Transcript of Relations Between National Research Investment Input and Publication Output: Application to an...
- Slide 1
- Relations Between National Research Investment Input and
Publication Output: Application to an American Paradox R. D.
Shelton Sponsored by a sabbatical from Loyola and NSF Grant
ENG-0423742
- Slide 2
- Slide 3
- Whats the Problem? European Paradox is the perceived failure of
Europe to cash in on its leadership of science publishing I use
"American Paradox" to mean the long term decline in U.S. share of
science publishing, despite world-leading investments in R&D,
particularly as that decline has become worse Why is todays subject
The methods are also useful for other countries; e.g. the EU is
developing the same syndrome
- Slide 4
- The American Paradox: Declining SCI Paper Shares Despite Huge
R&D Investments (GERD*) *Gross domestic expenditures on R&D
(OECD Constant $ PPP), paper share from NSI CD 2004. Paper Share %
NSF analyzed paradox for 6 years. GERD $Billions 320 160
- Slide 5
- EU Recently Developed the Same Syndrome Paper Share % GERD
$Billions 320 160
- Slide 6
- Outline 1.Is it just that the SCI is a rubber ruler? (Hint: no)
2.Which inputs are most important: a regression approach (Hint:
look for the money, but not just any money.) 3.A theory that can
account for U.S. (and EU) decline, and forecast the future (Hint:
it's simple, m = kw)
- Slide 7
- 1. Is it Just That the SCI is a Rubber Ruler? Hypothesis: In
response to criticism that it is biased toward English language and
the U.S., the Science Citation Index may have changed to be less
favorable to the U.S. This could account for some of the rapid
decline in U.S. publication since 1995. Method: For a sample of 8
fields, partition their journals into 2 sets: New journals added to
the SCI after 1994, and Old added before. Then measure U.S. share
of each journal to see if difference between Old and New is
significant
- Slide 8
- Example Analysis of U.S. Share in SCI
- Slide 9
- Sample of 8 Fields of Science Working paper at
http://itri2.org/Bpaper/current.doc
- Slide 10
- Conclusions from Bias Study In some fields (space science,
math, and microbiology) New journals were much less favorable to
the U.S. But, in some fields the opposite was true In aggregate
over 8 fields the change in bias was too small to account for the
sharp changes in national shares Most share differences between Old
and New journals are not statistically significant Therefore,
hypothesis is not proven, and is unlikely over 24 fields in the
National Science Indicator CD Thus the shifts in national shares
are real, and are probably not an artifact of the SCI database
- Slide 11
- 2. Since Total GERD Is Not Increasing U.S. Papers, Which Inputs
are Most Important? Multiple Linear Regression Can Help Identify
Which National Publication Systems (The Black Box) Resources In
Papers Out
- Slide 12
- A Simple Regression Scattergram for 1999 This example has only
one IV, of course. Regression Line: Papers99 = 4282 + 0.950 GERD99
Slope is about 1 paper / $1 million, and R = 0.983. U.S. and Japan
are far off the graph to the upper right
- Slide 13
- Multiple Linear Regressions: Drilling Down with 2 Independent
Variables (IV) per Level DV is papers out. Significance level is
0.05, but IV1 is always very significant (Sig. 0.000). R > 0.9
always. Level1. IV1 = GERD (Total national R&D) IV2 = Number of
Researchers (not significant) Level2. IV1 = Government Investment
part of GERD IV2 = Private Sector Investment (not significant)
Level3. IV1 = Civil Part of Government Investment IV2 = Government
Defense Investment (less significant and a smaller coefficient)
Details in table in Appendix
- Slide 14
- Regression Findings: In Producing Papers Research investments
are much more important than the number of researchers. Government
investments are much more important than those from industry.
Government investments in non-defense sectors are somewhat more
important than their investment in defense. For 2003 the ratio of
their IV coefficients has increased to 2. Not surprising to
bibliometricians, but regression quantifies these statements.
Causality cant be proven, but strong and weak associations can be
found.
- Slide 15
- Is This Why the EU Passed the U.S. in Science Publishing in
1995? (Greater Government Non-Defense R&D) May account for EU
gaining on U.S. before 1995, but we still need to find why loss of
U.S. share accelerated then, here shown as leveling of US Papers
curve
- Slide 16
- Conclusions from Regression Money is the key input variable,
but some types of investments are more effective U.S. Government
investments are not as efficient at producing papers as in the EU:
too much defense R&D, and not enough NSF These inefficiencies
may account for the EU passing the U.S. in papers, but are not
large enough to account for the recent sharp decline in U.S. paper
share
- Slide 17
- 3. A Theory That Can Account for U.S. Decline, and Predict its
(Bleak) Future Since we cant blame the recent sharp decline in U.S.
share on a rubber ruler from SCI or inefficient allocations by the
U.S., we have to cast a wider net overseas, in fact. We will look
at four regions: US, EU-15, "Asian Tigers" (China, Taiwan, S.
Korea, and Singapore), and rest of the world (ROW). Usually the
world will be the OECD countries, including affiliates.
- Slide 18
- More Detailed Model of Publication System (Inside the Black
Box) $ Inputs US EU AT ROW Papers Published National Research
Systems -- Fairly Independent Highly Interdependent Paper Selection
Journal Editors g1g1 p1p1 G (total)P (total) w i = g i /G GERD
share m i = p i /P Paper share
- Slide 19
- Search for a Model of System Since SCI does not increase P much
in a year, total papers is almost a zero-sum game, and is one for
share m i (on a fractional count basis) Thus competition with other
nations strongly influences U.S. success We know that the critical
input is money, but we need a U.S. series that declines to get a
strong positive R with declining U.S. paper share
- Slide 20
- US GERD has risen, but others have risen faster. Is the shape
of this curve familiar? An Input Series that Involves Both the
Investments of the U.S. and Others.
- Slide 21
- US Paper Share Depends More on Its Share of GERD Than Its Total
GERD
- Slide 22
- m i is share of papers published (fractional basis) w i is the
share of GERD for the OECD Group k i is a "constant" of
proportionality; it differs by country. k i is also the efficiency
of country i in producing papers per $1 million in GERD, normalized
by the OECD average efficiency. For data in a single year the
equation is an identity, but it is most useful over a range of
years when k i is approximately constant A Simple Model for Country
i m i = k i w i
- Slide 23
- Since 1998, k i Has Been Fairly Constant EU is about 40% more
efficient; AT is about 40% less efficient. Source for papers: NSF
S&EI 2006 (fractional counts)
- Slide 24
- Input Variable w i for m i = k i w i U.S. GERD share declined
because of aggressive investments by the Asian Tigers. ROW is
fairly flat. Since these add to 100%, U.S. and EU both lost GERD
share to AT. Can you forecast what is going to happen next?
- Slide 25
- Paper Share Fractional Count Lately, EU shows the same decline
as U.S.
- Slide 26
- The Future? Asian Tigers have announced plans to continue to
rapidly increase R&D investment. The American Competitiveness
Initiative is small compared to Asian plans--no net increase EU
seems to be faltering in meeting Lisbon goals Thus, U.S. and EU
investment shares will probably continue to be lost to the Asian
Tigers Based on the theory here, leadership of science publication
will probably continue to shift from the U.S. and EU to Asia But,
science itself is not zero-sum; a spirited competition helps
all
- Slide 27
- Appendix Also see review draft of text paper at:
http://itri2.org/Apaper/current.doc Complete citations are
there.
- Slide 28
- Detailed Regression Results: DV is papers unless stated
- Slide 29
- More Regression Findings: USG Allocations I cant resist
drilling down yet one more level, to the R&D funding by each
U.S. agency, using AAAS data U.S. government investment has been up
in recent years, but most of that is in defense. Even the
non-defense budget has been up some, but almost all that increase
is at National Institutes of Health. Except for NIH, non-defense
government research investment has been flat in constant dollars
since 1992. Since there has been very little increase in U.S.
publication in healthcare fields during this interval, this
resulted in a loss of publication share.
- Slide 30
- No Bang Per Buck: NIH Papers and Budgets (Papers in Psych.,
Clinical Meds, Pharma, Immune, Neuro.) Budgets are in constant
$
- Slide 31
- Now Use Model with Constant k i to See If It Could Have
Predicted Paper Share Decline NSI CD whole counts, actual GERD used
each year.
- Slide 32
- Accuracy of Model Using 1995 Value of k i Fractional count
basis, actual GERD shares used.
- Slide 33
- Slide 34