Coen’s estimates of total U.S. advertising expenditure from 1996 to 2006 are roughly comparable to estimates derived from Census data and IRS data. The Coen estimates are about 35% higher than estimates from Census data and about 8% lower than estimates from IRS data.
The Census-based advertising estimates come from annual surveys of establishments’ revenues. These surveys are benchmarked to the pentennial Census of Business, which is also establishment-based. The Coen ad expenditure figures exceed the Census figures most significantly for direct mail and miscellaneous categories. Establishments with primary businesses are other than direct-mail and miscellaneous advertising, but that also engage in these advertising activities, do not have their advertising revenue captured in the Census-based estimates. This measurement effect may account for some of the difference between the Census-based estimates and the Coen estimates.
The IRS-based advertising estimates come from advertising deductions submitted on businesses’ required annual tax returns. The IRS-based estimates included advertising deductions reported for all types of corporations, nonfarm sole proprietorships, and partnerships (estimated). For further analysis of IRS advertising data, see my analysis of small-business advertising. Advertising deductions on business tax returns may include some internal business expenses for advertising. Such expenses would not be included in the Coen figures.
All three data sources provide different sub-categories of advertising expenditure. All three sources are helpful for better understanding information industries.
Dear Douglas,
It seems like the Google Doc for the US Census data is down — is the data still available somewhere? Also, can you share the exact data series / source you used from the US Census? I’m trying to update the data to the present, but I can’t quite find the data source.
Whoops — it looks like the Google doc is actually up, sorry! But I would still like more info on how you calculate the Census totals — my question on that still stands.
I think the Service Annual Survey has changed and no longer provides a comparable series. For details on my calculations, see the embedded formulas in the Excel workbook version.