Draft Report to Congress on the Costs and Benefits of Federal Regulations

SUMMARY: The Office of Management and Budget (OMB) requests comments on

the attached draft report to Congress on the costs and benefits of

Federal regulations. The draft report is divided into an introduction

and four chapters. The introduction sets the context and provides the

background for the next four chapters. Chapter I presents OMB's best

estimate of the total costs and benefits of Federal regulatory programs

and discusses several retrospective studies of specific regulatory

programs to gain insight on how actual costs and benefits of

regulations may differ from the effects predicted prior to regulation.

Chapter II provides data on the costs and benefits of each of the

economically significant regulations reviewed by OMB under Executive

Order 12866 in the last year. Chapter III provides additional data on

the costs and benefits of the economically significant regulations

reviewed by OMB from April 1, 1995 through March 31, 1998. Chapter IV

discusses how OMB implemented last year's recommendations and presents

the Administration's proposal to restructure and deregulate the

electricity sector.

DATES: To ensure consideration of comments as OMB prepares this draft

report for submission to Congress on or before September 30, 1998,

comments must be in writing and received by OMB no later than September

16, 1998.

ADDRESSES: Comments on this draft report should be addressed to John F.

Morrall III, Office of Information and Regulatory Affairs, Office of

Management and Budget, NEOB, Room 10235, 725 17th Street, NW.,

Washington, DC 20503.

Comments may also be submitted by facsimile to (202) 395-6974, or

by electronic mail to [email protected] (Please note that the

``l'' in ``A1'' is the number one and not the letter ``l''.) Be sure to

include your name and complete postal mailing address in the comments

sent by electronic mail. If you submit comments by facsimile or

electronic mail, please do not submit them by regular mail also.

Electronic availability and addresses: This Federal Register notice

is available electronically from the OMB homepage on the World Wide

Web: http://www.whitehouse.gov/WH/EOP/OMB/html/fedreg.html.

FOR FURTHER INFORMATION CONTACT: John F. Morrall III, Office of

Information and Regulatory Affairs, Office of Management and Budget,

NEOB, Room 10235, 725 17th Street, NW., Washington, DC 20503.

Telephone: (202) 395-7316.

SUPPLEMENTARY INFORMATION: Congress directed OMB to prepare a report to

Congress on the costs and benefits of Federal regulations.

Specifically, under section 625 of the Treasury and Government

Appropriations Act, 1998 (Pub. L. 105-61), the Director of OMB is to

submit to Congress, no later than September 30, 1998, a report that, in

summary, provides (1) estimates of the total annual costs and benefits

of Federal regulatory programs, (2) estimates of the costs and benefits

of each rule that is likely to have a gross annual effect on the

economy of $100,000,000 or more in increased costs, (3) an assessment

of the direct and indirect impacts of Federal rules, and (4)

recommendations from OMB and a description of significant public

comments to reform or eliminate any Federal regulatory program that is

inefficient, ineffective, or is not a sound use of the Nation's

resources.

The attached document is a draft of this report to Congress. OMB is

to provide public notice and an opportunity to comment on the report

before it is submitted to Congress no later than September 30, 1998.

Issues for Comment

Accordingly, OMB seeks comment on all aspects of the attached draft

report, particularly comments and suggestions pertaining to the

following:

The validity and reliability of our new estimates of the

costs and benefits of regulations in the aggregate, as well as by

regulatory program or program element;

Our discussion of the methodological problems of

estimating the costs and benefits of existing rules, e.g., the baseline

and comparability problems and complications introduced by using

prospective studies to evaluate existing programs; and difficulties

reconciling quantitative and qualitative estimates of costs and

benefits;

Our review of several case studies of the costs and

benefits of existing regulations and the lessons we draw from them;

Any additional studies that might provide reliable

estimates or assessments of the annual costs and benefits, or direct

and indirect effects on the private sector, State and local government,

and the Federal Government, of regulation in the aggregate or of the

individual regulations that we discuss;

Our approach to estimating the costs and benefits of the

individual regulations issued between April 1, 1995, and March 31,

1998, that we discuss, and;

Programs or program elements on which there is objective

and verifiable information that would lead to a conclusion that such

programs are inefficient or ineffective and should be eliminated or

reformed.

Bruce McConnell,

Acting Administrator, Office of Information and Regulatory Affairs.

Draft Report to Congress on the Costs and Benefits of Federal

Regulations

Introduction

The Office of Management and Budget issued its first report to

Congress on the costs and benefits of Federal regulations on September

30, 1997. Section 625 of the Treasury and Government Appropriations

Act, 1998 (P.L. 105-61) directs OMB to issue a second regulatory

accounting report. The requirements of the report are the same as those

of last year. Section 625(a) directs the Director of the Office of

Management and Budget to submit to Congress, no later than September

30, 1998, a report that provides:

``(1) Estimates of the total annual costs and benefits of

Federal regulatory programs, including quantitative and non-

quantitative measures of regulatory costs and benefits;

``(2) Estimates of the costs and benefits (including

quantitative and non-quantitative measures) of each rule that is

likely to have a gross annual effect on the economy of $100,000,000

or more in increased costs;

``(3) An assessment of the direct and indirect impacts of

Federal rules on the private sector, State and local government, and

the Federal Government; and

``(4) Recommendations from the Director and a description of

significant public comments to reform or eliminate any Federal

regulatory program or program element that is inefficient,

ineffective, or is not a sound use of the Nation's resources.''

In last year's report we indicated that a complete accounting of

total costs and benefits of Federal regulation was a difficult

undertaking. The 1997 report was our effort to begin an incremental

process which we believe will lead to improved information on the

effects of regulations, and will help solve the many methodological

problems associated with this exercise. This year's report builds on

last year's work. In particular, we have additional data to

[[Page 44035]]

supplement our discussion of the aggregate costs and benefits of

regulation and expand our database of costs and benefits of individual,

major rules from one year (1997) to three years (1996 to 1998). In

addition, we have more experience in dealing with the methodological

problems.

One fact has not changed since the first report. There are still

enormous data gaps in the information available on regulatory benefits

and costs. Although accurate data is still sparse and agreed-upon

methods for estimating many effects are still lacking, we have made

significant progress in improving these estimates, especially for the

major rules of the last three years. As we stated last year, explicitly

quantifying and monetizing benefits and costs significantly enhances

our ability to compare alternative approaches to achieving regulatory

goals, ultimately producing more benefits with fewer costs. President

Clinton's Executive Order 12866, ``Regulatory Planning and Review,''

recognizes and incorporates this principle, requiring agencies to

quantify both costs and benefits to the best of their ability and to

the extent permitted by law. We continue to recognize that significant

regulatory costs and benefits may not be quantifiable, but may have to

be described in qualitative terms. All information, both qualitative

and quantitative, contributes to our understanding of the effects of

regulation.

This year's report presents new information on both the total costs

and benefits of regulation and the costs and benefits of major

individual regulations. We hope to continue this important dialogue to

improve our knowledge about the effects of regulation on the public,

the economy, and American society.

This document is a draft of our report. Section 625(b) requires the

Director of OMB to provide public notice and an opportunity to comment

on the report before it is submitted to Congress at the end of

September 1998. Furthermore, the final report is to contain a

description of significant public comments. Accordingly, we seek

comments on all aspects of this document, but in particular are

interested in comments and suggestions pertaining to the following:

The validity and reliability of our new estimates of the

costs and benefits of regulations in the aggregate, as well as by

regulatory program or program element;

Our discussion of the methodological problems of

estimating the costs and benefits of existing rules, e.g., the baseline

and comparability problems and complications introduced by using

prospective studies to evaluate existing programs;

Our review of several case studies of the costs and

benefits of existing regulations and the lessons we draw from them;

Any additional studies that might provide reliable

estimates or assessments of the annual costs and benefits, or direct

and indirect effects on the private sector, State and local government,

and the Federal Government, of regulation in the aggregate or of the

individual regulations that we discuss;

Our approach to estimating the costs and benefits of the

individual regulations issued between April 1, 1995, and March 31,

1998, that we discuss; and

Programs or program elements on which there is objective

and verifiable information that would lead to a conclusion that such

programs are inefficient or ineffective and should be eliminated or

reformed.

All comments received will be carefully considered in preparing the

final report that will be submitted to Congress.

The draft report is divided into four chapters. In accordance with

section 625(a)(1), chapter I presents our best estimate of the total

costs and benefits of Federal regulation. It builds on chapter II of

last year's report presenting updated and more detailed estimates of

the total annual costs and benefits of major Federal regulatory

programs.1 In particular, this year we present more

categories of regulatory costs and benefits than last year and use our

own estimates based on agency data of costs and benefits of individual

rules issued over the last three years (April 1, 1995 to March 31,

1998) to update the aggregate estimates. We also chose this year to

provide ranges of costs and benefits rather than point estimates to

emphasize the uncertainty embodied in the estimates.

---------------------------------------------------------------------------

\1\ Chapter I of last year's report discussed the role of

economic analysis in regulatory reform. We discussed the growth and

nature of regulation, the development of the U.S. regulatory

analysis and review program and the basic principles that should be

used in assessing regulatory costs and benefits. We did not repeat

that discussion this year but it is still useful for understanding

the context of this year's report. (See OMB 1997 or http://

www.whitehouse.gov/WH/EOP/OMB/html/rcongress.htm).

---------------------------------------------------------------------------

As we did last year, we use the study by Hahn and Hird (1991) for

the costs and benefits of regulations as of 1988, supplemented by an

Environmental Protection Agency (EPA) Cost of a Clean Environment

report to Congress (1990). We also use a new (1997) retrospective EPA

report to Congress (The Benefits and Costs of the Clean Air Act, 1970

to 1990). Because there are no studies comparable to the Hahn and Hird

or the EPA retrospective studies for the regulations issued after

1988,2 we use information about costs and benefits from

agency prospective regulatory impact analyses (RIAs) to account for the

major regulations that have been issued since 1988. In almost all

cases, the RIAs have been subject to notice and comment and have been

reviewed by OMB. This year we have systematically started to improve

the consistency of the agency estimates and to show monetized estimates

of benefits where appropriate and feasible. We have completed this

analysis for the last three years and plan to complete additional years

in the future.

---------------------------------------------------------------------------

\2\ EPA's Clean Air Act report covers effects through 1990.

However, for the annual estimates that appear in table 1 and in the

text, we have, in consultation with EPA staff, adjusted EPA's

estimates to reflect only effects as of 1988.

---------------------------------------------------------------------------

The new estimates range from $170 billion to $224 billion in annual

costs and $258 billion to about $3.55 trillion in annual benefits for

social, i.e., health, safety, and environmental regulation. Using the

ranges to reflect the substantial uncertainty in the estimates,

quantified (and monetized) net benefits could be as low as $34 billion,

or as high as $3.38 trillion. The main reason why these estimates are

different from last year, especially on the upper end of the range of

benefits, is that we have incorporated retrospective estimates from a

recent EPA report on the benefits and costs of the Clean Air Act. This

report, discussed in detail in chapter I, estimates the benefits of the

Clean Air Act at up to $3.2 trillion. Three new regulations also

included in the estimates (EPA's revised particulate matter and ozone

primary National Ambient Air Quality Standards and OSHA's respirator

rule) are estimated (using midpoints) to provide approximately $35

billion in benefits per year. While this information is useful, we

still believe that the limitations of these estimates for use in making

recommendations about reforming or eliminating regulatory programs are

severe. Aggregate estimates of the costs and benefits of regulation

offer little guidance on how to improve the efficiency, effectiveness,

or soundness of the existing body of regulations.

Chapter I also discusses the impacts of other types of regulation

and regulatory-like activities and reviews several estimates of the

aggregate costs of regulation as well as several retrospective case

studies. Estimates of

[[Page 44036]]

the impacts of economic efficiency losses, disclosure regulation,

economic transfers, tax compliance costs, Federal on-budget regulatory

expenditures, and the possible indirect effects of regulation on the

economy as directed by section 625(a)(3) are also presented and

discussed.

In fulfillment of section 625(a)(2), chapter II provides data from

the agencies on the costs and benefits of each of the economically

significant regulations reviewed by OMB under Executive Order 12866

over the period from April 1, 1997, to March 31, 1998. The data were

developed by the agencies as required by the Executive order. For the

most part, these data were subject to notice and public comment and

reviewed by OMB. We also examined the reports on major rules that GAO

provides to Congress for the independent agencies not subject to

Executive Order 12866; however, these generally were not of sufficient

detail or quality to provide much useful information for the purposes

of this report. Finally, this chapter also highlights examples where

agencies have done a particularly exemplary job of following the

guidance in the Best Practices 3 document, which is on our

web site at http://www.whitehouse.gov/WH/EOP/OMB/html/miscdoc/

riaguide.html.

---------------------------------------------------------------------------

\3\ OMB published in 1996 a document that describes ``Best

Practices'' for preparing the economic analysis called for by

Executive Order 12866 for significant regulatory actions. This

document represents the culmination of a two-year effort by an

interagency group to review the state of the art for economic

analyses required by the Executive order.

---------------------------------------------------------------------------

Chapter III provides estimates of the costs and benefits for the

economically significant/major rules issued between April 1, 1995 and

March 31, 1998, for which we were able to estimate costs and benefits.

The estimates that we present in chapter III for regulations issued

during these three years are either straightforward agency estimates,

or estimates that we calculated using a consistent methodology and

value estimates used by the agencies for other regulations or in some

cases found in the academic literature. We estimate annual costs of

major rules for these three years to be about $28 billion while annual

benefits range from $30 to $97 billion.

Chapter IV discusses how we implemented last year's recommendations

aimed at further developing the information, methodologies, and

analyses necessary for improving the efficiency, effectiveness, and

soundness of regulatory programs and program elements as required by

section 625(a)(4). We discuss how the agencies and OMB worked together

to improve the quality of the data and analysis found in the economic

impact studies submitted to OMB under Executive Order 12866, and in

particular how we promoted the use of the Best Practices guidance

document. Finally, also in fulfillment of section 625(a) (4), we

present a discussion of the Administration's proposal to restructure

and deregulate the electricity sector.

Chapter I: Estimating the Total Annual Costs and Benefits of

Federal Regulatory Programs

A. Overview

By using new data from agency regulatory impact analyses that

accompany regulations, this chapter builds on chapter II of last year's

report (OMB 1997) to present updated and more detailed estimates of the

total annual costs and benefits of Federal regulatory programs. We also

discuss and present quantitative estimates where available of indirect

impacts and other effects of regulation and related Government

policies. Finally, several retrospective studies of specific regulatory

programs are reviewed to gain insight on how the actual costs and

benefits of regulations may differ from the effects predicted prior to

regulation.

We respond to the comments we received on last year's report in

several ways. First, we present more details by regulatory program and

build on agency analyses to monetize benefits estimates. Second, we

review the analyses from independent agencies and present more

systematic data on the costs and benefits of economic regulation, tax

compliance costs, transfers, Federal regulatory expenditures, and

indirect impacts. Finally, our review of several important

retrospective studies responds to important methodological issues

raised regarding the use of prospective studies to estimate the costs

and benefits of existing regulations.

1. Estimation Problems

Before proceeding with our new estimates, we reiterate and

reemphasize the methodological concerns and caveats that were discussed

in last year's report. These concerns remain of critical importance. It

remains extremely difficult, if not impossible, to estimate the actual

total costs and benefits of all existing Federal regulations with any

degree of precision. There is a variety of estimation problems for both

individual estimates and aggregate estimates.

In order to estimate the impact of regulations on society and the

economy, one has to determine how things would have been if the

regulation had not been issued. In other words, what is the baseline

against which costs and benefits should be measured? With respect to

estimating total costs and benefits of all Federal regulations, the

baseline problem has several dimensions. First, what would have

happened in the absence of regulation can only be an educated guess

since it never happened. Furthermore, the greater the regulatory

change, the more problematic the exercise. For example, the assumptions

of welfare economics, upon which benefit-cost analysis is based, hold

only for marginal changes in economic activities. The larger the

changes, the less sure we are of the predictions. In other words, we

can be more confident in our estimates of the costs and benefits of a

small change in the level of automobile emissions permitted than in the

costs and benefits of all Clean Air Act regulations and still more

confident than in estimates of the costs and benefits of all

regulations issued by the Federal Government since the early 1900s. If

we use as a baseline a world with no regulation, one can reasonably

argue that the benefits of regulation must clearly swamp any likely

cost.

Even disregarding the problem of modeling large changes, there are

significant difficulties in determining the counterfactual or baseline

for individual regulations that one could begin to aggregate. One can

survey firms and other regulated entities on their expected compliance

costs either prospectively, before the regulation is implemented, or

retrospectively, after the regulation has gone into effect. For both

types of studies, the problem of potential estimation bias must be kept

in mind since regulators and regulatees may have different interests in

the outcomes. The problem of bias is potentially greater for

prospective studies because both the baseline and the regulatory

effects must be predicted while for retrospective studies only the

baseline or counterfactual must be predicted. In the ordinary course,

therefore, the best estimates of the costs and benefits of regulation

are likely to be retrospective studies done by individuals who do not

have vested interests, but do have reputations as objective analysts to

uphold.

To make matters even more complicated, a third type of study is

actually needed before recommendations can be made to eliminate or

modify regulatory programs. That is a hybrid study

[[Page 44037]]

somewhere between pure prospective and pure retrospective. The ideal

hybrid study would be a retrospective study of the existing regulation

with prospectively estimated costs and benefits of eliminating or

modifying it. A hybrid study is needed because ``sunk costs,'' such as

specialized capital costs and the cost of changing procedures already

in place, make the cost savings from eliminating regulation less than

the cost of complying with those regulations. Furthermore, on the

benefit side there appears to exist an asymmetry between giving someone

a benefit and taking it away. Studies have shown that people are

willing to pay less for a benefit than what they are willing to accept

in return for its loss. In other words, once people have attained safer

jobs or cars, or cleaner air or water, they appear willing to pay more

for keeping such benefits than they were willing to pay to attain them.

Very few studies of health, safety, and environmental regulation have

attempted to estimate the actual cost savings and benefit losses that

would result from reducing or eliminating an existing

regulation.4

---------------------------------------------------------------------------

\4\ Note that the problem of bias may be the greatest in this

case because often both the regulators and the regulatees will

prefer the status quo, i.e., regulation. This appears to be the

lesson from the Occupational Health and Safety Administration's

(OSHA) reconsideration of the cotton dust standard during the Reagan

Administration. After opposing the regulation at the proposal stage

during the Carter Administration, the industry did not support the

Reagan Administration's proposal to withdraw it. (See Viscusi 1992).

---------------------------------------------------------------------------

Further, virtually all of the studies of the costs of regulation

produced to date measure the expenditures of firms required by

regulation, whereas the cost to society of regulation should be

measured by the change in consumer and producer ``surplus'' associated

with the regulation and with any price and/or income changes that may

result (Cropper and Oates 1992). At one extreme, ignoring the consumer

surplus loss produced by a ban on the sale of a product understates

costs to society because although no compliance expenditures are

required, consumers can no longer buy the product. At the other

extreme, calculating compliance expenditures based on pre-regulation

output overstates costs because if the firm raises prices to cover

compliance costs, consumers will shift to other products and thereby

reduce their welfare losses (Cropper and Oats 1992, p. 722).

Another problem is the fact that many studies that we rely on for

cost and benefit estimates are dated. Over time the dynamic nature of

the economy may affect the estimation of both benefits and costs.

Technological improvements are often cited as the reason that predicted

costs of compliance often turn out to be less than actual costs (Office

of Technology Assessment 1995). Less well noted, however, is that

technological progress also takes place on the benefit side. For

example, medical progress can reduce the future benefits estimated for

health, safety and environmental regulations, just as productivity

improvements in manufacturing reduce the costs of compliance of some

regulations. New drugs or medical procedures can reduce the benefits of

regulations aimed at reducing exposure to certain harmful agents such

as an infectious disease. Regulations aimed at increasing the energy

efficiency of consumer products or buildings may see their expected

benefits reduced by new technology that reduces the cost of producing

energy. Furthermore, productivity improvements lead directly to higher

incomes, which lead people to demand better health and more safety.

Business responds to these demands by providing safer products and

workplaces, even in the absence of regulation. Individuals with rising

incomes may also purchase or donate land to nature conservancies to

provide ecological benefits. Yet, as on the cost side, the baseline

that is used is almost always the status quo, rather than what is

likely to be true in the future.

It is often difficult to attribute changes in behavior to specific

Federal regulations apart from the many other motivating factors. In

addition to overlapping Federal regulations, often from different

agencies, e.g., environmental issues may be regulated by the

Environmental Protection Agency (EPA), the Department of Agriculture

(USDA), the Department of Energy (DOE), the Department of the Interior

(DOI), the Department of Commerce (DOC) and the Department of

Transportation (DOT), state and local regulations also require

compliance. The tort system, voluntary standards organizations, and

public pressure also cause firms to provide a certain degree of public

protection in the absence of Federal regulation. As the General

Accounting Office (GAO) points out, determining how much of the costs

and benefits of these activities to attribute solely to Federal

regulation is a difficult undertaking (GAO 1996).

Adding to the complexity, the degree to which these other factors

cause firms and other regulated entities to provide safe and healthful

products and workplaces and engage in environmentally sound practices

changes over time, generally increasing with increasing per capita

incomes and knowledge about cause and effect. Thus, although the

National Highway Traffic Safety Administration (NHTSA) has

significantly increased the safety of automobiles, it is not likely

that if the agency's regulations were eliminated the automobile

companies would discontinue all the safety features that have been

mandated. Consumers are demanding safer cars and automobile companies

are concerned about product liability. This same phenomenon is taking

place in the environmental area. Environmentally responsible behavior

is good for the bottom line. Over time, this ``rising baseline''

phenomenon, if correct, should reduce the true costs and benefits of

health, safety, and environmental regulations. Estimates of the

aggregate costs and benefits of regulation that include unadjusted

estimates from aging studies are thus likely to overestimate the

current costs and benefits of those regulations.

Yet another problem may be termed the ``apples and oranges

problem.'' The attempts to aggregate the total costs and benefits of

Federal regulations have simply added together a diverse set of

individual studies. Unfortunately, these individual studies vary in

quality, methodology, and type of regulatory costs included. In

addition to using different assumptions about baselines and time

periods problems discussed above, the studies use different discount

rates, different valuations for the same attribute, and different

concepts of costs and approaches to dealing with uncertainty, to

mention a few. Furthermore, the possibility of interaction effects

between the tens of thousands of regulations is not addressed.

A final reason that any regulatory accounting effort has limits is

the lack of information on the effects of regulations on distribution

or equity. None of the analyses addressed in this report provides

quantitative information on the distribution of benefits or costs by

income category, geographic region, or any other equity-related factor.

As a result, there is no basis for quantifying distributional or equity

impacts.

2. Types of Regulation

Because there are so many different types of Federal regulations,

it is useful to break this heterogeneous body up into categories. As we

did last year we describe five commonly used categories.

Environmental. The true social cost of regulations aimed at

improving the quality of the environment is represented by the total

value that society places on the goods and services foregone as a

result of resources being diverted to environmental protection.

[[Page 44038]]

(EPA's Cost of a Clean Environment, pp. 1-2, 1-3.) These social costs

include the direct compliance costs of the capital equipment and labor

needed to meet the standard, as well as the more indirect consumer and

producer surplus losses from lost or delayed consumption and production

opportunities due to the higher prices and reduced output needed to pay

for the direct compliance costs. In the case of a product ban or

prohibitive compliance costs, almost all of the costs represent

consumer and producer surplus losses. Most of the cost estimates used

in this report do not include consumer and producer surplus losses

because it is difficult and often impractical to estimate the demand

and supply curves needed to do this type of analysis.

Further indirect effects on productivity and efficiency result from

price and output changes that spread through other sectors of the

economy. Estimates of compliance costs likely understate substantially

the true long-term costs of pollution control.5 The

estimates used in this report do not include these indirect and general

equilibrium effects.

---------------------------------------------------------------------------

\5\ See Jaffe, Peterson, Portney, and Stavins' survey (1995), p.

153.

---------------------------------------------------------------------------

The benefits of environmental protection are represented by the

value that society places on improved health, recreational

opportunities, quality of life, visibility, preservation of ecosystems,

biodiversity, and other attributes of protecting or enhancing our

environment. This value is best measured by society's willingness-to-

pay (WTP) for these attributes. Because most types of improvement in

environmental quality are not traded in markets, benefits must be

estimated by indirect means using sophisticated statistical techniques

or ``contingent valuation'' survey methods that generally make benefit

estimation more problematic than cost estimation.

Other Social. This category of regulation includes rules designed

to advance the health and safety of consumers and workers, as well as

regulations aimed at promoting social goals such as equal opportunity,

equal access to facilities, and protection from fraud and deception.

They are often lumped together with environmental regulation in the

category of ``Social Regulation.'' Social regulation is mainly

concerned with controlling or reducing the harmful or unintended

consequences of market transactions, such as air pollution,

occupationally induced illness, or automobile accidents. These

consequences are commonly called ``negative externalities'' and

regulation designed to deal with them attempts to ``internalize'' the

externalities. This can be done by regulating the amount of the

externality, e.g., banning a pollutant or limiting it to a ``safe''

level, or regulating how a product is produced or used. Social

regulation may also require the disclosure of information about a

product, service, or manufacturing process where access to inadequate

or asymmetric information may place consumers, citizens, or workers at

a disadvantage. The techniques and methodological concerns involved in

the estimation of the social costs and benefits generated by these

rules are similar to those involved in the estimation of costs and

benefits of environmental regulation discussed above. In the results

that we report below, we further break ``Other Social'' into three

categories: transportation, labor and other regulations. The third

category includes food and drug safety, energy efficiency, and quality

of medical care regulations.

Economic. Economic regulation restricts firms' primary economic

activities, e.g., their pricing and output decisions. It may also limit

the entry or exit of firms into or out of certain specific types of

businesses. Such regulations are usually applied on an industry wide

basis, e.g., agriculture, trucking, or communications. In the United

States, this type of regulation at the Federal level has often been

administered by ``independent'' commissions, e.g., the Federal

Communications Commission (FCC), the Securities and Exchange Commission

(SEC), or the Federal Energy Regulatory Commission (FERC), whose

members are appointed but not removable without good cause by the

President. The economic losses caused by this type of regulation result

from the higher prices and inefficient operations that often occur when

competition is prevented from developing.

The costs of such regulation are usually measured by modeling or

comparing specific regulated sectors with less regulated sectors,

estimating the consumer and producer surplus losses that result from

higher prices and lack of service, and estimating the excess costs that

may result from the lack of competition. In contrast to social

regulatory cost estimates, these are estimates of mainly indirect

costs.

Economic regulation may produce social benefits when natural

monopolies are regulated to simulate competition. Although Hahn and

Hird (1991) argue that the dollar amount of such efficiency benefits

are small in a dynamic and technologically vibrant economy, their

judgment is an educated guess based on a reading of recent history,

rather than the result of an empirical study. It appears to be based

largely on the widely accepted view that the U.S. economy has become

more competitive over time, with fewer long-lasting natural monopolies,

and on the observation that much of the motivation for economic

regulation is to enhance one group at the expense of another. But even

though monopoly power may not be long lasting in a dynamic U.S.

economy, it does exist at a given point in time.6

---------------------------------------------------------------------------

\6\ We are not including antitrust activities such as preventing

the formation of monopolies through mergers or anticompetitive

behavior in our definition of economic regulation. Clearly this type

of Government policy creates important social benefits.

---------------------------------------------------------------------------

Moreover, while Hahn and Hird (1991) define economic regulation as

including only regulation of entry, output, and prices, in practice

they appear to lump all Federal regulation of banking and other

financial institutions, as well as consumer protection regulation

through mandated disclosure requirements, into the ``economic

regulation'' category of their cost estimates. In our view, chartering,

branching, interest rate, and activity regulation are the only major

categories of banking regulation that conform to the definition of

economic regulation used here. The other categories are ``safety-and-

soundness'' regulation and ``consumer information and protection''

regulation, both of which fit more logically into the ``other social

regulations'' category used in this study (White 1991, pp. 32-33).

Consideration of this definitional issue is important because the type

and magnitude of benefits associated with the different categories of

banking regulation differ greatly. In particular, while costs may

exceed benefits for some types of economic regulation (entry, output,

and prices), safety-and-soundness regulation is essential to a well

functioning financial system and thus fully justifies the cost (White

1991), and the consumer protection regulation applicable to banking is

similar to consumer protection information for other industries where

there is general agreement that the benefits exceed the costs.

Transfer. As discussed in OMB's Best Practices document, transfers

are payments from one group in society to another and, therefore, are

not real net costs to society as a whole. Nonetheless, the consequences

for individuals can be very significant. One person's loss is another

person's gain. Examples of transfers include payments to Social

[[Page 44039]]

Security recipients from taxpayers and the higher profits that farmers

receive as a result of the higher prices consumers must pay for farm

products limited by production quotas. Our guidance document states

that transfers should not be added to the cost and benefit totals

included in regulatory assessments but should be discussed and noted

for policy makers.

Process. Process costs are the administrative or paperwork costs of

filling out Government forms such as income tax, immigration, social

security, procurement, etc. The majority of process costs is due to

program administration, Government procurement, and tax compliance,

which do not fall into either the social or economic regulatory

categories. Some of these, such as procurement costs, are reflected in

the Federal budget as greater fiscal expenditures and care must be

taken not to count them twice. Process costs can be viewed as part of

the costs of providing Government services or collecting revenues that

should be minimized for a given level or quality of service or revenue.

We break these types of costs into further categories and discuss their

effects in more detail below.

B. New Estimate of the Costs and Benefits of Existing Social

Regulations

Several commentators on last year's report called for more detail

on the costs and benefits of regulatory programs. It is important to

note that, as was the case last year, this section includes only

estimates of costs and benefits that have been quantified and

monetized. As we discuss elsewhere in this report, the fact that an

effect has not been monetized or quantified does not necessarily mean

that it is small or unimportant.

Last year we broke out costs and benefits of existing social

regulations into two categories: environmental and other social (OMB

1997, table 1). This year we have been able to further subdivide other

social into three categories: labor, transportation, and other social

regulation, mainly regulations from HHS, DOE, and USDA. We were able to

do this by further utilization of the results of the 1991 article by

Hahn and Hird and the 1996 book by Hahn as well as the Cost of a Clean

Environment report (EPA 1990), and by making new estimates of the costs

and benefits of regulations issued over the last three years (April 1,

1995 to March 31, 1998), which we derive in chapter III using data from

the Regulatory Impact Analyses submitted by the agencies to OMB under

E.O. 12866. We have also incorporated EPA's recently published report,

The Benefits and Costs of the Clean Air Act, 1970-1990 (EPA 1997),

hereafter referred to as the ``Section 812 Retrospective.'' In

addition, we examined data submitted to GAO by the independent agencies

over the last two years under the Congressional Review Act for major

rules. In order to estimate aggregate regulatory costs and benefits, we

combine three data sources covering three time periods--pre-1988, 1988

to 1994, and 1995 to 1998.

Since Hahn and Hird provide cost and benefit estimates for more

than two categories of social regulations, we were able to expand our

estimate detail from two categories last year to four this year. We

were limited to four categories because the cost data we relied upon to

fill the gap between the 1988 Hahn and Hird data and our cost and

benefit estimates starting in 1995, (from the 1996 OMB report, More

Benefits, Fewer Burdens) contain only the four categories listed above.

We also use additional information on the distribution of benefits that

we did not use last year. Last year we used Hahn and Hird's conclusion

that ``the net benefits of social regulation are positive but small''

(p. 253) to estimate that the costs and benefits of both environmental

and other social regulations were approximately equal. They came to

this conclusion by taking the midpoint of their ranges for costs and

benefits. However, as we pointed out last year, there is much

uncertainty associated with these estimates. Moreover, we were

criticized for presenting point estimates when ranges would have been

more appropriate (Hahn 1998). This year we have elected to present

ranges both for the base case and later for our estimates of the costs

and benefits of the regulations that have been issued since the base

period. Table 1 shows these cost and benefit estimates derived from

Hahn and Hird for the four regulatory program areas as of

1988.7 Table 1 also includes new estimates from the Section

812 Retrospective.8

---------------------------------------------------------------------------

\7\ We do not repeat the discussion of the derivation and the

qualifications of these estimates that appeared in last year's

report. We refer the reader to that discussion (OMB 1997 pp. 27-33)

for this information. Suffice it to say here that we realize, as

several commenters have pointed out, that there are gaps and

weaknesses in underlying studies that Hahn and Hird rely on for

their estimates and that not all the costs and benefits of social

regulation are captured in these estimates. We hope in future years

to fill in the gaps and use more accurate, up-to-date studies for

our estimates when such studies become available.

\8\ Table 1 (and all succeeding tables mentioned in the text)

can be found in sequential order at the end of this report.

---------------------------------------------------------------------------

The addition of the Section 812 Retrospective significantly changes

the upper bound benefit estimate for environmental regulation, i.e.,

more than 15 times the upper bound of the Hahn and Hird study. As we

outlined at the beginning of this chapter, there are a number of

critical estimation problems that must be confronted in developing

benefit and cost estimates. The available studies, such as the Hahn and

Hird study and the Section 812 Retrospective, also have had to confront

these problems and each study has had to make difficult choices. As a

result, there are advantages and disadvantages that attend each of

these studies. The EPA estimates of $378 million to $3.2 trillion per

year are substantially larger than the estimates presented by Hahn and

Hird. The Hahn and Hird estimates were based on a 1982 study by Freeman

that provided a synthesis of the available benefits literature. These

estimates do not reflect the benefits associated with Clean Air Act

initiatives in the 1980s, e.g., EPA's lead phasedown program. They also

do not reflect the recent literature suggesting an association between

exposure to fine particulate matter and premature mortality. In

addition, the 1982 Freeman estimates were based on actual air quality

improvements over the 1970s, i.e., they did not attempt to account for

the benefits associated with preventing degradation in air quality.

The Section 812 Retrospective estimates were developed through an

EPA Science Advisory Board peer review process. It presents a more

comprehensive set of the benefits and costs under the Clean Air Act

over the period from 1970 to 1990; for example, it includes regulatory

actions taken during the 1980s. In addition, these estimates also

include the benefits and costs of preventing any deterioration in air

quality and reflect the benefits and costs of all air pollution control

efforts, not just the Federal Clean Air Act. Our detailed discussion in

section D below presents a more complete description of the Section 812

Retrospective and identifies some key uncertainties and assumptions

underlying the benefit estimates that may have an important effect on

the magnitude of these estimates.

To get the costs of existing regulations as of 1997, last year's

report added to the 1988 base the costs of the major regulations

reviewed by OMB between 1987 and 1996 as estimated from the RIAs

agencies provided OMB under Executive Order 12866 and its predecessor

Executive Order 12291 (OMB 1996). To estimate benefits, last year we

used benefit/cost ratios for environmental and other social regulation

calculated from Hahn (1996), who estimated benefits and costs of

[[Page 44040]]

agency rules from 1990 to mid-1995, for a subset of our rules, to

estimate benefits that correspond to our rules. We then added that

total to the benefit estimate as of 1988 from Hahn and Hird. This year

we improve on that exercise by using benefit/cost ratios from Hahn

(1996) for environmental, transportation, labor, and other social

regulation to estimate benefits for rules issued between 1987 and

1995.9 For the rules issued from 1995 through the first

quarter of 1998, we used information from agency-supplied RIAs modified

for consistency with Best Practices as appropriate and extended to

provide more monetized estimates of benefits and costs using consensus

value estimates used by the agencies or found in the literature. These

calculations are shown and explained in chapter III. Our latest

estimates are shown in table 2.

---------------------------------------------------------------------------

\9\ Admittedly this is a crude estimation procedure because

Hahn's inventory of rules begins in 1990 and ours extends back to

1987. Consequently, we are assuming that the relationship between

costs and benefits that Hahn found for the later period extends back

three years. Still, we know of no other approach to fill this gap in

the data until RIAs for these years are re-examined.

---------------------------------------------------------------------------

Table 3 combines the results from tables 1 and 2 to present our new

estimates for the existing costs of social regulation as of the first

quarter in 1998. It shows that health, safety and environmental

regulation produces between $34 and $3.38 trillion of net benefits per

year.

We must underline the uncertainty of these estimates. They are

useful primarily for drawing general conclusions about categories of

regulations that should be corroborated by additional data and

analysis. As specific values, however, they are fraught with

uncertainties. As discussed above, the baseline, apples and oranges,

and other methodological problems significantly reduce the likelihood

that these findings are robust. In addition to these problems, we are

also concerned that as the aggregate categories are divided into

smaller parts, the accuracy of the estimates may weaken because it is

less likely that randomly distributed errors in the data and analysis

even out. Furthermore, one must be doubly careful about drawing

conclusions from these results because these estimates are average

benefits and costs for aggregates of existing regulations, not the

incremental costs and benefits that are required to be able to make

reliable recommendations to improve specific regulatory programs or

regulations. Also note that these estimates are a combination of the

1988 baseline estimates, which are mostly from retrospective studies,

and the 1988 to 1998 estimates that are from the prospective studies

for individual rules. How well the cost and benefit estimates of

prospective studies predict actual costs and benefits is a question

that has not been answered. In section D of this chapter, we review the

evidence from several case studies that might shed light on this

question. Where we can make direct comparisons between prospective and

retrospective analyses, we find that both costs and benefits were

sometimes overestimated by prospective studies. In other instances,

costs were underestimated.

Finally regarding the utility of these estimates for making

recommendations for changes in regulatory programs, it bears repeating

that the actual costs and benefits of a regulation or regulatory

program are not the appropriate calculation. Rather, before a

recommendation is made to repeal or modify a regulation or regulatory

program, the necessary question is: ``What would be the incremental

costs and benefits of repealing the regulation or regulatory program.''

C. Other Regulatory Impacts

Despite the weaknesses in the estimates of the costs and benefits

of social regulation, the estimates of the costs and especially the

benefits of the other types of regulation are even more problematic. In

last year's report, we made the assumption that the costs and benefits

of fundamentally different types of regulations and government policies

could be aggregated and displayed in one table, with caveats. In doing

this, however, we were adding regulatory programs together that had

quantified costs and unquantified benefits with regulatory programs

that had quantified costs and quantified benefits. We also added

together the direct compliance costs of social regulation with the

indirect, mostly consumer surplus, losses of economic regulation.

However, direct compliance costs may have significantly different long

run effects than indirect consumer surplus losses. We have concluded

this year that such totals are more misleading than helpful, even with

extensive explanation of the absent benefit estimates and the apples

and oranges and other problems. To prevent confusion, this year we are

presenting the estimates separately in table 4.

Table 4 presents a list of the other types of regulation or

regulatory-like activities. In some cases we do not agree that these

activities are true regulations or should be considered in the same

category with what we have classified as social regulation. However,

this wide range of activities was noted by several commenters who urged

us to include them in this year's report. Table 4 also lists costs and

benefits, and is followed by a discussion of each.

1. Efficiency Losses From Economic Regulation

In last year's report, we presented an estimate that the efficiency

costs of economic, i.e., price and entry, regulation amounted to about

$71 billion. This is based on an estimate by Hopkins (1992) of $81

billion, which we adjusted downward by $10 billion to account for the

deregulation and increase in competition that has occurred in the

financial and telecommunications sectors since Hopkins' estimates were

made in 1992. Our estimate has recently been corroborated by analysis

in a recent, comprehensive two volume Organization for Economic

Cooperation and Development (OECD) report, OECD Report on Regulatory

Reform (OECD 1997), which attempts to estimate the benefits of further

economic deregulation of five sectors of the economy (electricity,

airlines, trucking, telecommunications, and retail and wholesale

distribution) for five countries (the U.S., Japan, Germany, France, and

the U.K.). Adding up any remaining benefits from deregulating these

sectors and using a macroeconomic model to simulate the economy-wide

effects on GDP, the OECD estimated that U.S. GDP would increase by 0.9

percent from these actions. This estimate implies that the current

costs of regulation in these sectors is $68 billion (0.9 percent of

1996's GDP of $7.6 trillion). Although the two estimates are not

strictly comparable, because our estimate of $71 billion includes

import restrictions and the OECD estimate does not and our estimate is

only for Federal regulation and the OECD estimate includes State and

local as well as National, the two estimates are close enough to be

mutually supportive.

There appear to be no reliable quantified estimates of the total

benefits of economic regulation. We pointed out last year that price

regulation of natural monopolies does have the potential to provide

consumer surplus benefits. However, most economists believe that few

natural monopolies, except perhaps in local distribution markets, have

long staying power because of the globalization of markets and rapidly

changing technology. Over time both the benefits and costs of

regulation (assuming regulation does not change) are eroded by changes

in technology and adaptive behavior, i.e., the rising baseline

phenomenon discussed above.

[[Page 44041]]

The static welfare benefits of economic regulation are not likely to be

long lasting in a dynamic world. The OECD report also implies that few

benefits are produced by sectoral entry restrictions. The report points

out that the loss of universal service may be a concern, but states

that methods besides regulation, e.g., targeted subsidies, can be

adopted to provide services to worthy entities less able to pay full

costs. In table 4 we enter under the benefits of economic regulation

the term ``expected to be small.''

Last year, we received comments from several independent economic

regulatory agencies suggesting that we had not emphasized the potential

benefits of economic regulation enough. The comments made good points.

Economic regulatory agencies are producing significant benefits.

However, these benefits do not flow from their imposing new

restrictions on entry. Rather, the benefits stem from their efforts to

open up markets and promote competition, which often means preempting

State competition or correcting past mistakes. In other words, some

agencies view the reduced costs created by deregulating as a benefit of

regulation. The correct view is determined by the baseline. Is the

baseline the existing patchwork of State and Federal regulation, which

has produced artificially constructed telecommunications and financial

services firms, or the more competitive environment that most likely

would have existed if we had not had these restrictions? There is no

inconsistency in saying that economic regulation has produced few

significant benefits, as Hahn and Hird (1992) state in summarizing the

consensus view of economists on this subject, and saying that economic

regulatory agencies are currently providing important benefits to

society by promoting competition.

The OECD study points out the important role that regulators have

in smoothing the transition toward a more competitive environment.

Regulators must carefully consider the issues of stranded capital

costs, unemployment, and universal service as competition is

introduced. However, the long run benefits of reform appear to have

been worth the transitional costs. The OECD study points out that the

US's regulatory reform efforts have already produced major benefits,

especially compared to the other major industrial countries. The study

estimates that the average GDP gain for the other seven countries from

deregulation of the five sectors would be 4.7 percent, ranging from 3.5

percent for the U.K. to 5.6 percent for Japan. The 4.7 percent of GDP

estimate would be equivalent to $360 billion if applied to U.S. GDP.

The study also points out that a significant portion of the 0.9 percent

remaining benefits for the U.S. is likely to be achieved by regulatory

reform efforts already underway because of the Telecommunications Act

of 1996 and the early State efforts at electricity restructuring.

Clearly economic deregulation does not imply that the economic

regulatory agencies' jobs are done.

2. Disclosure Regulation

A second type of regulation often mixed in with economic regulation

is information disclosure. There is a strong consensus among economists

that regulations requiring the disclosure of information about the

price and quality of products and services can produce significant

benefits for consumers and improve the functioning of markets when this

information would not otherwise be available. Our estimate, based on

burden-hour calculations for the independent regulatory agencies, e.g.,

SEC, FCC, FTC, reported in OMB's Information Collection Budget for FY

1998 (272 million hours) and Hopkins' opportunity costs of time

estimate ($26.50 per hour), is that disclosure costs are about $7

billion. Although benefits have not been quantified, we expect that

they are significantly greater than $7 billion.

3. Transfers From Economic Regulation

Economic regulation often produces income transfers from one group

to another. These transfers are not social costs or benefits; they

neither create new net benefits for society nor reduce society's scarce

resources. Consequently benefit-cost analysis is not appropriate or

meaningful for evaluating transfer programs. As the Best Practices

document makes clear, distributional analysis, which should be part of

the economic assessment, is the proper method of analyzing transfers.

Table 4 includes an estimate for transfers based on the Hopkins

approach that assumes that the transfers created by economic regulation

are about twice the economic efficiency loss. The estimate is $140

billion (two times $70 billion), which we enter in both the costs and

benefits columns.

Although as one commenter pointed out (Hopkins 1997), transfers may

be associated with real lobbying costs, this fact of life does not

justify equating transfer costs with social costs. Lobbying goes on for

all sorts of Government policies including expenditure, tax, and

regulatory policies whether they exist or not, which are impossible to

measure separately. For example, lobbying goes on in an attempt to

impose regulations that do not now exist and therefore have no

efficiency costs. In this case, the multiple of two times the

efficiency loss would estimate social costs of zero. The best approach

to including these types of costs is by directly estimating the costs

of lobbying rather than using a multiple of economic efficiency losses.

Once that is done it is not clear how to evaluate the social benefits

of lobbying, which clearly produces benefits because at least some

amount of lobbying, i.e., citizen participation, is a necessary part of

a democratic government.

4. Tax Compliance

Last year we stopped short of including tax compliance costs and

transfer costs in the totals. Although we were criticized for that

(Hopkins 1997 and Dudley and Antonelli 1997), other commenters (Hahn

1998) agreed with us that such data should be reported, but not

included in the totals. As we pointed out in last year's report, a

major reason for not including tax compliance costs in our totals,

despite their real nature and obvious concern to the public, is that it

would be misleading to add these types of costs to the totals without

accounting for the fact that taxes are necessary for the basic

functions of government. Cost-effectiveness analysis, not benefit-cost

analysis, is the appropriate way to evaluate the efficiency of tax

policy. In Table 4, we present an estimate of the paperwork costs of

the tax code by multiplying the number of hours of tax preparation time

required to file tax forms (5.3 billion in FY 1997) according to OMB's

Information Collection Budget (OMB 1998) by an estimate of the

opportunity costs of the average hour spent on the forms ($26.50) based

on Hopkins (1991). That cost estimate is $140 billion. While we do not

have quantitative estimates of the aggregate benefits of tax

compliance, they are undoubtedly very large. Tax compliance is

necessary for the whole range of services the government provides.

5. Federal Budgetary Expenditures

Several comments also suggested that we report the Federal

budgetary costs of regulation. These Federal expenditures include the

costs of developing and issuing regulations and enforcing them once

they are on the books. For many years, the Center for the Study of

American Business at Washington University has compiled Federal

Expenditures for the Regulatory Agencies of the U.S. Government.

Douglas, Orlando, and Warren (1997) have produced the latest estimates.

[[Page 44042]]

Table 4 presents these estimates for both social and economic

regulation.10 For benefits, we reproduce the quantified

estimate of the net benefits for social regulation as shown above in

table 3 and summarize the earlier discussion of qualitative benefits of

economic regulation.

---------------------------------------------------------------------------

\10\ Note that they do not consider the Internal Revenue Service

to be a regulatory agency and therefore do not include it in their

estimates. Their approach is consistent with ours and inconsistent

with Hopkins (1997).

---------------------------------------------------------------------------

6. Welfare Effects

A final category of regulatory effects, which several commenters

suggested we include in our estimates, is the indirect or full welfare

impacts of regulation. The estimates presented above for social

regulation are mostly estimates of direct compliance costs. However, as

our Best Practices document points out, the proper concept of the cost

of regulation is the best estimate of the value of the opportunity

foregone as a result of the imposition of the regulation. The

opportunity costs are likely to be greater than direct compliance

costs. In addition to the consumer surplus losses that result when

compliance costs drive up prices and reduce consumption of the goods

and services produced by the regulated entity, there may be secondary

effects on other markets, which reduce consumer welfare. The effects

result because regulation increases the overall costs of consumption

relative to output and reduces investment and productivity. These

effects can only be estimated with a computable general equilibrium

model that traces the myriad interrelationships that make up the modern

economy. Unfortunately the results of these models are highly dependent

on model specifications, which are not transparent to outside reviewers

making it difficult to determine the reasonableness of model

estimates.11

---------------------------------------------------------------------------

\11\ See Hahn and Hird, pp. 244-246, for a discussion of these

problems and several others.

---------------------------------------------------------------------------

The two most well known models that have been used to estimate the

general equilibrium effects apply to environmental regulation. These

models find that by 1990 the social welfare effects were about twice

the direct compliance cost effects (Hazilla and Kopp 1990 and Jorgenson

and Wilcoxen 1990). In table 4 we present this estimate for

environmental regulation but not for workplace and product regulation.

The reasons are that the estimates were made for environmental

regulation and there is no theoretical reason why the effect should be

the same for the two types of regulation. This is because the benefits

of environmental regulation generally flow to third parties not

involved in the production of the regulated product, while the benefits

of workplace health and safety regulation and product safety and

energy-efficiency regulations mostly flow to parties that are part of

the transaction (workers and consumers of the product). This factor

causes the costs to the regulated firms to be less than the direct

compliance costs because firms will likely eventually reap at least a

portion of the benefits of the regulation through lower employee costs

for workplace regulation and higher product quality for product safety

and energy-efficiency regulation. If the actual costs of compliance to

firms are less than the estimated direct compliance costs, the general

equilibrium effects will also likely be smaller.

The general equilibrium or secondary effects of the regulation on

the benefit side are less well understood than they are for the cost

side. But as discussed in last year's report, the health and safety

benefits of regulation, in particular, should result in indirect

welfare benefits for the economy. Because a healthier and longer-living

population is likely to have a longer time horizon and more optimistic

outlook, it is also likely to work more years more productively and

save and invest more. These effects could very well expand economic

activity and increase the standard of living significantly, especially

in the long run.

D. Lessons Learned from Studies of Federal Regulation

A review of several studies of the costs and benefits of regulation

offers insights into both the actual effects of regulations and into

the problems that attend any estimation of their benefits and costs.

Below we discuss the two key studies underlying our estimate of the

aggregate benefits and costs of environmental regulation and a new

study by Robert Hahn of 106 regulations using prospective estimates of

costs and benefits published by the agencies at the time the final

rules were issued (Hahn forthcoming). We also review two additional

retrospective studies that compare the actual and predicted costs and

benefits of regulation.

First, as noted earlier, EPA recently published its Section 812

Retrospective study of the costs and benefits of the Clean Air Act, as

required by section 812 of the Clean Air Act of 1990. It estimated that

the present value of benefits of the Clean Air Act regulations issued

between 1970 and 1990 is $22.2 trillion (central estimate, 1990$).

Publication of the Section 812 Retrospective provides an opportunity to

compare it with the Hahn and Hird study, which served as the basis for

our estimates in last year's report.

Hahn's study expands on his earlier one, which we used in section 2

in our aggregate estimate to cover the years 1987 to 1994 (Hahn 1996).

The 106 final regulations with both costs and benefits in the new study

were issued between 1982 and mid-1996 by EPA, OSHA, NHTSA, HHS, HUD,

and USDA. Hahn uses consensus estimates to value reduced units of

pollution and increased life-years to calculate benefits of health,

safety and environmental regulation. He takes as given the quantity

estimates of benefits and the monetized estimates of costs found in the

agency-produced regulatory impact analyses. He also converted to

constant 1995 dollars and used a 5 percent discount rate to put costs

and benefits in a consistent present value framework. Hahn estimated

that the net present value of benefits of the 106 regulations is about

$1.6 trillion. However, he also found that not all agency rules

provided net benefits. In fact, less than half of all final rules

provided benefits greater than costs. The main reason for his large

estimate of net benefits and relatively poor performance for many

individual regulations was that a few rules provided most of the net

benefits. NHTSA's automatic restraints in cars and EPA's lead phasedown

in gasoline provided just over 70 percent of total net benefits (Hahn

forthcoming, p. 15).

1. EPA's Retrospective Report to Congress on the Benefits and Costs of

the Clean Air Act

EPA's Section 812 Retrospective represents the culmination of a

six-year effort by EPA. The Section 812 Retrospective also reflects, as

required by section 812, peer review by an independent, external panel

of economists, health scientists, and environmental scientists known as

the Science Advisory Board Council on Clean Air Act Compliance Analysis

(Council). The Council provided detailed review and guidance throughout

each step of study design, implementation, and report drafting. The

quality and reliability of the Section 812 Retrospective was addressed

by the Council in its review closure letter by stating that the Council

``finds that the Retrospective Study Report to Congress by the Agency

is a serious, careful study and employs sound methods along with the

best data available.'' 12 The Council further concluded that

the Section 812

[[Page 44043]]

Retrospective's findings are ``consistent with the weight of available

evidence.'' 13

---------------------------------------------------------------------------

\12\ SAB Council, letter to EPA Administrator Browner, July 8,

1997, p. 1.

\13\ Ibid.

---------------------------------------------------------------------------

The Section 812 Retrospective presents estimates of monetized

benefits ranging from $6 to $50 trillion (present value in 1990$) over

the period from 1970 through 1990, with a central estimate of $22

trillion. Over this same period, the Section 812 Retrospective

estimated direct compliance expenditures of roughly $0.5 trillion. The

estimated net monetized benefits for the 1970 to 1990 period range from

$5.1 to $48.9 trillion dollars, with a central estimate of $21.7

trillion. The Section 812 Retrospective also notes that the monetized

benefits estimate may understate benefits because a number of benefit

categories were not quantified and/or monetized, e.g., air toxics

effects and ecosystem effects. Table 5 presents the non-monetized

benefits listed by the Section 812 Retrospective.

While the findings of the Section 812 Retrospective suggest that

the aggregate historical benefits of the clean air regulatory programs

substantially exceed the aggregate costs, the Section 812 Retrospective

itself provides the following cautionary note on page ES-10:

Finally, the results of the retrospective study provide useful

lessons with respect to the value and limitations of cost-benefit

analysis as a tool for evaluating environmental programs. Cost-

benefit analysis can provide a valuable framework for organizing and

evaluating information on the effects of environmental programs.

When used properly, cost-benefit analysis can help illuminate

important effects of changes in policy and can help set priorities

for closing information gaps and reducing uncertainty. Such proper

use, however, requires that sufficient levels of time and resources

be provided to permit careful, thorough, and technically and

scientifically sound data-gathering and analysis. When cost-benefit

analyses are presented without effective characterization of the

uncertainties associated with the results, cost-benefit studies can

be used in highly misleading and damaging ways. Given the

substantial uncertainties which permeate cost-benefit assessment of

environmental programs, as demonstrated by the broad range of

estimated benefits presented in this study, cost-benefit analysis is

best used to inform, but not dictate, decisions related to

environmental protection policies, programs, and research.

In terms of our charge under section 625(a), we must also consider

these new benefit and cost estimates in developing an overall estimate

of the benefits and costs of Federal regulation. The magnitude of EPA's

benefit estimate, $22 trillion over the 1970 to 1990 period, is very

large. The expected value of the estimated monetized benefit for 1990

is $1.25 trillion per year. This represents approximately 20 percent of

total 1990 Gross Domestic Product and is comparable in magnitude to

total 1990 U.S. expenditures on nondurable goods. There are several

important elements of the analysis in the Section 812 Retrospective

which deserve further discussion in order to understand the basis for

the benefit estimates over the 1970 to 1990 period.14

---------------------------------------------------------------------------

\14\ ``A final, brief interagency review, pursuant to Circular

A-19, was organized in August 1997 by the Office of Management and

Budget and conducted following the completion of the extensive

expert panel peer review by the SAB Council. During the course of

the final interagency discussions, it became clear that several

agencies held different views pertaining to several key assumptions

in this study as well as to the best techniques to apply in the

context of environmental program benefit-cost analyses, including

the present study. These concerns include: (1) The extent to which

air quality would have deteriorated from 1970 to 1990 in the absence

of the Clean Air Act, (2) the methods used to estimate the number of

premature deaths and illnesses avoided due to the CAA, (3) the

methods used to estimate the value individuals place on avoiding

those risks, and (4) the methods used to value non-health related

benefits. However, due to the court deadline the resulting concerns

were not resolved during this final, brief interagency review.

Therefore, this report reflects the findings of EPA and not

necessarily other agencies in the Administration.'' See Section 812

Retrospective, p. ES-2.

---------------------------------------------------------------------------

(a) Establishing a baseline. The Section 812 Retrospective uses as

a counter-factual ``baseline'' the modeled air quality in the United

States over the 1970 to 1990 period for a scenario in which control

technology and requirements are frozen at the levels mandated in 1970.

It assumed that no additional air pollution controls would have been

imposed by any other level of government or voluntarily initiated by

private entities after 1970. The Section 812 Retrospective acknowledges

that this is an obvious oversimplification and that, in fact, State and

local governments as well as private initiatives were responsible for

an important fraction of the estimated benefits and costs over the

period from 1970 to 1990.15 At the same time, it notes that

the Federal CAA played an essential role in achieving these results and

leaves to others the question of parsing out the precise fraction of

costs and benefits attributable to the Federal CAA.16

---------------------------------------------------------------------------

\15\ Section 812 Retrospective, pp. 2-3.

\16\ Ibid, p. 3.

---------------------------------------------------------------------------

Because the modeled baseline includes significant growth in

population, car and truck travel, and economic activity, there is a

marked deterioration in baseline air quality over the period from 1970

to 1990. While there is no direct sensitivity analysis of alternative

baselines, the available documentation for the ``no control'' scenario

suggests that a substantial fraction of the estimated benefits are

attributable to the degradation in modeled air quality from 1970

levels, rather than the result of an improvement in air quality from

the levels that existed in the United States in 1970.17

---------------------------------------------------------------------------

\17\ Of course, any change in the baseline scenario would also

require revision of the cost estimates. The Section 812

Retrospective specifically notes that the ``no control'' scenario

avoids the difficulties of sorting out the fraction of costs

required to maintain an alternative baseline, such as maintaining

air quality at 1970 levels. See Section 812 Retrospective, pp. 2-3.

---------------------------------------------------------------------------

In any event, considerable uncertainty necessarily surrounds ``what

would have happened'' over this 20-year period, rendering all attempts

to construct aggregate benefit and cost estimates somewhat speculative.

(b) Key benefit categories. The Section 812 Retrospective developed

monetized benefit estimates for ten benefit categories, including

mortality, hospital admissions, chronic bronchitis, soiling damage, and

visibility. (See table 6.) As indicated by table 6, the monetized

benefit estimates associated with reducing exposure to fine particulate

matter (PM) account for 90 percent of the total estimated benefits. The

discussion below discusses three key elements in developing benefit

estimates associated with reductions in PM levels.

(i) Uncertainties in magnitude and causation. The Section 812

Retrospective describes some elements of the uncertainty in the

estimates of health risks, focusing on those elements of uncertainty

that are most readily quantifiable. For example, it addresses specific,

quantifiable elements of the uncertainty in the benefits estimates

through the use of a ``Monte Carlo'' analysis. It also presents a

thoughtful, qualitative discussion of some of the uncertainties

associated with the estimated mortality risk--for example, the effect

of an historical trend in particulate matter levels and the effect of

intercity movement of population on the concentration-response

relationship.

The Section 812 Retrospective offers little discussion, however, of

the uncertainty associated with the critical question of the causal

relationship between fine particulate matter levels and mortality. It

observes that the Clean Air Scientific Advisory Committee has pointed

out that a causal mechanism has not been clearly established. It

concludes that ``the well-established correlation between exposure to

elevated PM and premature mortality is sufficiently compelling to

warrant an

[[Page 44044]]

assumption of a causal relationship and derivation of quantitative

estimates of a PM-related mortality.'' 18

---------------------------------------------------------------------------

\18\ Ibid., p. 34.

---------------------------------------------------------------------------

The preamble to EPA's 1996 proposal to revise the National Ambient

Air Quality Standard for Particulate Matter (PM NAAQS) discusses at

greater length the difficulties associated with the interpretation of

specific concentration-response relationships, pointing out that it is

the most problematic issue in conducting risk assessments for PM-

associated health effects. These include: 19

---------------------------------------------------------------------------

\19\ 61 FR 65650. The preamble to the final rule reaffirms these

concerns by citing the proposal and a more complete discussion in

the criteria document (chapters 10-13) and the staff paper (chapter

IV). See 62 FR 38655 and 38656.

---------------------------------------------------------------------------

(1) The absence of clear evidence regarding mechanisms of action

for the various health effects of interest;

(2) Uncertainties about the shape of the concentration-response

relationships; and

(3) Concern about whether the use of ambient PM2.5 and

ambient PM10 fixed-site monitoring data adequately reflects

the relevant population exposures to PM that are responsible for the

reported health effects.

(ii) Timing of effects. The Section 812 Retrospective assumed that

reductions in ambient PM concentrations yield contemporaneous

reductions in the mortality and chronic health risks associated with

long-term exposure. Given that the concentration-response relationships

in the underlying study are presumptively thought to be the result of

long-term exposure, the assumption of a contemporaneous response--that

is, a zero lag in the response--represents only one end in a range of

possibilities. It is quite possible, however, that there is a lag in

the changes in the risk of chronic health effects and mortality with

changes in exposure to particulate matter. Other researchers (World

Health Organization, 1996) have assumed the effect of particulate

matter exposure does not begin until 15 years of exposure.20

The incorporation of a latency period can have an important effect on

the benefits estimate. The adoption of an alternative latency

assumption of 15 years, for example, would reduce the estimated present

value of the mortality benefits by a factor of two, given the discount

rate of five percent used in the Section 812 Retrospective.

---------------------------------------------------------------------------

\20\ Section 812 retrospective, p. D-17.

---------------------------------------------------------------------------

(iii) Valuation of changes in health risk (``benefits transfer'').

The Section 812 Retrospective also highlights the difficulties of

transferring estimates from other settings to value the projected

benefits of a regulatory initiative, e.g., changes in mortality risk.

In valuing changes in mortality risk, EPA reviewed 26 studies to

develop an estimate of the ``value of a statistical life'' based on the

willingness-to-pay (WTP) of individuals to avoid small increases in

mortality risk. Using a Weibull distribution to fit the estimates from

these 26 studies, the Section 812 Retrospective estimated a mean value

of $4.8 million per statistical life (with a standard deviation of $3.2

million in 1990).21 This estimate reflects a WTP of $5 for a

reduction in mortality risk of one in a million.

---------------------------------------------------------------------------

\21\ Section 812 Retrospective, p.44.

---------------------------------------------------------------------------

This estimate is derived from studies involving very small changes

in mortality risk. However, the changes in mortality risk associated

with changes in particulate matter exposure estimated in the Section

812 Retrospective are roughly 10 to 100 times greater than the changes

associated with these valuation studies. When the marginal valuation of

$5 for a one in a million change in mortality risk is applied to the

``no control'' scenario where modeled baseline mortality risk is on the

order of 1 in a 1000, the resulting WTP estimates for changes in

mortality risk represent a large share of each household's annual

budget, i.e., household ability to pay. Since the total outlay for risk

reduction represents a large share of the household budget, this

situation is very different from that examined by the 26 valuation

studies where the WTP estimates were a small fraction of household

budgets.

(c) Hahn and Hird's estimate for environmental benefits. For its

environmental benefit estimate, the Hahn and Hird assessment relied on

an analysis by Freeman conducted in the late 1970s (Freeman,

1982).22 The Freeman analysis largely represented a

synthesis of the best existing work of the 1970s. The analysis

estimates air pollution control benefits for the year 1978, and water

pollution control benefits for the year 1985. Hahn and Hird adjust the

Freeman estimates to account for inflation; but these adjustments do

not reflect other changes--for example, additional regulations--in the

air pollution control program between 1978 and 1988 and in the water

pollution program control between 1985 and 1988. For water pollution

control benefits, the Freeman analysis may still represent the most

comprehensive estimate available. There are, however, several elements

of the Freeman analysis that deserve further discussion in order to

understand the strengths and limitations of the benefit estimates used

by Hahn and Hird.

---------------------------------------------------------------------------

\22\ See Hahn and Hird (1991 pages 253, 273; Portney (1990)

pages 54-60; Freeman (1990 in Portney (1990) page 123.

---------------------------------------------------------------------------

(i) Establishing a baseline. As noted elsewhere in this report,

choice of an analytic baseline can be difficult, since many options are

available, and the preferred baseline may be unworkable due to the

inadequacy of available data. In the Freeman analysis, different

baselines were chosen for the air and water benefits analyses.

The Freeman analysis evaluated the improvement in ambient air

quality between 1972 and 1978, and did not consider the deterioration

in air quality that might have occurred in the absence of air pollution

regulations.23 In effect, the counterfactual baseline was

assumed to be the level of air quality in 1972. As a result, the air

quality improvements that were analyzed were much smaller than those

incorporated in the CAA Section 812 Retrospective (EPA, 1997).

Furthermore, the baseline used for the air benefits analysis was not

consistent with that used for Freeman's cost analysis, which estimated

all air pollution control costs.

---------------------------------------------------------------------------

\23\ Implicitly, the Analysis assumed increased state, local,

and private initiatives great enough to offset air quality

deterioration due to increased economic activity, population growth,

and vehicle-miles-traveled (VMT) by automobiles and trucks during

the 1972 to 1978 period.

---------------------------------------------------------------------------

The baseline used for the water analysis, on the other hand,

assumed changing population and recreational participation rates

between 1972 and 1985. The baseline used for the water benefits

analysis was consistent with that used for Freeman's water pollution

control cost analysis.

(ii) Key benefit categories. Freeman's air pollution benefits

analysis developed monetized benefit estimates for six categories:

human health (mortality), human health (morbidity), soiling and

cleaning, vegetation, materials, and property values. Approximately two

thirds of the monetized benefits were for human health improvements,

primarily reduced mortality incidence, due to reductions in ambient air

concentrations of particulate matter and sulfur oxides. His analysis

does not include any estimate of the benefits arising from reductions

in airborne lead (Pb) concentrations, which were a significant source

of air pollution control benefits found by later studies. The

discussion below addresses 3 key factors to bear in mind when

interpreting the primary benefit category, i.e., reduced mortality,

found in the air benefits estimates of his analysis.

[[Page 44045]]

1. Uncertainties in Magnitudes of Physical Effects: The Freeman

analysis surveys seven studies from the 1970s which developed a dose-

response relationship between particulate matter and human

mortality.24 Based on these studies, Freeman provides a

range of possible results, with a ``best-guess'' estimate assumed to be

at the midpoint of the range. Since 1978, a number of additional

epidemiological studies have been completed on the relationship between

particulate matter and human mortality rates. It does not reflect the

advances in knowledge achieved in the 1980s and 1990s.

---------------------------------------------------------------------------

\24\ Freeman (1982), pages 63-66. Five of the seven studies

relied on the statistical work by Lave and Seskin from 1970, 1973,

and 1977.

---------------------------------------------------------------------------

2. Timing of Effects: The Freeman analysis assumed that reductions

in ambient PM concentrations yield contemporaneous reductions in the

mortality risks associated with exposure to PM. If one were to assume,

for example, a significant lag, e.g., many years, between changes in

exposure and changes in risk, then the mortality benefit estimates

would be reduced.

3. Valuation of Changes in Health Risk: The Freeman analysis

assumed a value per statistical life (VSL) of $2.4

million.25 Since 1978, there have been significant

additional contributions to the economic literature on the value of

mortality risk. After considering these more recent studies, the

Section 812 Retrospective adopted a midpoint of $4.8 million ($1990) as

a better estimate on the population's willingness-to-pay for reductions

in mortality risk. Use of an alternative valuation for mortality risk

would have a significant effect on the aggregate benefit estimate in

the Freeman analysis.

---------------------------------------------------------------------------

\25\ Freeman (1982), page 68. The estimate of $1 million in 1978

is converted to 1996 using the CPI.

---------------------------------------------------------------------------

Freeman's water pollution benefits analysis developed monetized

benefits estimates for four categories: recreation, nonuse, commercial

fisheries, and diversionary uses. Approximately half of the monetized

benefits are attributable to recreation. This analysis is based on a

number of studies carried out in the 1960s and 1970s, with benefits

projected forward to reflect projected population and recreational

participation rates in 1985. However, these estimates do not include

benefits associated with the reduction in toxic loadings in waste water

discharges, even though Freeman's cost estimates include ``substantial

costs for the control of discharges of these substances'' (Freeman,

1982). Benefits of non-point source pollution control also were not

included. Benefits to new and existing recreational users for hiking,

picnicking and nature observation that might result from improvements

in water quality were also omitted because of the absence of data for

these activities.

(d) Summary assessment of Section 812 Retrospective. The discussion

above illustrates the difficulty, which we emphasize throughout this

report, of developing aggregate estimates of the benefits and costs of

major Federal regulatory programs. The results obtained in both the

Section 812 Retrospective and the Freeman analysis used by Hahn and

Hird appear to be sensitive to choices made concerning the baseline for

the analysis and the translation of the reduction of air pollution into

human health benefits.

2. Two Other Retrospective Studies

In general, retrospective studies are likely to provide more

accurate results than prospective studies because there are fewer

unknowns to deal with. Prospective studies must estimate what will

happen as a result of a proposed regulation and compare it with what

would happen without the regulation (the counterfactual). Retrospective

studies only need to measure the actual and estimate the

counterfactual. Below we discuss several case studies from the

literature that compare retrospective studies with their respective

prospective studies. NHTSA recently completed the third in a series of

studies of its 1983 center high-mounted stop lamp regulation. In brief

the studies found that although benefits exceeded costs, costs had been

underestimated by a factor of two and that the effectiveness of the

rule had been over estimated by a factor of seven in the prospective

study. The second case study examines eight regulations issued by OSHA

between 1974 and 1989 by drawing on an Office of Technology Assessment

(1995) report and a book by Viscusi (1992) that examined the cost

estimates and actual impacts of various OSHA regulations. The case

studies reveal that in some cases the agency overestimated expected

costs compared to the actual and in other cases it underestimated them.

The OTA study itself concluded that the agency had a tendency to

overestimate costs because of unanticipated improvements in compliance

technology after the regulations were issued. However, as in the NHTSA

example, the agency also appears to have overestimated the

effectiveness of its rule, if not the benefits.

(a) The Center High-Mounted Stop Lamp Case. A comparison of NHTSA's

prospective with its retrospective analyses of its Center High-Mounted

Stop Lamp (CHMSL) 26 regulation illustrates how the benefits

and costs of a rule can be substantially different in practice than

what one would have expected based solely on the prospective

work.27 It further illustrates that early post-rule

estimates may differ substantially from long-term estimates. In the

case of the CHMSL rule, the Final Regulatory Impact Analysis (FRIA) in

support of the rule made what appeared to be an overwhelming case that

the rule would generate very large net benefits. The FRIA was based on

substantial amounts of experimental data and for many years served as a

model of an RIA that consistently employed sound benefit-cost analysis

principles. Nevertheless, when compared with NHTSA's long-term

evaluation, the FRIA overestimated the actual effectiveness (though not

the consequent benefits) of CHMSLs by a factor of more than seven and

underestimated the cost by a factor of more than two. Despite these

revelations, however, the analyses continue to confirm that the rule

generates positive net benefits, though not nearly as large as what one

might have expected at the time the rule was proposed or even based on

the early post-rule analyses.

---------------------------------------------------------------------------

\26\ CHMSLs are the ``third tail light'' found on all new cars

beginning with the 1986 model year. The purpose of CHMSLs is to

reduce the time it takes for following drivers to react when drivers

in front of them put on their brakes, allowing them to stop sooner

and thereby avoid crashes (or reduce the speed at which impact

occurs).

\27\ Over the years, NHTSA has conducted a total of five

distinct analyses of its rule. These include two prospective

analyses (preliminary and final regulatory impact analyses) and

three retrospective analyses.

---------------------------------------------------------------------------

(i) 1980 and 1983 Regulatory Impact Analyses. In early 1981 NHTSA

proposed to require CHMSLs. At that time the agency estimated in its

Preliminary Regulatory Impact Analysis (PRIA) that the rule would

reduce rear-end collisions by 35 percent (see table 7). NHTSA estimated

this would lead to 1,511,000 fewer crashes per year once the entire

passenger-car fleet was so equipped. NHTSA also estimated that an

additional 1,339,000 crashes per year would be less severe than they

otherwise would have been. The combined value of the savings in

property damage would range from $1.3 to $2.3 billion per year. In

addition, the PRIA estimated the rule would prevent 66,000 injuries and

533 fatalities per year. NHTSA estimated the cost of the proposal at

$49 million per year. Thus the analysis of the proposal held out the

[[Page 44046]]

promise of very large net benefits in property damage reductions

alone.28

---------------------------------------------------------------------------

\28\ Since the costs occur when the vehicles are manufactured

and the benefits occur over the lifetime of the vehicle, it is

inappropriate simply to subtract annual costs from benefits. Even

after discounting, however, the PRIA estimates would yield net

benefits of between $600 million and $1.3 billion annually in

property damage alone.

---------------------------------------------------------------------------

NHTSA completed its FRIA and published the final rule in 1983. In

response to comments it received on the proposal and in light of some

new evidence of the effectiveness of CHMSLs, NHTSA revised several

components of its benefit estimates downward. The FRIA also included a

somewhat refined cost estimate. The FRIA estimated the effectiveness of

CHMSL at 33 percent. In order to provide a more ``conservative''

estimate of the benefits, NHTSA applied this effectiveness rate to a

smaller proportion of rear-end crashes than in the PRIA.29

In the FRIA, NHTSA also assumed a lower value of damage per crash

avoided ($510 vs. $1,116 in the PRIA). The result of these and other

related adjustments was estimates of 902,500 fewer crashes, $434

million in reduced property damage, 40,000 fewer injuries and no

estimate of reduced fatalities.

---------------------------------------------------------------------------

\29\ For example, the estimate excluded rural accidents, which

account for nearly one quarter of all accidents, because the test

fleets were driven in urban areas only thus leaving NHTSA with no

evidence that CHMSLs would be effective in rural settings. As NHTSA

later discovered, the actual effectiveness was about the same

between urban and rural settings.

---------------------------------------------------------------------------

The effectiveness estimates were based on three separate

experimental studies for which CHMSLs had been installed on fleets of

taxis and telephone company passenger cars. The three studies covered

over 3,000 vehicles and over 150 million vehicle miles. Nevertheless,

as early as 1980, NHTSA recognized the possibility that the

effectiveness estimate based on experimental studies may overstate the

true effectiveness of CHMSLs if there is a ``novelty'' effect which

caused following drivers to react more quickly than they would once

CHMSLs became commonplace. The effectiveness estimate was critical to

the decision to go forward with the rule because it underlies all

components of the benefit estimates. To its credit, NHTSA committed at

the time it proposed the rule to reassess the effectiveness after the

fact, if NHTSA adopted a CHMSL requirement in a final rule.

(ii) 1987, 1989, and 1998 retrospective studies. Since the rule

became effective with the 1986 model year, NHTSA has conducted three

analyses with the benefit of hindsight. The most important results of

these studies are that: (1) The effectiveness of CHMSLs is considerably

lower than NHTSA estimated in the PRIA and FRIA; (2) the effectiveness

has fallen over time, though it now appears to have stabilized; (3)

actual costs are about double those estimated in the RIAs; and, most

importantly, (4) despite these findings, the rule still generates net

benefits.

In 1987, NHTSA conducted a preliminary evaluation of the

effectiveness of production CHMSLs.30 It found an

effectiveness of about 15 percent. Thus, even though the CHMSLs were

installed in a small percentage of cars nationwide, i.e., when any

``novelty effect'' would most likely occur, effectiveness was less than

half of the estimates in the RIAs.

---------------------------------------------------------------------------

\30\ This study did not attempt to evaluate the benefits in a

broader sense or the costs.

---------------------------------------------------------------------------

In 1989, NHTSA conducted the second of its retrospective studies.

This study was based on 1987 data, by which time about one-fourth of

the passenger car fleet was equipped with CHMSLs. By this time, the

estimate of effectiveness had fallen again, to about 11 percent.

Despite the drop in estimated effectiveness and a corresponding

reduction in the number of accidents prevented compared with the FRIA,

the estimated benefits of CHMSLs increased. The number of injuries

prevented rose to between 79,000 and 101,000 and the estimate of

property damage prevented increased to $774 million per year. At that

time, NHTSA also concluded that CHMSLs were unlikely to prevent any

fatalities. The reasons for the increase in the benefits estimate

despite the reduction in effectiveness is due to three factors: (1) The

retrospective estimate includes all accidents (not just urban ones);

(2) the injury reduction estimate was based on actual crashes whereas

the estimates in the RIAs were modeled based on estimates of the

reduced speeds at which crashes that weren't avoided would occur; and

(3) the actual value of property damage given an accident was much

higher than NHTSA assumed in the FRIA. In other words, had NHTSA used

the same methodology and data for the FRIA and the retrospective, each

of the benefit categories would contain a value of about one-third of

what the FRIA reported, as the difference in effectiveness rates would

suggest.

Earlier this year, NHTSA completed its long-term study of the

benefits and costs of CHMSLs.31 This most recent estimate of

the effectiveness of CHMSLs is 4.3 percent. NHTSA does not expect it to

fall further since it has remained steady throughout the last seven

years of data NHTSA has analyzed (1989 to 1995). Part of the decline in

effectiveness between the 1989 study and this one is attributable to a

further refinement in NHTSA's methodology which more accurately

controls for vehicle age, which is a factor in rear-end crashes. (Had

NHTSA used the same methodology in the 1989 study, the effectiveness

would have been about 8.5 percent, rather than 11.3 percent, and the

corresponding benefits would have been proportionately lower.) Thus,

the long-term effectiveness of CHMSLs is about one-eighth of NHTSA's

original estimate, while the costs are more than double. Even so, these

estimates imply that the rule continues to produce net benefits, though

not nearly as large as what NHTSA estimated prospectively.

---------------------------------------------------------------------------

\31\ In the early 1990s, NHTSA extended the CHMSL requirement to

include ``light trucks,'' i.e., minivans, sport-utility vehicles,

and pickup trucks, which comprise about 40 percent of the fleet. The

estimates in the long-term study include the effects on these

vehicles as well. However, in order to facilitate comparisons with

NHTSA's previous estimates which pertained to cars only, all

aggregate estimates in this study have been reduced by 40 percent to

reflect the effects on cars only.

---------------------------------------------------------------------------

The FRIA included an aggregate cost estimate of $70 million ($7 per

vehicle) in each of the first two years and $40 million ($4 per

vehicle) each year thereafter. The retrospective analyses estimated the

cost at $89 million (about $9 per vehicle) per year, or more than twice

the long-term cost estimate in the FRIA.

(iii) Lessons learned from CHMSLs. These analyses confirm what many

believe: that benefits and costs are difficult to estimate

prospectively. In this instance, the RIAs overstated the effectiveness

of CHMSLs despite the advantage of substantial data from field

experiments. The estimates of benefits in the FRIA were not nearly as

large as those estimates presented in the PRIA. Nevertheless, the FRIA

estimates overstated the effectiveness of the rule by a factor of more

than seven. The changes in effectiveness estimates over time suggest

that it is important to re-evaluate the effects of regulations,

particularly where behavioral responses to the regulation may evolve

over time.

With respect to cost, even though the only cost component was a

fairly simple piece of hardware, the FRIA estimate was less than half

the actual cost. It is interesting that, in their comments on the

proposed rule, the three domestic manufacturers estimated costs in the

$8 to $15 range. The low end of this range was lower than NHTSA's

actual (long-term retrospective) estimate and the high end was only

slightly further from actual costs than the FRIA estimate.

[[Page 44047]]

(b) Eight OSHA cases. The Office of Technology Assessment was asked

by Congress in 1992 to examine how well OSHA had estimated the impacts

of the regulations it had issued. OTA attempted to answer this question

by comparing OSHA's prospective analysis of impacts with actual

outcomes for a selective set of regulations. Although OTA did not

directly attempt to estimate actual benefits, in some cases they can be

inferred from the discussion and in other cases other information

sources, e.g., Viscusi 1992, can be used. Because of funding

constraints, three of the eight cases--vinyl chloride, cotton dust, and

ethylene oxide--were chosen because existing studies had already been

done. For the other five, new retrospective studies were commissioned.

The eight cases examined exhibited a variety of outcomes. Table 8,

based on our analysis of the report's findings as well as other

information, shows that costs and benefits were both over-and

underestimated and that benefits were sometimes overestimated by OSHA

in its prospective analyses of the impacts of the rules. The 1974

regulation of vinyl chloride is often cited as an example of an agency

overestimating costs, although to be fair to OSHA the cost estimate was

supplied by industry and OSHA at that time did not conduct its own

economic analyses of prospective regulations. When cotton dust was

issued four years later, the agency was conducting economic analyses

for major rules. Cotton dust is also often cited as an example of the

agency overestimating compliance costs. OSHA, itself, contracted for a

retrospective study of the regulation five years after the rule was

issued but before the final controls took effect. The study found that

OSHA had earlier overestimated actual capital costs by a factor of five

(Viscusi 1992). The later study also found that benefits had also been

overestimated by at least two fold because of mistakes in methodology

and overcounting of the number of exposed individuals.

In the secondary lead smelters case, also issued in 1978, OSHA

underestimated costs and overestimated benefits. The OTA report (p. 62)

points out that as of 1995 secondary lead smelters were not able to

comply with the engineering controls requirement to reduce air-lead

levels to the permissible exposure limit because compliance was

economically infeasible, i.e., costs had been underestimated. However,

smelters had found less expensive and more direct ways than engineering

controls to reduce blood-lead levels, the key health indicator and

performance goal. In other words, reducing air-lead levels through

engineering controls was not needed to attain the sought-after health

benefits. The benefits of engineering controls had been overestimated.

In the 1984 ethylene oxide regulation of hospitals, OTA found that

OSHA had underestimated the costs of ventilation equipment but that

hospitals had little trouble complying with the standard by other

means. OTA found that overall hospitals spent more than expected, but

that was because they brought exposure levels down significantly below

the regulated level. On average, the agency had estimated costs about

right.

The agency appears to have overestimated costs by about a factor of

two for metal foundries in its 1987 regulation of formaldehyde because

firms used low-formaldehyde resins rather than the predicted

ventilation controls to attain compliance.

The next three case studies were for safety standards and the

findings are difficult to summarize. The OTA study did not directly

estimate costs or benefits for grain handling but found that the

standard was economically feasible. The PSDI power presses and powered

platforms rules were actually attempts at deregulation. In both cases

the cost savings that were predicted failed to materialize because

firms did not take advantage of the newly offered flexibility,

presumably because the agency had underestimated the costs and/or

overestimated the benefits of the flexibility. (See OTA 1995 p. 62.)

Looking at this evidence, OTA concluded that OSHA tended to

overestimate costs because new technology was often developed between

the time the analysis was done, which in several cases was several

years before the final rule was issued, and the compliance date. The

report recommended that the agency consider the dynamic nature of

technology including the possibility of ``regulation-induced

innovation'' in order to set lower compliance levels (p. 11). However,

there is an opportunity cost to forcing innovation that is being

neglected. The resources that are directed at reducing compliance costs

by developing new technologies have to be pulled from other projects,

which presumably the company thought had a larger potential for payoff.

Since adding another constraint to the economic system is not likely to

increase the overall rate of technological progress for the economy,

``regulation-induced innovation'' is not likely to be the ``win-win''

situation that the report suggests (p. 53).

Taken as a whole, these retrospective studies show that OSHA has

both underestimated and overestimated costs, sometimes by large

amounts. At the same time, in instances where there are clear data,

OSHA appears generally to have overestimated benefits. Although there

are important cases of overestimating costs because technological

progress and learning-by-doing over time reduced expected costs, it is

not clear that agencies should compensate for this tendency by reducing

costs estimates. These same factors may also lead to a tendency to

overestimate benefits.

Chapter II: Estimates of Benefits and Costs of This Year's

``Economically Significant'' Rules

A. Scope

In this chapter, we examine the benefits and costs of ``each rule

that is likely to have a gross annual effect on the economy of

$100,000,000 or more in increased costs,'' as required by section

645(a)(2). We have included in our review those final regulations on

which OIRA concluded review during the 12-month period April 1, 1997,

through March 31, 1998. This ``regulatory year'' is the same time

period we chose for last year's report. We chose this time period to

ensure that we covered a full year's regulatory actions as close as

practicable to the date our report is due, given the need to compile

and analyze data and publish the report for public comment. In

addition, we thought it would be useful to adopt a time period close to

that used for the annual OMB report required by the Unfunded Mandates

Reform Act of 1995.

The statutory language categorizing the rules we are to consider

for this report is somewhat different from the definition of

``economically significant'' in Executive Order 12866 (section

3(f)(1)). It also differs from similar statutory definitions in the

Unfunded Mandates Reform Act and subtitle E of the Small Business

Regulatory Enforcement Fairness Act of 1996--Congressional Review of

Agency Rulemaking. Given these varying definitions, we interpreted

section 645(a)(2) broadly to include all final rules promulgated by an

Executive branch agency that meet any one of the following three

measures:

Rules designated as ``economically significant'' under

section 3(f)(1) of Executive Order 12866

Rules designated as ``major'' under 5 U.S.C. 804(2)

(Congressional Review Act)

Rules designated as meeting the threshold under title II

of the Unfunded Mandates Reform Act (2 U.S.C. 1531-1538)

[[Page 44048]]

This year we also include a discussion of major rules issued by

independent regulatory agencies, although we do not review these rules

under Executive Order 12866. This discussion is based on data provided

by these agencies to the General Accounting Office (GAO) under the

Congressional Review Act.

During the regulatory year selected, OIRA reviewed 33 final rules

that met the criteria noted above. Of these final rules HHS submitted

10; EPA nine; USDA five; DOI and DOE two each; DOL, DOT, DOJ, and VA

one each. In addition three agencies, DOL, HHS, and Treasury, worked

together to issue one common rule. These 33 rules represent about 14

percent of the 230 final rules reviewed by OIRA between April 1, 1997,

and March 31, 1998, and less than one percent of the 4,720 final rule

documents published in the Federal Register during this period.

Nevertheless, because of their greater scale and scope, we believe that

they represent the vast majority of the costs and benefits of new

Federal regulations during this period.

1. Overview

As noted in chapter I of last year's report, Executive Order 12866

``reaffirms the primacy of Federal agencies in the regulatory decision-

making process'' because agencies are given the legal authority and

responsibility for rulemaking under both their organic statutes and

certain process-oriented statutes, such as the Administrative Procedure

Act, the Unfunded Mandates Reform Act, and the Small Business

Regulatory Enforcement Fairness Act. The Executive order also reaffirms

the legitimacy of centralized review generally and in particular review

of the agencies' benefit-cost analyses that are to accompany their

proposals. The Executive order recognizes that in some instances the

consideration of benefits or costs is precluded by law. For example,

the primary National Ambient Air Quality Standards under the Clean Air

Act are to be health-based standards set by EPA solely on the basis of

the scientific evidence. A variation is the Occupational Safety and

Health Act, where health standards must be based on reducing

significant risks to the extent doing so is economically and

technologically feasible. However, the Executive order requires

agencies to prepare and submit benefit-cost analyses even if those

considerations are not a factor in the decision-making process. Again,

it is the agencies that have the responsibility to prepare these

analyses, and it is expected that OIRA will review (but not redo) this

work. The costs and benefits identified may be attributable solely to

the regulation in question, where the agency has substantial

discretion, or they may in fact be attributable just as much to the act

of Congress that they are implementing.

Reviewing for this report the benefit-cost analyses accompanying

the 33 final rules listed in table 9, we found, as we did last year, a

wide variety in the type, form, and format of the data generated and

used by the agencies. For example, agencies developed estimates of

benefits, costs, and transfers that were sometimes monetized, sometimes

quantified but not monetized, sometimes qualitative, and, most often,

some combination of the three. Generally, the boundaries between these

types of estimates are relatively well defined.

2. Benefits and Costs of Economically Significant/Major Final Rules

(April 1997 to March 1998)

(a) Social Regulation. Of the 33 rules reviewed by OIRA, 22 are

regulations requiring substantial additional private expenditures and/

or providing new social benefits.32 (See table 9). EPA

issued nine of these rules; USDA three; HHS three; DOI and DOE two

each; DOT and DOL one each; and HHS/DOL/Treasury jointly issued one

rule. Agency estimates and discussion are presented in a variety of

ways, ranging from a purely qualitative discussion, e.g., the benefits

of EPA's toxics release inventory rule, to a more complete benefit-cost

analysis, e.g., DOE's energy conservation standards for refrigerators

and freezers.

---------------------------------------------------------------------------

\32\ The other 11 are ``transfer'' rules.

---------------------------------------------------------------------------

(i) Benefits analysis. Agencies monetized at least some benefit

estimates in a number of cases including: (1) USDA's $2.41 billion over

15 years from the effects of its environmental quality incentives

program on net farm income, pollution damage reductions, and wildlife

enhancements; (2) EPA's $12 to $57 million per year in terms of better

water quality from its pulp and paper effluent guidelines rule; and (3)

DOE's $7.62 billion over 30 years in energy savings from its energy

efficiency rule for refrigerators and freezers.

Of the 22 (non-transfer) rules listed in table 9, agencies

monetized all the benefit estimates that they were able to quantify in

eight cases. In five cases, agencies provided some of the benefit

estimates in monetized and quantified form, but did not monetize other,

important components of benefits. DOE's two energy efficiency rules

monetized the value of energy savings and quantified, but did not

monetize, the power plant emission reductions associated with the

reduced energy consumption. DOL's respiratory protection rule monetized

the out-of-pocket savings associated with its estimate of injury and

illness reductions, but monetized neither the other aspects of those

injuries and illnesses (such as pain and suffering) nor the fatalities

avoided.

In three cases, agencies provided quantified but not monetized

benefit estimates. These included: (1) HHS's 297 to 1306 life-years

extended as a result of its organ transplant rule; (2) EPA's 593,000

tons of nitrogen oxide emission reductions per year from its highway

heavy-duty engines rule; and (3) EPA's annualized emission reductions

of 385,000 tons of nitrogen oxides, 6,000 tons of hydrocarbons and

4,000 tons of particulate matter from its locomotives rule.

Finally, in six cases, agencies reported neither monetized nor

quantified benefit estimates. In many, though not all, of these cases,

the agency provided a qualitative description of benefits. For example,

HHS' animal feed rule discusses the potential benefits of avoiding an

outbreak of ``mad cow'' disease, but does not estimate the probability

of such an episode. EPA's analysis of its expansion of its toxic

release inventory reporting rule includes a qualitative discussion of

making these data available to the public.

(ii) Cost analysis. In 19 of the 22 cases, agencies provided

monetized cost estimates. These include such items as: USDA's estimate

of $1.65 billion over 15 years for its environmental quality incentives

program; DOL's estimate of $111 million per year for its respiratory

protection rule; and EPA's estimate of $37 billion per year to achieve

full attainment of its revised primary National Ambient Air Quality

Standard for particulate matter. For three deregulatory rules--USDA's

Sonoran pork and Argentinian beef rules and EPA's PCB disposal rule--

agencies' monetized cost estimates were small or zero.

For the remaining three rules, the agencies did not estimate costs.

These included DOI's two migratory bird hunting rules and NHTSA's light

truck fuel economy rule.

(iii) Net monetized benefits. Thirteen of these 22 rules provided

at least some monetized estimates of both benefits and costs. Of those,

six have positive net monetized benefits, that is, estimated monetized

benefits that unambiguously exceed the estimated monetized costs of

[[Page 44049]]

the rules. For example, DOE's energy conservation standards for

refrigerators and freezers will generate an estimated net benefit of

$4.18 billion (present value) through 2030. EPA's PCB disposal rule

will result in an estimated net benefit of about $161 million per year.

Four rules resulted in negative net monetized benefits. These included

DOL's respiratory protection rule and EPA's medical waste incinerator

rule. Two rules resulted in monetized benefit estimates that were

sufficiently uncertain as to include both possibilities (net benefits

and net costs). For example, EPA's pulp and paper hazardous air

pollutant rule was estimated to generate between $925 million in net

benefits and $1.165 billion in net costs. Finally, one rule (USDA's

Sonoran pork rule) was estimated to have $0 benefits and $0 costs.

(iv) Rules with quantified effects of less than $100 million per

year. Seven of the rules in table 9 are classified as economically

significant even though they have no quantified effects that exceed

$100 million in any one year. These deserve comment:

USDA (2 Rules)--Importation of Pork from Sonora, Mexico, and Beef

from Argentina: In 1997, USDA began implementing a new general policy

allowing, under certain conditions, the importation of animal products

from certain regions of countries shown to be free of pests. This

policy was promulgated by rule on October 28, 1997 (62 FR 56000,

56027), but was not designated as major because the Department

concluded that analysis of the benefits and costs of the general policy

was infeasible. Instead, the Department undertook to perform such

analyses on each significant action implementing the general policy:

Because this framework will not be fully implemented until we

receive a new request to allow the importation of animals or animal

products into the United States, and because we do not know the

number or sources of requests we will receive in the future, we

cannot estimate the economic impact of this rule as stipulated in

E.O. 12866. We are therefore committed to performing a risk

assessment and cost-benefit analysis on a case-by-case basis for

each request we receive in the near future. [62 FR 56010]

The individual rulemakings concerning the importation of pork from

Sonora, Mexico, and beef from Argentina represent the first two

applications of this general regionalization policy and were analyzed

as if they were ``major'' pursuant to this departmental commitment.

HHS--Substances Prohibited in Animal Feed: FDA estimated that this

rule will cost $53 million per year. It did not attempt to estimate the

benefits to be expected from the rule because it was unable to estimate

the probability of an outbreak of Bovine Spongiform Encephalopathy

(``mad cow disease''). However, FDA did estimate that the consequences

of an outbreak, should one occur, would be substantial. It estimated

the losses from the destruction of exposed livestock would be about

$3.8 billion.

DOI--Migratory Bird Hunting (2 Rules): These are unusual rules in

that they are permissive rather than restrictive; that is, migratory

bird hunting is prohibited absent these annual regulations which allow

hunting, setting bag limits and other controls on both early and late

season hunts. Thus the rules permit such spending rather than requiring

the expenditure of private resources. DOI reports that the National

Survey of Fishing, Hunting, and Wildlife Associated Recreation

indicated that expenditures by migratory bird hunters (exclusive of

licenses, tags, permits, etc.) totaled $686 million in 1991. Based on

this estimate, DOI estimated expenditures for duck hunters would be

over $400 million per year in 1995. However, this figure is not in the

commonly used sense a social benefit.

DOE--Room Air Conditioners: This rule was proposed as part of a

substantially larger rulemaking that included seven other types of

household appliances, such as water heaters, fluorescent lamp ballasts,

and mobile home furnaces. Energy efficiency standards for all eight

combined clearly would have been economically significant. Even though

the monetized effects of this rule are less than $100 million in any

year, the annualized energy savings benefits (about $60 million per

year) are substantial. This fact, combined with the rule's history led

to the decision to maintain the ``economically significant''

designation.

DOT--Light Truck CAFE: Each year, DOT must establish a corporate

average fuel economy (CAFE) standard for light trucks, including sport-

utility vehicles and minivans. (DOT also sets a separate standard for

passenger cars but is not required to revisit the standard each year.)

For the past three years, however, appropriations language has

prohibited NHTSA from spending any funds to change the standards. In

effect, it has frozen the light truck standard at its existing level of

20.7 miles per gallon (mpg) and has prohibited NHTSA from analyzing

effects at either 20.7 mpg or alternative levels. Although benefits and

costs are not estimated, DOT's experience in previous years indicates

that they may be substantial. Over 5 million new light trucks are

subject to these standards each year, and the standard, at 20.7 mpg, is

binding on several manufacturers. Some are just above the standard and

at least one is currently below 20.7 mpg. Because of these likely,

substantial effects, we designated the rule as economically significant

even though analysis of the effects was prohibited by law.

(b) Transfer Regulations. Of the 33 rules listed in table 9, 11

were rules necessary to implement Federal budgetary programs. The

budget outlays associated with these rules are ``transfers'' to program

beneficiaries. Of the 11, two are USDA rules that implement Federal

appropriations language regarding home day care meal programs and

agricultural policies; seven are HHS rules that implement Medicare and

Medicaid policy; one is a DOJ rule regarding immigration policy; and

one is a VA rule regarding compensation of veterans who have

cardiovascular disabilities.

(c) Major rules for independent agencies. Several commenters

suggested that last year we omitted a major category of costs and

benefits: the costs and benefits of major rules from the independent

agencies. The General Accounting Office (GAO) is required to submit

reports on major rules to the Committees of Jurisdiction in both houses

of Congress under the congressional review provisions of the Small

Business Regulatory Enforcement Fairness Act (SBREFA), including rules

issued by agencies not subject to Executive Order 12866 (the so-called

independent agencies). We reviewed the information on the costs and

benefits of major rules contained in the GAO reports for the period

April 1, 1996 to March 31, 1998. According to the GAO reports, five

independent agencies issued 41 major rules during this period. The

agencies are listed in table 10 along with a summary of the kinds of

information provided by the agencies as summarized by GAO.

Table 10 clearly reveals that the independent agencies provide

relatively little quantitative information on the costs and benefits of

regulations for major rules, especially compared to the agencies

subject to E.O. 12866. Indeed, according to a recent GAO report,

Regulatory Reform: Major Rules Submitted for Congressional Review

During the First 2 Years, (April 24, 1998), the independent agencies

themselves reported doing benefit/cost analyses for only eight, or 18

percent, of the 44 major rules they submitted to GAO during this

period. That compares to 72 out of 78 rules, or 92 percent, that

[[Page 44050]]

GAO examined for the agencies subject to Executive Order 12866. Table

10 also shows that 12 of the 41 rules, or 29 percent, from independent

agencies in our sample, which were all in the GAO sample, included some

discussion of benefits and costs even though in some cases the agencies

reported that they did not do a benefit cost analysis. However, table

10 also reveals that only four of the 41 regulations had any monetized

cost information and only one had any monetized benefit information.

The one rule in table 10 that estimated both benefits and costs was

an SEC rule amending the Investment Advisors Act of 1940 to exempt

certain types of investment advisors from the prohibition of SEC

registration as investment advisors. The SEC estimated benefits of $7

million and costs of $930,000.The three other rules for which costs

were estimated are the SEC's rule allowing electronic storage for

brokers or dealer reporting, which the industry estimated would reduce

costs by $160 million per year; a Federal Reserve Board (FRB) bank

holding regulation that would reduce paperwork burden by $1.3 million

per year; and an FCC regulation that requires that phones in most

public facilities be hearing aid compatible with volume controls, which

was estimated to increase the costs of a phone by from 50 cents to a

dollar.

The only estimate of costs or benefits of approximately $100

million was the industry-supplied estimate of $160 million savings for

the SEC's broker/dealer reporting rule. Since we have used a criterion

of using only agency or academic peer reviewed estimates, we conclude

that the 41 GAO reports contain no information useful for estimating

the aggregate costs and benefits of regulations.

3. Best Practices and RIAs

Based on a review of the 21 agency cost-benefit analyses for the

period from April 1, 1996 to March 31, 1997, last year's report

concluded that we need better information in order to determine whether

proposed regulations produce the greatest net benefits. Based on a

review of 22 additional agency analyses for the year from April 1, 1997

to March 31, 1998, that conclusion still stands. Nevertheless, agencies

are making significant efforts to apply the Best Practices principles

in their RIAs. Below we discuss several examples of agencies'

application of these principles to their analytical work.

Serious deviations from Best Practices on any one criterion can

dramatically diminish the usefulness of the analysis, or worse, lead to

analytical results that distort the facts and ultimately result in

regulatory decisions that are far from optimal. Because of the

importance of ``getting it right,'' we thought it would be instructive

to select several criteria from the Best Practices document and discuss

some examples of how agencies properly applied them in their regulatory

analyses:

Quantification and monetization of estimates and treatment

of qualitative estimates

Determination of a consistent and reasonable baseline

Evaluation of regulatory options

Treatment of bias and uncertainty

Treatment of future streams of benefits and costs

(i) Quantification, monetization and treatment of qualitative

estimates. All monetized estimates are, by definition, given in dollars

and (unless there are overlapping effects of rules that are not

accounted for) permit ready comparison and aggregation. Monetized

estimates of effects are what is most generally considered the basis of

benefit-cost analysis. Even when such figures are available, however,

care must be taken when interpreting them because they depend for

comparability on a number of distinct elements. Specifically, monetized

estimates consist of: (1) The dollar value itself; (2) the base year of

the dollar used; (3) the initial year in which the effects occur; (4)

the final year after which the effects disappear; and (5) the discount

rate used to convert future into current values (or vice versa).

Quantified estimates may take the form of a variety of different

units, but they share in common a numeric measure. Generally,

quantified estimates of benefits, costs, and transfers must be

interpreted with the same elements noted above in mind. The most

important difference, of course, is that quantified estimates are

expressed in units other than dollars. Such estimates may be aggregated

only if they are presented in the same or similar units. Also, a

quantified estimate should identify the applicable time period, e.g.,

tons of pollution controlled per year, number of endangered species

protected from extinction per decade. Quantified estimates that lack

reference to the time periods to which they apply may be highly

misleading, and should be converted to similar time periods to be

comparable. Indeed, even when estimates of a similar type include

explicit reference to their underlying time periods, care must be taken

when aggregating or comparing them because of the risk of summing

estimates based on different time periods or inconsistent base years.

In contrast, qualitative estimates may not have any units at all,

or they may be expressed in units that do not lend themselves to simple

comparisons. As has often been observed, it is more frequently the case

that costs are monetized and that benefits are more often quantified or

presented in qualitative form. Qualitative effects should be evaluated

in terms of their uniqueness, reversibility, timing, and geographic

scope and severity. These effects are the most difficult to interpret,

and this may lead some to give them short shrift. The fact that an

effect has not been monetized or quantified does not, however,

necessarily mean that it is small or unimportant.

Qualitative effects must be used with care for other reasons as

well. Because they tend to be general and descriptive, they may be

broader than the incremental effects of the particular regulation being

analyzed. For example, in developing a rule designed to address a

particular safety problem, an agency may describe the extent of the

problem--that is, so many persons injured per year from this particular

cause. While important in estimating the benefits of the rule, this

figure itself is not a benefit estimate unless and until it is linked

to the likely effectiveness of the proposed rule. Finally, qualitative

estimates cannot be aggregated at all because they do not contain units

that permit arithmetic operations. In addition, not infrequently they

fail to contain relevant information about the period of time during

which they apply.

(ii) Baseline. One of the criticisms often cited in evaluating RIAs

is the failure to use a consistent baseline against which to estimate

both benefits and costs, or the failure to adopt a baseline that

reflects current and future conditions (including current regulatory

requirements). Using inconsistent or incorrect baselines will lead to

biased estimates of benefits and/or costs. When this happens, the

analysis may incorrectly make one or more of the various regulatory

options appear reasonable or vice versa.

The Best Practices document states that the baseline should be the

best assessment of the way the world would look absent the proposed

regulation. In addition, when more than one baseline appears reasonable

or the baseline is very uncertain, the agency may choose to measure

benefits and costs against multiple alternative baselines as a form of

sensitivity analysis.

In its analysis of the cost impacts for the final PCB disposal

rule, for example, EPA considered three alternative baselines

reflecting different interpretations of existing regulatory

[[Page 44051]]

requirements. EPA's preferred baseline scenario reflects EPA policy as

it has evolved over the period since 1979 when EPA published an earlier

final rule with regard to PCBs generally (although it does not reflect

the special circumstances associated with the disposal of PCB-

contaminated ship hulls). A second baseline reflects a literal

interpretation of the 1979 rule; a third alternative, the ``special

circumstances'' baseline, reflects current EPA policy because the Navy

is already disposing of ship hulls in a manner consistent with the new

rule. Using these alternative baselines, EPA estimates that the final

PCB rule would yield net cost savings ranging from $150 million for the

special circumstances baseline to $740 million for a literal

interpretation of the 1979 rule. The use of multiple baselines is

informative because it illustrates that changes in EPA policy in

implementing regulations can have a substantial effect on the cost of a

regulatory program. In this case, in the years after EPA adopted a

final disposal rule in 1979, changes in EPA policy--especially allowing

the disposal of automobile ``shredder fluff'' in municipal landfills--

have operated to reduce the cost of the program by more than $500

million per year.

(iii) Regulatory options. The analysis should consider the most

important alternative regulatory options in addressing the problem.

Failure to do so may give the selected option the appearance of being

the best alternative when in fact there are one or more others that

result in higher benefits and/or lower costs and thus greater net

benefits. It is critical that the alternatives analyzed be reasonable.

Analyzing bogus or ``straw man'' options only exacerbates the problem.

The analysis might consider, for example, the use of performance-

based standards, different levels of stringency, differential standards

for different parts of the regulated population, and differential

approaches for assuring compliance. If the proposed regulation is

composed of a number of distinct provisions, it is important to

evaluate the benefits and costs of the different provisions separately.

Particularly in the case of alternative levels of stringency, the

analysis should estimate the incremental benefits and costs of each

option as compared with the next-less-stringent option.

DOE's final rule setting new energy efficiency standards for

refrigerators and freezers, for example, includes analysis of a

comprehensive set of options. For each of eight classes of

refrigerators, e.g., top-mounted freezer with automatic defrost, DOE

estimated the benefits and costs of at least 12 alternative levels of

performance standards. For one class, DOE analyzed 28 options. This

extensive analysis of alternatives provided DOE with a very rich array

of information on the relative effects of alternative standards. For

example, DOE's analysis of over 20 alternative performance standards

for one class of top-mounted refrigerators enabled it to select an

option that resulted in per-unit net benefits more than $200 greater

than for the least attractive option considered in the analysis.

(iv) Bias and uncertainty. The analysis should address areas of

uncertainty and potential bias. The analysis should also provide a

clear discussion of the assumptions underlying the analysis and address

the uncertainties that attend these assumptions. Sensitivity analysis

helps to identify the truly critical assumptions, thereby enabling the

analysts to focus their efforts on further refinements to the analysis

in those areas.

The Best Practices document states that where benefit or cost

estimates are heavily dependent on certain assumptions, it is essential

to identify these assumptions explicitly and to carry out sensitivity

analyses based on alternative plausible assumptions.

EPA's analysis for the two rules revising primary National Ambient

Air Quality Standards (NAAQS) for ozone and particulate matter (PM)

presents a plausible range for the benefits estimates; the range

reflects alternative assumptions with respect to the estimates for

specific benefit categories (EPA, RIA for PM and ozone primary NAAQS,

pp. ES-9 and 10). For example, the analysis presents high and low ozone

benefit estimates which reflect differences in the treatment of the

possible effect of ozone on premature mortality. Similarly, the

analysis presents high and low PM benefit estimates to reflect

differences in the treatment of a possible threshold below which PM

would have little or no effect on premature mortality.

(v) Future streams of benefits and costs. As discussed above, care

must be taken in comparing estimates of effects to assure that they are

presented in a comparable time frame. This requires consideration of

several factors: (1) The initial year in which the effects occur; (2)

the final year after which the effects disappear; (3) the discount rate

used to convert future into current values (or vice versa); and (4) the

format in which the value is presented.

Format means the characterization of the monetized or quantified

effects over time. In the rules on which we are reporting, we found

that agencies used a variety of formats:

(1) Annualized values;

(2) Present values;

(3) Constant annual values; and

(4) Other or unknown formats.

From the perspective of benefit-cost analysis, annualized and

present value formats are always preferred because they permit

aggregation and comparisons within and across regulatory actions.

Constant annual values are slightly less desirable insofar as they

require the additional step of discounting to permit such aggregation

and comparison. Constant annual values are typically found in monetized

cost estimates involving Federal budget outlays, and in quantified

benefit estimates where agencies have chosen not to discount.

Aggregation and comparison within and across regulations generally

cannot be performed without a common discounting methodology. Where an

agency's estimation methodology follows an unknown format, further

research needs to be performed to ascertain how to convert or

reconstruct annualized or present value estimates.

The analysis should present a schedule of the stream of benefits

and costs where there is a variation in benefits and costs over time or

where they occur in different years, e.g., where there is a delay in

the timing of benefits relative to the costs. These streams of benefits

and costs should either be discounted to yield ``present value''

estimates or ``annualized'' to provide an estimate of annual benefits

and costs in a typical year so that they can be considered in a

comparable time frame. Failure to do so will bias the analysis in favor

of alternatives that deliver benefits later or impose costs sooner.

The Best Practices document refers to OMB Circular A-94 as the

basic guidance on discount rates for regulatory analyses. As noted in

the A-94 guidance, agencies may also present sensitivity analyses using

other discount rates (with a justification for using these alternative

rates).

For example, EPA's analysis of its final rule setting both effluent

limits for wastewater discharges and air toxic emission limits for pulp

and paper mills developed present value estimates using discount rates

of three and seven percent for benefit and cost streams over a 30 year

period (EPA, Economic Analysis * * *, October 1997, pp.10-3 and 10-4).

EPA phased in the recreational benefits over a two-year period (full

value in year three and thereafter) and the health benefits over a five

year period (full value in year six and thereafter). On the cost side,

EPA

[[Page 44052]]

assumed the capital costs would be incurred in years one and twenty-one

with operations and maintenance costs incurred in the second through

thirtieth years. The analysis adopted the 7 percent discount rate in

accordance with OMB guidance and used 3 percent, reflecting the social

rate of time preference, to reflect the sensitivity of these estimates

to alternative discount rates. The benefit estimates (including the

lower absolute value of the bound negative benefit estimate) are

roughly 50 percent larger and the costs are roughly 40 percent larger

using a 3 percent discount rate vis-a-vis a 7 percent discount rate.

4. GAO Report

A review completed by GAO looked at how well the regulatory impact

analyses for 20 economically significant health, safety, and

environmental regulations issued between July 1996 and March 1997

followed our Best Practices guidelines (GAO 1998). For example,

according to GAO, five of the 20 rules examined did not discuss

alternatives, six did not assign dollar values to benefits, and one did

not assign dollar values to costs--all practices recommended by our

guidance (GAO, 1998). In addition, GAO found that the analyses differed

in their treatment of assumptions and uncertainty. For example,

agencies used various discount rates that ranged from 2.1 percent to 10

percent, and for the six analyses that used an estimate for the value

of a statistical life, the estimates ranged from $1.6 million to $5.5

million. GAO does point out, however, that the Best Practices guidance

does allow agencies flexibility to vary the assumptions to fit the

circumstances of the specific rules, although GAO also points out that

in many cases the agencies do not explain why they varied from Best

Practice recommendations.

On a more positive note, GAO also reported that according to agency

officials, 12 of the 20 analyses were used to help identify the most

cost-effective of several alternatives or to cost-effectively implement

health-based regulations and that seven of the remaining analyses were

used to define the scope and timing of implementation, document and

defend regulatory decisions, and reduce health risks at feasible costs.

Only one of the analyses played almost no part in regulatory decisions,

and that was because the statute was too prescriptive to leave any

discretion in implementing the regulation.

As we stated last year:

Although considerable progress has been made in providing micro

data in advance of regulatory proposals and in developing the Best

Practices guidance, further progress is needed to continue improving

regulatory decisions. Specifically, we need to ensure that the

quality of data and analysis used by the agencies improves, that

standardized assumptions and methodologies are applied more

uniformly across regulatory programs and agencies, and that data and

methodologies designed to determine whether existing regulations

need to be reformed are developed and used appropriately.

Chapter III: Estimates of Benefits and Costs of ``Economically

Significant'' Rules, April 1995-March 1998

In last year's report, we recommended that OIRA continue to develop

a data base on benefits and costs of major rules. This chapter seeks to

respond to that recommendation by presenting the available benefit and

cost estimates for individual rules from April 1, 1995 through March

31, 1998. The summary of agency estimates for final rules from the

current year (April 1, 1997 to March 31, 1998) is presented in chapter

II, table 9. The summary of agency estimates for final rules from the

preceding two years (April 1, 1995 to March 31, 1997) is presented in

tables 17 and 18.

In assembling agency estimates of benefits and costs, we have:

(1) Applied a uniform format for the presentation of benefit and

cost estimates in order to make agency estimates more closely

comparable with each other, e.g., provided the benefit and cost streams

over time, annualized benefit and cost estimates, etc., and

(2) Monetized quantitative estimates where the agency has not done

so, e.g., converted tons of pollutant per year to dollars.

The adoption of a format that allows the presentation of agency

estimates so that they are more closely comparable also allows, at

least for purposes of illustration, the aggregation of benefit and cost

estimates across rules. At the same time we caution the reader that

agencies have used different methodologies and valuations in

quantifying and monetizing effects and we have attempted to be faithful

to the respective agency approaches. In this chapter, we also aggregate

benefit and cost estimates for those Federal rules with significant

quantified benefit and cost estimates.

As noted in chapters I and II, the substantial limitations of the

available data on the benefits and costs for this set of rules raise

significant obstacles to the development of a meaningful aggregate

estimate of benefits and costs for even a single year's regulations.

For example, in many cases agencies identified important benefits of

their rules that were not quantifiable. In such cases, we necessarily

omitted them from the monetized estimates we develop in this chapter.

To the extent that these benefits are substantial, the monetized

estimates will understate the total value of the benefits. The

discussion below addresses other limitations in the data and outlines

the steps we have taken in an effort to overcome some of them.

A. Monetized Benefit and Cost Estimates for Individual Rules

First, we have only included in this chapter those major rules with

quantified estimates of benefits and costs. These include six rules

from the 1995/96 period, 15 rules from the 1996/97 period, and 13 rules

from 1997/98 period. We have excluded 13 rules without quantified

estimates of either benefits or costs. (See table 11.) Six additional

rules listed in table 12 have also been excluded from further

discussion because only quantified cost estimates were available and/or

there were only relatively small benefit and cost estimates.

Second, for some of the remaining rules, agencies quantified

estimates of significant effects, but did not assign a monetized value

to these effects. Some of the quantified effects--for example, small

changes in the risk of premature death or serious injury--are

frequently identified as outcomes for a variety of rules. In a number

of instances, though, agencies did assign monetized estimates to these

outcomes.

Differences in valuation across rules are often critical,

particularly in comparisons between and among individual rules or

programs. Furthermore, the different approaches in the quantification

and monetization of these effects across agencies result in an ``apples

and oranges'' problem in aggregating estimates; in particular, where

effects have been quantified, but not monetized, the different

quantitative effects cannot be summed because they are not expressed in

common units. In order to address this problem, this section takes the

additional step of assigning a monetized value in order to provide a

more consistent set of estimates in those cases where agencies only

quantified significant effects. We have not, however, attempted to

quantify or monetize any qualitative effects identified by agencies

where the agency did not at least quantify them.

Agencies have, over the years, taken, and continue to take, several

different approaches toward rules that affect small risks of premature

death. In some cases, such as FDA's tobacco rule, agencies have

quantified and monetized

[[Page 44053]]

these effects in terms of ``quality-adjusted statistical life years.''

In other cases, such as FRA's roadway worker protection rule, agencies

have quantified and monetized these effects in terms of statistical

lives. In still other cases, such as HHS's organ procurement rule and

NHTSA's air bag depowering rule, agencies have quantified risks of

death in terms of life-years or lives, but have not monetized them.

Finally, in some cases, such as FDA's animal feed rule, the agency did

not develop a quantified estimate of the rule's mortality effects.

Estimates for the value of a statistical life varied across

agencies. For the tobacco rule, FDA estimated benefits based on a value

of $2.5 million per statistical life. For the roadway worker rule, FRA

used $2.7 million per statistical life. For the upper-bound estimates

of EPA's ozone and PM NAAQS rules, the agency used $4.8 million per

statistical life; and for its mammography rule, FDA also used $5

million per statistical life.33 Similarly, agency estimates

for the value of a statistical life-year have also varied. FDA used

$116,500 per life-year for its tobacco rule; EPA used $120,000 per

life-year to produce its lower-bound estimates of benefits in its ozone

and PM NAAQS rules; FDA used $368,000 per life-year in its mammography

rule. As a general matter, we have deferred to the individual agency's

judgment in this area. In cases where the agency both quantified and

monetized fatality risks, we have made no adjustments to the agency's

estimate.

---------------------------------------------------------------------------

\33\ There is a relatively rich body of academic literature on

this subject. The methodologies used and the resulting estimates

vary substantially across the academic studies. Based on this

literature, agencies have developed estimates they believe are

appropriate for their particular regulatory circumstances.

---------------------------------------------------------------------------

In cases where the agency provided only a quantified estimate of

fatality risk, but did not monetize it, we have monetized these

estimates in order to convert these effects into a common unit. For

example, in the case of HHS's organ donor rule, the agency estimated,

but did not monetize, statistical life-years saved, although it

discussed HHS's use of $116,500 per life-year in other contexts. We

valued those life-years at $116,500 each. For NHTSA's air bag

depowering rule, we used a value of $2.7 million per statistical life.

In cases where agencies have not adopted estimates of the value of

reducing these risks, we used estimates supported by the relevant

academic literature. For DOL's respirator rule, for example, we used $5

million per statistical life. As a practical matter, the aggregate

benefit and cost estimates are relatively insensitive to the values we

have assigned for these rules because the aggregate estimates are

dominated by the FDA tobacco rule and EPA's rules revising the ozone

and PM primary NAAQS. Finally, we did not attempt to quantify or

monetize fatality risk reductions in cases where the agency did not at

least quantify them.

B. Valuation Estimates for Other Regulatory Effects

The following is a brief discussion of our valuation estimates for

other types of effects which agencies identified and quantified, but

did not monetize.

Injury. For the air bag depowering rule, we adopted the

Department of Transportation approach of converting injuries to

``equivalent fatalities.'' These ratios are based on DOT's estimates of

the value individuals place on reducing the risk of injury of varying

severity relative to that of reducing risk of death. For the two OSHA

rules we used a ratio of 20 injuries per equivalent fatality.

Change in Gasoline Fuel Consumption. We valued reduced

gasoline consumption at $.80 per gallon pre-tax.

Reduction in Barrels of Crude Oil Spilled. We valued each

barrel prevented from being spilled at $2,000. This reflects double the

sum of the most likely estimates of environmental damages plus cleanup

costs contained in a recent published journal article (Brown and

Savage, 1996).

Change in Emissions of Air Pollutants. We used estimates

of the benefits per ton for reductions in hydrocarbon, nitrogen oxide

(NOX), sulfur dioxide (SO2), and fine particulate

matter (PM) presented in EPA's Pulp and Paper cluster rule (October,

1997). These estimates were obtained from the RIA prepared for EPA's

July, 1997 rules revising the primary NAAQS for ozone and fine PM. We

note that in this area, as in others, the academic literature offers a

number of methodologies and underlying studies to quantify the

benefits. There remain considerable uncertainties with each of these

approaches. For each of these pollutants, we used the following values

(all in 1996$) for changes in emissions: 34

\34\ Where applicable, the lower (higher) end of the value

ranges in all of the tables throughout this report reflect the lower

(higher) values in these ranges.

---------------------------------------------------------------------------

Hydrocarbons: $519 to $2,360/ton;

Nitrogen Oxides: $519 to $2,360/ton;

Particulate Matter: $11,539/ton; and

Sulfur Dioxide: $3,768 to $11,539/ton.

Third, in order to make agency estimates more consistent, we

developed benefit and cost time streams for each of the rules. Where

agency analyses provide annual or annualized estimates of benefits and

costs, we used these estimates in developing streams of benefits and

costs over time. Where the agency estimate only provided annual

benefits and costs for specific years, we used a linear interpolation

to represent benefits and costs in the in-between years. In the case of

EPA's Federal test procedure rule, for example, the analysis reported

emission reductions for only four years, i.e., 2005, 2010, 2015, and

2020. We used linear interpolation to provide benefit and cost streams

over the intervening years.

In addition, agency estimates of benefits and costs cover widely

varying time periods. For example, EPA's analysis for the pulp and

paper effluent guidelines rules developed annualized benefit estimates

for a stream of benefits over 30 years. Annualized cost estimates for

this rule were based on installation of control equipment in the first

year with full replacement of the control equipment in year 21 at the

end of the 20-year useful life for the control equipment and operating

and maintenance costs after the first year. USDA's analysis of the

conservation reserve program provided annual benefit and cost estimates

for the five-year period from 1997 to 2002. On the other hand, DOE's

analysis of energy conservation standards for refrigerators and

freezers evaluated a much longer time frame from 2000 to 2030, and

EPA's analysis of its rule setting emission standards for new

locomotives used a time frame of forty years (2000 to 2040).

These differences in the time frames evaluated reflect specific

characteristics of individual rules. The short time frame of USDA's

conservation reserve program rule reflects, for example, the five-year

legislative cycle of the farm bills. On the other hand, the longer time

frames of DOE's refrigerators and freezers rule and EPA's new

locomotives rule reflect the relatively long period required for

turnover of the existing stock of equipment and replacement with

equipment meeting the new standards. Because there are substantial

differences in the time frame of analysis for these rules, we have

decided--with the one exception of DOT's air bag depowering rule--to

treat the benefit and cost streams as though all of these rules are in

place through the year 2050. We made the one exception to this approach

for DOT's air bag depowering rule because the rule automatically

terminates at the end of five years. We believe that this is a

reasonable treatment of the benefit and

[[Page 44054]]

cost streams because a number of these rules will not achieve their

full effect for many years into the future. In addition, major

regulatory programs tend to be long-lived and, thus, the adoption of a

longer time horizon appears to be appropriate. This approach holds the

baseline constant and does not consider, of course, the potential

effect of a ``rising baseline'' as a result of technological change,

cultural changes, etc. (See discussion in chapter I.)

Finally, we have not made any changes to agency monetized

estimates. To the extent that agencies have adopted different monetized

values for effects, e.g., different values for a statistical life, or

different discounting methods, these differences remain embedded in

tables 13 through 15. Any comparison or aggregation across rules should

also consider a number of factors which the presentation in tables 13

through 15 does not address. First, for example, these rules may use

different baselines in terms of the regulations and controls already in

place. In addition, these rules may well treat uncertainty in different

ways. In some cases, agencies may have developed alternative estimates

reflecting upper- and lower-bound estimates. In other cases, the

agencies may offer a midpoint estimate of benefits and costs, and in

some cases the agency estimates may reflect only upper-bound estimates

of the likely benefits and costs. Also, in order for comparisons or

aggregation to be meaningful, benefit and cost estimates should

correctly account for all substantial effects of regulatory actions,

including potentially offsetting effects, which may or may not be

reflected in the available data.

C. Aggregation of Benefit and Cost Estimates Across Rules

In table 16, we aggregated the estimates for individual rules from

tables 13 through 15 by year. This approach yields ex ante estimates of

the benefits and costs that Federal agencies expected from major rules

issued in each of the last three years.

We have several important observations to offer on these aggregate

estimates. First, the 1996 HHS rule placing restrictions on the sale of

tobacco and EPA's 1997 rules revising the NAAQS for ozone and

particulate matter dominate the annualized and present value aggregates

presented in table 16. Changes in estimation methodology for these

rules, as reflected by the ``plausible range'' adopted by the analysis

for the EPA NAAQS rules for ozone and particulate matter, will have a

marked effect on the aggregated benefit and cost estimates for the

rules published over the period

This text is long and has been trimmed here. Open the source document for the complete record.