Journal of Agncultural Economics - Volume 54 Numher 3 - September 1999 -Pages 512-535

The Normative Analysis of Agricultural Policy:

A General Framework and Review
David S. Bullock, Klaus Salhofer and Jukka Kola

T
he paper provides a general framework of normative agn‘cultural policy analysis m’thin
which it is possibb to unajg,fwty years of literature. It is shown that the literature has gone
from examining a v e r ~small set of simple policies to a much broader set of policies thai
combine policy instruments simultaneously. In ranking alternative policies agn‘cultural economists
have hied to consider distributive equity. While at a glance it may seem that many dlferent methods
have been used to consider distributive equity, we show that in general all these methods can be traced
back to three alternative methods based on welfarism and Pareto efficiency.

1. Introduction
In his seminal work in thisJourna1, Josling (1974) reviewed and systematised economists’
attempts to analyse policies. “His ‘framework’ has in many ways become the ‘dominant
paradigm’ in which [agricultural] policy is discussed” and “still echoes through the
subsequent literature” (Peters, 1995, p. xix). While Alston and Carter (1991), Swinnen
and van der Zee (1993). de Gorter and Swinnen (1994), and Brooks (1996) very recently
have discussed the rapid developments in positive policy analysis, our paper stresses the
normative side of policy analysis, by which we mean research using welfare economics to
evaluate or rank agricultural policies. Over twenty years after Josling, we extend his work
and provide a general framework of normative policy analysis. With this framework we
attempt to unify forty years of literature on the normative analysis of agricultural policy
and provide a “big picture” of the development and accomplishments of this area of
research.
We assert that normative policy analysis is built upon three principles of social value
judgement: welfarism, the Pareto principle, and distributive equity. We present a general
W David S. Bullock is an Associate Professor in the Department of Agncultural and Consumer
Economics, University of IHinois; Klaus Salhofer is a Lecturer in the Department of Economics, Politics
and Law, Universidt fir Bodenkultur Wien (University of Agncultural Sciences Vienna) and Jukka
Kola is a Professor in the Department of Economics and Management, University of Helsinki. This
paper has benefited from comments of participants of the Annual Meetings of the American
Agricultural Economics Association in Toronto, from seminars at the University of California Davis
and the University of Agricultural Sciences Vienna, as well as from two anonymous referees. Research
w;is partly conducted while Klaus Salhofer was a Visiting Scholar at the University of Illinois and at the
University of California. He wishes to thank the Department ofAgncultura1 and Consumer Economics
in Urbana/Champaign and the Department of Agricultural and Resource Economics in Davis for
their hospitality. He also gratefully acknowledges partial finding from the Austrian Science Fund
Project No.Jl4790EK.Seniority of authorship is shared equally by the first two authors.
 Normative Analysis ofAgn'cultura1 Policy: Framework and Review 513

framework of normative policy analysis based on welfarism. Within this framework we

show how a government's capacity to affect welfare is constrained by resource scarcity,
technology and the economic behaviour of individuals. Policy analysis is limited by
researchers' abilities to identify and model the capacity of government to affect welfare.
We argue that a major development in this area was achieved by researchers exploring
the limits of how government can affect welfare, marked by the Pareto frontier. We
discuss how in efforts to get a complete social preference ordering, many researchers
have assumed various forms of distributive equity criteria, either adding these criteria as
constraints to the social welfare function (SWF), or directly incorporating these criteria
in the functional form of the SWF. Our framework provides a clearer view of the
obstacles policy analysts face, and enables us to discuss directions for future research in
normative policy analysis.

2. A General Framework of Policy Analysis Based on Welfarism, and

Related Research Challenges
Government can influence an economic system in many ways. Alternative government
actions imply alternative outcomes for individuals and hence society. The goal of
normative policy analysis is to obtain a social ordering of these alternative policies or the
implied policy outcomes (social states). A basic value judgement criterion (VJC)
commonly assumed in normative policy analysis is welfarism (Sen, 1977), defined as,

(vC.1) (The ranking of social states depends solely on the welfare of individuals.)'

Under welfarism no additional information (about individual liberty, for example) is

needed to rank policy outcomes. Social values, such as liberty, count only through the
contribution they make to individual welfare. Though not undisputed (see the
discussion between Sen (1979, 1981) and Ng (1981)), this basic value judgement
criterion is widely agreed upon among economists.
Following this welfaristic view of society, normative policy analysis is frequently
conducted by modelling government as having some number m policy instruments with
which to affect the policy outcome which is a function of the welfare of all n individuals
in society (Bullock, 1994). Formally, let x = ( x l , x2, ..., x,)be a vector of policy
instrument variables available to government. For example, policy instrument variable x1
could be an import tariff, x2 a production subsidy, x3 an environmental regulation, etc.
Each of these policy instrument variables can take on different specific values, and we
denote a specific policy instrument value with a superscript, e.g. $is an import tariff of
$0.25 per unit, xris an import tariff of $0.53 per unit, xtis a production siibsidy of $0.50
per unit, etc. If government does not use a policy instrument we denote it by superscript
0, e.g., xp is an import tariff of $0.00 per unit. A specific government policy is described
A x2,
by the values of all available policy instruments, e.g. x* = (XI, A ...,x:). One policy often
of interest is "non-intervention," here denoted by xo = (xp, x,! ...,xz), which is the policy
of simply not using any of the available instruments.
I Here no distinction is made between individual utility and individual welfare. For discussion of the
differences behveen these two concepts see Ng (1979, pp. 7-12).
514 D.S. Bullock, K SaLhoferandj. Kola

Each government policy affects the policy outcome u which is a vector of the welfare of
all n individuals in society, u = (ul, u2, ..., un). For tractability policy analysts usually
aggregate individuals into social groups and often analyse only the effect on some groups
(partial analysis) rather than all groups (general equilibrium). The agricultural
economics literature has often focused on the welfare of farmers, which we denote u1.
We use u2, u3, ..., u, to denote the welfare of social groups of nonfarmers, e.g.,
consumers, taxpayers, input suppliers, etc. Different policies imply different welfare
levels for social groups and hence a different policy outcome, e.g. xA implies uA= (uf,
B B
u;, ..., ut), XB implies UB = (ul,u2, u?, and the nonintervention policy xo implies the
"b' a 0
free market policy outcome uo = (ul, u2, ..., un).
Though government has various policy instruments to derive various policy outcomes,
what government can do in affecting welfare is limited by the realities of economic
markets.' In economic models limits imposed by economic market realities are implicit
in the assumed functional relations (typically, for example, supply and demand
functions) and in the model parameters (typically, for example, supply and demand
elasticities). Ultimately, the functional relations and the parameters reflect the wdy
economists model human behaviour (with preferences and maximising behaviour) and
the technological relationship between scarce resources and production. Let b = (bl, bp,
..., b,) be a vector of z model parameters. Let f(.) = (fl(.),f2(.), ..., f,(-)) be a vector of y
functional relations describing the economic system. Furthermore let g(.) = (gl (.), g2(.),
..., gn(.)) be a vector of welfare measures (typically producer surplus, consumer surplus,
compensating variation, etc.) . Groups' welfares are usually functions of government
policy, market conditions, the functional relations of the economic system and the used
welfare measures: u = (u1, up, ..., un) = (gl(f(x, b)), g2(f(x, b)), ..., gn(f(x,b ) ) ) = (hl(x,
b), hp(x, b), ..., h,(x, b)) = h(x, b). Assuming some specific values of market parameters,
some specific functional forms of the relations describing the economic system, some
specific measures of welfare, and a specific policy, a specific modelled policy outcome
can be obtained.
To illustrate this general framework assume a simple single market model with linear
inverse supply and demand functions, that the only policy instrument available to
government is a production quota QI and that we can divide society into two groups,
farmers (group 1) and non-farmers (or consumers, group 2) whose welfare can be
A
measured by producer surplus and consumer surplus. Hence, x = (xl) = (Q),

f(.) = P = bl + b2Q S(-):P = b3 t b4Q) and

(D(-):

Assigning specific values to the market parameters b = (bl, b2, b3, b4) = (8, -1, 0, 1 ) and
assuming a specific quota policy, one can derive a specific policy outcome. For example,
A
if Q =x f = 2, then uA = ( u t 4); A
(hl(x$ b), h2(xl,b))= ([bl - b3]x? + [b2 - b4/2] [xfl?,
- [b2/2][~:]~) = (8, 8). If Q = x1=3, then uB = (21/2, 9/2).
Given the normative contents of this study we discuss only how government is constrained by economic
market realities, not politics.
 Nonnative Analysis of Agricultural Poliq: Framework and Reuiew 515

Government's capacity to affect welfare is also limited since government can choose
only from a limited set of policies, for not all values of x are technically feasible. It makes
little sense, for instance, to think about a negative import quota, or about a per-unit
production subsidy greater than the gross domestic product. Given some vector of policy
instruments x = (xl, x2, ..., xm) modelled as available, we will denote X E Rm as the
model's set of technicallyfeasibkpolicies.' Often an analyst does not consider the effects of
all the policies in his or her model's set of technically feasible policies. We will denote the
set of examined policies as X', where of course X' E X.
Given all this, all studies in normative policy analysis face five challenges: (i) to develop
f(.), a model of the economy; (ii) to estimate the model's market parameters b; (iii) to
obtain a welfare measure h(.);* (iv) to choose a set of policies to be examined, X'; and
(v) to apply value judgements that rank or evaluate the examined policies. Challenge (i)
is of course the focus of economic modelling in general and (ii) the study of
econometrics.
Numerous developments have taken place in both areas during the last forty years,
many of which agricultural economists have adopted in their applied work. Examples of
efforts to address (i) are advances made in noncooperative game theory (see Sexton,
1994a, 1994b for a survey). Examples of efforts in (ii) are Deaton and Muellbauer's
(1980) estimating of demand systems rather than single equations, and Christensen el
d ' s (1971) more flexible descriptions of production technology. Challenge (iii) is the
focus of applied welfare economics and the techniques of estimating the welfare effects
have been developed in various directions, e.g., taking into account multi-market effects
(Just and Hueth, 1979;Just el al., 1982; Thurman and Wohlgenant, 1989; Bullock, 1993;
Thurman, 1993; Br5nnlund and Kristrom, 1996), noncompetitive market structure (Just
et al., 1979; Wong, 1989; McCorriston and Sheldon, 1994; Peterson and Connor, 1995),
and the presence of risk and uncertainty (Just et al., 1977; Konandreas and Schmitz,
1978; Wright, 1979; Helms, 1985; Larson, 1988; Fraser, 1992)s
In the current study we leave aside challenges (i). (ii), and (iii), and focus on challenges
(iv) and (v). In the next two sections we discuss how the agricultural economics literature
covering the effects of government policy on welfare has developed largely by
broadening X', the set of examined policies. In section 5 we systematise the approaches
used by agricultural economists to rank alternative policies.

3. Finding the Welfare Effects of a Policy:Mapping from Policy Space to

Welfare Space
The Simplest Case: Mapping when X' is Discrete
The basic framework most often used in the agricultural economics literature .of
mapping from policy space to social groups' welfare space follows the pioneering work
Technically feasible policies need not be politically feasible (Bullock, 1994, footnote 3).
* Welfare measure h ( . ) is a reduced form of welfare measure g(.). These forms are related by the identity h ( . )
= g(f(.)).We refer to g(f(.)) when we want to emphasise that in mapping from policy space X to welfare space
an economic model f ( . ) consisting of supply and demand relationships is often used. But for ease of notation,
throughout the rest of the paper we will usually simply refer to the welfare measure's reduce form, h ( . ) .
For an interesting review of empirical approaches to the measurement of welfare in general see Slesnick (1998).
516 D.S.Bullock, K Salhofi andJ. Kolo

Figure 1 Using Supply and Demand Curves to Map From Policy Space to Social Groups’ Welfare Space

IP
g 1 CE$= area gde

PSA

I.
7
xc =<x;,x;>

x A = XA,XO
xo =(xp,x;)

of Marshall (1920). This basic framework uses an econometrically estimated model of

agricultural markets f(x, b) to examine a set of policies X which is discrete, for example
X’ = (xA,xB, xc, xot. Hence, in this example normative comparison is made among four
separate policies, one of them the non-intervention policy.* Typically, geometric areas
behind estimated supply and demand curves (producer and consumer surpluses) are
used to obtain the values of h(xi, b) with i = A, B, C, 0.2
In terms of our general framework this standard procedure of normative policy analysis
can be represented in Figure 1 for the example of m = 2 policy instruments (xl is a target
price and x 2 a fertiliser tax/subsidy) and n = 2 social groups (farmers and consumers-
taxpayers or non-farmers). The left-hand panel shows the policy instrument space. Four
policies, xA, xB, xc, xo are depicted. Policy xA sets the target price at some positive level
A D
x t and does not use the fertiliser tax/subsidy (i.e., sets it at x20 = 0), so x* = (xl,x2). Policy
B
xB does not use the target price and sets some positive level x2 for the fertiliser tax, so xB
= (xy, 4). Policy xc sets the target price at a positive level xy and simultaneous1 sets a
c xq).
c Non-intervention policy, xo = (xl, 0 8
positive fertiliser tax xf, so xc = (xl, x2) uses
neither of the two instruments. We will call policies like xA and XB that use only one
instrument simple policies. We will call policies like xc that use multiple instruments
simultaneously combined policies.
Using a supply curve and a demand curve the welfare effects of each examined policy
are estimated by geometric areas in price-quan tity space (in the supply-demanddiagram
A
in the middle panel). In the case of policy xA, for example, a target price of x1 results in

A
surplus of PSA = area xlcf, x
a consumer price of e, Marshallian consumer sur lus of CSA = area gde, pI-oducer
and taxes of TXA = area xlcde. Similarly, the welfare effects of
Numerous studies in normative policy analysis try to evaluate the social cosu of a policy. In terms of our
framework these papers compare the non-intervention policy xo to an actual or hypothetical government
interventionary policy, e.g. x*.
* Though there exist other techniques for obtaining a welfare measure h(.) that do not calculate areas behind
demand and supply curves (see, for example, Chipman and Moore, 1980; McKenzie, 1983; Comes, 1992;
Martin and Alston, 1994), using geometric areas is still most common among agricultural economists (Alston
and Larson, 1993).
 Nonnative Analysis of Agn'cultural Policy: Framework and h i m 51 7

all other policies of the discrete set X' = (xA, xB, xc, xo) can be estimated from geometric
areas in the supply-demand diagram. Using the h(x, b) so estimated and some additional
value judgement criteria described later, it is possible to obtain a social preference
ordering of such a discrete set of examined policies.
Griliches (1958), Nerlove (1958), and Wallace (1962) introduced this basic framework
to analyse discrete sets of simple agricultural policies. Josling (1969) and Dardis and
Dennison (1969) conducted early studies of the welfare effects of a discrete set of
combined agricultural policies. Since then, various studies have conducted normative
policy analysis by comparing the welfare effects of a discrete set of combined policies
(e.g., Otsuka and Hayami, 1985; Lichtenberg and Zilberman, 1986; Babcock et al., 1990;
Constantine et aL, 1994).
Josling (1974) recommended mapping the pricequantity space model typified by the
middle panel of Figure 1 into social groups' welfare space to gain further insights into
the choice of policy instruments. The, right-hand panel of Figure 1 shows the welfare
space, of two social groups (farmers, consumer-taxpayers) and what government can and
cannot do to social groups' welfare if it has available policies xA,xB,xc and xo. The point
uA = h(xA,b) = ( P g , C@ - TXA), is found by the calculation of geometric areas in the
middle panel. Similarly, points LIB = h(xB, b), uc = h(xc, b), and uo = h(xo, b) can be
calculated using the supply and demand model as well, and mapped as shown. Using the
concept of welfarism as a starting point for the judgement of policies suggests that
Josling's basic idea was an important one: since policies should be judged by their
(welfare) outcomes, examination of these policy outcomes u*, uB, uc, and uo in welfare
space is a natural step in judging among policies xA, xB, xc, and xo, as will be discussed
in more detail later.

Broadening Ihe Set of Examined Policies: X' Continuous

In general, a model's set of technically feasible policies X is continuous, meaning that if
A A A A A
a policy (x,, x2, ..., x 3 can be imposed, then so can some (xl , + ~2 , ..., xn + E ~ ) ,
+ E ~ x2
where E ~E, ~ ...,
, E, are arbitrarily small numbers in absolute value. Recognising that when
X is discrete only a very partial view of what is technically feasible for government is
provided, Josling (1974) observed that by continuously changing the level of the
instrument of a simple policy a curve could be mapped in social groups' welfare space to
provide a broader picture of government's opportunities and constraints when using a
single policy instrument. Gardner (1983) took up Josling's basic idea and presented it in
a more systematic framework, callingJosling's curves su$~lustransformation curves (STCs).
The mapping procedure which produces Josling-Gardner STCs assumes that X' is made
up of line segments in Rm, as is illustrated in the left-hand panel of Figure 2 for the case of
two policies (xl is a production quota and x2 a target price). Point e shows the non-
intervention policy (xl,0 x2).1
0 Function h(x, b) maps point e onto the non-intervention
policy outcome uo at point E in the right-hand panel which again is the welfare space of
Note that at non-intervention policy (xy, x i ) shown by point e in Figure 2 is not (0,O).This is because under
non-intervention the production quota is a positive number: x i > 0.That is, the amount farmers are permitted
to produce is set high enough to be non-binding, that is the policy does not effect the quantity produced. At
some number x t c 4 the quota would become binding and at x i = 0 no production would be allowed.
518 D. S. Bullock, R Salhofer andJ Kola

Figure 4 Surplus TransformationCurves and the Pareto Frontier

xE@) (Set of Pareto

efficient policies) P@) (Pareto frontier)
X (Set of technically
feasible licies)

A 0
two groups (farmers and non-farmers). Point d shows a simple policy (xl, xz) which is
mapped by h(x, b) onto point D in the right-hand panel. The thick line segment between
e and d shows a set of simple policies in which x1 is used and x2 is not used. Using h(x,
b) to map line segmented onto social groups’ welfare space creates a curve passing
through points E and D, possibly shaped like STC(x1, xg, b) in the right-hand panel.
STC(xl, x,! b) is the curve made up of points (hl(xl, $,
b), h2(x1, x,: b)) generated
parametrically by changing x2 continuously from xp to XI (point d), maintaining x2 = x8
all the while. Similarly, STC(xY, x2, b) is the locus of points (hl(x9, x2, b), h2(x:, x2, b ) )
generated parametrically by changing x1 continuously from xi to x2, B maintaining x1 = x1 0

all the while. STCs usefully depict the welfare effects of policy instruments, and this is the
principal advantage of the Josling-Gardner procedure over earlier procedures that
simply provided an analysis of the effects of discrete policies. That is, Josling-Gardner
enable us to answer whether, for example, we can find a level at which to set a production
quota instrument such that its outcome is more desirable than setting a production
subsidy at some level. Examples of studies using this framework to study the welfare
effects of policy instruments are Thomson and Harvey (1981). Just (1984). Gardner
(1983, 1985, 1987, 1991), Williams and Wright (1991, pp. 378 - 383), Isosaari (1993),
Maier (1993a, pp. 126-143, pp. 216225), and Wright (1993).
In the early 199Os, several researchers extended Gardner’s (1983) approach to show
how multiple STCs can be combined to study the welfare effects of combined policies
(Bullock 1992; de Gorter el al., 1992; Gardner, 1992). Since government often combines
several policy instruments simultaneously, this extension of Gardncr’s framework is quite
 Normative Analysis ojdgricultural Policy:Framework and h i m 519

useful for the empirical analysis of actual policies. This procedure is illustrated in Figure
2, where point g in the left-hand panel represents a combined policy (XI, c x2).
c and point
C in the right-hand panel shows the policy outcome of this policy. Because two policy
C x a , the welfare effects of the combined polic can be
instruments are used in policy (xl,
0 0 C J
traced along two STCs; changing policy from non-intervention (XI, x2) to (XI, x2) in the
left-hand panel changes welfare from point E to point F along STC(xl,&, b) in the right-
hand panel. Then changing policy from (xy, 4)
to (xy,x$) keeping x1 = xy constant in
the left-hand panel changes welfare from point F to point G along STC(x7, x2, b) in the
right-hand panel. Kola (1991, 1993), Maier (1993b) and Lothe and Garcia (1996) also
depicted STCs for combined policies.
A complete view of how government can affect social groups’ welfare is also represented
in Figure 2. Given the set of technically feasible policies X in the left-hand panel and the
constraints imposed by market conditions, F(b) in the right-hand panel represents the
model’s set of technically feasible p o k y outcomes. F(b) is the mapping of the model’s entire
set of technically feasible policies X onto the model’s social groups’ welfare space
(Bullock, 1995, p. 1239). Following Bullock (1994) F(b) is defined as

F(b) = ( U I u = h(x, b), x E XI. (1)

Note that sets of simple policies, such as line segments ec and ed, are subsets of X. Since
the STCs result from mapping subsets of X onto social groups’ welfare space, then the
STCs must be contained in F(b). Note also that many technically feasible policy
outcomes, such as point C, can only be obtained by way of combined policies.

4. The Pareto Principle and Pareto Efficiency

Having analysed government’s abilities to affect the policy outcome, and given the basic
value judgement criterion of welfarism, we can discuss how to rank feasible policies using
additional value judgement criteria. A value judgement criterion commonly accepted
among economists is the Pareto principle. According to this value judgement criterion, a
policy xAis preferred (or Pareto superior) to a policy xB if xA makes at least one person (or
in this context group) better off than he or she is under xB,while no one is made worse
off.’ That is, under the Pareto principle

(W.2) for any xA, xB E X, xA F xB e hi(xA,b) 2 hi(xB,b), i = 1, 2, ..., n,

with at least one inequality strict

A policy x* is said to be Pareto eflicient (or Pareto optimal) if there is no technically feasible
policy Pareto superior to x*. Formally, x* is Pareto efficient if x* E X, and there exists no
Defining the Pareto principle for groups rather than individuals of course requires some strong assumptions
about preferences and endowmenu of the individuals condensed within a group, but is nevertheless common
practice in applied work.
520 D.S. Bullock, K SalhofeandJ Koh

other x’ E X that satisfies hi(x’,b) 2 hi(x*, b) i = 1, 2,. . ., n, with at least one inequality
strict. As in Figure 2,we denote a model’s set of Pareto efficient policies by XE(b) = ( x I
x is Pareto efficient], where the Puretojiontier P(b) = {u I u = h(x, b), x E XE(b)) is the
result of using h(x, b) to map XE(b) onto social groups’ welfare space. Clearly P(b), the
“north-east’’boundary of the set of feasible policy outcomes F(b), is of interest to policy
analysts.
In the agricultural economics literature of the past decade, the importance of the Pareto
frontier as a limit to how government can affect welfare has been considered by several
researchers in independent work.’ Bullock (1991, 1996) developed a technique for
finding Pareto efficient policies and policy outcomes for the general m-policy instrument,
n-social group model. Bullock (1991) proved formally that a policy x* is Pareto efficient
if and only if it solves simultaneously the n constrained maximisation problems:

max (hi(x,b): hj(x, b) 2 hj(x*, b)) i = 1, 2, ..., n; j = 1, 2, ..., n ; j # i. (2)

XE x

Bullock (1994, 1996) noted how the envelope theorem implies that the Pareto frontier
envelopes all Josling-Gardner STCs, and how at points along the Pareto frontier all STCs
are tangent to a common hyperplane.*
Just (1984, pp. 58, 130) and Alston and Hurd (1990) used a graphical technique to
derive a Pareto frontier for a simple model of two social groups (farmers and consumers-
taxpayers) and two policy instruments (a production quota and a production subsidy).
Alston et ul. (1993) formalised the approach ofJust (1984) and Alston and Hurd (1990)
by solving a problem in which two policy instruments are chosen to maximise consumer-
taxpayer welfare subject to a predetermined farmers’ welfare. The problem they solved
may be written in our paper’s notation as,

which necessitates maximisation of the welfare of non-farmers subject to the constraint

of a predetermined level of farmers’ welfare upre. As the authors note (footnote 2, p.
516), their approach may be thought of as a procedure for defining an “efficient surplus
transformation curve” (a Pareto frontier) for a given set of available policy instruments.
A s proved in the appendix Bullock’s (1991, 1996) method of solving n constrained
1 Alston and Carter (1991) and Bullock (1994) discuss the importance of the Pareto frontier to a wide range
of models in political economy. In particular, models in the tradition of Becker (1983) assume that government
redistributes welfare eficiently and hence the observed policy outcome is o n the Pareto frontier. Given this
importance, some of the developments discussed in this section in exploring the limits of government policy
were accomplished in studies with primarily positive intent, e.g., studies trying to rationalise observed policy
choices.
2 This is closely related to Bullock and Salhofer’s (1998b) finding that an optimal combination of In
instruments is always at least as desirable as a policy using only a subset of those m instruments for every
welfaristic set of value judgement criteria that give a full ranking of the set of technically feasible policies.
 Nonnative Analysis of Agricultural Policy: Framework and Review 521

optimisation problems simultaneously is equivalent to simpler methods proposed by

Alston et al. (1993) of solving a single constrained maximisation problem only if the
solution to their problem is unique.
Maier (1993b), Bullock (1994, 1996) and Bullock and Salhofer (1995) actually derive
Pareto frontiers for the case of two policy instruments. Salhofer (1996) derives a Pareto
frontier for the case of four instruments. Other recent studies deriving theoretical or
empirical optimal combinations of two or more policy instruments and hence points on
Pareto frontiers (compare pointj and J in Figure 2) are Gallagher (1988), Gardner (1988,
1992,1995),Innes and Rausser (1989), McCorriston and Sheldon (1991), Salhofer (1997),
Guyomard and Mahe (1994), Moschini and Sckokai (1994), Swinnen (1997), Alston and
Spriggs (1998), Bullock and Salhofer (1998a), and Swinnen and de Gorter (1998).

5. Completing the Social Preference Ordering: Distributive Equity and

Social Welfare Functions
The Pareto principle is a weak criterion for value judgement. Indeed, this weakness
accounts for its wide acceptance as a tool for establishing a social preference ordering of
policies. For the Pareto principle does not establish a complete ordering; under the
Pareto principle policy xA is (Pareto) non-comparable to xB if hi(xA,b) > hi(xB,b ) , for at
least one i E { 1, 2, . . ., n), and hj(xA, b) < hj(xB, b ) , for at least one j E (1, 2, . . ., n). To
obtairl a complete social preference ordering of X, additional value judgement criteria
must be employed. There is little doubt that human behaviour is affected by equity
considerations (Sen, 1987, 1992). In efforts to complete the social preference ordering,
agricultural economists have assumed various forms of distributive equity criteria, either
imposing these criteria as constraints to a SWF, or directly incorporating these criteria in
the functional form of the SWF.

A Bergson-Samuelson social welfare function W assigns numerical values to policy

outcomes: W: u + R. Since the arguments of a SWF are social groups’ welfare levels u,
clearly SWFs are welfaristic constructs, consistent with (VJC.1). By using a SWF one can
obtain a complete social preference ordering of X, since W assigns a number to every
technically feasible policy outcome (W: F(b) + R). A policy x* which results in a higher
(equal, lower) SWF level W is socially superior (equal, inferior) to policy XB with a lower
(equal, higher) SWF level: [x*$ xB u W(h(xA,b) $ W(h(xB, b)].’ Under the social
welfare function criterion, a policy x* is said to be socially optimal if it maximises the SWF
given the constraints outlined in the last section, that is, if it solves y W ( h ( x , b ) ) , or
equivalently, axW(u). Provided that the SWF is assumed increasing in u, x (i.e.,
. society is
TF(P).
assumed to bene it if the welfare of any social group is increased without decreasing the
I In the orthodox theory of economic policy (Tinbergen, 1952; Theil, 1958), social welfare is a function of
economic indicators, such as the rate of economic growth, the rate of employment, the external trade balance,
etc. However, such targets are not themselves ends but are only indicators of policy success. The ends of policy
are to influence the welfare of individuals, and hence our framework covers the “orthodox”view. I n the case of
agriculture, officially stated policy objectives are manifold, such as “LO promote agricultural efficiency and the
optimum utilisation of factors of production,” “to assure a fair farmer income,” “to maintain vigorous and
pleasant rural communities” or “to conserve the natural environment” (Winters, 1987, p. 291). Again, such
objectives are desirable because they contribute to individual welfare.
522 D.S. Bullock, L SauOfer and J. &la

welfare of any other social group), then if x* maximises the S W , x* is Pareto efficient
(Varian, 1992, p. 333), but not necessarily vice versa.
In applied work the most common specific functional form of a Bergson-Samuelson
SWF and hence the most common value judgement criterion used to derive a complete
ranking of policy outcomes is a utilitarian or Benthamite SWF:W(u,, u2, ..., un) = u1 + up
t ... + u, = hl(x, b) + h*(x, b) t h,(x, b). The value judgement criterion implied by the
Benthamite SWF is,

(VJC.3-1) xAt.xB if W(h(xA,b)) > W(h(xB,b ) ) , with W(h(x, b))

= hl(x, b) + h2(X, b)+ ... + h,(x, b)
>.

Policy xA is preferred to policy xB if xA gives a higher social welfare.

Instead of using the maximum sum of welfare to find x* it is also common to use the
minimum social cost (or deadweight loss) value judgement criterion (Hotelling, 1938;
Harberger, 1954), where social costs (SC) are defined as: SC(x, b) = - [Aul t A u ~t ... t
, - h2(xo, b)] + ... + [h,(x, b) - hn(xo, b ) ] ) .
Au,] = - ([hl(x, b) - hl(xo, b)] + [ h 2 ( ~b)
However, it is easily shown that minimising SC implies the same social preference
ordering as does maximising the Benthamite SWF since - SC(x, b) = h, (x, b) t h2(x,b) t
... t h,(x, b) - E, where E = h,(xo, b) + h,(xO, b) + ,.. + h,(xO, b) is the sum of non-
intervention welfare levels, and is a constant.
While the utilitarian value judgement criterion (VJC.3-1), completes the social
preference ordering of policies, ranking policy options by summing welfare levels is based
on the assumption that increasing the welfare of a wealthy person by one unit is of equal
social value as is increasing the welfare of a poor person by one unit. Hence, (VJC.3-1)
has been criticised by many noted agricultural economists over a long period of time for
failing to consider distributive equity (Nerlove, 1958; Josling, 1974; Rausser, 1982;
Gardner, 1983; Just, 1984). However, (VJC.3-1) remains the most often used value
judgement criterion (e.g., Otsuka and Hayami, 1985; Lichtenberg and Zilberman, 1986;
Leu et al., 1987; Murphy et al., 1993).

Placing Constraints on a S W to Implement Equity Considerations

Nerlove (1958) and Wallace (1962) were among the first to use welfare economics to
assess agricultural policies. Also, because they were partly dissatisfied with (VJC.3-1), they
were the first to depart from the utilitarian value judgement criterion. Nerlove (p. 223)
wrote: “If we take as an axiom that government programmes are designed to benefit the
producer, the benefit to producers resulting from a programme becomes an important
magnitude.” Nerlove and Wallace first tried to capture this by fixing the instrument levels
so that all examined policies guaranteed the same predetermined “fair” producer price
level (or equally, the same increase in price level compared to the non-intervention
price). Wallace recognised that comparing policies with the same predetermined “fair”
price may be misleading if the goal of a policy is to increase total farm revenue. Hence
 Nonnative Analysis of Agn’cultural Policy: Framework and Review 523

he suggested comparison of policies that guarantee the same predetermined level of total
farm revenue. In the last part of his paper Wallace argued that “[nleither of the two bases
for comparison are proper if one assumes the goal of agricultural policy to be one of
increasing farmers’ disposable income” (p. 589), and ”[plerhaps a more relevant basis for
comparing two plans is for equal changes in producer surplus” (p, 586). This set of value
judgement criteria can be stated in very general terms as

(VJC.3-2a)
(xA+xB if W(h(xA,b)) > W(h(xB,b)), with W(h(x, b ) ) =
h,(x, b) + h2(x, b)+ ... + h,(x, b) and hi(xA, b) =

the predetermined welfare level of group

where in Wallace’s. case i = 1. Hence, Wallace imposed distributive equity while

attempting to maintain greater consistency with the welfarism criterion since producer
price (and also total producer revenues) are probably poorer measures of producer
welfare than is the geometric area “producer surplus”.
Examples of studies using (VJC.3-2a) are Josling (1974), Gallagher (1988), de Gorter
and Meilke (1989), Alston and Hurd (1990), Alston el ul. (1993), Gisser (1993),
Blandford and Dewbre (1994), Moschini and Sckokai (1994), OECD (1995), Schmitz and
Vercammen (19951, Salhofer (1996, 1997), and Swinnen and de Gorter (1998).
When n - 1 of the n welfare levels are predetermined, the same ranking is derived when
these predetermined welfare levels are combined with the Pareto principle instead of
with a S W .Hence, we could have equivalently described (VJC.3-2a) as

(vJC.3-2b) x*> XB if xA is Pareto superior to xB, and

n-1
hl(xA, b) = hi(xB, b)= upre, for i = 1, 2, _._,
I

This is illustrated in Figure 3 for the case of two social groups. Only some policies result
in the predetermined level of farmers’ welfare. Calling the set of such policies Xpw(b) E
X, using the h(x, b) function to map Xpw(b) onto social groups’ welfare space results in
a line segment like the one labelled Fpw(b) E F(b). Hence, a full ranking of all policy
outcomes in Fpw(b) is obtained whether the Pareto principle (as in (VJG3.2b)) or the
Benthamite SWF (as in (VJG3.2a)) is applied.
Nerlove (1958) first used relative social cost (RSC) defined as RSC(X, b) = SC(x, b)/Aui,
, - h,(xO, b)l + ... [hn(x,b) - hn(xo, b ) l ) / ([hj(x, b)
= ([hl(x, b) -h,(xo, b)l + ( [ h 2 ( ~b)
- hi(xo, b)], where in his case i = 1. It can be shown that GaTdner’s measure of average
transfer efficiency (ATE) defined as ATE(x, b) = [Aui/ZAu,l = [hj(x, b) - hi(xo,
j=l
b)][Z,[h.(x, b) - hj(x0, b)], where in his case i = 1 a n d j = 2, ...,n implies the same ranking
1 J
as drlove’s RSC since ATE(x, b) = l/(RSC(x, b) - 1).
 524 D.S.Bullock, X Salhofer and J.Kola

Figure 3 Social Welfare and the Predetermined Welfare Level and Predetermined Welfare Ratio Criteria

\
of technically fe policy outcomes)

Josling (1969) and Dardis and Dennison (1969) first compared policies using RSC
combined with the constraint of a predetermined level of farmers’ welfare. It is easily
shown that minimising the RSC (with predetermined welfare levels of groups i) implies
the same social preference ordering as does minimising SC, and hence as does
maximising a SWF, since Aui = u y - ui0and hence is a constant. So,the ranking derived
from minimising RSC subject to a predetermined welfare of group i is identical to that
derived from (VJC.32a).
Josling (1974) as well as Gardner (1983, pp. 228-229) also recommended a
predetermined ratio of groups’ welfare levels value judgement criterion, very generally
defined as

(VJC.3-3a)
xA>xBif W(h(xA,b)) > W(h(xB, b)) with W(h(x, b)) =
hi(xA,b) hi(XB,b)
hl(x, b) + h2(x, b) + ...+ h,(x, b), and - = rPre,
hj (x‘, b) hj ( xB, b)
1 the predetermined ratio of welfare levels between groups I

where in Josling and Gardner’s case i = farmers and j = non-farmers. With this set of value
judgement criteria they sought to find the highest attainable point on a fixed ray through
the origin (such as at uA in Figure 3).
 Normative Analysis of Agricultural Policy: Fratnaowk and h i m 525

Placing Distributive Equity ConsiderationsDirectly into the S W

Just (1984) showed that (VJC.3-3a) can aIso be represented by a SWF with right-angled
social indifference curves (SICS). Such a SWF can be expressed by a Leontief-type
function. The value judgement criteria (VJC.33a) can also be expressed as

(VJC.3-3b)
xA+xBif W(h(xA,b ) ) > W(h(xB,b ) ) with W(h(x, b ) ) =

min(&hl(x, b), e,h,(x, b), .-.,Bnhn(x, b)l

where ei/Bj = rPre is the welfare distribution ratio between group i and group j . If the
welfare ratio rPre = 1 for all i, j, then (VJC.3-3b) is the Rawlsian maximin criterion
(Tuomala, 1990).
Nerlove (1958) and Dardis (1967a, 1967b) used RSC alone to judge policy outcomcs.
Hence their value judgement criterion can be expressed as

i
(vJC.3-4)
xA>xB if W(h(xA, b)) > W(h(xB, b)) with W(h(x, b)) =
, t ... t hn(x, b) - E
hl(x, b) t h 2 ( ~b)
hi(x, b) - hi(x0, b)

where i = 1. More recent applications of (VJC.3-4) are Cramer et al. (1990), Harvey and
Hall (1990), Sarwar and Fox (1992), Lothe and Garcia (1996) and Karagiannis et al.
(1997).
There are shortcomings to using the RSC value judgement criterion to rank policies.
First, the SWF is not defined for at any policy outcomes such that hi(x, b) = hi(xo, b ) , and
this of course includes the non-intervention policy outcomes. Second, if we try to rank
the policy outcomes for the whole feasible set, we get a transitivity problem since all SICs
intersect (Figure 4). Third, for all points East of the non-intervention point E the SWF
may judge a Pareto inferior point as more socially valuable than a Pareto superior point.
For example,uB is Pareto superior to uA, but is on a lower-valued SIC. Thus, the RSC is
not necessarily consistent with the Pareto principle.
Just (1984) and Paarlberg (1984) judged policies by a linear weighted SWF:

Note that in the political economy literature this weighted social welfare function is interpreted as a political
preference function with the weighu being determined by the relative political power of social groups.
526 D. S. Bullock, K Salhofer and J. %la

F i e4 Using the Relative Social Cost Value Judgement Criterion

Iu1 F(b) (Set of technically feasible policy outcomes)

SIC=
2

SIC==
1

Gardner (1985, 1988, 1991, 1992, 1995), Innes and Rausser (1989), Innes (1990), de
Gorter et al. (1992), Alston and Spriggs (1998), and Swinnen and de Gorter (1998)
performed other applications of (VJC.3-5)

6. A Research Agenda for the Future: the Statistical Analysis of Policy

Now that techniques to thoroughly explore government’s capacity to affect social groups’
welfare have been developed, the agenda for the future is to attain empirical results in
which we can place more confidence. This necessitates, of course, further developments
in economic modelling, econometric techniques and welfare measurement-
developments which economists have long pursued and will continue to pursue. A new
direction of research, however, will be with the pursuit of a more statistical analysis of
policy. For true values of the market parameter vector b, which must be known to map
policies into policy outcome space where they can best be judged, are rarely known.
Instead usually a “point estimate” of market parameters is obtained using econometric
methods, along with accompanying standard errors.2 Since in general the welfare
1 Beside (VJC.3-2a) and (VJC.3.5) Swinnen and de Gorter (1998) also consider the case where government
maximises political support. We don’t discuss this decision making model here because it clearly implies a self-
interested government and hence is not a normative approach.
2 Sometimes a sensitivity analysis is conducted by assuming different ”plausible” values of b to investigate the
sensitivity of policy analysis results to changes in b.
 Normative Analysis of Ap‘cultural Poliq: Framework and Review 527

measures h(x, b) are non-linear functions of the market parameters, they are also
random variables. Clearly, it is desirable to have knowledge of the distribution of the
estimated welfare measurements rather than to have point estimates only.

Different approaches to estimating standard deviations or confidence intervals for

welfare measurements, mainly in the context of consumer welfare and the valuation of
environmental goods, have been developed and used. The traditional approach is to
linearise the welfare measures using Taylor series approximations and to construct an
estimate of their standard deviations (Hausmann, 1981; Kealy and Bishop, 1986;
Bockstael and Strand, 1986). But the use of Taylor series approximations of welfare
measurements has been criticised for several reasons (Graham-Tomasi et al., 1990; Kling,
1991; Krinsky and Robb, 1991).

Very recently computer-intensive methods have been introduced in the agricultural

economics literature to obtain a fuller picture of the statistical properties of welfare
measurements. Bootstrapping procedures (Efron, 1979; Freedman and Peters, 1984) can
be used to estimate the distribution of h ( . ) by resampling either from the empirical error
distribution or from the data set directly (Efron and Tibshirani, 1993, chapter 9.5). A
different approach was suggested by Krinsky and Robb (1986, 1990, 1991). Following
their procedure one can estimate the distribution of h ( . ) by assuming that the parameter
A
estimate b is normally distributed, then using a Monte Carlo approach taking a great
number of random drawings of b from this normal distribution, and substituting them
into h ( . ) to estimate its distribution. The advantage of the Krinsky-Robb pLocedure is that
it is less computer intensive, while bootstrapping creates a distribution of b by itself rather
than assuming a normal distribution. Applications of bootstrapping and/or Krinsky-Robb
procedures in the context of consumer welfare and environmental valuation are
Adamowicz et al. (1989a, 1989b), Kling (1991), Chen and Cosslett (1998), and Griffiths
(1999). Applications to agricultural markets and the welfare of several social groups
include Tremblay and Tremblay (1995), Alston et al. (1998),Jeong et al. (1999a), and
Jeong et al. (1999b).

Numerous policy studies do not actually estimate their model parameters, but rather
assume “plausible” values of the parameters, often based on a literature review, and
justified by either the nonexistence of (reliable) data or time constraints. Hence, in this
case the analyst usually has only one or a few point estimates of each parameter available.
A procedure by which to derive a distribution of h ( . ) is to assume each parameter bi can
lie within a plausible range rather than being a point estimate, and to assume some
distribution of this range, based on information available.’ Then, as with the Krinsky-
Robb procedure, the distribution of h(.) is derived by taking random drawings from this
distribution of b. Davis and Espinoza (1998) as well as Griffiths and Zhao (1999) discuss
how in general to utilise this procedure to derive distributions for price and quantity
changes in a market displacement model. Zhao et al. (1998) further develop the same
techniques in the context of welfare measures. Salhofer (1999) uses similar procedures
with a simulation model.

1 Usually with only a view point estimate available from the literature review one will assume either a normal
distribution around the mean of the assumed range or a uniform distribution.
528 D. S.Bullock, K SalhoferandJ. Kola

7. Concluding Comments
We have shown that the researcher conducting normative policy analysis faces five
challenges: (i) modelling of the economy; (ii) estimation of model parameters; (iii)
estimation of welfare; (iv) choosing a set of policies to be examined; and v) applying value
judgements that rank the examined policies. Here we have focused on developments in
the literature that have addressed the last two challenges.
To address challenges (iv) and (v), in line with the modern manner of presentation of
microeconomic theory we have presented a general framework of normative policy
analysis based on set theory. Our proposed framework helps capture the progress in the
literature over the past 40 years in new and useful ways. We have used our framework to
identify three different spaces pertinent to the normative analysis of policy: policy
instrument space, market (pricequantity) space, and policy outcome (welfare) space,
and to show that what policy analysts generally do is map from policy instrument space to
policy outcome space via pricequantity space.
Using our framework we also show that the literature has gone from examining a very
small discrete set of simple policies to a much broader (often continuous) set of policies
that combine policy instruments simultaneously. We show that the literature’s gradual
expansion of the set of examined policies has led to a corresponding gradual expansion
of the examined feasible set of policy outcomes, from policy outcomes of a few specific
policies, to Josling-Gardner surplus transformation curves, to submanifolds of feasible
policy outcomes and corresponding Pareto frontiers. Our framework clarifies these issues
to the point where one begins to wonder why it took so long for the literature to progress
from the discrete policy analysis of the 1950s, 60s, and 70s to the current analysis of the
entire- feasible set of policies. We believe that part of the reason it took the literature so
long to progress is because there has not existed a unifying framework from which to
understand the literature as a whole. It is our hope that this paper helps fill this unifying
role, aiding in sorting out, categorising, and so understanding how and why the literature
has developed.
Given the importance of the objective of redistribution in agricultural policy,
agricultural economists have often departed from the traditional utilitarian value
judgement criterion of minimising social costs, and have tried to incorporate equity
considerations. While at a glance it may seem that many different methods have been
used to consider distributive equity, we show that in general all the work done so far can
be traced back to three alternative methods consistent with the more basic value
judgements of welfarism and Pareto efficiency: (i) maximising a utilitarian SWF subject
to predetermined welfare levels of social groups; (ii) maximising a utilitarian SWF subject
to predetermined welfare ratios of social groups (or, equivalently, maximising a Leontief-
type SWF); and (iii) maximising a weighted linear SWF. In many studies, the underlying
value judgement criteria are not immediately obvious, which can make comparison of
different contributions of the literature more difficult than necessary. After completing
this lengthy literature review, we feel strongly that researchers conducting normative
analysis of policy should state straightforwardly their value judgement criteria, in the
form of a SWF with or without constraints. Given the applied nature of agricultural policy
research, equity considerations have been incorporated in very simple ways only. A good
 Nonnative Analysis of Agn'mltural Policy: Framework and Review 529

starting point for considering newer developments of the literature on fairness might be
Chavas (1994).
The literature reviewed in this study analyses traditional commodity market
interventions. In the last decade environmental and resource policies have become
increasingly important sources of intervention in agricultural markets. In general, as long
as one accepts the basic value judgement of welfarism, the theoretical normative analyses
of traditional agricultural policies and environmental and resource policies lie on similar
foundations. From a normative point of view both should be introduced if they improve
the social state, where the measure of such improvement is a function of the welfare of
all individuals. The effects of any such policies are usually best discussed in reference to
welfare (policy outcome) space (Antle, 1991; Gardner, 1991). Unfortunately, since
environmental quality is a public good which is usually not traded in markets, mapping
environmental policy into policy outcome space is often a good deal trickier than is
mapping commodity policy into policy outcome space. Nevertheless, the theoretical
advantages of discussing the effects of policy by examining policy outcome (welfare)
space hold for the study of environmental policy just as well as they hold for the study of
commodity policy.

References
Adamowicz, W. L., Fletcher, J. J. and Graham-Tomasi, T. (1989a). Functional Form and the Statistical Properties
of Welfare Measures, Amm'canJournal of Ap'cultural Economicc, 71, 414-4421.
Adamowicz, W. L., Graham-Tomasi, T. and Fletcher, J. J. (1989b). Inequality Constmined Estimation of
Consumer Surplus, CanadianJournal OjAgn'cuhral Economics, 37,407420.
Alston, J. M. and Carter, C. (1991). Causes and Consequences of Farm Policy, Conlemporary Policy Issues, 9, 107-
120.
Alston, J. M., Carter, C. A. and Smith, V.H. (1993). Rationalising Agricultural Export Subsidies. Ammicanjournal
of Ap'mltural Economics, 75, 1000-1009.
Alston, J. M.,Chalfant, J. A. and Piggott, N. E. (1998). Advertising and Consumer Welfare. Unpublished
Manuscript, Department of Agricultural and Resource Economics, University of California, Davis.
Aston, J. M. and Hurd, B. H. (1990). Some Neglected Social Costs of Government Spending in Farm
Programmes, Amm'canJournal of Agn'cultural Economics, 72, 149-156.
Alston, J. M. and Larson, D. M. (1993). Hicksian vs. Marshallian Welfare Measures W h y Do We Do What We Do?,
Ametican Journal of Agricultural Economics, 75, 764-769.
Alston,J. M. and Spriggs, J. (1998). Endogenous Policy and Supply Management in a PostGATT World,
CanadianJournal of Economics, 31, 220-239.
Antle, J. M. (1991). Farm Policy Reform and the Environment, AmericanJournal ofA~'eiculturalEconomics,73,917-
921.
Babcock, B. A, Carter, C. A. and Schmitz, A. (1990). The Political Economy of U. S. Wheat Legislation, Economic
Inquiry,28,335353.
Becker, G . S. (1983). A Theory of Competition Among Pressure Groups for Political Influence, QuarfedyJournal
of Economics, 98,371400.
Blandford, D. and Dewbre, J. (1994). Structural Adjustment and Learning to Live without Subsidies in OECD
Countries, Amm'canJournal of Agncultural Economicz, 76, 1047-1052.
Bockstael. N. and Strand, I. (1986). The Effect of Common Sources of Regression Error on Benefit Estimates,
Land E m a r i a , 63. 11-20.
Briinnlund, R. and Krisuom, B. (1996). Welfare Measurement in Single and Multimarket Models: Theory and
Application, Amen'canJournal ofAgn'culturn1Economics, 78, 157-165.
530 D. S.Bullock, K SalhoferandJ. Kola

Brooks, J. (1996). Agricultural Policies in OECD Countries: What Can We Learn from Political Economy
Models, Journ@l of Agricultural Economics. 47, 366389.
Bullock, D. S. (1991). Pareto Optimal Income Redisuibution and Political Preference Functions: An
Application to EC Common Agricultural Policy. Unpublished Manuscript, Department of Agncultural
Economics, University of Illinois.
Bullock, D. S. (1992). Redistributing Income Back to European Community Consumers and Taxpayers through
the Common Agricultural Policy, AmericanJournal of Agricultural Economics, 74, 59-67.
Bullock, D. S. (1993). Welfare Implications of Equilibrium Supply and Demand Curves in an Open Economy,
AmmicanJmrnal of Agn'culfural Economics, 75, 52-58.
Bullock, D. S. (1994). In Search of Rational Government: What Political Preference Function Studies Measure
and Assume, American J o u m d OfAgricultural Eunzmnics, 76,347-361.
Bullock, D. S. (1995). Are Government Transfers EfXcient? An Alternative Test of the Efficient Redistribution
Hypothesis, Journal of Political Economy, 103, 12361274.
Bullock, D. S. (1996). Pareto Optimal Income Redistribution and Political Preference Functions: An
Application to EC Common Agricultural Policy, in Antle, J. and Sumner, D. (eds.), D. Cak Johnson on
AgticuUurc, Vol. 2: Essays on Agn'culfurcrlPolicy in Honor o/D. Cab Johnson, Chicago, 11: University of Chicago
Press.
Bullock. D. S. and Salhofer. K (1995). Is Government Efficient? An Illustration from US Agricultural Policy, Die
Bo&nhultur; 46, 379-391.
Bullock, D. S . and Salhofer, K (1998a). Measuring the Social Costs of Suboptimal Combination of Policy
Instruments - A General Framework and an Example, Agriculfural Economics, 18, 249-259.
Bullock, D. S. and Salhofer, K (1998b). A Note on the Efficiency of Income Redistribution with Simple and
Combined Policies, Agncultural and Rcsourcc Economics Mew, 27, 266269.
Chavas, J.-P. (1994). Equity Considerations in Economic and Policy Analysis, Amnican Journal of Ap'cultural
E C O ~ O W I76,
~ C1022-1033.
S,
Chen, H. 2. and Cosslett, S. R (1998). Environmental Quality Preferences and the Benefit Estimation in
Multinomial Probit Models: A Simulation Approach, Amnican Journal of Agricultural Econaics, 80,512-520.
Chipman, J. S. and Moore, J. C. (1980). Compensating Variation, Consumer's Surplus, and Welfare, American
Economic Review, 70, 933-949.
Christensen, L. R., Jorgenson, D. W. and Lau, L. J. (1971). Conjugate Duality and the Transcendental
Logarithmic Production Function, Econmfnca. 39, 255-256.
Constantine, J. H.. Alston, J. M. and Smith, V. H. (1994). Economic Impacts of the California Onevariety Cotton
Law,Journal ofPolitical Economy, 102,951-974.
Comes, R. (1992). Dualilg and Modern Ewnomics, Cambridge: Cambridge University Press.
Cramer, G. L., Wailes, E. J., Gardner, B. L. and Lin, W. (1990). Regulation in the U.S. Rice Industry, 195689,
Ama'can Journal ofAgncultura1 Economics, 72, 10561065,
Dardis, R. (1967a). The Welfare Cost of Grain Protection in the United Kingdom, Journal ofFarm Economics, 49,
597-609.
Dardis, R (1967b). Intermediate Goods and the Gain from Trade, RNinu ofEmom'cs and Slalisrics, 49,502-509.
Dardis, R and Dennison, J. (1969). The Welfare Cost of Alternative Methods of Protecting Raw Wool in the
United States, AmmicanJournal of Agn'cultural Economics, 51, 303-319.
Davis, G. C. and Espinoza. M. C. (1998). A Unlfied Approach to Sensitivity Analysis in Equilibrium Displacement
Models, American Journal of Agn'culfuralEconomics, 80, 868-879.
Deaton, k and Muellbauer, J. (1980). An Almost Ideal Demand System, American Economic RNinu, 70, 341-367.
de Gorter, H. and Meilke, K D. (1989). Efficiency of Alternative Policies for the EC's Common Agricultural
Policy, American Journal of Agriculhcral EcoMmics, 71, 592-603.
de Gorter, H., Nielson, D, J. and Rausser G. C. (1992). Productive and Predatory Public Policies: Research
Expenditures and Producer Subsidies in Agriculture, Amerkan Journal of Agrinclhcsol Eammnics, 74, 27-37.
de Gorter, H. and Swinnen, J. F. M. (1994). The Economic Polity of Farm Policy,Journal ofAgriculfura1Economics,
45, 312-326.
 Normative Analysis of Agricultural Policy:Framework and Reyiew 531

Efron, B. (1979). Bootstrap Method: .bother Look 3t the Jackknife, AnnaLc of Sfahtics, 7, 1-26.
Efron, B. and Tibshirani. R J. (1993). An Infrodudon to fhc Boorstrap, New York, Ny: Chapman & Hall.
Fraser, R. W. (1992). The Welfare Effects of Deregulating Producer Prices, Amm'can Journal of Agn'culfural
Economics, 74, 21-26.
Freedman, D. and Peters, S. (1984). Bootstrapping a Regression Equation: Some Empirical Results, Journal of
fhe A m d c a n Statistical Association, 79,97-106.
Gallagher, P. (1988). The Grain Sector of the European Community: Policy Formation, and Implications for
Trade, A m ' c a n Journal oJAgticuuUra1Economics, 70,767-778.
Gardner, B. L. (1983). Efficient Redistribution through Commodity Markets, Amen'can Journal ofAgn'culfura1
Economics, 65, 225-234.
Gardner, B. L. ( 1985). Export Subsidies are Still Irrational, Ap'culfuralEconomics Research, 37(1), 17-19.
Gardner, B. L. (1987). TheEconaUu ofAgricultura1 Policies, New York, Ny: McGraw-Hill.
Gardner, B. L. (1988). Export Policy, Deficiency Payments, and a Consumption Tax, Agriculfural Economiu
Research, 40(1), 3841.
Gardner, B. L. (1991). Redistribution of Income Through Commodity and Resource Policies, in Just, R. E. and
Bockslael, N. (eds.), Commodity and Resourcs Policies in Agricultural S y s f m ,Berlin: Springer Verlag.
Gardner, B. L. (1992). A i m Su@Orts and Optimal Spending on Agricultural Research, Working Paper No. 92-20,
University of Maryland, Department of Agricultural and Research Economics.
Gardner. B. L. (1995). Rationalising Agricultural Export Subsidies: Comment, Amm'can Journal of Agn'culfural
Economics, 77, 205-208.
Gisser. M. (1993).Price Support, Acreage Controls, and Efficient Redistribution, Journal of PolificalEconomy, 101,
584-61 1.
Graham-Tomasi. T., Adamowicz, W. L. and Fletcher, J. J. (1990). Errors of Truncation in Approximation to
Expected Consumer Surplus, Land Economics, 66,50-55.
Griffiths, W. E. (1999). Estimating Consumer Surplus: Comments on "Using Simulation Methods for Bayesian
Econometric Models: Inference Development and Communication," Econometric h i e u s , 18, 75-87.
Griffiirtis, W. E. and Zhao, X. (1999). A Unified Approach to Sensitivity Analysis in Equilibrium Displacement
Models: Comment, Amm'canJournal OjAgricultural Economics, forthcoming.
Griliches. 2. (1958). Research Costs and Social Returns: Hybrid Corn and Related Innovations, Journal of
Political Ewnomy, 66,419-431.
Guyomard, H. and Mahe, L P (1994). Is a Productinn Quota Pareto Superior to Price Support Only?, European
Mew of Agricultural Economics, 21, 31-56.
Harberger, A. C. (1954). Monopoly and Resource Allocation, Amm'can Economic h i m , 44,77-92,
Harvey, D. and Hall, J (1990). PSEs. Producer Benefits and Transfer Efficiencies of the CAP and Alternatives,
in van der Noort, P. C. (ed.), Costr and BneJfs ojAgn'culfura1Policies andPmjecfs,Kiel: Wissenschaftsverlag
Vauk.
Hausmann, J. (1981). Exact Consumer's Surplus and Deadweight Losses, American Economic h i m , 71,663-676.
Helms, L. J. (1985). Errors in the Numerical Assessment of the Benefits of Price Stahilisation, AmericnnJournal
of Agricultural Economics, 67, 93100.
Hotelling, H. (1938). The General Welfare in Relation to Problems of Taxation and of Railway and Utility Rates,
Econometn'ca, 6, 242-269.
Innes, R. D.(1990).Uncertainty, Incomplete Markets and Government Farm Programs, S m t k n Journal of
Economics, 57, 47-65.
Innes, R. D. and Rausser G. C. (1989). Incomplete Markets and Government Agricultural Policy, Amen'can,
Journal oJAgricultura1 Economics, 71, 915-931.
H. (1993). Policy Choice under Impetjieci Compefifionw'fhan Applicafion fo fhe Finnish Sugar Markef,Pellervo
Isosaan,
Economic Research Institute (F"Publications
) 14, Espoo.
Jeong, K-S., Bullock, D. S.,and Garcia, P. (1999a). Testing the Efficient Redistribution Hypothesis: An
Application to Japanese Beef Policy, Amm'canJournal of Agricultural Economics, 81, 408423
532 D.S. Bullock, K Salhofm and J. Koh

Jeong, KS.,Garcia, P., and Bullock, D. S. (1999b).A Statistical Method of Welfare Analysis Applied to Japanese
Beef Policy Liberalisation, Journal of Policy Modelling, forthcoming.
Josling, T. 11969).A Formal Approach to Agricultural Policy,Journal ofAgricultural Economics, 20, 175-195.
Josling, T. (1974).Agricultural Policies in Developed Countries: A Review,Journal of Agrincltural ofEconomics, 25,
229-264.
Just, R. E. (1984).Automatic Adjustment Rules For Agn'cultural Poliq Confrob,AEI Occasional Papers, Washington
DC: American Enterprise Institute for Public Policy Research.
JUSL,R E., and Hueth, D. L. (1979).Multimarket Welfare Measurement, Anoican EconomicRevim, 69,947-954.
Just, R. E., Hueth, D. L. and Schmitz, A. (1982).Applied Welfare Economics and Public Polig, Englewood Cliffs:
Prentice-Hall.
Just, R. E., Lutz, E., Schmitz, A. and Turnovsky S. (1977).The Distribution of Welfare Gains from International
Price Stabilization under Distortions, Amm'canJuurnal of Agricultural Economics, 59, 652-661.
Just, R E., Schmitz, A. and Zilberman, D. (1979).Price Controls and Optimal Export Policies under Alternative
Market Structures, American Economic Review, 69, 706714.
Kangiannis. G., Katranidis, S. and Velentzas, K (1997).Redistribution and the CAP Efficiency in the Greek
Cotton Industry, Indian Journal of Agn'cullural Economics, 52, 782-790.
Kealy, M. and Bishop, R. (1986).Theoretical and Empirical Specification in Travel Cost Demand Studies,
AmnicanJournal of Agn'culhtral Economics, 69,660-667.
Kling, C. L. (1991).Estimating the Precision of Welfare Measures, Journal of Environmental Economics and
Illanagement, 21, 244259.
Kola,J. ( 1991 ). Production Control inFinnish Agnculture:Determinants of Control Poliq and Quantihztiue and Econ~mic
EfJcimq of Dairy Restrictions, Helsinki: Agricultural Economics Research Institute (MTTL) Publications 64.
Koh, J. (1993). Efficiency of Supply Control Programmes in Income Redistribution, European Reuiew of
Agn'cullural Economics. 20, 183198.
Konandreas, P. A. and Schmitz, A. (1978).Welfare Implications of Grain Price Stabilization: Some Empirical
Evidence for the United States, Amm'mnJournal of AgnkulfuralEconomics, 60, 74-84.
Krinsky, I. and Robb, A. (1986).On Approximating the Statistical Properties of Elasticities, Review ofEconomiu
and Statistics, 68, 715-719.
Krinsky, [. and Robb,A. (1990).On Approximating the Statistical Properties of Elasticities: A Correction, Review
of Economics and Statistics, 72, 189-190.
Krinsky, I. and Rohb,A. (1991).Three Methods for Calculating Statistical Properties for Elasticities, Empirical
Economics, 16,1-11.
Larson, D. M. (1988). Exact Welfare Measurement for Producers Under Uncertainty, American Jarrnal of
Agricullural Economics, 70,597-603.
Leu. G.-J.M., Schmitz, A. and Knutson, R. D. (1987).Gains and Losses of Sugar Program Policy Options.
A d c a nJournal of Agricultural Economics, 69,591-602.
Lichtenberg, E. and Zilberman, D. (1986).The Welfare Economics of Price Supports in U.S. Agriculture,
Amm'can Economic Review, 76, 1135-1141.
Lothe, S. E. and Garcia, R. J. (1996).Evaluation AltmMtive Policiesfor the Norwegian Dairy Sectm Paper Presented
at the VIII Congress of the EAAE, Edinburgh, 3-7September 1996.
Maier. L. (1993a).The Costs and Benefits of US Agnnrltural Policies with Imperfect Competition in Food Manufacturing,
New York, Ny: Garland Publishing.
Maier, L. (1993b).The Relative Transfzr Efjicimcy of Six Ap'cultural Support Policies for a Small Exporting Country,
Discussion Paper 25-W-93, Department of Economics, Politics, and Law, University of Agricultural Sciences
Vienna.
Marshall, A. (1920).Principles of Economics, 8th edition, London Macmillan
Martin, W. J. and Alston, J. M. (1994).A Dual Approach to Evaluation Research Benefits in the Presence of
Trade Distortions, Amen'canJournal of Agn'cultural Economiu, 76, 2635.
McCorriston. S.and Sheldon, 1. M: (1991).Government in Imperfectly Competitive Agricultural Input Markets,
American Journal of Agn'cultural Economics, 73,621-632.
 Normative Analysis of Agn'cultural Policy: Framework and Review 533

Swinnen,J. F. M and de Corter, H. (1998). Endogenous Commodity Policies and the Social Benefits from Public
Research Expenditures, AmericanJournal of Agncultural Economics, 80, 107-115.
McCorriston, S. and Sheldon I. M. (1994). Selling Import Quota Licenses: The US Cheese Case, Ammiran
Journal oJAgricullura1Economics, 76,818-827.
McIknzie, G. W. (1983). Measuring Economic Weyare; New Metlrorls, Cambridge: Cambridge University Press.
Moschini. G., and P. Sckokai (1994). Efficiency of Decoupled Farm Programmes Under Distortionary Taxation,
Ammican Journal U/AgrinrlturalEconomics, 76,362-370.
Murphy, J. A., Furtan, W. H. and Schmitz, A. (1993). The Gains from Agricultural Research under Distorted
Trade, Journal of Public Economics, 52, 161-172.
Nerlove, M.(1958). The Dynamic of Supply, Baltimore, OH: Johns Hopkins Press.
Ng. Y-K (1979). Welfare Economics: Introduction and Development of Basic Concepts, London Macmillan.
Ng, Y-K (1981).Welfarism, a Defense Against Sen's Attack, TheEconomicJournal, 91, 527-530.
OECD (1995). Adjustment in OECD Agriculture: Issues and Policy Responses, Paris.
Otsuka, K and Hayami, Y (1985).Goals and Consequences of Rice Policy in Japan, 1965-80, AmericanJournal of
Agnnrltural Economics, 67, 529-538.
Paarlberg, P. L. (1984). When Are Export Subsidies Rational, Agncultural Economics Research, 36(1), 1-7
Peters, G. H. (1995). Agticultural Economics (Internafional Library of Critical LVrifings in Economics), Aldershot :
Edward Elgar.
Pererson. E. B. and Connor, J. M. (1995). A Comparison of Oligopoly Welfare Loss Estimates for U.S. Food
Manufacturing, A m k m Journal oJAgricultura1 Emumics, 77, 300-308.
Rausser, G. C. (1982). Political Economic Markets: PERTs and PESTS in Food and Agriculture, American Journal
of Agn'cullural Ecwtmi~ics,64, 822-833.
Salhofer, K (1996). Efficient Support Policy for a Small Country Using Optimal Combined Policies, Agncuhml
Economics, 13, 191-199.
Salhofer, K (1997). Efficiency of Income Redisfn'httion Through A6mcuhral Policy: A WelJare Economic Analysis.
Frankfurt: Peter Lang: European University Studies.
Salhofer, K (1999). Distributive Leakages fmm Ap'cultural Supporf. Some Empirical Euidence l;om Austria, Paper
Presented at the 1999 Annual Meetings of the Southern Agricultural Economics Association in Memphis,
Tennessee.
Sarwar, G. and Fox, G. (1992). An Evaluation of the Redistributive Efficiency of Alternative Crow Benefit
Payment Policies in Western Canada, Review of AgnculfuralEconomics, 14, 187-204.
Schmitz, A. and Vercammen, J. (1995). Efficiency of Farm Programs and Their Trade-Distorting Effects,
Rausser, G. C. (ed.), GATT-Negotiations and the Political Economy of Policy Reform, Berlin: Springer
Verlag.
Sen, A. K. (1977). On Weights and Measures: Informational Constraints in Social Welfare Analysis, Economdrica,
45, 1539-1572.
Sen, A. K (1979). Personal Ut es and Public Judgments: Or What's Wrong with Welfare Economics?, The
EconomicJournal, 89, 537-558.
Sen. A. K (1981). A Reply to Welfarism, a Defense Against Sen's Attack, The EconomicJournal, 91, 531-535
Sen, A. K (1987). On Ethrcs and Economics,Oxford: Blackwell.
Sen, A. K (1992). Inequalify Reexamined, New York, NY Russell Sage
Sexton, R. J. (1994a). A Survey of Noncooperative Game Theory with Reference to Agricultural Markets: Part
1. Theoretical Concepts, Review ojkfavketing and Agn'culturalEconomics, 6 2 , 11-28.
Sexton, R. J. (1994b). A Survey of Noncooperative Game Theory with Reference to Agricultural Markets: Part
2. Potential Applications to Agriculture, Review of Markeling and Agncultural Economics, 62, 183200.
Slesnick, D. T. (1998). Empirical Approaches to the Measurement of Welfare, Journal oJEconomic Liferafure,36,
21082165.
Swinnen, J. F. M. (1997). Income Distribution and the Efficiency of Government Transfers, Unpublished
Manuscript, Department of Agricultural Economics, Katholieke Universiteit Leuven.
534 D.S. Bullock, X SalhofeandJ. Koh
Swinnen, J. F. M. and van der Zee, F. A. (1993).The Political Economy of Agricultural Policies: A Survey,
European Review ofAp’cullUra1 Ecamnics, 20,261-290.
Theil, H. (1958).E c o d c F o r e c a r t and Pohq, 2nd Revised Edition. Amsterdam: North-Holland.
Thomson, K J. and Harvey, D. R. (1981).The Efficiency of the Common Agricultural Policy, European Revlew o/
Agricultural Economics,8, 57-83.
Thurman, W. N. (1993).The Welfare Significance and Nonsignificance of General Equilibrium Demand and
Supply Curves, AcblicFinance Quarterly, 21,449469.
Thurman, W. N. and Wohlgenant, M.K. (1989).Consistent Estimation of General Equilibrium Welfare Effects,
Ampn’canJournal ofAp’cultura1Eummnics, 71,1041-1045.
Tinbergen, J. (1952).On the Themy ofEconomic Policy, Amsterdam: North-Holland.
Trernblay, C. H. and Trernblay V. J. (1995).The Impact of Cigarette Advertising on Consumer Surplus, Profit.,
and Social Welfare, Confempary Economic Policy, 13, 113-124.
Tuomala, M. (1990).Optimal Income Taxation and Rcdistn’bution, New York,Ny: Oxford University Press.
Varian,H. R. (1992).Microeconomic Analysis, New York, NY Norton.
Wallace, T. D. (1962),Measures of Social Cosu of Agricultural Programs,Journal ofFamEconomicc, 44,580-594.
Williams, J. C. and Wright, B. D. (1991).Stmnge and Commodity Markets,New York, Ny: Cambridge University
Press.
Winters, A. L. (1987).The Political Economy of the Agricultural Policy of Industrial Countries, European Review
of Agricultural Economics, 14,285-304.
Wong, C . M. (1989).Welfare Implications of Price Stabilisation with Monopolistic Trade, Ammican Journal oJ
Agncuilural Economics, 71, 4354.

Wright, B. D. (1979).The Effects of Ideal Price Stabilization: A Welfare Analysis under Rational Behaviour.
Journal of P o l i f i d Economy, 87, 1011-1033.
Wright, B. D. (1993).Dynamic Perspectives on Agricultural Policy Issues, American Journal oJ Agricultural
,

Economics,75, lllP1125.
Zhao, X., Griffiths, W. E., Griffith, G. R. and Mullen, J. D. (1998).Probability Distributions for Economic
Surplus Changes: The Case of Technical Change in the Australian Wool Industry, Unpublished
Manuscript, Department of Econometrics, University of New England, Armidale.

APPENDIX
Prove: 1 Let x* be a unique solution to

maxhi(x, b) s.t. hj(x, b) 2 hj(xA, b) j = 1, ..., i - 1, i t 1, ..., n.

XE x

Then x* is Pareto efficient.

Proof: Suppose not. Then x* is a unique solution to (l), but x* is Pareto inefficient: Since x+ is Pareto
inefficient, there exists some xB in X such that h,(xB, b) 2 hl(x*, b), hz(xB, b) 1 hP(x*,b), ..., h,(xB, b) 1 h,(x*,
b), where at least one of these n inequalities holds strictly. Since x* solves (1). we know that hi(x*,b) 1 hi(xA.
b), hj(x*, b) 2 hj(xA, b), j = 1, ..., i - 1, i + 1, ..., n. Therefore, there are two possibilities:

Possibility 1: h;(xB, b) > h;(x*, b) 1 hi(XA, b)

h,(xB, b) > hj(x*, b) 2 hi(x*, b) j = 1, ..., i - 1, i t 1, ..., n.
Note that the proof is for a maximisation problem with inequality constraints and hence includes the
formulation by Alston et al. (1993),which has equality constraints.
 N m t i v e Analysis of Ap'cultural Policy: Framauark and h i e w 535

(That is, the inequality holds strictly for i). But Possibility 1 implies that x* is not a solution to ( I ) , which
contradicts our original assumption.

Possibility 2: hi(xB, b) = hj(x*, b) 2 hi(X*, b)

-
hj(xB, b) 2 hj(x*, b) b hi(xA, b) j = 1, .... i 1, i + 1, ...,n.

(That is, the inequality does not hold strictly for i). But Possibility 2 implies that x* solves ( I ) , so does xB, which
contradicts our assumption that x* is a unique solution to ( I ) . Since both Possibility 1 and Possibility 2 lead to
contradictions of our original assumptions, the proof is complete.