in server investment in 2010 while the vendors were reporting double-digit revenue growth.  Similarly, BEA's reduced level of server investment numbers in 2010 would set the stage for its reporting much stronger 2011 and the first half of 2012 growth in server investment (15% to 18% throughout these six quarters) than the server vendors themselves reported in revenues (6% growth in Q2 2011 and declines since then).   
  • Did the BEA miss the rise of Apple as a seller of PCs and tablets to business in 2010 and 2011?  As with servers, the BEA's numbers on business investment in personal computers show little or no growth in 2010, and strong growth in 2011.   According to the BEA, business investment in personal computers grew by 1% to 2% in the last three quarters of 2010 after falling by 5% in Q1 2010, but have been growing by 27% to 33%  per quarter since then.  Our data on PC vendors' US revenues showed almost the reverse -- 18% to 32% quarterly growth in 2010, but 7% to 18% in 2011 and the first two quarters of 2o12.  But much of the growth in our PC vendor revenue growth is due to sales of Apple Macs and iPads (which we count in the PC category, as we were told by BEA that they do as well).  Since Apple until 2010 had a small and shrinking portion of the corporate PC market, it is possible that BEA missed the rapidly growing portion of corporate PC spending that started to flow to Apple starting in 2010.   
  • Is BEA missing the rise of SaaS software?  BEA's software investment numbers are broken down into three types of software: pre-packaged software, custom software, and own-account software.  Our understanding from reading through BEA's description of its data sources is that its data for pre-packaged software comes from the US Census Bureau's Services Surveys of US software publisher revenues, with adjustments for exports and imports of software.  Similarly, its data for pre-packaged software comes from the Census Bureau's survey of computer systems design and related services (NAICS 5415), which is the industry category many large vendors of business software classify themselves.  As long as the software vendors in these categories were selling licensed software, than both their license and maintenance revenues captured by the Census Bureau's services surveys would flow into the BEA data on business investment in software.  However, SaaS subscription revenues almost never get counted by firms as capital investments.  If BEA decided to exclude these SaaS subscription revenues from its calculations of business investment in software, that would help explain why the 4% 2010 and 8% 2011 growth of its software investment in these two categories was so much less than the 10% growth in 2010 and the 12% growth in 2011 in the US revenues of the large traditional and large SaaS vendors that we track.
  • Is BEA lagging in capturing the shift from internally built software to commercial software?  Own account software represents over  a third of the BEA's total for business and government investment in software.  According to BEA's descriptions of its data sources, it estimates own account using a model that includes data on IT department staff devoted to software development and their average compensation.  When we prepare our software estimates, we exclude own account software, because we are counting its implicit value in our data on IT staff costs.  BEA's own data shows that the portion of own-account software in total business software investment has shrunk from 38%-39% in 2004 to 2009 to 37% in 2010 and 36% to 2011.  And BEA only releases these numbers with a one-year lag -- the latest available is for 2011, and that was released in mid-2012.  So the 2012 quarterly data on business investment in software could contain an even lower level of own account software and higher proportions of prepackaged and custom software than was the case in 2011.  That would mean that the growth in software investment as we at Forrester measure it would be higher than the BEA's reported growth in business investment in software.

These are fairly esoteric tech investment accounting issues.  But I bring them up because they matter to Forrester's tech market forecasts. Our sizing and forecasting methodology rests on government data like the BEA's.   We base our forecasting models on these data sources, and we compare our forecasts with updated data from these sources to see where our forecasts turned out to be accurate or off base.   Put another way, they keep us honest.  But when this source data starts to deviate from other data sources that we track in large and significant ways, that becomes a problem.  We will continue to base our forecasts off of these data sources, because in my view that is critical to the integrity of our numbers.  Still, I think it is helpful to our clients to learn when we have questions about those data sources -- especially when they are subject, as BEA data is, to large annual revisions -- and when we think they may overstate or understate reality.  In this case, I suspect that the tech market condition was actually better than the BEA data is portraying it in 2010 and 2011, not as positive as the BEA is saying so far in 2012, and likely to be better in 2013 than the BEA data will show in coming quarters.