Is
Macroeconomics for Real?*
by
Kevin D. Hoover
Department of Economics
University of California
Davis, California 95616-8578
Tel. (530) 752-2129
Fax (530) 752-9382
E-mail kdhoover@ucdavis.edu
First Draft, 25 July 1994
Revised, 22 June 1999
*I am grateful to Uskali Mäki and Thomas Mayer for valuable comments on the first version of this paper, and to D. Wade Hands for some pointers on the literature on supervenience. This research was supported under National Science Foundation grant No. 9311930.
Is macroeconomics for real? KEVIN D. HOOVER
Despite lip-service to methodological individualism, the true basis for the program of microfoundations for macroeconomics is shown to be a commitment to ontological individualism and a denial that macroeconomic aggregates exist externally and objectively. This article shows that macroeconomic aggregates are not reducible, even in principle, to the microeconomic entities that underlie them. The relationship between microeconomic and macroeconomic entities is one of supervenience - although a supervenience that does not underwrite reductionism. The ontological reality of macroeconomic aggregates is supported by an appeal to Ian Hacking’s argument from manipulability. It is shown that macroeconomic aggregates can be used as instruments of manipulation in contexts other than the ones that define them theoretically. A second argument appeals to the fact that macroeconomic aggregates are fundamental in idealized economic models and that the success of idealizations depends on their isolating the ontological essence of the subject of the idealization.
Is Macroeconomics for Real?
All are keeping a sharp look-out in front, but none
suspects that the danger may be creeping up from behind. This shows how real the island was.
J.M.
Barrie Peter Pan
Children
are often thought to be peculiarly honest - witness the story of "The
Emperor's New Clothes." My title
comes from a group of my academic children:
first-year graduate students. I
teach a mandatory class in macroeconomic theory to graduate students in both an
economics department and an agricultural economics department. The students in agricultural economics are
typically more interested in crop patterns or natural resources - relentlessly
microeconomic topics - than in unemployment, GDP growth, or interest
rates. Each year at least one student,
who I assume comes from the agricultural economics department, writes on the
anonymous class evaluation something like this: "If macroeconomics were real economics - which it is not! -
this would have been a good class."
What is one to say to the honest and piercing doubts of an academic
child?
The
idea that macroeconomics stands in need of a microfoundational base is a
commonplace among economists. I shall
argue that what motivates this belief are principally ontological concerns,
naively, but pointedly expressed, in my students’ questions about the reality
of macroeconomics. I shall argue that
ontological reduction of macroeconomics to microeconomics is untenable. Thus, while the program of microfoundations
may illuminate macroeconomics in various ways, it cannot suceed in its goal of
replacing macroeconomics.
To
begin at the beginning, it might help to define the key terms. "Macroeconomics" is sometimes
thought to be the economics of broad aggregates, and "microeconomics"
the economics of individual economic actions.
Although he did not use the terms "microeconomics" and "macroeconomics,"
John Maynard Keynes (1936, pp. 292-3) drew a related distinction: microeconomics is the theory of the
individual industry or firm; macroeconomics is the theory of output and
employment as a whole. As Maarten
Janssen (1993, ch. 1) shows, these alternative definitions cut in somewhat
different ways. They are, however,
similar enough for present purposes, since any quantification of output and
employment as a whole, is bound to
involve broad aggregates.
Macroeconomics is thus that area of economics that treats of GDP,
unemployment, interest rates, the flow of financial resources, exchange rates,
and so forth.
Uskali
Mäki (1994) offers a careful discussion of realism in relation to economics
that might help to define the term "real" in the title. Mäki distinguishes between ontological
realism, which raises questions about what there is, and semantic realism,
which raises questions about the connection between language and what there
is. Semantic realism can be analyzed
further, but the central claim of this essay is ontological: macroeconomic
aggregates exist. More precisely,
and again using Mäki's terminology, I will argue that macroeconomic aggregates
exist externally (i.e., independently
of any individual human mind) and objectively (i.e., unconstituted by the
representations of macroeconomic theory).
This claim, if it can be sustained, undermines one of the central
rationales for the program of microfoundations for macroeconomics that since at
least the mid-1940s has dominated thinking among economists.
I. The Program of Microfoundations
Before
this century, the most common definition of economics was epitomized by Alfred
Marshall (1920, p. 1): "[A] study
of mankind in the ordinary business of life; [economics] examines that part of
individual and social action which is most closely connected with the
attainment and with the use of the material requisities of
well-being." This definition,
which is reasonably hospitable to macroeconomics, has now largely been
supplanted by that of Lionel Robbins (1935, p. 16): "Economics is the science which studies human behaviour as a
relationship between ends and scarce means which have alternative
uses." On Robbins's definition,
economics must be fundamentally about the individual.
Modern
macroeconomics developed in the wake of Keynes's General Theory of Employment Interest and Money (1936). Typical elements of Keynes's analysis were
the consumption function, which related aggregate consumption to aggregate
national income, the investment function, which related aggregate investment to
the general rate of interest, and the liquidity preference function, which
related the aggregate stock of money to aggregate national income and the
general rate of interest. It is easy to
understand that in a profession committed to Robbin's definition of economics,
such aggregate relationships were at best rather unappealing way-stations on
the path to an individualist economics.
The program of microfoundations, as it has developed over the past fifty
years, aims to explain all macroeconomic properties of the economy - in principle, at least - by reference to
the behavior of rational economic actors such as postulated by microeconomics.[1]
Approval
of the program of microfoundations is almost universal among economists. Those economists who have reflected on the
matter at all deeply typically associate microfoundations with methodological
individualism (e.g., Janssen 1993, pp. 26-29 passim ; Boland 1979, chs. 2 and 5)[2]. Blaug (1992, p. 44) defines methodological
individualism as the principle that ". . . asserts that explanations
of social, political, or economic phenomena can only be regarded as adequate if
they run in terms of the beliefs, attitudes, and decisions of
individuals."
Methodological
individualism is a doctrine about explanation.
Despite lip-service to it, it is not widely practiced by
economists. The reason is what I have
elsewhere labelled the "Cournot problem" after its lucid, early
formulation by Augustine Cournot (1838/1927, p. 127), the 19th-century mathematician
and economist: there are too many
individuals (firms and consumers) and too many goods to be handled by direct
modelling.[3] Mark Blaug (1992, p. 46) observes that few explanations of
macroeconomic phenomena have been successfully reduced to their
microfoundations, so that insistence on microfoundations would eliminate
explanations of macroeconomic phenomena tout
court. Even Lucas (1987, pp.
107-108), an important advocate of the program of microfoundations, holds up
only the hope of the elimination of
distinction between microeconomics and macroeconomics.
The
commitment of economists to methodological individualism is thus not grounded
in successful applications. Rather it
appears to be based on an instinctive belief in ontological individualism:
the doctrine that all that exists fundamentally for the economy are
individual economic actors. Lucas and
his fellow new classical economists have promoted representative-agent models,
a class of models in which the mathematical methods of microeconomic optimal choice
are applied to a single individual who takes national income as his budget
constraint and whose choices are taken to represent the aggregate choices of
the economy, because they appear to achieve the reduction of macroeconomics to
microeconomics as required by the program of microfoundations for
macroeconomics. A. P. Kirman (1992)
severely criticizes the representative-agent model, not because it aspires to
methodological individualism, but because it fails to fulfill the necessary
conditions for perfect aggregation, so that the representative agent in the
models fails to represent actual individuals successfully. Methodological individualism remains the
goal. Similarly, David Levy (1985)
argues that complete methodological individualism is impossible, because, given
imperfect information, individual economic actors must make reference to
collective entitites as part of their own decision-making processes. Nevertheless, Levy (1985, p. 107)
writes: "These collectives have no
real existence but are simply the product of theories." While defending macroeconomics against the
strong claims of the program of microfoundations, Blaug (1992, p. 45)
nevertheless writes: ". . .
ontological individualism is trivially true. . ."
It
is important to understand that there are some senses in which neither the
methodological nor the ontological individualist denies the existences of
aggregates, collectives, or wholes. No
one denies that GDP calculations can be made and reported, or even that GDP may
have some locally stable relationship to unemployment or average interest rates
or some other aggregate. Similarly, no
one denies the existence of social organizations such as governments or firms
(in the sense that talk of governments and firms conveys meaning). What is typically denied is that such aggregates
or organizations are among the fundamental units from which economic reality is
constructed.
Hayek
(1979, ch. 4) argues that such entities are secondary, and that the role of a
social science is "compositive" - that is, that it must explain these
entities as arising from the fundamental individual components.[4] Hayek (1979, ch. 6) denies that the wholes that social science
explains through compositive methods are subject to scientific laws. He holds up the attempt to connect them through
laws as an example of Whitehead's "fallacy of misplaced
concreteness." He writes: ". . . the wholes about which we speak
exist only if, and to the extent to which, the theory is correct which we have
formed about the connection of the parts which they imply, and which we can
explicitly state only in the form of a model built from those
relationships" (Hayek 1979, p. 98).
Hayek thus argues that aggregates exist, but derivatively rather than
fundamentally, and that in Mäki's terminology they do not exist objectively
(i.e., unconstituted by the representations of theory).[5] Still even Hayek does not endorse
practicable methodological individualism, stressing the importance of a
reduction to microfoundations in
principle and himself citing the Cournot problem (Hayek 1979, pp. 74-75,
esp. fn. 8).
II. Is Macroeconomics Ontologically Problematic?
One might concede the main point of the last section - namely, that the
drive for microfoundations is driven by ontological individualism - yet not
believe that any interesting metaphysical issue is involved, because the
ontology of economics is too well understood by common sense to pose any
serious puzzle. Mäki (1994, p. 11), for
example, contrasts folk economics
with scientific economics, arguing
that scientific economics merely presents modifications of the "'ontic
furniture' of the general folk views of 'man' and society." He lists some types of possible
modifications: selection, abstraction,
idealization, exaggeration, projection, aggregation. But he maintains that none of these modifications or combinations
thereof ". . . accomplishes a major departure from the ontic furniture of
the ordinary realm. No new kinds of entities or properties are
introduced" (Mäki 1994, p. 12).
Mäki illustrates his point with a discussion of the business firm in
standard neoclassical analysis, and concludes ". . . that folk economics
and neoclassical economics have real business firms as their shared referent
even though they represent these firms differently."
Mäki's
case for the "ontological commonsense realism" of economics in which
the ontic furniture poses no special challenges to the understanding is more
persuasive for some parts of economics than others. I want to argue that for some important macroeconomic aggregates,
it is not particularly cogent, and that some such aggregates do not share a
referent with folk notions. The case
can be illustrated with reference to the related notions of "real
GDP" (or "real income" or "real output," these terms
being used almost interchangeably) and the "general price level."
On
any interpretation, macroeconomics takes a larger view of the economy and deals
with aggregates, which are, in turn, constructed from characteristics of
individual economic actors. It is
helpful to distinguish two types of aggregates.
First
are what we might call natural
aggregates: simple sums or
averages. Examples of natural
aggregates are the level of total employment or the average rate of interest on
six-month commercial paper. I call
these natural aggregates because they are measured in the same units (i.e.,
they have the same dimensionality) as the individual components that they
comprise and, therefore, preserve a close analogy with those individual
components. Employment, for instance,
is measured by the number of workers or the number of man/hours at both the
level of a single individual and in aggregate.
The rate of interest on a bond and the average rate of interest on a
group of similar maturity bonds are both expressed as a percentage yield per
unit time.
A
second type of aggregate, what we might call synthetic aggregates, are important for macroeconomics. I call these synthetic because they are
fabricated out of components in a way that alters the structure of the
components, so that they are dimensionally distinct from the components and so
that there is no close analogy (despite there sometimes sharing a common name)
with the components. The nature of the
synthesis is well illustrated by the general price level. The notion of a general price level aims to
capture a pre-scientific insight:
"a dollar just ain't what it used to be!"; "when I was a
lad a penny would buy what a quarter does these days." To capture this insight, one would like to
have some notion of the average level of prices. A simple average will not of course work: (10¢/orange + 20¢/apple + $27,948/Volvo
stationwagon)/3 does not convey any useful information. One cannot add apples and Volvos, as they
say.
It
might be argued that any sort of an average is altogether the wrong way to
start. What is really wanted is an
estimate of the price of money itself, and not the average price of goods. The price of money would, like the price of
oranges, be a single number. The
general price level could be defined to be its inverse (1/pm). Since relative prices of goods change
because of changes in the conditions of demand and supply, there would be, at
best, a rough proportionality between individual prices and the general price
level. Indeed, it would permit one to
isolate which changes in individual prices were the result of "real"
factors and which of monetary factors.
This
approach, however, does not do justice to the pre-scientific insight, for it
does not provide us with a concept or a measurement of prices that is
independent of highly particular and highly inadequate theoretical models. To see this consider, how one would actually
determine the value of pm?
One might, for instance, write down a complete Walrasian general
equilibrium model in which commodities were expressed in natural quantities,
prices in terms of money, and all assets denominated in money were valued using
pm. Aside from the
impracticality of formulating and solving such a model for an actual economy
(the Cournot problem again), it is well-known that pm might not be
determinable in such a model, or if it is, might not be unique (Hahn 1965;
Samuelson and Sato 1984; see Hoover 1988, ch. 5, section 1 for a simple
exposition). The essence of the problem
is that the real quantity of money (i.e., the useful services that it
provides), unlike the real quantities of apples or Volvos, depends
fundamentally on the price of other goods.
In adjusting the prices of various goods, including money, unique
convergence may not be possible, because each time prices adjust to remove an
excess demand or supply, the quantity of money changes - possibly in a
discontinuous manner - which can increase rather than diminish some of the
excess demands or supplies.
This
problem in the foundations of monetary theory has yet to be satisfactorily
resolved. But what if it had been? It would tie the notion of the general price
level extremely closely to a particular theoretical analysis. The measurement of pm would be
"derived" rather than fundamental (see Ellis 1966, ch. 8). The pre-scientific notion is not tied to
such a derivation. That would not pose
any special problem if the general price level derived in this way correlated
closely with numerous other theoretical methods of deriving it, which in turn
correlated reasonably with the pre-scientific sense of a general rise in
prices.[6] Temperature provides an example of what is wanted (Ellis 1966,
ch. 6). The notion of hotter and colder
is pre-scientific. The first attempts
to provide quantification relied on some presuppositions - e.g., the linearity
of the expansion of the various fluids used in thermometers - but were not tied
to particular theories. Temperature
measures can now be derived from particular theoretical understandings - e.g.,
from the kinetic theory of gases. Such derived
measures show considerable consilience with the pre-scientific notions of
hotter and colder and with the atheoretical measurement systems. They permit the extension of temperature
scales beyond ordinary experience - e.g., to the measurement of the temperature
of the sun - but retain their independence from particular theories because of
the consilience of measurements derived from different theoretical starting
points. In contrast, the measurement of
the general price level remains at the atheoretical stage in which the makers
of price index numbers, Laspeyres, Paasche, and Fisher, are the economists'
Fahrenheit and Celsius.
The
point of raising these difficulties in measuring the general price level is not
that the existence of aggregates is tied to their measurement. Rather it is that the difficulties in
measuring them help to expose what problematic entities they are, and undermine
the appeal of seeing them as close analogues of their components (particular
prices, particular goods, and so forth).
The disanalogies can be made clearer through a more detailed examination
of the general price level.
The
fundamental insight of the index number is that one can avoid some of the
dimensional nonsense of averaging disparate prices by averaging percentage
rates of change instead. A simple
average, however, does not capture the commonsense feeling for the degree of
price change. A change in the price of
gasoline should count for more than, say, a change in the price of caviar in
measuring the change in the general price level. How to weight various price changes turns out to have an
irreducible degree of arbitrariness.
In
general the percentage change in the general price level, indicated as p (where p is the logarithm of the general price level), is related to
the individual underlying prices as
(1)
where pj is the price of good j for j =
1, 2, . . . n. Now, the properties that
theoretically restrict the functional form of f(.) are very weak:
1. 
2. 
.
3. 
.
Conditions 2 and 3 together imply an obvious
corrollary:
4. if 
then 
Condition 1 says that if each price increases
(decreases) the general price level must itself increase (decrease), and that
the general price level cannot change if no individual price changes.
Conditions 2 and 3 say that the general price level cannot increase by more
than the largest nor decrease by less than the smallest individual price
change. Condition 2 says that if every individual price changes
equiproportionally, so must the general price level. An infinite number of functions fulfill these conditions, and the
range of consistent changes in the general price level, given a fixed set of
underlying price changes is wide. In
practice, price indices are generally linear
, where 
. (2)
For m <
n, this formulation recognizes the practical fact that price indices are based
on samples of selected goods. The
weights wj in these indices are chosen in practice to capture the
pre-scientific sense of the amount of a price rise. Two common weighting schemes with rationales in microeconomic
consumer theory are the Laspeyres index, which chooses the weights to reflect
the share of each good in base period consumption, so that the general price
level effectively measures the changing cost of a fixed bundle of goods, and
the Paasche index, which chooses weights to reflect the share of each good in
current period consumption, so that some compensation is made for substitution
from relatively more expensive to relatively cheaper goods in the face of
changes in relative prices.
Neither
index is "correct;" there are an infinite number of indices lying
between the two; and economists have from time to time argued the case for
other indices with different weighting schemes.[7] The non-uniqueness of the price index is important for the point
of this essay. It is a fundamental
property. A price index is an attempt
to quantify the pre-scientific insight that the value of money changes. The different admissible price indices are
not, however, approximations to some true underlying general price level. The general price level is in some
fundamental sense non-scalar, although there is no currently acceptable
scientific refinement that captures that fact.[8] No similar indefiniteness attaches to any of the prices of the
underlying individual goods.
The
change in the general price level, p, may be integrated over time to generate a price level (
). The constant of
integration (c) permits us to choose an arbitrary base usually Pt =
1 or Pt = 100, where P = exp(p), for some desired base period t (other
base values are less common, but not unknown).
P differs from the particular price of, say, a Volvo, not just in its
intrinsic indefiniteness, but also in its dimensions. The dimensions of the price of a Volvo are dollars/Volvo; the
dimensions of P are period-t dollars/base-period dollars. The dimensions of P are not the dimensions
of the price of any good. They appear
to be the inverse dimensions of the price of money, taking base-period money to
be the numeraire. Given the
indefiniteness of the price index, however, it is evident that the price of
money is unlike the price of other goods, and represents a substantial
departure from pre-theoretic notions of price.
The price index is used to normalize the price of particular goods,
thereby to decompose individual price changes into a common or general element
and a "real" or relative (to the index) element. That this operation is not obvious to
commonsense will be evident to anyone who has taught elementary economics or
read policy analysis by non-economists.
Real
GDP is another important example of a synthetic aggregate. Considered as national income, nominal GDP
adds up the incomes of each individual in the economy and is an obvious
extension of the accounting framework for business or personal income. In a major innovation in economic analysis,
the national accounting framework since the 1930s establishes the three-way
identity between the sum of all incomes, the value-added in production, and the
value of all final goods and services.
That these other methods of computing GDP have obvious commonsense
analogues is less clear. If final goods
(i.e., goods that are not inputs into other production processes) are indicated
by Qj, then nominal GDP is
(3)
The dimension of income is dollars/unit time. Money provides the common unit that is
essential if disparate goods are to be added.
If
some or all of the prices of individual goods increase, it is obvious that
nominal GDP could increase without any of the individual quantities
changing. In the utilitarian framework
that underlies economics this is an undesirable characteristic, because the
measure of income has changed without the underlying utility, which is assumed
to be generated by the quantities of the goods themselves, changing. It is clearly desirable to correct for
changing prices. The usual way to do
this is to compute real GDP as
. (4)
Real
GDP is often treated as the analogue of an individual good. It does not, however, have the dimensions of
a real good. Rather its dimension is
base-period dollars (not dollars/unit good).
Real GDP is a derivative measurement.
One gets a different measurement for it for every different admissible
price index. It inherits the
fundamental fuzziness of the general price level.
The
analogy of real GDP to an individual good is suggested to some by the
possibility of perfect aggregation. If,
for example, relative prices are constant (i.e., pj/pk is
constant for all j and k), then 
(where the t in the subscript indicates the base time period
t) can be normalized by choosing the units for the Qj,ts so that
each Pj,t = 1. Then
nominal GDP at time n can be written
(5)
In this case, conditions 1 to 4 above insure that P
is unique. Some conclude therefore that
in this limited case, one can treat the summation in the right-hand side of
equation (5) as a natural aggregate quantity analogous to an individual
quantity. The conditions for constant
relative prices are almost certainly never fulfilled, but even if they were the
summation is not analogous to an individual quantity. The general price level P in (5) still has the dimension period-n
dollars/period-t (i.e., base period) dollars.
To sum heterogeneous goods, they must still be converted to a common
denominator, and in this case, the summation still has the dimensions of
period-t dollars. This would be more
perspicacious if (5) were written as
(6)
where the subscripted numeral 1 is a place holder
for the dimensional conversion.
The
general price level and real GDP are the most important aggregates in
macroeconomics. There are many
others. Each mixes the characteristics
of simple and synthetic aggregates to different degrees. Average interest rates were cited above as
an example of a simple aggregate, but when averaging is across nonhomogeneous
maturities and risk classes, interest rates too are complicated by the
fundamental problems of index numbers.
Aggregation of employment across skill or quality levels faces similar
considerations. There are other
derivative quantities as well. The real
rate of interest is defined to be the market interest rate less the percentage
change in the general level of prices (p). Like real GDP, the real
rate of interest inherits the fundamental fuzziness of the general price
level.
The
history of quantitative economics demonstrates that even the use of simple
averages represented a difficult conceptual leap. On the best interpretation what is accepted to common sense is
relative. To treat synthetic aggregates
as mere extensions of commonsense notions appears in comparison to make a
category mistake. Despite their
deceptively related names, there is no simple analogy between the general price
level and individual prices or between quantities of individual goods and real
GDP.
III. The
Supervenience of Macroeconomics on Microeconomics
Synthetic
aggregates, at least, are not direct extensions of folk ontology. It is clear, however, that, if their
independent reality is to be demonstrated in a sense more fundamental than that
one can always calculate them according to some algorithm, we must first show
that such aggregates cannot be reduced to properties of individual economic
actors. Aggregates are in fact
calculated; they clearly do not exist in a separate Platonic realm; and
ontological individualism has immediate appeal, because we all have first-hand
experience as economic actors. Any
account of the autonomy or nonreducibility of macroeconomic aggregates must
account, therefore, for the relationship of the individual to the aggregate.
Macroeconomic
aggregates I believe supervene upon
microeconomic reality. What this means
is that even though macroeconomics cannot be reduced to microeconomics, if two
parallel worlds possessed exactly the same configuration of microeconomic or
individual economic elements, they would also possess exactly the same
configuration of macroeconomic elements.
It is not the case, however, that the same configuration of
macroeconomic elements implies the same configuration of microeconomic
elements.
Biology
provides analogies and disanalogies for economics. Alexander Rosenberg (1985, ch. 4, section 8, ch. 6, section 3, passim) applies the notion of
supervenience to the relationship of functional biology (macro) to molecular
biology (micro). To take one example,
hemoglobin is an element in functional explanations of the operations of the
cardio-pulmonary and circulatory systems of higher animals (Rosenberg 1985, ch.
4, section 2). At the molecular level,
hemoglobin is not a single chemical, but a family of chemicals. To be hemoglobin at the macro level, a
molecule must possess nine particular proteins at critical junctures in the
molecular structure. Across different
species the approximately 140 remaining proteins which the hemoglobin molecule
comprises vary considerably. Similarly,
Rosenberg argues that Mendelian genetics supervenes on a molecular base. Mendelian genetics uses a conceptual scheme
that is not easily mapped onto molecular features, but nevertheless identical
molecular configurations produce identical genetic behavior.
The
notion of supervenience was initially suggested in the philosophy of mind as a
method of retaining the dependence of the mental on the physical, while at the
same time denying psychophysical laws (see Kim 1978, p. 153). Rosenberg draws on the analysis of Jaegwon
Kim (1978). For Kim, supervenience is a
relationship between two distinct realms of properties (or relations). Consider Ij, which is a
conjuction of properties in the micro realm in which every one of the
properties or its complement form one of the conjuncts.[9] Each Ij is
then a complete characterization of a possible micro state, and the disjunction
of every Ij defines every possible micro state. Consider the Aj, constructed mutatis mutandis for the macro
state. A family of macro properties is
supervenient on a family of micro properties when any objects which share the
same micro properties necessarily share the same macro properties. Kim (1978, pp. 152-153) shows that one can
derive the following relationship:
I1
I2 . . . In Ak, for any Ak, (7)
where the Ih, h = 1, 2, . . ., n, are a
subset of the Ij.[10] Of relation (7), Kim (1978, p. 153) says: ". . . I don't see how such
generalizations could fail to be lawlike."
Using
Kim's analysis, Rosenberg argues against the autonomy of Mendelian
genetics. The conceptual scheme of
Mendelian genetics (the macro level) does not map easily into the conceptual
scheme of molecular biology (the micro level).
Mendelian genetics permits explanation of phenomena not easily
explainable directly from the molecular level.
However, Mendelian genetics fails to account for some phenomena within
its scope. Rosenberg argues that
Mendelian genetics supervenes on molecular biology, and that molecular biology
is the more scientifically advanced, more fundamental and autonomous
theory. Mendelian genetics is reducible
in principle (that is the upshot of Kim's analysis in (7) above), but it
retains heuristic power, because something like the Cournot problem prevents
the practicable application of molecular biology to some phenomena in which
Mendelian genetics is relatively successful.
Can
a similar argument be applied to economics?
I do not think so. Rosenberg
(1992, p. 129) himself has argued that the intentional character of
microeconomics limits its scientific development: microeconomic ". . . theory's prediction and explanation of
the choices of individuals [cannot] exceed the precision and accuracy of
commonsense explanations and predictions with which we have all been familiar
since prehistory." In fact, macroeconomic
explanation and prediction is not only often better, but may have more scope
for improvability. An electric supplier
could not say when Mary Smith will switch on her oven, but it may know pretty
precisely how many kilowatts it must supply at a given time, based on an
aggregate analysis of past behavior.
Insurance companies know that whether an individual is, say, a smoker or
obese matters probabilitistically to his chances of dying. But the company would go broke trying to
predict individuals' precise dates of death.[11]
It
is important to remember that it is not macroeconomic theory that supervenes on
microeconomic theory, but macroeconomic reality that supervenes on
microeconomic reality. The disabilities
of microeconomic theory thus prove, at most, that there can be no automatic
presumption that microeconomics is more basic, because more successful, and
that macroeconomics is merely heuristic.
The critical relationship is the reducibility in principle suggested by
(7) above. To begin to undermine
reducibility in the case of macroeconomics, it helps to note a crucial
disanalogy with biology. Reduction
appeals to biologists because it removes scientifically suspect teleological
explanation common in evolutionary biology and other functional accounts. The aim of reduction in economics, however,
is precisely the opposite:
macroeconomics appears mechanical and dehumanized, and the point of the
program of microfoundations is reintroduce human decision-making as an
explanatory element. The point is to
recover intentionality.
Kim's
analysis posits two levels of properties that are (semantically at least)
distinct and then investigates how they must be related if one set is
supervenient on the other.
Intentionality at the microeconomic level undermines the distinctness of
microeconomic properties from macroeconomic properties. Levy's argument (see section I above) that
individual economic actors will invariably make reference to social wholes and
aggregates is even more fundamental than he imagines. In evaluating the future individuals must form expectations about
real prices and real quantities.
Independently of the uncertainty of the future, the Cournot problem implies
that it is impracticable to solve good-by-good, price-by-price, period-by-period
planning problems in all their fine detail.
The best that one is practically able to do is to work with
aggregates. The information on which
these are based is fundamentally monetary.
Economic actors must use estimates and expectations of the general price
level and real interest rates to form any practical assessment of their
situations.
Hayek
(1979, p. 62) writes:
. . . in the social sciences it is necessary to draw
a distinction between those ideas which are constitutive
of the phenomena we want to explain and the ideas which either we ourselves or
the very people whose actions we have to explain may have formed about these phenomena and which are not
the cause of, but theories about, the social structures.
What Levy's argument demonstrates is that Hayek is
mistaken, that how people theorize about the economy is constitutive of macroeconomic phenomena.[12] Since people cannot theorize about certain sorts of phenomena
without appealing to macroeconomic categories - that are not themselves reducible
to microeconomic categories - the Cournot problem introduces analytical
constraints, not only in practice, but in principle as well. The distinctiveness of the properties at the
microeconomic and macroeconomic levels is breached, undermining Kim's analysis,
because complete characterizations of the microeconomic must include
characterizations of the macroeconomic on the part of individual agents.
To
challenge the applicability to economics of the reductionism in principle,
implicit in Kim's analysis and in Rosenberg's application of it to biology,
does not challenge the notion that macroeconomics supervenes on
microeconomics. Kim's analysis is
epistemological: it argues that there
must be laws that would permit us to draw connections between the micro and
macro levels. The point here is
ontological: even though macroeconomics
cannot be reduced to microeconomics as the program of microfoundations
suggests, the elements of macroeconomics could not exist without the substrate
of microeconomic individuals.
IV. Two Arguments for the Reality of
Macroeconomics
So
far we have argued that the ontological status of macroeconomic entities is
problematic in the sense that, like other entities posited by scientific
theories, they are not part of our commonsense ontic furniture. Furthermore, the nature of the relationship
through which the elements of macroeconomics supervene on the elements of
microeconomics precludes direct reduction of the macroeconomic to the microeconomic,
even in principle. If macroeconomic
entities exist, they cannot be said therefore to exist only derivatively,
despite their supervenience on microeconomic entities. It remains to argue directly for the
existence of macroeconomic entities.
The
first argument is based on the argument from manipulability championed by Ian
Hacking (1983, esp. pp. 22-24):
"If you can spray them, then they are real." Hacking argues that convincing evidence of
the reality of the electron is found in experiments aimed at detecting the
existence of free quarks, in which niobium balls are charged by
"spraying" them with electrons.
The general point is that an entity defined by a scientific theory has
real existence when procedures that are best understood as using the entity
referred to by the theory as a tool to manipulate other parts of the world,
such as in laboratory experiments. I
take this argument to be related to the "no-miracles" argument for
the reality of scientific entities.[13] The best explanation of why theories are predictively successful,
including successfully predicting the consequences of using them to design
experimental or engineering manipulations of the world is that the entities
posited by them in fact exist - anything else would be an inexplicable miracle.
It
is common to denigrate the empirical success of economics (see e.g., Rosenberg
1992, pp. 18, 56, 112, 238, passim). It is true that economics does not have the
precision of physics or chemistry, although it is less clearly inferior to
meterology, geology, climatology, and parts of biology - to name just a few of
the less exact, but nevertheless scientific disciplines. The reputation of economics for predicting
poorly arises partly because people seek unconditional forecasts ("what
will happen tomorrow?") while economic theories typically predict only
conditionally ("tomorrow X will happen if Y happens"). Quantified economic relations are at best
locally stable: the precise estimate of
the price elasticity of demand for Volvos changes with changes in the range of
alternative brands and models, with changes in the proportion of academics to
the total population, and with changes in other background conditions. Nevertheless, qualitatively stable relations
are well established: e.g., demand
curves slope down (i.e., when the price of Volvos rises, sales of Volvos
fall). And there is often enough local
stability that useful quantitative assessments are possible. Can irreducible macroeconomic aggregates be
manipulated as well?
The
answer seems to be clearly yes.
Consider the follow example.
Almost every macroeconomic theory predicts that sufficiently large
expansions of government expenditure will change (probably increase) nominal
GDP and the general level of prices.[14] Different theories differ in their precise
understanding of the mechanisms.
Similarly, no macroeconomic theory disputes the ability of the Federal
Reserve to use its ability to supply or remove reserves from the banking system
to set the level of the Federal funds rate (the rate at which one commerical
bank borrows from another overnight).
The empirical evidence in support of these propositions is also
overwhelming. Now consider two
irreducibly macroeconomic aggregate entities:
the real rate of interest (i.e., the market rate of interest less the
percentage change in the aggregate price level (p)) and the yield curve (an aggregate relation portrayed as a graph of
market interest rates against time to maturity of the associated bonds). Both the real rate of interest and the yield
curve are synthetic aggregates and both are entities with causal powers in some
economic theories. Every macroeconomic
theory that I know predicts that actions that increase the general price level
or the Federal funds rate will shift the yield curve upwards in the short-run. And, at least if the changes are
unanticipated, increases in the general price level will reduce the level of
the real interest rate. The empirical
evidence for these effects is overwhelming, and indeed are easily confirmed by
anyone willing to read the Wall Street
Journal regularly for a month. Just
like the electron, some macroeconomic aggregates can not only be controlled,
but can be used to manipulate other macroeconomic aggregates.
The
second argument is related to the first.
Nowak (1980) and others have argued that the principal method of
constructing scientific theories is idealization. Nowak's (1980, p. 29) paradigm idealization
statement is:
If
G(x) and p1(x) = 0 and . . . pk-1 = 0 and pk =
0 then
F(x)
= fk(H1(x), . . ., Hn(x)), (8)
where Hi (i = 1, . . ., n) denote primary factors and the pj (j
= 1, . . . , k) denote secondary factors. An idealized theory is one which picks out
the primary factors by setting the secondary factors to extreme values: zero or , represented here, without loss of generality, as pj =
0.
Were
G(x) a known and exhaustively complete theory of the phenomenon within its
explanatory range such that one could accurately specify each of the secondary
factors that were set aside, then the distinction between primary and secondary
factors would in fact be be unclear, because our complete knowledge of G(x)
would allow us for example to replace (8) with
If
G(x) and H1(x) = 0 and . . . Hn-1 = 0 and Hn =
0 then
F(x)
= fn(p1(x), . . ., pk(x)). (9)
In the case of either (8) or (9), releasing the
idealizing conditions (pk = 0 or Hi = 0) allows us to recover the complete
theory, G(x). Idealization has been
reduced to a fancy name for an arbitrary selection of ceteris paribus conditions or to a formal nesting relationship for
theories.
Hoover
(1994) argues that in an empirical context, the method of idealization has
power only if we recognize that not all of the idealizing conditions can be
explicitly stated. The claim to
distinguish between primary and secondary factors is then a claim that the primary factors are the essence of the matter.
Idealized theories thus aim to identify, isolate and relate the real
essences or causally effective capacities of economic reality.[15] The success of such an idealized theory then amounts to an
ontological claim for its primary factors.
That
Keynesian macroecononomics could be cast as an idealization that employs
macroeconomic aggregates essentially is beyond doubt. The major competitor to Keynesian macroeconomics today, new
classical macroeconomics, trades on an explicitly microfoundational
approach. There is, however, less here
than meets the eye. The currently most
popular new classical macroeconomic theory is embodied in the
real-business-cycle model (see Hoover 1995b and forthcoming). The proponents of this representative-agent
model would like it to be interpreted as an idealization from a complete
Walrasian general equilibrium model of the economy in which distributional
issues are idealized out of the model so that what remains is a one-agent,
one-good, one-price representation of the economy. This would work if the Walrasian model (the analogue for G(x) in
Nowak's schema) were both true and known in detail. At least the second condition is false, which undermines the
evidential basis for the first condition.
Empirically,
far from isolating a microeconomic core, real-business-cycle models, as with
other representative-agent models, use macroeconomic aggregates for their
testing and estimation. Thus, to the
degree that such models are successful in explaining empirical phenomena, they
point to the ontological centrality of macroeconomic and not microeconomic
entities. The appeal to the methods of
microeconomics do not in this case amount to the successful implementation of
the program of microfoundations, for they are but the simulacrum of
microeconomics. The relationship
between models that are microeconomic in form and their macroeconomic empirical
implementation is metaphorical. The
nature of metaphorical connection deserves futher exploration. It is enough for the present purpose to
understand that, at the empirical level, even the new classical
representative-agent models are fundamentally macroeconomic in content.
V. The Contingency of Economic Reality
Unless one is committed to a certain kind of apriorism, then what we
judge to be real depends on experience, on the nature of our theorizing, and
the success of our scientific and everyday practices. The best guess of a scientist in 1789 would have been that
phlogiston was real, although most scientists would today say that it is
not. The point is not that reality is
constituted by our theorizing: for a
realist it is not only the case that phlogiston is not real today, but that it was not real in 1789. The point is rather that our knowing whether
or not something is real is a scientific fact like other scientific facts,
which are established by argument and evidence, and about which we may be mistaken. There is, therefore, no timeless, certain
answer to the question in the title of this paper. What I have sought to show in this essay is that the nature of
microeconomics and macroeconomics - as they are currently practiced -
undermines the prospects for a reduction of macroeconomics to
microeconomics. Both microeconomics and
macroeconomics must refer to irreducible macroeconomic entities. These macroeconomic entities occupy
ontologically independent places in economic theory. To the degree that such theories are empirically successful, the
best account of these macroeconomic entities is that they are real.
References
Blaug, Mark. (1992) The Methodology of Economics: Or How Economists Explain, 2nd edition. Cambridge: Cambridge University Press.
Boland, Lawrence. (1979) The Foundations of Economic Method. London: George Allen and Unwin.
Cartwright, N. (1983). How the Laws of Physics Lie. Oxford: Clarendon Press.
Cartwright, N. (1989). Nature's Capacities and Their Measurement. Oxford: Clarendon Press.
Cournot, Augustin. (1838/1927) Researches into the Mathematical Principles of the Theory of Wealth, Nathaniel T. Bacon (trans.). New York: Macmillan.
Ellis, Brian. (1966) Basic Concepts of Measurement. Cambridge: Cambridge University Press.
Friedman, Milton. (1955) "Leon Walras and His Economic System: A Review Article," American Economic Review, 45(5), 900-909.
Hacking, Ian. Representing and Intervening: Introductory Topics in the Philosophy of the Natural Sciences. Cambridge: Cambridge University Press.
Hahn, Frank H. (1965) "On Some Problems of Proving the Existence of Equilibrium in a Monetary Economy," in Frank H. Hahn and F.P.R. Brechling (eds.) The Theory of Interest Rates. London: Macmillan.
Hayek, Friedrich von. (1979) The Counter-revolution of Science: Studies in the Abuse of Reason. Indianapolis: Liberty Press.
Hoover, Kevin D. (1988) The New-classical Macroeconomics: A Skeptical Inquiry. Oxford: Blackwell.
Hoover, Kevin D. (1994) "Six Queries About Idealization in an Empirical Context," Poznan Studies in the Philosophy of Science and the Humanities, 38, 43-53.
Hoover, Kevin D. (1995a) "Why Does Methodology Matter for Economics? A Review Article," Economic Journal, 105, 715-734.
Hoover, Kevin D. (1995b) “Facts and Artifacts: Calibration and the Empirical Assessment of Real Business Cycle Models,” Oxford Economic Papers, 47, 24-44.
Hoover, Kevin D. (forthcoming) “Quantitative Evaluation of Idealized Models in the New Classical Macroeconomics,” in Nancy Cartwright and Martin Jones (eds.) Correcting the Model. Poznan Studies in the Philosophy of Science and the Humanities. Amsterdam: Rodopi.
Janssen, Maarten. (1993) Microfoundations: A Critical Inquiry. London: Routledge.
Keynes, John Maynard. (1936) The General Theory of Employment Interest and Money. London: Macmillan.
Kim, Jaegwon. (1978) "Supervenience and Nomological Incommensurables," American Philosophical Quarterly, 15(2), 149-156.
Kirman, A.P. (1992) "Whom or What Does the Representative Agent Represent?" Journal of Economic Perspectives, 6(2), 117-136.
Lachmann, Ludwig. (1976) "On the Central Concept of Austrian Economics: Market Process," in Edwin G. Dolan (ed.) The Foundations of Austrian Economics. Kansas City, MO: Sheed and Ward.
Levy, David M. (1985) "The Impossibility of a Complete Methodological Individualist: Reduction When Knowledge is Imperfect," Economics and Philosophy, 1(1), 101-108.
Lucas, Robert E., Jr. (1987) Models of Business Cycles. Oxford: Blackwell.
Mäki, Uskali (1992) "On the Method of Isolation in Economics," Poznan Studies in the Philosophy of Science and the Humanities, 25, 289-310.
Mäki, Uskali (1994) "Scientific Realism and Some Peculiarities of Economics," in R.S. Cohen et al. (eds.) Realism and Anti-Realism in the Philosophy of Science, Boston Studies in the Philosophy of Science, Vol. 169. Dordrecht: Kluwer 1996, pp. 425-445.
Marshall, A.
(1930) Principles of Economics: An
Introductory Volume, 8th ed.
London: Macmillan.
Mises, Ludwig von. (1943) "'Elastic Expectations' and the Austrian Theory of the Trade Cycle," Economica NS 10(3), 251-252.
Mises, Ludwig von. (1949/1966) Human Action: A Treatise on Economics, 3rd. edition. Chicago: Henry Regnery.
Nowak, L. (1980). The Structure of Idealization: Towards a Systematic Interpretation of the Marxian Idea of Science. Dordrecht: Reidel.
Robbins, L. (1937). An Essay on the Nature and Significance of Economic Science. London: Macmillan.
Rosenberg, Alexander. (1985) The Structure of Biology. Cambridge: Cambridge University Press.
Rosenberg, Alexander. (1992). Economics: Mathematical Politics or Science of Diminishing Returns? Chicago and London: Chicago University Press.
Samuelson, Paul A. and Ryuzo Sato. (1984) "Unattainability of Integrability and Definiteness Conditions in the General Case of Demand for Money and Goods," American Economic Review, 74(4), 588-604.
Footnotes