Ba

White Paper вЂ" Pew Center on Global Climate Change 1

White Paper

Insights Not Numbers:

The Appropriate Use of Economic Models

By

Janet Peace and John Weyant

April, 2008

White Paper вЂ" Pew Center on Global Climate Change 2

Insights Not Numbers:

The Appropriate Use of Economic Models1

by

Janet Peace and John Weyant

April, 2008

Executive Summary

Economic modeling has played a prominent role in the climate-change policy debate as

stakeholders have sought to understand the impacts and assess the costs of different strategies for

reducing greenhouse gas (GHG) emissions. Models are an invaluable tool for exploring

alternative policy choices and for generating insights about how the economy is might respond to

different types and forms of regulation. They cannot, however, predict future events, nor can

they produce precise projections of the consequences of specific policy.

Every model uses its own set of assumptions, definitions, structure and data вЂ" its results

ultimately depend on these attributes and choices. A proper understanding of economic models,

their uses and limitations, is therefore critical in furthering a constructive debate about options

for climate policy. As a starting point we highlight three general observations about the use of

economic models:

• While economic models have become increasingly sophisticated, forecasting the future

remains inherently uncertain. The longer the time horizon of the analysis, the larger the

uncertainties involved.

• Model results are strongly dependent on input assumptions and on the structure of the

model itself. Critical assumptions and structural biases are not always readily apparent to

the outside observer.

• What is left out of a model can be as important as what goes in. Whether a model

accounts for the benefits (or avoided costs) of climate mitigation, technological change

1 The maxim “insights, not numbers” has a long and illustrious history starting with Hamming (1962) who argued

that “insights not numbers" constitute the purpose of computing. The same maxim was subsequently applied by

Geffrion (1976) in the context of mathematical programming and by Huntington, et al. (1982) in the context of

mathematical modeling. We are also indebted to William Hogan who made the link to the Geoffrion piece and

Richard Richels for occasionally reminding us what our objectives in modeling ought to be. These ideas probably

all build on the work of W. Edwards Demming in the 1950s who, without ever explicitly using the phrase, surely

implied that insights, not numbers are the purpose of statistical quality control.

White Paper вЂ" Pew Center on Global Climate Change 3

spurred (or “induced”) by climate policy, or the “recycling” of revenues generated

through certain policies can have large effects on the results.

Many of the cost analyses published over the last decade rely on general equilibrium models that

use complex systems of mathematical equations and large amounts of data to simulate the

workings of the economy. Comparisons across multiple studies suggest that several categories

of assumptions are especially important in driving model results:

(1) specific features of the policy or policies being analyzed (including the degree of

flexibility allowed in meeting the emissions constraints);

(2) reference case (or baseline assumptions) about how the economy and environment will

perform in the absence of the policy;

(3) flexibility in the economyвЂ"that is, the ease with which consumers/producers can adapt

to emissions limits;

(4) pace and magnitude of technological change/innovation; and

(5) treatment of benefits (or avoided costs) from climate-change mitigationвЂ"what benefits

are included and how.

A detailed comparison of results from two modeling initiatives sponsored by the Pew Center

reveals that cost estimates can differ widely as a result of structural characteristics and

assumptions embedded in the model, even where other key parameters (such as the policy being

analyzed and base-case projections of future emissions) are the same. For example, the

responsiveness (or elasticity) of various components of the economyвЂ"including assumptions not

only about how readily low-carbon alternatives will be substituted for carbon-intensive goods

and services, but also about how readily individuals make trade-offs between consumption and

leisure are critical assumptions. A model which assumes a highly responsive