[GOAL] Re: How many researchers are there?

[Arthur Sale]

Thanks Stevan. Unfortunately that does not answer the question I posed, but
a different question which is not relevant in the research being undertaken
to see the adoption of OA practices amongst researchers, as opposed to the
application of OA to articles which you have ably handled.

 

As you have recognized, it only needs one author of a multi-author article
to make the whole paper OA; however as we approach 100%, single-author
articles will require that sole author to make his or her paper OA. (The
question is irrelevant to Gold OA because all authors jointly agree to make
the article OA, once.) It would be an interesting study to see amongst Green
OA, whether the rate of making articles OA improves as the number of authors
does. Hypothesis A: it will, but not linearly. Secondly one could look at
the number of times an article is OA (ie the number of OA copies there are
on the Internet). Hypothesis B: this measure should increase with the number
of authors, though probably not linearly. Zipf's law is more likely in these
cases as earlier-listed authors are probably the more likely to take OA
action. Is your crawled data capable of being re-interpreted this way?

 

I propose to do the following:

 

(1)   Estimate the total number of papers P published per year y, Py

(2)   Estimate the average number of authors per paper for this corpus, m.

(3)   Compute m x Py = N, an estimate of the number of active researchers.

 

The expected errors in N are:

.       The value of Py is not certain - neither ISI nor Scopus are
complete. This leads to an under-estimate.

.       Not all researchers publish every year. This means that the number
of researchers is again under-estimated. 

.       Some researchers publish more than once per year. This is
double-counting and results in an over-estimate. ISI or Scopus may be able
to provide disambiguated estimates from their databases.

.       Unfortunately aggregating the number of years causes both the above
errors to change - the first reducing, the second increasing. I have seen
statements to the effect that an active researcher publishes at least once
every three years, so the effective limit is 3 successive years. 

 

Still, the information will be interesting and perhaps useful.  It may be
useful to do a pilot study in a single institution. Australian universities
have complete citation databases of their publications, so it may be
possible to check this type of data for a single institution. If it is a big
one, the data may extrapolate.

 

Best wishes

 

Arthur

 

From: goal-bounces at eprints.org [mailto:goal-bounces at eprints.org] On Behalf
Of Stevan Harnad
Sent: Sunday, 1 January 2012 8:52 PM
To: Global Open Access List (Successor of AmSci)
Subject: [GOAL] Re: How many researchers are there?

 

Some suggestions:

 

(1) Estimate the total number of papers P published per year y, Py, rather
than the number of researchers.

 

(2) Start with the Thompson-Reuters-ISI-indexed (or SCOPUS-indexed) subset.

 

(3) For Py, sample the web (Google Scholar) to see what percentage of it is
freely available (OA).

 

Our latest rough estimate with this method, using a robot, is about 20%.

 

(Using estimates of the number of researchers, if the margin of error for
the total is 1M - 10M then the margin of error for the percentage OA would
be 10% - 100%, which is too big. Using known, published papers as the
estimator also eliminates the multi-author problem.)

 

Cheers, Stevan

 

On 2011-12-31, at 6:25 PM, Arthur Sale wrote:





I am trying to get a rough estimate of the number of active researchers in
the world. Unfortunately all the estimates seem to be as rough as the famous
Drake equation for calculating the number of technological civilizations in
the universe: in other words all the factors are extremely fuzzy.  I seek
your help. My interest is that this is the number of people who need to
adopt OA for us to have 100% OA. (Actually, we will approach that sooner, as
the average publication has more than one author and we need only one to
make it OA.

 

To share some thinking, let me take Australia. In 2011 it had 35
universities and 29,226 academic staff with a PhD. Let me assume that this
is the number of research active staff. The average per institution is 835,
and this spans big universities down to small ones. Australia produces
according to the OECD 2.5% of the world's research, so let's estimate the
number of active researchers in the world (taking Australia as 'typical' of
researchers) as 29226 / 0.025 = 1,169,040 researchers in universities. Note
that I have not counted non-university research organizations (they'll make
a small difference) nor PhD students (there is usually a supervisor listed
in the author list of any publication they produce).

 

Let's take another tack. I have read the number of 10,000 research
universities in the world bandied about. Let's regard 'research university'
as equal to 'PhD-granting university'. If each of them have 1,000 research
active staff on average, then that implies 10000 x 1000 = 10,000,000
researchers.

 

That narrows the estimate, rough as it is, to

         1.1M  < no of researchers < 10M

I can live with this, as it is only one power of ten (order of magnitude)
between the two bounds. The upper limit is around 0.2% of the world's
population.

 

Another tactic is to try to estimate the number of people whose name
appeared in an author list in the last decade. Disambiguation of names rears
its ugly head. This will also include many non-researchers in big labs, some
of them will be dead, and there will be new researchers who have just not
yet published, but I am looking for ball-park figures, not pinpoint
accuracy. I haven't done this work yet.

 

Can we do better than these estimates, in the face of different national
styles?  It is even difficult to get one number for PhD granting
universities in the US, and as for India and China @$#!

 

Arthur Sale

University of Tasmania, Australia

 

 

_______________________________________________
GOAL mailing list
GOAL at eprints.org
http://mailman.ecs.soton.ac.uk/mailman/listinfo/goal

 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.ecs.soton.ac.uk/pipermail/goal/attachments/20120102/b1a1219e/attachment-0001.html