Why Online Retail Sales Are Much Larger Than US Census Data Report
Global Economics Group; University of Chicago Law School; University College London
Market Platform Dynamics
Massachusetts Institute of Technology (MIT) – Sloan School of Management
January 28, 2016
US Census data are widely used to show the level and growth rate of online retail sales. However, based on a detailed review of the data reported by Census, Census’s technical documentation for collecting those data, the survey instrument used by the Census to collect the data, and e-mail correspondence with a senior Census official, it appears likely that those data significantly understate online retail sales for two reasons. First, it appears that the online sales of large retailers are not included in the three-digit NAICS code figures reported by the Census Bureau. Second, it appears that the Census Bureau does not count the substantial online retail sales of firms not classified as retailers. It is possible to derive conservative estimates of the amount of online sales not included by the Census Bureau under the assumption that the Census Bureau follows its published methodology. Conservative estimates based on available data indicate that online retail sales were $83 billion higher in 2013 than those reported by the Census Bureau; that is, actual online retail sales were 32 percent higher than those reported by the Census. The percentage undercount is similar for earlier years, so that Census-reported growth rates seem roughly correct. The undercount it is likely to persist as a consequence of the methods used by Census to collect and report online retail sales data. It is not possible to present these estimates as definitive, however, because the Census Bureau has made inconsistent statements concerning its methodology for estimating online retail sales, which Census has declined to help us reconcile, and because it is not possible for the Census Bureau to share details on the inclusion of data for large retailers with online sales.
Why Online Retail Sales Are Much Larger Than US Census Data Report – Introduction
Like many researchers and other observers of the retail sector, we have long taken the U.S. Bureau of the Census data on e-commerce at face value. Our recent work on the reinvention of the retail industry in the U.S., however, led us to look much more closely at the Census data and the importance of online sales. We saw an apparent contradiction between the Census data, which showed that online commerce had barely dented brick-and-mortar sales two decades after the start of the commercial Internet, and the massive disruptions and dislocations that have been taking place in physical retail.
Examination of data on Wal-Mart made it clear that there was a problem with the Census estimates. Wal-Mart is a general merchandise store and, as Morningstar confirms, is thus classified in NAICS (North American Industry Classification System) industry 452. In 2013, Census data shows that there were $88 million of U.S. e-commerce sales by general merchandise stores. But Wal-Mart has a very large online retail business. It had global e-commerce sales in 2013 of more than $10 billion, of which a significant portion, likely more than $6 billion, occurred in the U.S. The Census data are collected and reported at the firm level, as we discuss further below. Wal-Mart has an NAICS code of 452. Therefore, the only place where Wal-Mart’s online sales would be reported based on the methodology that the Census Bureau says they follow is under NAICS 452. When presented with this discrepancy, the Census Bureau confirmed by e-mail that there is an undercount in e-commerce.
That has led us to ask what other firms were missing from the Census e-commerce statistics and to inquire more broadly into other possible ways in which the Census Bureau figures might not provide accurate measures of US online retail sales. This paper reports what we discovered and shows that online retail sales in the U.S. are considerably higher than reported by the Census Bureau.
The Census Bureau data understate online retail sales in two major ways. First, many large retailers that make most of their sales in offline stores (so-called physical retailers), like Wal-Mart, are not included in the relevant NAICS category. According to the Census Bureau, “large retailers such as you have inquired about are required to include their E-commerce receipts in with their brick and mortar store receipts rather than in the E-commerce receipts” All of their sales are therefore counted as offline sales. Second, significant online retail sales are made directly by companies such as Apple that are not primarily retailers and are thus not classified in an NAICS retail industry.
The understatement is quite large. We have estimated, using data collected by Internet Retailer, that among the top 1000 online retailers, there were 201 physical store retailers that had online sales in the U.S. of $76 billion in 2013. That figure is $36 billion higher than the total online retail figure of $40 billion reported by the Census Bureau for the NAICS codes that correspond to physical stores. Because Internet Retailer does not cover smaller physical retailers with online sales, we can be confident that Census missed at least $36 billion in online retail sales. We have also estimated that the Census online retail sales figures did not count another $22 billion of retail sales made by companies classified as something other than retailers in 2013.
As a result, online retail sales as a percentage of all retail sales for 2013 was at least 7.4 percent compared with the Census figure of 5.8 percent. If we exclude from total retail sales the total sales made in NAICS code 447, which is Gasoline Stations, online retail sales as a percentage of all retail sales for 2013 was at least 8.5 percent compared with the Census figure of 6.7 percent. Table 1 reports our estimates of online retail sales for 2010 through 2013 and shows the Census figures for comparison.
See full PDF below.