Critique of Limits to Growth
Posted: Wed Feb 28, 1996 10:12 am
Re Wayne Wakelands comments on Bill Nordhaus:
The article Wayne is refering to is "World Dynamics: Measurement
without data" and appeared in the Economic Journal in Dec 73.
It is a very interesting article, not least for the many errors and
misunderstandings it contains. Nordhaus is a distinguished economist,
and has done excellent work in some areas. This work, however, is one
of the least civil and most rude pieces of ad hominem invective Ive
ever seen in an academic journal. The article does little to advance
the cause of civilized discourse for the purpose of greater
understanding, something I was taught was important as a foundation for
the advancement of scientific knowledge as well as just being good
manners.
More to the point, the substance of the article contains numerous errors
and misunderstandings. Wayne refers to the claim, advanced in the
paper, that the relationship between some variables in the world2 model
is "backwards", i.e. has the wrong slope compared to the data. This is
indeed what Nordhaus claims, but his analysis is flawed. The particular
relationship has to do with the impact of food per capita on birth
rates. In world2 this is an upward sloping curve (no food, no births;
low food: high infant mortality; lots of food: the relationship
saturates at something close to the biological potential). Nordhaus
shows that as food per capita has risen in the developed world, birth
rates have fallen (which is true). The problem is confounding of
causation with correlation: food per capita has risen along with income
per capita, and the birth rate has fallen. But it is not the greater
access to Big Macs that reduces births, but the higher costs and lower
benefits of children in an industrial society compared to an agrarian
society that drives birth rates down (this is the essence of the theory
of the demographic transition). Rising income increases food
consumption; rising income also reduces desired family size and thus
births. Hence birth rates and food per capita are negatively
correlated. The causal relationship between food per capita and live
births, however, is positively sloped, as Forrester assumes (No food, no
live births - no doubt about it). What Nordhaus did was to compare the
structural assumptions of the model to the aggregate behavior of the
real system. In fact, the behavior of the world model is consistent
with the behavior of the real system. If you run the model with, say,
the initial conditions for the United States, it generates the same
negative correlation between food and births seen in the actual data.
Nordhaus made some good points in the article, but the claim that some
model relationships are "backwards" is false and based on erroneous
analysis.
Interested readers should refer to the rebuttal of the Nordhaus article,
Forrester, J. W., Mass, N. J., & Low, G. W. (1974). The Debate on World
Dynamics: A Response to Nordhaus. Policy Sciences, 5(2), 169-190.
Model critiques cut both ways: readers might be interested to see the
modeling errors and hidden technological optimist assumptions in
Nordhaus well-known model of global climate change, documented in Tom
Fiddamans paper from the 1995 System Dynamics conference in Tokyo. The
errors include non-conservation of mass (in this case carbon) and DT
error; the technological optimist assumptions include a high discount
rate on utility (the welfare of future generations is not worth as much
as the welfare of people today: your grandchildren simply dont
count), exogenous population growth, lowball estimates of future
economic growth, high estimated costs of GHG abatement with no
technological advance in abatement but high technological advance for
aggregate productivity, and so on. The lesson for us as modelers is
that we are all blind to the underlying axioms of our own world-view,
and must take great care to expose our biases and assumptions in our
modeling work. This is why the system dynamics community has always put
such great effort into documentation (the world2 and world3 models are
among the most heavily critiqued models ever published in large measure
because all equations, assumptions, and data were fully documented and
the simulations were fully reproducible -- shocking as it may seem, this
is not the norm in economics or the social sciences).
I strongly object to Waynes unsubstantiated attribution about "the
tendency of S-D modelers to make assumptions that are not backed up with
data." We have worked hard to find ways to confront our models with
data. Good system dynamics work is solidly grounded in data of all
types. Waynes statement is simply false as a characterization of the
field of system dynamics, and it is highly debatable even regarding the
world models. Lets keep unsubstantiated generalizations out of the
discourse on these issues and focus on documented specifics.
John Sterman
Sloan School of Management
MIT, E53-351
30 Wadsworth Street
Cambridge, MA 02142
phone: 617/253-1951 fax: 617/258-7579 e-mail: jsterman@mit.edu
The article Wayne is refering to is "World Dynamics: Measurement
without data" and appeared in the Economic Journal in Dec 73.
It is a very interesting article, not least for the many errors and
misunderstandings it contains. Nordhaus is a distinguished economist,
and has done excellent work in some areas. This work, however, is one
of the least civil and most rude pieces of ad hominem invective Ive
ever seen in an academic journal. The article does little to advance
the cause of civilized discourse for the purpose of greater
understanding, something I was taught was important as a foundation for
the advancement of scientific knowledge as well as just being good
manners.
More to the point, the substance of the article contains numerous errors
and misunderstandings. Wayne refers to the claim, advanced in the
paper, that the relationship between some variables in the world2 model
is "backwards", i.e. has the wrong slope compared to the data. This is
indeed what Nordhaus claims, but his analysis is flawed. The particular
relationship has to do with the impact of food per capita on birth
rates. In world2 this is an upward sloping curve (no food, no births;
low food: high infant mortality; lots of food: the relationship
saturates at something close to the biological potential). Nordhaus
shows that as food per capita has risen in the developed world, birth
rates have fallen (which is true). The problem is confounding of
causation with correlation: food per capita has risen along with income
per capita, and the birth rate has fallen. But it is not the greater
access to Big Macs that reduces births, but the higher costs and lower
benefits of children in an industrial society compared to an agrarian
society that drives birth rates down (this is the essence of the theory
of the demographic transition). Rising income increases food
consumption; rising income also reduces desired family size and thus
births. Hence birth rates and food per capita are negatively
correlated. The causal relationship between food per capita and live
births, however, is positively sloped, as Forrester assumes (No food, no
live births - no doubt about it). What Nordhaus did was to compare the
structural assumptions of the model to the aggregate behavior of the
real system. In fact, the behavior of the world model is consistent
with the behavior of the real system. If you run the model with, say,
the initial conditions for the United States, it generates the same
negative correlation between food and births seen in the actual data.
Nordhaus made some good points in the article, but the claim that some
model relationships are "backwards" is false and based on erroneous
analysis.
Interested readers should refer to the rebuttal of the Nordhaus article,
Forrester, J. W., Mass, N. J., & Low, G. W. (1974). The Debate on World
Dynamics: A Response to Nordhaus. Policy Sciences, 5(2), 169-190.
Model critiques cut both ways: readers might be interested to see the
modeling errors and hidden technological optimist assumptions in
Nordhaus well-known model of global climate change, documented in Tom
Fiddamans paper from the 1995 System Dynamics conference in Tokyo. The
errors include non-conservation of mass (in this case carbon) and DT
error; the technological optimist assumptions include a high discount
rate on utility (the welfare of future generations is not worth as much
as the welfare of people today: your grandchildren simply dont
count), exogenous population growth, lowball estimates of future
economic growth, high estimated costs of GHG abatement with no
technological advance in abatement but high technological advance for
aggregate productivity, and so on. The lesson for us as modelers is
that we are all blind to the underlying axioms of our own world-view,
and must take great care to expose our biases and assumptions in our
modeling work. This is why the system dynamics community has always put
such great effort into documentation (the world2 and world3 models are
among the most heavily critiqued models ever published in large measure
because all equations, assumptions, and data were fully documented and
the simulations were fully reproducible -- shocking as it may seem, this
is not the norm in economics or the social sciences).
I strongly object to Waynes unsubstantiated attribution about "the
tendency of S-D modelers to make assumptions that are not backed up with
data." We have worked hard to find ways to confront our models with
data. Good system dynamics work is solidly grounded in data of all
types. Waynes statement is simply false as a characterization of the
field of system dynamics, and it is highly debatable even regarding the
world models. Lets keep unsubstantiated generalizations out of the
discourse on these issues and focus on documented specifics.
John Sterman
Sloan School of Management
MIT, E53-351
30 Wadsworth Street
Cambridge, MA 02142
phone: 617/253-1951 fax: 617/258-7579 e-mail: jsterman@mit.edu