Next: Secondary redistribution Up: Results Previous: How Do the

How many people get documents or fun with numbers

What do the survey results tell us about the number of primary recipients? We can either look at the respondents as an adjustable skewed sample of this population or extrapolate from the reported activity of the distributors. for reasons already mentioned, we assume in the first approach that response rates vary across respondents, according to their sources for the documents. It seems reasonable to assume that the response rate for subscribers to mailing lists is bracketed by the response rates of subscribers to Clinton-Info and the healthplan lists. Note the healthplan list subscribers have two response rates: to the customers surveys (7.4 %) and the mailing list query (3.2%). Using the latter rate to the number of respondents. The latter rate together with the number of respondents who indicated receiving a document-summaries mailing list (from ESUSDA) predicts to a total subscription of 2900, quite close to a list administrator's rough estimate of 3000.

For retrieval from FTP and gophers, we have two pairs of data points: the number of respondents reporting retrieval from the gophers at Texas A&Mand University of Maryland and estimates by administrators at these sites of average daily visits to the White House document files. Similarly, there are two pairs of data points for commercial forums, reports by users of America on Line and Genie, and the estimates of daily visits by the respective system operators. The relevant response rates that we use are accordingly

Data points for bulletin boards and newsgroup retrievals have less grounding. Although there is concentration around one particular source in each case, Fedworld and alt.politics.clinton, respectively, we do not know the percent of BBS users without email and the consequent effect on the response rate. (This lack had effect in two ways: most obviously, persons lacking email could not respond to the survey and, second, some BBS operators did not post the survey, because they feared that users would upload the surveys and ask them to mail it on their behalf.) For the sake of estimating a lower bound, we assume that response rate from newsgroup readers was no worse than from server retrievers, and assume for bulletin boards a rate no better than half that of the commercial forums. This falls in line with information from the FedWorld operator re: the percent of his BBS users lacking email (60%) and the comparison of his estimate of daily users of White House documents at FedWorld (2400) with number of respondents (6) who are users there.

We also make certain allowances for the multiple listing of sources and errors, such as mistaking a mailing-list for a newsgroup, by reducing the estimates. Given these assumptions, the break out is

The adjusted total, which includes an additional 1%for those reporting receipt in hardcopy and which deflates by 30sources (based on responses to USAGE) is 39000.

In extrapolating from distributors responses to the question in the follow up surveys about the size of their daily audience, we use an average number of daily recipients per distributor of 200. We assume that each distributor designated by the response to the USAGE survey constitutes a separate entity. Note also we do not need to adjust for users' having multiple sources, since the number of regular users is larger than that of daily users. The estimate under this approach is about 33000. We are in the same ballpark.

Next: Secondary redistribution Up: Results Previous: How Do the
Sun Jun 5 21:21:16 EDT 1994