PresentationPDF Available

Global warming: An anti-scientific movement

Authors:
Global warming:
An anti-scientific movement
J. Scott Armstrong
The Wharton School, University of Pennsylvania
jscottarmstrong.com
Kesten C. Green
International Graduate School of Business,
University of South Australia
kestencgreen.com kesten@me.com
Bucharest Dialogues
Expert Knowledge, Prediction, Forecasting: A Social Sciences Perspective
20 November 2010
File: AGW political-Bucharest 2010-R6
1
Purpose of the talk
We do not want to convince you, nor can
we. We only want to explain how we
reached some conclusions about global
warming, and how we believe the
problem should be addressed.
Our expertise is limited to forecasting
methodology.
We will provide you with the sources of
evidence that we relied upon.
2
Science is a process:
Our working definition
1. Objective
a) multiple reasonable hypotheses (vs.
advocacy)
b) evidence-based procedures
c) decentralized funding: market place of ideas
2. Replicable
a) disclosed: full & understandable disclosure of
methods and data
b) corrected: errors corrected
3
Global warming:
An anti-scientific political movement
J. Scott Armstrong
The Wharton School, University of Pennsylvania
jscottarmstrong.com
Kesten C. Green
International Graduate School of Business,
University of South Australia
kestencgreen.com kesten@me.com
Bucharest Dialogues
Expert Knowledge, Prediction, Forecasting: A Social Sciences Perspective
20 November 2010
File: AGW political-Bucharest 2010-R6
4
Multiple Hypotheses
Which sciences progress, asked Chamberlin
in 1890?
Those that follow the method of multiple
reasonable hypotheses.
Chamberlin, T. C. (1890).
Platt (1964).
5
Few scientists use multiple
reasonable hypotheses
Example: In “marketing science,” only 13%
of the studies used the method of multiple
hypotheses. Similar results for
“management science.
Armstrong, Brodie & Parsons, (2001)
6
Is it valid to use scientists opinions
to make forecasts?
Research over nearly 80 years has
shown that experts‟ unaided opinions*
have no value for forecasting for
problems that involve…
high uncertainty
complex situations
poor feedback
* Unaided by scientific forecasting principles
7
8
Can climate scientists make useful
climate forecasts?
“Seer-sucker theory”:
Armstrong (1985) Literature review: ”No matter how
much evidence exits that seers do not exist, suckers
will seek seers.
Tetlock (2005) evaluated:
• 82,361 forecasts
• made over 20 years
• by 284 professional commentators and advisors
on politics and economics
Expertise did not lead to better forecasts than those
from students or from using simple rules.
What about climate models?
They are expressions of expertsopinions and
judgments on poorly understood phenomena.
They produce scenarios* not forecasts.
Scenarios create the impression that events
are more likely than they are.
Scenarios are not valid for forecasting.
* Scenarios are elaborate stories about the future that are told in the
past tense.
Source: Gregory & Duran (2001)
9
Scientific Forecasting
Scientific forecasting (as contrasted to
forecasting by scientists) depends upon
evidence-based procedures.
Such procedures were developed by the
interdisciplinary team of 39 forecasting
experts in the late 1990s, culminating in
Principles of Forecasting.
There are currently 140 principles provided at
ForPrin.com with an open invitation for
challenging, revising or adding principles.
Example: “Statistical fit is not related to forecast
accuracy for time-series data.
10
1111
Audit of IPCC forecasting procedures
(Are they valid?)
IPCC “scenariosof global temperature
change used improper procedures.
Our forecasting audit* showed that
IPCC authors violated 72 of the 89
relevant forecasting principles.
*Green & Armstrong (2007).
12
Conditions for long-term climate
forecasts
1. Climate is complex.
2. Much uncertainty*:
causes of changes are disputed,
existence and direction of feedbacks not clear,
causal factors difficult to forecast (e.g., energy from the
sun),
data are subject to error.
3. No clear-cut long-term trends
* Evidence: Idso & Singer (2009).
800,000-year Record of Antarctic
Temperature Change: No clear-cut trend
Green 2009: Public policy lessons from history 13
Method selection
Selection of methods should use evidence-
based findings as contrasted to best fit or
even best ex ante forecasts.
Use of the Forecasting Method Selection
Tree* led us to use of the naïve (no
change) model
* ForPrin.com
14
Why the no-trend model is appropriate
Complex situations with high uncertainty call for
simple methods
The no-trend (or no-change) model violates:
(1) Use all relevant data &
(2) combining.
Can be corrected, but gains in accuracy would
be be minor.
Simplicity seems counter-intuitive. Hogarth,
R. (in press). “When simple is hard to
accept.
15
1616
Validation test
of the IPCC forecasts
1992 IPCC report’s 0.03oC/year linear
projection
We tested IPCC vs. no-change model
for 1851 through 2008 (simulated ex
ante)*
* IPCC model has an advantage because the trend
in these data were known to the IPCC modelers.
1717
Design of validation test
Used UK Hadley Centre‟s best estimate” of global mean
temperatures from 1850 through 2007 (HadCRUt3).
Absolute errors calculated as forecast vs. actual
Forecast for up to 100 subsequent years on rolling
horizon:
157 one-year-ahead forecasts
58 hundred-year-ahead forecasts
10,750 forecasts across all horizons
1818
IPCC performance 1851-1975 *
IPCC/No-change error ratio** < 1 means forecast errors
are smaller (better) than no-change errors
IPCC/No-change Error
Ratio n
Rolling (1-100 years) 7.7 10,750
1-10 years 1.5 1,205
41-50 years 6.8 805
91-100 years 12.6 305***
* Green, Armstrong & Soon (2009).
** A.k.a. Cumulative Relative Absolute Error or CumRAE
*** Covers only 1941 through 2007
Correlations with Global Mean
Temperature
19
Series Correlation
Atmospheric CO21850-2008 0.86
U.S. Price Index 1850-2009 0.85
NOAA* expenditure 1970-2006 0.82
No change model 0.00
*National Oceanic and Atmospheric Administration
Fit not related to forecast accuracy
Results from this validation study
consistent with research on time-
series forecasting:
correlation Error
with temp ratio*
Naïve model 0.00 1.0
IPCC 0.86 7.7
* Averaged over all forecast horizons
20
Effects of global warming: Good, or bad?
When the CO2causes warming theory was
proposed, in the early-1900s (by the Swedish
Nobel Prize winner, Svante Arrhenius), he
expected that the effects to be beneficial.
Many assessments of effects, but we found no
scientific forecasts.
21
Ill-conceived attempts to forecast
the effects of GW
Our audit of two government reports on
effects of global warming on polar bear
numbers showed that they followed only
13% of relevant forecasting principles.
Rather than a sharp decrease, we forecast a
modest short-term increase in the polar bear
population.
22
Global warming movement is not based
on scientific method: Our ratings
1. Objective
a) multiple hypotheses?
No. Single hypothesis and appeal to precautionary principle.
b) Validated methods
No. Violated 72 forecasting principles
Long-term IPCC forecast errors 12 times larger than
for an evidence-based method.
c) Decentralized
No. Large government and business control over funds and reports
2. Replicable
a) full & understandable disclosure of methods and data?
No. Refusal to share; ClimateGate; Polar bears
b) correction of errors?
No. ClimateGate; Polar bears; New Zealand official temps.
23
Conclusions with respect to
the forecasting process
There are no scientific forecasts of:
(1) manmade global warming, or
(2) net harmful effects due to warming, or
(3) net beneficial effects from proposed policies.
Forecasts of dangerous manmade global warming are
the product of an anti-scientific political movement.
Green & Armstrong (2010) forecasts the outcome of this
movement
24
Solution:
Remind people of personal standards
Funding from sources associated with different biases.
Biases should be fully disclosed.
Asking people how they would act as “individuals”
completely eliminated the problem in the Panalba study.
25
Why do people in organizations do
unethical things?
One reason is “obedience to authority.” (an
extension of Milgram‟s experiments on obedience to
authority where no money was involved).
Panalba Case: When groups role-played the Board
of Directors of the Upjohn pharmaceutical company,
none removed a dangerous drug that had no
competitive benefits. 79% took strong action to
block the FDA.
Armstrong (1977)
26
Solution:
Make standards explicit
Ask each individual to sign a statement that
they have followed all of the steps involved
in the scientific method.
We would be happy to do so for our
research.
Would IPCC scientists? We could ask them.
[A prediction problem: % “yes”?]
27
Would science improve if scientists
signed a code of conduct?
The Ten-commandments study
Chance to cheat on a test involving self-reported
scores.
Most control group subjects cheated.
What percent of those first asked to write the Ten
Commandments cheated?
Zero
The MIT honor code study: Similar study, as half
signed a statement that they knew that the MIT
honor system applied.
Zero
Source:Mazar, Amir & Ariely (2008) 28
Solution: Decentralized research
Allow scientists to discover and study problems on
their own or to obtain private sponsorship.
Proper forecasting for climate change could easily
be done by individuals. It would not be expensive.
If money is needed, prizes could be offered.
For example, we had no funding for our research
to date (with the exception of about $3,000 each
for the initial polar bear audit).
29
Evidence on decentralized research
Adam Smith‟s answer to why Scottish professors
“invented everything” while professors in England
did little.
Centrally funded research and the associated
problems with peer review, stamp out objectivity
and productivityand creativity.
Milton Friedman and many others agreed, and
experimental studies reach the same conclusion.
Kealey (1996)
30
Solution: Funded researchers must
agree to use scientific procedures
Present researchers with the procedures for
scientific forecasting.
Exceptions must be empirically supported
and agreed upon in advance.
What‟s to disagree with?
Examples of commonly violated principles:
• Full-disclosure of data and procedures
• Replication
31
Solution: Eliminate government
funding for research
Why? Because it:
introduces bias and inefficiency.
stifles innovation. (Committees do not
invent things that create wealth.)
Evidence: Kealey (1996).
32
Solution: Publish all research
(No censorship)
No need to publish based on peer
review. Technology has solved the
problem.
Peer review to decide on publication
is contrary to free speech and is
detrimental to scientific progress.*
*Evidence: Armstrong (2001); Armstrong & Pagell (2003)
33
Summary of Recommendations
for scientific research
1. Make people aware of their personal standards
2. Make standards explicit
3. Make funding conditional on use of scientific
procedures
4. Decentralize research
5. Eliminate government funding
6. Publish all findings (No censorship)
34
“The Precautionary Principle”
It is a political principle holding that if a government is
persuaded that even a small risk of a risk of a high cost
eventuality exists, there is no need for a rational analysis.
Contrary to scientific analyses of costs and benefits.
Brings to mind the slogan on the Ministry of Truth building
in George Orwell‟s 1984: “Ignorance is Strength.
Scientific forecasting suggests appropriate policy in this
case is “don‟t just do something, stand there!
For more see Green & Armstrong (2008)
35
3636
No-change model forecast errors
ResearchGate has not been able to resolve any citations for this publication.
ResearchGate has not been able to resolve any references for this publication.