Description
The modern minimum-wage controversy derives from a recent literature in empirical labor economics, which argues, foremost among several claims, that moderate increases in mandated minimum wages do not lead to adverse employment outcomes for low-wage workers.
The Very Idea of Applying Economics:
The Modern Minimum-Wage Controversy
and Its Antecedents
Thomas C. Leonard
Disagreements among economists rarely breach the academy’s walls.
The recent minimum-wage controversy is a signi?cant exception. At
?rst glance, all the sound and fury is a puzzle.
1
Minimum-wage effects,
at least for current U.S. magnitudes, are small potatoes.
2
Of several
more important policy concerns—entitlement reform, health insurance,
CPI calculation, for example—none has generated anything like the
vituperation that has characterized the minimum-wage controversy. But
if the minimum-wage controversy is not especially important to the
economy, it is very important to economics, and, thereby, to the status
of economics as a policy science.
Correspondence may be addressed to Professor Thomas C. Leonard, Department of Econom-
ics, Fisher Hall, Princeton University, Princeton, NJ 08544; e-mail: [email protected]. I
am grateful to Jeff Biddle, Bob Goldfarb, an anonymous referee, and participants in the 1999
HOPE conference on the history of applied economics for helpful comments on an earlier
draft of this material.
1. In the wake of Card and Krueger 1995, one Nobel laureate intoned, “I tremble for my
profession” (Miller 1996). Another made reference to “camp-following whores” (Buchanan
1996). A third eminence called Card and Krueger’s main interlocutors, David Neumark and
William Wascher, “heroes” (Barro 1996).
2. Even a 27 percent increase in the U.S. minimum wage (from $3.35 to $4.25, in 1990 and
1991) affected only about 8 percent of the labor force, and these workers were concentrated
in the retail trade and food service sectors. Ignoring employment effects, Card and Krueger
estimate that $5.5 billion was transferred to low-wage workers, only 0.2 percent of total U.S.
annual earnings (1995, 277). Card and Krueger estimate that if ?rms had passed along all of
the labor cost increases to consumers, prices would have increased (once and for all) only
about 0.3 percent (393).
118 Thomas C. Leonard
The reason, I want to argue, is that the core of modern economics—
neoclassical price theory—is seen to be at stake. In particular, minimum-
wage research has come to be seen as a test of the applicability of
neoclassical price theory to the determination of wages and employment.
The modern minimum-wage controversy is not just a technical quarrel
over the sign and magnitude of wage-elasticity coef?cients; it is the
latest chapter in a longstanding methodological dispute over whether
and in what domains neoclassical price theory can be said to properly
apply. One does not make opposition to minimum wages a disciplinary
litmus test, unless there is something this important at stake.
3
The modern minimum-wage controversy engages economists be-
cause it brings to light methodological issues ordinarily submerged in
the course of normal science. I present a history of the economics of min-
imum wages designed to illuminate two such issues: (1) does prevailing
theory explain and predict the determination of wages and employment,
and (2) by what means do economists decide when explanation and
prediction are successful?
The history—really three vignettes—is designed to show, ?rst, that
the current debate is a modern-dress version of an old family argument
in economics. Second, it provides some evidence for my claim that the
minimum-wage controversy is not merely a substantive disagreement
among labor specialists, but also a recurring touchstone for method-
ological debate about the application of economics.
What Is the Modern Minimum-Wage
Controversy About?
The modern minimum-wage controversy derives from a recent literature
in empirical labor economics, which argues, foremost among several
claims, that moderate increases in mandated minimumwages do not lead
to adverse employment outcomes for low-wage workers.
4
Many of these
?ndings are collected or summarized in David Card and Alan Krueger’s
3. George Stigler said in 1976 that “one evidence of professional integrity of the economist
is the fact that it is not possible to enlist good economists to defend . . . minimum wage laws”
(Stigler 1982, 60).
4. Card and Krueger say that their “strongest and most important ?nding” is the absence of
disemployment effects from moderate increases in minimum wages (1995, 387). The quali?er
“moderate” is important. No economist believes that ?xing a wage ?oor at $100 per hour
would not create disemployment. Card and Krueger also recognize that the intensity of the
minimum-wage debate is out of proportion to its economic importance (1995, 395).
The Minimum-Wage Controversy 119
in?uential 1995 book, Myth and Measurement: The New Economics of
the Minimum Wage (hereafter cited as Myth).
The current controversy arises because “the new economics of the
minimum wage” is at odds with neoclassical price theory, which pre-
dicts disemployment, and with a generation of time-series econometric
research, which consistently ?nds evidence of disemployment effects.
Myth’s ?ndings are also at odds with what most U.S. economists, es-
pecially labor economists, believe (Whaples 1996). Card and Krueger
understand this; indeed, they expressly conceive of their project as a
refutation of neoclassical price theory’s predictions and of the econo-
metric evidence adduced in support (1995, 396–97).
Following the contemporary emphasis, I focus on employment and
welfare effects: do minimum-wage increases disemploy low-wage
workers, and do minimum-wage increases, on balance, help low-
wage workers? Later I take up the less emphasized but essential matter
of distribution: To what extent do workers affected by minimum-wage
legislation live in poor or near-poor households?
5
The Neoclassical Theory of
Minimum-Wage Effects
The neoclassical theory is familiar. Mandated minimum wages are
treated as any other price ?oor. If the mandated wage exceeds the
market-clearing wage, then, given a downward-sloping labor demand
curve, ?rms reduce the quantity of labor demanded. The magnitude of
the reduction depends upon the wage increase and the wage elasticity
of labor demand, but the direction of change is unambiguous. Welfare
effects also depend upon the wage elasticity of demand. Some workers
will receive the higher wage and are better off. Other workers, those
whose product is worth less than the new minimum, will be laid off, or
will work fewer hours. If the quantity of labor refers to employment,
then the wage gains of those who keep their jobs must be traded off
against the wage losses of those who lose their jobs.
6
5. Poor and near-poor families are those with incomes at or below 1.5 times the U.S.
poverty level.
6. Suppose, for example, that the wage elasticity of demand is ?0.2 and that the minimum-
wage increase is 10 percent. This implies that 2 percent of workers (if L is employment) will
lose their jobs, and that 98 percent will get a 10 percent raise, meaning that affected low-
wage workers collectively realize income gains of 7.8 percent [(0.98)(0.1) ?(.02)(1.00)]. By
implication, minimum wages will help the class of affected workers when the wage elasticity
120 Thomas C. Leonard
Disemployment is a logically necessary result if one assumes that
(1) ?rms maximize pro?ts and (2) the low-skilled labor market is com-
petitive, that is, that ?rms have no monopsony power (Kennan 1995).
Neoclassical theory can accommodate a weakening of assumption 2.
If labor markets are monopsonistic, ?rms exercise their market power
and pay a wage lower than workers’ marginal revenue product (MRP
L
).
Monopsonistic ?rms offer higher wages to attract newhires, which leads
to costlier labor expenditures (more pay to the inframarginal workers)
and inef?ciently low employment.
In principle, then, a judiciously chosen minimum wage (set between
monopsony wage and MRP
L
) can increase employment and ef?ciency.
In practice, however, labor economists traditionally regard monopsony
as an empirical curiosum (Brown 1988, 134 n. 1).
7
Individual low-wage
?rms typically employ a small percentage of the low-wage work force
(Brown, Gilroy, and Kohen 1982, 489), so monopsony power is seen as
limited in practice.
8
Neoclassical economics is less able to accommodate a weakening
of the pro?t-maximization assumption, which is more fundamental.
With few exceptions (see, for example, Prasch 1996), contemporary
economists maintain the pro?t-maximization assumption, generally re-
verting to monopsony models derived from ef?ciency-wage explana-
tions.
If some ?rms pay more than the going wage, and pro?t maximization
is assumed, then the only remaining explanation is that some ?rms ?nd it
pro?table to pay above the going rate, which de?nes an ef?ciency-wage
explanation (Krueger and Summers 1988). Ef?ciency-wage accounts
propose that wages can exceed MRP
L
when (1) higher wages bring
forth greater worker productivity, a reversal of the traditional causality,
is less elastic than ?1.0. But there are two important caveats. First, not all affected workers
get a 10 percent raise; those between the old and new minima get a smaller increase. Second,
wage elasticities generally derive from studies that calculate employment effects for entire
demographic groups, not just for the minimum-wage workers within those groups. Thus,
an elasticity estimate such as ?0.2 measures the effect of minimum-wage increases on the
employment of some group (say, teenagers), which includes higher-wage (hence unaffected)
workers, and thereby underestimates the employment effects on the lowest-wage workers.
(These points are due to Neumark, Schweitzer, and Wascher 1998, 3–4).
7. Robert Whaples ?nds that “six out of seven labor economists believe that ‘few American
workers are employed in monopsonistic labor markets’” (1996, 726).
8. Even with monopsony power, the range of employment-increasing wage increases is
limited. A minimum set above the MRP
L
at the monopsonist’s pro?t-maximizing output also
results in disemployment.
The Minimum-Wage Controversy 121
(2) there is an informational asymmetry such that ?rms cannot readily
monitor worker effort, and (3) ?rms have high recruitment and turnover
costs (the last is emphasized by Card and Krueger 1995, 373–79). Since
pro?t maximization and perfect competition entail disemployment with
higher minima, it is not surprising that, historically, both assumptions
recur in arguments against and for the minimum wage.
History of Minimum-Wage
Legislation and Its Economics
The idea of administered wages is ancient. Wage payments to laborers
were conventional and noncontractural long before the industrial revo-
lution and the concomitant rise of freely contracting labor. But the idea
of administered wages as a legal means of improving the lot of the urban
working poor has its origins in nineteenth-century reform movements,
especially in the United Kingdom. Minimum wages were connected to
mid-Victorian-era trade union movements and were an important item
on the Fabian socialist agenda to “democratize” unregulated industries
and improve working conditions (Webb and Webb [1897] 1920).
9
The ?rst laws setting wage ?oors, as against a ?xed wage or an output-
price peg, have their origin in Australasia. The District Conciliation
Boards of NewZealand, established in 1894 principally for arbitration of
labor disputes, had statutory authority to set minimum wages (Douglas
1919, 701). The ?rst independent legal agencies designed speci?cally
for setting minimum wages were established by the 1896 Factory and
Shops Act, enacted by the state of Victoria to ?x minimum wages in
“sweated” industries (701).
10
Britain passed the Trade Boards Act of
1909, which likewise set minimum wages in several industries deemed
to be “sweating.”
9. Aminimumwage means only that workers cannot legally accept a wage belowit. Higher
wages are legal. How and in what settings the minimum is to be ?xed is more ambiguous, as
the compelling but vague term living wage suggests. See Pigou 1922.
10. De?nitions of “sweating” vary. Reporting on the 1888–90 investigation of “sweated
industries” by a select committee in the British House of Lords, Beatrice Webb offers the
following de?nition: “Lord Derby and his colleagues decided that sweating was no particular
method of remuneration, no particular form of industrial organization, but certain conditions
of employment—viz., unusually low rates of wages, excessive hours of labor, and unsani-
tary workplaces. When we get any one of these conditions in an exaggerated and extreme
form, . . . then we may say, that the labor is sweated, and that the unfortunates are working
under the sweating system” (cited in Holcombe 1910, 574).
122 Thomas C. Leonard
The late-nineteenth-century minimum-wage movement in the United
States was originally voluntarist in spirit. Organizations like the Na-
tional Consumers League hoped, by information provision and moral
suasion, to persuade consumers to avoid ?rms that paid especially low
wages to workers—a minimum wage was seen as a moral rather than
a legal standard (Kelley 1912). State legislation overtook their efforts.
Massachusetts (1912) was the ?rst, followed a year later by Wisconsin,
Minnesota, Washington, Oregon, California, Colorado, Nebraska, and
Utah (see Millis 1914).
Some laws empowered supervisory boards to determine minimum
wages; others directly ?xed minima by statute. In all cases, the states’
laws, unlike their British Commonwealth precursors, applied to women
and minors only. Until the advent of a federal minimum in 1938, men
were deemed outside such paternal protection, in light of the U.S. Con-
stitution’s explicit protection of the right to free contract.
11
Stretching back into the classical era is the question of whether a
general theory—that of free markets and how they determine price and
quantity—can be applied to matters of wage and employment deter-
mination. In other words, is it the case, as Adam Smith said, that the
market for men is like the market for goods, so that there exists a “labor
market” in which wages are determined as prices are in other markets?
At least as early as John Stuart Mill, we see a further application of
this application: what does the general “free market” model, as applied
to the workers and employers’ behavior, tell us about the effects of a
particular policy measure, the wage ?oor?
Mill’s 1848 political economy text has a recognizably modern ar-
gument on employment effects. Ignoring the peculiarities of Mill’s
wages-fund doctrine, which he had yet to abandon, the outline of the
neoclassical disemployment argument is present 150 years ago. Said
Mill:
If law or opinion succeeds in ?xing wages above this [competitive]
rate, some laborers are kept out of employment; as it is not the
intentions of the philanthropists that these should starve, they must
be provided for. (1848, 432)
Minimum wages will disemploy some workers, says Mill, and some-
thing must be done to help the newly disemployed. But, whatever is done
11. For a brief history of the dramatic political events leading up to the Fair Labor Standards
Act (FLSA), see Grossman 1978.
The Minimum-Wage Controversy 123
to help—public assistance, for example—will create a disincentive to
work.
Mill points to the now familiar ef?ciency cost of all interventions—
the tension between alleviating poverty and providing incentives to
work. If laid-off workers receive a guaranteed income, then there is
no incentive to work unless the minimum wage is well above this level.
But if the minimum is so set, previously viable jobs are eliminated,
which is wasteful (Kennan 1995).
The disemployment argument is in place, as is a familiar theme of
British political economy: unintended policy consequences. The sundry
social reform movements of nineteenth-century Britain generally re-
garded Mill’s late classical economics as a kind of vulgar laissez-faire,
a mere rati?cation of the raw excesses of Victorian industrial capitalism.
But Mill was an ardent social reformer. He quarreled not with the goal
of meliorating adverse free-market outcomes, but with the ef?cacy of
policy instruments such as a legal minimum.
Henry Sidgwick (1886) shared Mill’s skepticism regarding the net
bene?ts of government interventions, especially utopian socialist
schemes (see Hutchison 1953, 53–54). And Sidgwick was at pains to
distinguish British political economy from more truly laissez-faire ap-
proaches, such as Claude-Frédéric Bastiat’s. British political economy
does not argue that low-wage workers get what they deserve, but that
well-intentioned interventions may do more harm than good. Says Sidg-
wick:
You can make it illegal to employ a man under a certain rate of
wages, but you cannot secure his employment at that rate, unless
the community will undertake to provide for an inde?nite number of
claimants work remunerated at more than its market value; in which
case its action will tend to remove, to a continuously increasing extent,
the ordinary motives to vigorous and ef?cient labor. (1886, 631)
In an 1897 article by Alfred Marshall, the disemployment argu-
ment is also quite clear: “[The academic economist] must for instance
analyse the methods which people are tempted to take for securing a
high minimum wage . . . and must show which of them will have indi-
rect effects that will cause workingmen as a whole a loss greater than
the bene?t. . . . ordinary people do not see that the means most com-
monly advocated . . . would impoverish all” (1897,128). Here, in an ad-
dress to the inaugural meeting of the Cambridge Economics Club (29
124 Thomas C. Leonard
October 1896), we also ?nd Marshall elevating ef?ciency as the badge
of the academic economist. The economist must ?ght conventional wis-
dom on minimum wages, says Marshall, not because one disagrees with
the goal of alleviating poverty among the working poor, but because
minimum wages do more harm than good (1897, 129).
The Marshallian theory of minimum wages ?nds its most thorough
and lucid exposition in a 1907 essay by H. B. Lees Smith; it is a re-
markable paper written in partial response to minimum-wage arguments
in Sidney and Beatrice Webb’s Industrial Democracy ([1897] 1920). I
present Lees Smith’s argument truncated and translated into modern
argot, but it bears reading in the original.
At the outset, Lees Smith assumes that productivity is independent of
cost: “we may assume for the present that the ef?ciency of both wage-
earners and employers is the same before and after the enactment of the
minimum” (1907, 504). He notes that statutory coverage may be partial
or total, taking partial coverage ?rst.
If the minimum is actually binding, higher labor costs must be funded
fromreduced expenditures on other inputs, higher output prices, or lower
pro?ts (504). Take reduced expenditures ?rst. If other inputs such as
skilled labor are substitutes, and they generally are, then a binding min-
imum will increase, not decrease, their prices and expenditures thereon,
which eliminates other input costs as a means of funding higher costs
of low-skilled labor (505).
If, alternatively, output prices are increased to fund higher low-skilled
labor costs, then output demand will fall, unless such demand is perfectly
inelastic. And, if ?rms and workers are already optimizing, the reduction
in output quantity demanded will entail an inef?cient reallocation of
inputs to other uncovered sectors (506). Thus, present in Lees Smith is
the law of demand and the assumption of pro?t maximization:
A rise in the price of the commodity will lead to a decrease in the
demand for it. This means that a certain amount of the labor and
capital hitherto employed in the production of the commodity must
seek other occupations. But this is likely to involve both the laborers
and the employers in a loss. Employers and laborers were presumably
occupied where, with such choices as they had, they thought they
would do the best for themselves. . . . It is no kindness to the workers
in a trade to merely turn them out. (506; emphasis added)
The Minimum-Wage Controversy 125
Finally, increased wages may be funded out of pro?ts. But if pro?ts
are diminished by higher labor costs, capital will migrate to more re-
munerative sectors. This will occur unless the covered sector’s output
market is so uncompetitive that supranormal pro?ts can fund the higher
labor costs, a claim nowhere made by minimum-wage proponents.
Thus, each possible alternative entails low-skilled labor being dis-
employed. With universal coverage, the same outcome obtains, only
low-skilled labor becomes unemployed or leaves the labor force. Con-
cludes Lees Smith: the disemployment verdict of “pure theory” is clear.
Lees Smith fully anticipates the neoclassical argument, as exempli?ed
by Stigler 1946.
I do not wish to suggest that political economists were uniformly
opposed to minimum wages. Not only is it anachronistic to assume
that economics was, around 1900, a uniform or even professionalized
discipline, but, all along the way, there were those who denied that
late-classical and then Marshallian models applied to labor markets.
F. D. Longe (1894), who was one of the ?rst to challenge Mill’s wages-
fund doctrine, dissented, as did a number of American Institutional-
ists, notably John R. Commons, who helped draft the ?rst Wisconsin
minimum-wage statute.
Perhaps the most conspicuous proponents of minimum wages in this
era are Beatrice and Sidney Webb. Sidney Webb (1912) offers an espe-
cially useful counterpoint to Lees Smith 1907.
Webb argues that, for the least productive workers, wages are not de-
termined by labor productivity. He says, “The wages of the lowest grade
of labor are ?xed, not by ‘worth’ in any sense—not even the possible
‘value in exchange’—of the individual laborer, but (as we nowadays
sadly concede) largely by the urgent necessities of the ‘marginal’ man,
or, rather, the ‘marginal’ woman” (1912, 990). By implication, either
low-wage ?rms are not pro?t maximizing, or they ?nd it pro?table,
along ef?ciency-wage lines, to hire workers whose productivity is very
low.
Webb makes recourse to both claims. At one point he says that “a
Legal Minimum positively increases the productivity of the nation’s
industry, by ensuring the surplus of unemployed workmen shall be ex-
clusively the less ef?cient workmen . . . [which] is plainly not the case
under ‘free competition’” (979). For this to be true, ?rms cannot be, ex
ante, pro?t maximizing, so Webb is offering, in rough form, an early
version of the “shock-therapy” argument.
126 Thomas C. Leonard
Arguing in the alternative, Webb also makes an ef?ciency-wage ar-
gument. “To put it plumply,” he says, “if the employers paid more, the
labor would be worth more” (990). Why might higher productivity fol-
low higher wages? At one point Webb argues that affected ?rms will
be obliged to innovate—“[the minimum wage] acts as a constant incen-
tive to the further improvement of the manual laborers, the machinery
and the organizing ability used in industry”—implicitly assuming, as
against Lees Smith, that factor substitution would not, on balance, work
against low-skilled workers (985). Webb also argues that if the minimum
were to increase wages above subsistence, laborers will be physically
stronger and thus more productive. This latter claim requires that ?rms
had failed to perceive that a wage above subsistence would increase
pro?tability.
What we observe, then, is that Webb takes issue with the key ideas in
Lees Smith’s account, not least pro?t maximization and perfect compe-
tition. Webb also argues, at least implicitly, against the law of demand.
Where Lees Smith insists that higher output prices will entail lower
output quantity demanded, Webb suggests that “the community [con-
sumers] would willingly pay more for it, and yet consume as much, or
nearly of much of it as it now does” (990).
I do not wish to suggest that Webb and Lees Smith would have de-
scribed their disagreement solely in terms of pro?t maximization and la-
bor market structure. The hope, rather, is that this sketch suf?ces to show
that the leading contemporary issue in minimum-wage economics—em-
ployment effects—is already being debated in familiar terms around the
turn of the century. The disemployment account is recognizable in Mill
150 years ago, and, in the sense of considering the labor-market condi-
tions under which disemployment does and does not obtain, is almost
fully formed 90 to 100 years ago.
What is missing from the minimum-wage arguments of this era is
empirical evidence. The evidence for the successful application of eco-
nomics to labor markets is entirely theoretical—that is, based on appeal
to reason and logic. Clearly, U.S. data became available only follow-
ing the states’ minimum-wage legislation. But what is striking through
much of the twentieth century is the extent to which the application of
economics to labor markets remains almost entirely theoretical. George
Stigler’s landmark paper, for example, presents only a few stylized facts
for evidence, admitting that “no precise estimate of the effects of the
minimum wage upon aggregate employment is possible” (1946, 361).
The Minimum-Wage Controversy 127
The mid-twentieth-century debate recapitulates many of the old ar-
guments about whether the general theory of competitive markets, now
full-blown neoclassical price theory, is applicable to labor markets. But,
here our second theme emerges: the role of empirical evidence as the
arbiter of which theory (if any) is indeed applicable. The marginalist
controversy and the economics of minimum wages come together in the
person of Richard Lester.
Minimum Wages and the Marginalist Controversy
Richard Lester, the labor economist to whom Card and Krueger dedicate
their book, argued that ?rms do not always maximize pro?ts and that
labor markets generally are not competitive. Lester’s conclusion was
based on surveys of manufacturing CEOs who did not know marginal
costs and revenues, and on wage surveys, which found a wide vari-
ance in wages paid to production workers of roughly equal productivity
(Lester 1946a, 1946b). Lester read ignorance of marginal magnitudes
as implying that ?rms do not maximize pro?ts; and he read the unex-
plained wage variation as suggesting that ?rms do not take wages as
given.
Lester was not the ?rst to argue that ?rms do not maximize pro?ts,
12
but his work generated a volley of neoclassical responses, an exchange
which has come to be known as the marginalist controversy. Some re-
sponses were directed at Lester’s challenges explicitly (Machlup 1946,
1947; Stigler 1947) and others more indirectly (Alchian 1950; Friedman
1953).
13
It was Lester’s investigation of labor markets, and his strong
conclusions, that gave rise to a withering exchange on the nature of eco-
nomic theory and on the role of empirical evidence in testing theories
and their implications.
That labor research was the occasion for the methodological dispute
we ordinarily associate with the marginalist controversy is not coin-
cidence. Lester and other institutionally minded labor economists of
the mid-century era—especially John Dunlop, Clark Kerr, and Lloyd
12. R. L. Hall and C. J. Hitch (1939,18) interpret their interviews of businessmen as
arguing that ?rms do not equate marginal revenues and marginal costs, but use rules of thumb
for marking up average costs.
13. Lester’s rejoinder to Machlup 1946 (itself a reply to Lester 1946a) and reply to Stigler
1946, the neoclassical locus classicus, are found in Lester 1947.
128 Thomas C. Leonard
Reynolds (DKLR)—practiced a kind of labor economics that differed
signi?cantly from the neoclassical labor economics that later predom-
inated (Freeman 1989, 317–42). The DKLR method was built around
institutional expertise, detailed case studies, and surveys of management
and workers. They surely resented what they regarded as neoclassical
theory-tropism: relative indifference to empirical work, and hostility to
survey questionnaires in particular (Machlup 1946).
Related, and perhaps more important, DKLRrejected the neoclassical
model’s assumptions of pro?t maximizing and perfect competition, at
least for empirical purposes.
The labor market is not a bourse; wage rates are typically quoted
prices. (Dunlop 1944, cited in Freeman 1989, 330)
The labor market is by nature an imperfect instrument. . . . all kinds
of wage distortions exist even under non-union conditions. (Reynolds
and Taft 1956, 168, cited in Freeman 1989, 319)
Reasoning about labor markets as though they were commodity mar-
kets seems to be an important explanation for erroneous conclusions
on such matters as minimum wages. (Lester 1947, 146, responding
to Stigler 1946)
A [competitive market–determined wage structure] is a useful norm
for theoretical speculations but an unstable departure point for em-
pirical studies. (Kerr 1977, 46, cited in Freeman 1989, 318)
In short, DKLR disagreed with the neoclassical presumption that labor
markets, especially the determination of wages and employment, operate
no differently than do goods markets.
The controversy following Lester’s challenge had methodological
consequences that reverberate still. Fritz Machlup’s 1946 response to
Lester 1946a was important less for its argument that Lester’s survey
methodology was sloppy, than for its methodological claim that pro?t
maximization is not a descriptive claim but a useful ?ction. Managers
may not know marginal magnitudes, nor need they be calculating on
all conceivable margins. But their habits and routines can be thought of
as learned strategies that themselves derive from a goal of pro?t maxi-
mization. Methodologically, it is as if ?rms maximize pro?ts (Machlup
1946).
The Minimum-Wage Controversy 129
Alchian (1950) added a natural selection story to the gap that Machlup
opened between intent and effect. Whether or not ?rms intend to maxi-
mize pro?ts, Alchian argued, a Darwinian business environment ensures
that surviving ?rms will be maximizing pro?ts, as underperforming
?rms are weeded out by competition. It is a competitive output market
structure that creates pro?t maximizing behavior, albeit as a by-product
rather than as the necessary goal of ?rms.
Milton Friedman’s famous paper (1953) borrowed elements from
both Machlup 1946 and Alchian 1950. With Machlup, Friedman em-
braced the as if method, which he in?ated into a grander claim: theo-
retical assumptions need not be true. While it is clearly unrealistic to
assume that billiard players solve systems of differential equations when
making shots, it is useful for prediction, says Friedman, to pretend that
they do. Friedman goes so far as to make “unrealistic” (read: false) as-
sumptions into a theoretical virtue: “The more signi?cant the theory, the
more unrealistic the assumptions” (1953, 14).
From Alchian, Friedman carries over the idea that an agent’s intent is
unimportant; economics should be seen not as a theory of motivational
states but of market outcomes. Friedman concedes that ?rms need not
intend to maximize pro?ts, but boldly argues that this does not matter
methodologically. Theories are to be judged not by the truth or falsity
of their assumptions, but, rather, by the success of their predictions. The
assumption that ?rms consciously optimize is false, he concedes, but
the prediction that only pro?t-maximizing ?rms will survive in the long
run is empirically accurate.
Methodologically, the realism-of-assumptions debate is something of
a muddle. Our focus is on the role of empirical results in adjudicating
among competing theories. Much turns on whether one sees assump-
tions as restricting the applicability of the theory—a domain assump-
tion—or whether one sees assumptions’ truth or falsity as negligible for
theoretical purposes, such as prediction (Musgrave 1981).
If we regard the assumption of competitive labor markets as a domain
assumption, then the disemployment prediction obtains only when the
assumption is true, that is, only when labor markets are competitive.
Stigler’s reply to Lester 1946a at one point reverts to just this domain-
assumption interpretation (1947, 155).
14
Alternatively, one can predict
14. Joan Robinson, parenthetically commenting on the minimum-wage debate, connects
the matter of assumptions to the Marshallian-era version of the controversy: “This [monopsony]
analysis may throw light on the dispute between Marshall and Mr. and Mrs. Webb upon the
130 Thomas C. Leonard
disemployment regardless of whether the competitive-market assump-
tion is true or not, a negligibility assumption. If the power of ?rms to
set wages below MRP
L
is negligible, then the falsity of the assumption
will not much matter because, ceteris paribus, the disemployment pre-
diction will still be borne out. This is Friedman’s interpretation of how
assumptions work in economics.
Both readings have important implications for application. Since
Stigler agrees that assumptions constrain the applicability of theory,
then it is clearly important that assumptions not be too restrictive; oth-
erwise, there is nothing to apply to. But Stigler, of course, does not
really attend to the question of whether maximization and competition
are overly restrictive assumptions. As we have seen, his own contribu-
tion offers only limited empirical evidence. And he is quite dismissive
of Lester’s attempt to empirically test the key assumptions.
In any event, it was Friedman’s reading that carried the day:
economists need not worry about the truth or falsity of assumptions; only
predictive accuracy matters. By implication, empirical work should be
con?ned to testing the predictions of theory.
15
Testing of assumptions,
of the kind that Lester has undertaken, can be dismissed as methodolog-
ically beyond the pale.
Although Friedman prevailed, his methodological injunction—em-
pirically test theoretical predictions—went mostly unheeded. With few
exceptions (e.g., Peterson 1957), the relative lack of mainstream empir-
ical work on minimum-wage effects persists well into the 1970s. Robert
Goldfarb’s survey (1974) of the minimum-wage literature ?nds that the
empirical support for disemployment effects, especially for adults (ages
twenty and up), who have generally composed the majority of low-wage
workers, is quite limited. With the exception of teenage employment, it
is not hyperbole to suggest that, in the Nixon era, the basis for profes-
sional priors was essentially unchanged from that of three generations
earlier.
‘marginal productivity theory’ of wages (see [Marshall’s] Principles, 705). It seems to have
arisen because Mr. and Mrs. Webb failed to realized the implications of the assumptions
of perfect competition, while Marshall failed to recognize the extreme unreality of those
assumptions” (Robinson [1933] 1950, 300–301 n. 2).
15. I do not take up Friedman’s implicit claim that theoretical components are separable,
that is, that there is a meaningful distinction between theoretical assumptions and theoretical
implications.
The Minimum-Wage Controversy 131
Empirical Evidence on Employment
Effects of Minimum Wages
The empirical literature grew rapidly with the expansion of time-series
econometric techniques. Roughly twenty-?ve papers are reviewed in
Charles Brown, Curtis Gilroy, and Andrew Kohen’s (1982) survey.
Goldfarb (1981) reports empirical progress, especially with regard to
distribution effects.
Most empirical papers in this period make use of a regression with
something like the following speci?cation:
Y = f (MW, D, X
1
. . . X
n
),
where Y is labor force status (e.g., labor force participation ratio,
employment-population ratio, unemployment rate), MW is some mea-
sure of the minimum wage,
16
D is a business cycle variable to account
for changes in aggregate economic activity, and X
i
is an exogenous
variable controlling for labor supply, school enrollment, participation
in the armed forces, and so on (Brown, Gilroy, and Kohen 1982, 497).
Some speci?cations add a time trend as well (see Mincer 1976). There
are different functional forms to the speci?cations, but nearly all studies
use Current Population Survey (CPS) data.
Brown, Gilroy, and Kohen summarize the literature as follows:
teenage employment-rate elasticities are on the order of ?0.1 to ?0.3,
but “the lower half of that range is to be preferred.” For young adults
(ages 20–24), the effect of the minimum wage on employment is “neg-
ative and smaller” than that for teens. But Brown, Gilroy, and Kohen
are careful to say that “uncertainty about the effects on [older] adults
is a serious gap in the literature . . . ,” echoing Goldfarb 1974 (1982,
524). These conclusions generally comport with contemporaneous sur-
veys and collections (Rottenberg 1981; Report of the Minimum Wage
Study Commission 1981; Eccles and Freeman 1982).
The uncertainty surrounding adult employment effects notwithstand-
ing, the disemployment prediction seems comfortably consolidated by
the 1980s (Alston, Kearl, and Vaughan 1992). There may have been
16. “The key variable, minimum wage, has generally been measured by the ratio of the
nominal legal minimum wage to average hourly earnings weighted by coverage, as devised
by Kaitz (1970). Ratios of minimum wage rates to average hourly earnings are calculated for
each industry, weighted by the proportion of workers covered. These are combined into an
index in which the weight for each industry ration is the number of persons employed in the
industry as a percentage of total employment” (Brown, Gilroy, and Kohen 1982, 499).
132 Thomas C. Leonard
quibbles about the exact magnitude of disemployment effects, but none
about whether they existed (Deere, Murphy, and Welch 1995, 232).
Indeed, there was very little empirical work on minimum wages under-
taken during the mid-eighties. What changed?
First, the few new time-series estimates which did incorporate data
from the eighties seemed to suggest that employment elasticities with
respect to minimumwages were becoming more inelastic. Second, while
the nominal value of the federal minimum wage stayed at $3.35 per
hour between 1981 and 1990, several states raised their minima, which
created a research opportunity for cross-sectional studies. Finally, the
increase in income inequality during the 1980s spurred political action
for redistributive measures. In the nineties, minimum-wage research
revived, and with dramatic consequences.
The “New” Economics of Minimum Wages
and Its Critics
I emphasize two lines of attack that Card and Krueger mount on the
neoclassical account. First, they present cross-sectional evidence from
four studies—most famously from the New Jersey–Pennsylvania (NJ-
PA) fast food experiment—which ?nds that the employment elasticity
of the minimum wage is positive, or at least nonnegative. Of Myth’s
seven analyses of employment effects, one shows zero effect and six
show positive effects, but only two of these are signi?cantly different
from 0 at the 0.05 level (1995, 388).
Second, and less well known, Card and Krueger review the estab-
lished econometric literature and judge it altogether inadequate. They
conclude that previous ?ndings are unreliable, tainted by publication
bias and speci?cation search. Paul Osterman calls this aspect of Myth “a
damning indictment of how labor economics has been practiced over the
past three decades” (1995, 839). Card and Krueger offer their own empir-
ical approach—“natural experiment” and “model-free empiricism”—as
evidentially superior to earlier econometric studies.
Card and Krueger’s Evaluation
of Previous Empirical Work
Card and Krueger (CK) review earlier empirical work on minimum-
wage effects and judge it inadequate and tendentious. CK replicate and
The Minimum-Wage Controversy 133
extend some earlier work, ?nding that employment elasticities attenuate
when more recent years are included. A meta-analysis of previous time-
series estimates concludes that earlier work is corrupted by publication
bias and speci?cation search.
CK argue that time-series estimates, over time, involve larger and
larger samples, which, ceteris paribus, should increase t -statistics. If
the additional data in newer studies are independent of the older data,
“then a doubling of the sample size should result in an increase in the
absolute value of the t -ratio of about 40 percent” (1995, 187).
17
If the t -
statistics do not increase with sample size, then, CK suggest, something
is ?shy. Using the time-series studies reviewed in Brown, Gilroy, and
Kohen 1982, CK ?nd that the t -statistics are, in fact, declining in sample
size (188).
18
CK also conduct a related meta-analysis that compares the relation-
ship between the estimated coef?cients of employment elasticity with
their respective standard errors, for the same set of time-series studies.
If the estimated coef?cients (b) are truly unbiased, CK argue, then there
should be no systematic relationship between them and their respec-
tive standard errors (se). If, on the other hand, editors will only publish
studies with robust test statistics (t > 2, say), then this publication bias
should be re?ected in the meta-analysis, given that t = b/se. Plotting
“implicit” standard errors against the coef?cients, CK ?nd a rather good
?t (for b = 2se), with only two or three exceptions, suggesting a problem
with independence (1995, 191).
From this meta-evidence, CK offer three possible explanations. One
explanation is that structural change made the employment effects of
minimum wages decline over time. But structural change calls into
question the validity of time-series work to begin with. The alternative
17. Card and Krueger note the lack-of-independence problem that occurs with time-series
data more generally, but point out that previous authors themselves assume independence
following adjustments for serial correlation in errors and similar procedures.
18. Actually, they plot the absolute value of the t -statistic against the square root of
the degrees of freedom, or (n ? k)
1/2
. The rationale is explained by Kennan 1995. The
t -statistic in a regression with n observations on k explanatory variables satis?es the equa-
tion t
2
(n ? k) = r
2
/(1 ? r
2
), where t is the t -statistic of a regressor and r is the par-
tial correlation coef?cient between this regressor and the dependent variable. As the sam-
ple size increases, r converges to its population value, assuming stationarity of the partial
correlation, so t must grow at the same rate as the square root of n ? k (Kennan 1995,
1963).
134 Thomas C. Leonard
explanations are speci?cation search and publication bias. CK argue
that the early literature was contaminated by a preference (on the part
of data-mining researchers and biased editors) for statistically signi?-
cant results that comported with the disemployment predictions of the
neoclassical model.
The econometric speci?cations which became standard were chosen,
CK suggest, because they produced the desired negative sign and got
the t -statistics above 2. Later researchers tended to reuse these speci?-
cations. Because true statistical signi?cance was overstated in the earlier
research, however, more recent work (presumably not so susceptible to
temptation) discovered weaker effects (1995, 192–93). Three of Finis
Welch’s studies are singled out by CK as examples of this econometric
malpractice.
Natural Experiments and “Model-Free”
Empiricism
Myth is notable for its emphasis on natural experiments (NE) and, more
generally, the idea of “model-free” empirical work. CK take the NJ-
PA study as their most important result, precisely because they believe
that its setting is closest to the ideal of the randomized controlled ex-
periments of the “hard” sciences. They imply that an NE, even if an
imperfect substitute for a randomized controlled experiment, is eviden-
tially superior to traditional econometric techniques.
The New Jersey–Pennsylvania (NJ-PA) study compares employment
responses in the fast-food market in New Jersey and eastern Pennsylva-
nia to increases in the minimum wage (New Jersey raised its state min-
imum from $4.25 to $5.05, while Pennsylvania’s remained at $4.25).
The “experiment” compares the employment effects on affected work-
ers against the Pennsylvania “control” group. If all differences between
New Jersey and Pennsylvania remain unchanged before and after the
minimum-wage increase, then any employment (E
it
) “difference in
differences” (DD) can be attributed to the minimum-wage effects, as
follows:
DD = (E
NJ2
? E
PA2
) ? (E
NJ1
? E
PA1
),
where period 1 is February 1992 (before the new minimum takes effect
in April 1992) and period 2 is November 1992. Neoclassical theory
predicts that, ceteris paribus, employment in New Jersey should fall
The Minimum-Wage Controversy 135
relative to employment in Pennsylvania, that is, that DDwill be negative.
CK, however, ?nd it to be positive.
19
CK sometimes make extravagant claims regarding the signi?cance
of this ?nding, even as they are careful to acknowledge the relative
weaknesses of NEs. First, NEs do not randomly assign subjects into
treatment and control groups, which is essential for attributing any dif-
ferential outcomes to the minimum wage exclusively. Minimum-wage
legislation is the outcome of a political process, and states that enact
minimum wages may differ from those that do not; they may have more
robust economic growth, for example (1995, 23). Second, NEs are not
“blind.” Subjects know when they are (or are not) being treated. Third,
NEs require ceteris paribus. Any current differences between treatment
and control groups must persist in the absence of the minimum-wage
intervention (e.g., both groups should exhibit identical rates of growth)
(23).
CK are con?dent that these departures from classical experiments are
small relative to the advantage NEs provide over traditional econometric
approaches. The NE approach is “model free” in that its results do not
hinge on a particular econometric speci?cation or, therefore, upon the
theoretical assumptions that underwrite a given speci?cation. And, at
least in principle, NEs can reduce the incentive for speci?cation search,
the temptation to ?t the model to the data.
Replies to Card and Krueger
The replies to Myth are too numerous to document fully, a measure of
its in?uence. Professional critics have focused less on the nature than
on the execution of CK’s empirical methods, especially in the famous
NJ-PA study.
The most serious replies are as follows: (1) the survey is poorly de-
signed as re?ected in the noisy data that result (Kennan 1995; Welch
1995); (2) the study is not a proper NE, particularly with respect to the
control group (Hamermesh 1995; Kennan 1995; Welch 1995); (3) re-
sults are short run only (Hamermesh 1995); (4) employment increases
should mean higher output and lower output prices, but CK ?nd evi-
dence of higher output prices (Brown 1995); and, most conspicuously,
19. Employment is measured in full-time equivalents (FTEs), where a part-time worker is
counted as 0.5 FTE. The results: 2.7 = [21.0 ? 21.2] ? [20.4 ? 23.3].
136 Thomas C. Leonard
(5) results cannot be replicated and are contradicted when superior data
are employed (Neumark and Wascher 1995).
20
Low-wage employers may well have already adjusted to the New
Jersey minimum-wage increase before it took effect in April 1992, given
that the statute passed in 1990. Or their adjustment may have occurred
after the second survey. And the replication by David Neumark and
William Wascher (1995) does contradict CK’s original work.
21
And
CK’s most important result—that employment in New Jersey increases
relative to employment in Pennsylvania—obtains principally because
employment fell in Pennsylvania, the control. It is, John Kennan argues,
“like having a drug trial in which the drug has no effect but the placebo
makes people sick” (1995, 1958).
Conclusion: The Very Idea of Applying Economics
The “new” economics of the minimum wage is not really new at all.
Although Card and Krueger do not mention it, their method of natu-
ral experimentation has direct empirical antecedents in a little-known
statistical literature developed in the 1910s and 1920s. There are large
repositories of “before and after” employment data collected by the U.S.
Bureau of Labor Statistics and by various state agencies.
22
And Card and
Krueger’s empiricism and their monopsony explanation have acknowl-
edged af?nities with the DKLR approach. But two things do distinguish
the contemporary controversy: (1) the substantive emphasis on employ-
ment effects and (2) the methodological emphasis on empirical work as
the arbiter of successful theoretical application.
As the debate over minimum wages increasingly stresses its role as a
test of the applicability of price theory to labor markets, there is a con-
comitant emphasis on employment effects. Reactions by nonspecialists
to Myth are almost entirely in this vein. Said one Nobel laureate:
20. A more detailed account is available in Leonard 1997, 268–73.
21. Card and Krueger 1998 replies to Neumark and Wascher 1995.
22. Kennan 1995 instances a remarkably detailed study of minimum-wage effects in Ore-
gon (Obenauer and von der Nienburg 1915), which begins, “Does the establishment of a legal
minimum wage for women do more harm than good to the women it is designed to assist?” and
which presents employment and wage data for tens of hundreds of affected girls and women.
The BLS published eleven bulletins on minimum-wage issues prior to Obenauer and von der
Nienburg 1915, and there are detailed data in numerous state publications, such as National
Industrial Conference Board 1927.
The Minimum-Wage Controversy 137
Just as no physicist would claim that “water runs uphill,” no self-
respecting economist would claim that increases in the minimum
wage increase employment. Such a claim, if seriously advanced, be-
comes equivalent to a denial that there is even minimal scienti?c
content in economics, and that, in consequence, economists can do
nothing but write as advocates for ideological interests. (Buchanan
1996)
Labor specialists are less intemperate but equally preoccupied with dis-
employment. Myth examines many consequences of minimum wages,
but CK themselves make central their own ?ndings and their critique of
previous econometric work on employment effects.
The emphasis on employment has crowded out a crucial, and answer-
able, policy question: Do minimum wages reduce poverty? A desire to
help the working poor, of course, lies at the root of minimum-wage
policy. An obviously relevant question is, Are affected workers poor?
(Stigler 1946).
23
The surprising answer is, generally, no. The reasons for
a positive but weak relationship between minimumwages and household
income are well understood and well documented (Gramlich 1976).
The relationship is weak when low-wage workers are members of
nonpoor households (teens, say), or when households have nonwage
income, or when households have no workers at all. Low-wage workers
are more likely to be in poor families than are all workers, but the great
majority of low-wage workers are not poor. Ralph Smith and Bruce
Vavrichek (1987) found that 75 percent of affected workers in 1985 were
neither poor nor near-poor. CK’s own distribution analysis ?nds that
only 30 percent of affected workers live in poor and near-poor families
(1995, 280). In other words, over two-thirds of income redistributed by
minimum-wage increases—even assuming zero disemployment—will
go to individuals in families neither poor nor near-poor.
Myth concludes, “We believe that our ?ndings suggest a reorientation
of policy discussions away from the ef?ciency aspects of the minimum
wage and toward distributional issues” (393). But, at the same time,
CK are obliged to say that the minimum wage is a “blunt instrument”
for redistributing income to poor families and ?nd that “the effect of
23. Not all economists accept alleviation of poverty as the goal of minimum wages. Some
argue that ?attening the income distribution is worthwhile for its own sake, and others sug-
gest that progressive income redistribution stimulates aggregate demand, owing to a higher
propensity to consume in lower-income households (Prasch 1996).
138 Thomas C. Leonard
the minimum wage on the overall poverty rate of adults is statistically
undetectable” (1995, 285, 280; see also Burkhauser, Couch, and Witten-
berg 1996).
24
Even if one believes that minimum-wage increases do not
disemploy low-wage workers, or is prepared to suffer disemployment as
a cost of redistribution, it is signi?cant that minimum wages do so little
for the goals of poverty reduction or progressive income redistribution.
The evidence on distribution—where the data are comparatively reli-
able and uncontroversial—has been completely overshadowed by the
emphasis on whether the disemployment prediction obtains.
The methodological emphasis on empirical work as the arbiter of suc-
cessful theoretical application is also new to the contemporary period.
Friedman (1953), we have seen, made testing of theoretical predictions
the paramount goal of economic science, but only in the contemporary
controversy does empirical work actually assume this role. CK are ex-
plicit: they regard their work not merely as four papers contributing to
a large empirical literature, but as “a simple and direct test of the kind
of theoretical reasoning that economists routinely apply to other, more
complicated phenomena” (1995, 397).
Some are prepared to renounce Myth on a priori grounds. Evidence
that says “water runs uphill” is to be rejected on its face. But labor
is the quintessentially empirical ?eld in economics. Labor economists
take pride in their willingness to submit theoretical claims to the tribunal
of empirical evidence. In other corners of economics, such as macro-
economics, says Richard Freeman, “the evidentiary base is suf?ciently
diffuse to allow economists with different priors to reach different con-
clusions—to make a lawyer’s case as it were” (1995, 831). But, Free-
man argues, not so in minimum-wage research, where changes occur
in discrete exogenous jumps, and evidence permits far less scope for
“disparate interpretation” (831).
Nearly all labor specialists in the controversy, critics and supporters
alike, take for granted (at least nominally) that attempts to apply theory
must be tested by empirical evidence. CK’s valedictory to Myth says that
“although many economists may disagree with our interpretation of the
?ndings in this book, we hope that they will at least agree on the value of
testing the implications of standard economic theory, and on the validity
of our empirical approach” (1995, 399). Their hope has been met. I can
24. A minimum wage can help only the employed. Card and Krueger (1995, 305) estimate
that about two-thirds of adults in poor families do not work at all. Burkhauser, Couch, and
Wittenberg (1996, 551) estimate that fully one-quarter of poor families have no working adults.
The Minimum-Wage Controversy 139
locate no serious objection, even among CK’s strongest critics, to the
idea of empirical evidence as testing theory or to the evidential virtues
of natural experiments. But a methodological ambivalence persists. If
all agree that the data should be decisive, no one agrees on when the
data are decisive.
I conjecture that the ambivalence regarding empirical testing has two
sources. One is methodological. To answer the key substantive ques-
tion—do Myth’s results refute price theory’s disemployment predic-
tion?—one must know the answer to a prior methodological question:
what evidence constitutes a refutation? If Myth really offers “a simple
and direct test” of neoclassical price theory, then one must know what
counts as a simple and direct test. Without methodological argument,
still less consensus, there are no ready means for evaluating CK’s claim.
The other source of ambivalence derives from a residual unease with
the robustness of the empirical evidence on disemployment. The evi-
dence for adults, we have seen, was never as strong as that for teenage
workers. Furthermore, CK’s most contentious claim—that the entire
empirical literature before circa 1990 is contaminated by econometric
malpractice—has gone virtually unrebutted. The silence is fairly deaf-
ening. The only paper to directly take issue with CK’s extraordinary
indictment is Neumark and Wascher 1996.
25
Finally, good evidence on minimum-wage effects is hard to obtain,
even given the relatively advantageous NE setting, as Kennan 1995
points out. Even ignoring legal noncompliance and survey measurement
error, there are thorny problems. First, employers can respond to wage
increases by reducing nonpecuniary compensation—health insurance,
training, pleasant work environment, scheduling ?exibility, discounts
on goods and services. This simply rearranges the composition of the
worker’s pay packet and should not affect employment (Kennan 1995).
Second, employers might reduce hours without reducing employment,
so measuring employment will miss changes in labor demand. Third,
affected workers already have highly volatile employment patterns. The
noise in the employment data makes it dif?cult to disentangle changes
25. Neumark and Wascher (1996) examine Myth’s meta-analysis of the time-series litera-
ture and argue that CK’s results are better explained by structural change than by publication
bias leading to speci?cation search. Parameter values changed, they say, producing spuri-
ous evidence of publication bias. Neumark and Wascher do not consider the implications of
structural change for the validity of time-series evidence.
140 Thomas C. Leonard
caused by statutory minima from those caused by confounding fac-
tors—the proverbial needle-in-the-haystack problem (Kennan 1995).
All of these empirical challenges are well known to labor researchers.
But they are troubling. Methodologically, the belief that empirical ev-
idence “decides” when theory applies obviously has less force when
the evidence tends to be equivocal. Substantively, if the evidence for
minimum-wage effects is equivocal—when, according to its producers,
it offers the best of empirical opportunities—then what does this imply
about the quality of evidence in economics more generally?
Herein lies a basic disciplinary tension. Modern labor economics is
empirical. Indeed, in some respects, Myth can be read as an empiricist
manifesto—go forth and test, and if the data are inconvenient for the
theory, so much the worse for theory. But modern labor economics is
also, unlike its mid-century variant, ?rmly neoclassical (Freeman 1989,
318). Price theory, like its predecessors, has the goal of producing a
general theory of economic behavior, which places the determination of
wages and employment comfortably within its purview—markets are
markets are markets. And a general theory that does not apply in one
domain is not a general theory. These twin disciplinary ambitions—to
be an empirical policy science, and to have a general theory of economic
behavior—can clearly come into con?ict, especially when, as with min-
imum wages, compelling evidence meets core theoretical belief.
The economist’s dilemma is as follows: do I reject the evidence or
reject the application of theory? Most respondents to Myth have chosen
the ?rst option—some on principle, but most on grounds that the new
evidence is too fragile to justify abandoning core doctrine. Still others
argue that more evidence is required. Kennan says:
Myth and Measurement’s lasting contribution may well be to show
that we just don’t know how many jobs would be lost if the minimum
wage were increased to $5.15, and that we are unlikely to ?nd out
using more sophisticated methods of inference on the existing body
of data. What is needed is more sophisticated data. (1995, 1964)
More sophisticated data are surely desirable. But the con?icting ambi-
tions of economics will continue to manifest themselves in recurring
controversy, at least until the data are so sophisticated that economists
routinely answer methodological questions such as, What evidence
would convince you that your theory is refuted?
The Minimum-Wage Controversy 141
References
Alchian, Armen. 1950. Uncertainty, Evolution, and Economic Theory. Journal of
Political Economy 58.3:211–21.
Alston, Richard, J. R. Kearl, and Michael Vaughan. 1992. Is There a Consensus
among Economists in the 1990s? American Economic Review Papers and Pro-
ceedings 82.2:203–9.
Barro, Robert. 1996. Higher Minimum Wage, Higher Dropout Rate. Wall Street
Journal, 11 January, A14.
Brown, Charles. 1988. Minimum Wage Laws: Are They Overrated? Journal of
Economic Perspectives 2.3:133–45.
———. 1995. Comment for review symposium on Myth and Measurement: The
New Economics of the Minimum Wage, by David Card and Alan Krueger. Indus-
trial and Labor Relations Review 48.4:828–30.
Brown, Charles, Curtis Gilroy, and Andrew Kohen. 1982. The Effect of the Mini-
mum Wage upon Employment and Unemployment. Journal of Economic Liter-
ature 20.2:487–528.
Buchanan, James. 1996. Commentary on the Minimum Wage. Wall Street Journal,
25 April, A20.
Burkhauser, Richard, Kenneth Couch, and David Wittenberg. 1996. “Who Gets
What” from Minimum Wage Hikes: A Re-estimation of Card and Krueger’s
Distributional Analysis in Myth and Measurement: The New Economics of the
Minimum Wage. Industrial and Labor Relations Review 49.3:547–52.
Card, David, and Alan Krueger. 1995. Myth and Measurement: The New Economics
of the Minimum Wage. Princeton, N.J.: Princeton University Press.
———. 1998. AReanalysis of the Effect of the NewJersey MinimumWage Increase
on the Fast-Food Industry with Representative Payroll Data. NBER Working
Paper no. 6386. Cambridge, Mass.: NBER.
Deere, Donald, Kevin Murphy, and Finis Welch. 1995. Employment and the
1990–1991 Minimum-Wage Hike. American Economic Review Papers and Pro-
ceedings 85.2:232–43.
Douglas, Dorothy. 1919. American Minimum Wage Laws at Work. American Eco-
nomic Review 9.4:701–32.
Dunlop, John. 1944. Wage Determination under Trade Unions. New York: Augustus
M. Kelley.
Eccles, Mary, and Richard Freeman. 1982. What! Another Minimum Wage Study?
American Economic Review Papers and Proceedings 72.2:226–32.
Freeman, Richard. 1989. Labor Markets in Action. Cambridge: Harvard University
Press.
———. 1995. Comment for review symposium on Myth and Measurement: The
New Economics of the Minimum Wage, by David Card and Alan Krueger. Indus-
trial and Labor Relations Review 48.4:830–34.
Friedman, Milton. 1953. The Methodology of Positive Economics. In Essays on
Positive Economics. Chicago: University of Chicago Press.
142 Thomas C. Leonard
Goldfarb, Robert. 1974. The Policy Content of Quantitative Minimum Wage Re-
search. Industrial Relations Research Association Proceedings: 27:261–68.
———. 1981. The Context of Recent Research. In The Economics of Legal Mini-
mum Wages, edited by Simon Rottenberg. Washington, D.C.: American Enter-
prise Institute.
Gramlich, Edward. 1976. Impact of MinimumWages on Other Wages, Employment,
and Family Incomes. Brookings Papers on Economic Activity, no. 2:409–51.
Grossman, Jonathan. 1978. Fair Labor Standards Act of 1938: Maximum Struggle
for a Minimum Wage. Monthly Labor Review 101.6:22–30.
Hall, R. L., and C. J. Hitch. 1939. Price Theory and Business Behavior. Oxford
Economic Papers, no. 2:13–45.
Hamermesh, Daniel. 1995. Comment for review symposium on Myth and Mea-
surement: The New Economics of the Minimum Wage, by David Card and Alan
Krueger. Industrial and Labor Relations Review 48.4:835–38.
Holcombe, A. 1910. The British Minimum Wages Act of 1909. Quarterly Journal
of Economics 24.3:574–77.
Hutchison, Terrence. 1953. A Review of Economic Doctrines 1870–1929. Oxford:
Clarendon Press.
Kelley, Florence. 1912. Minimum-Wage Laws. Journal of Political Economy
20.10:999–1010.
Kennan, John. 1995. The Elusive Effects of Minimum Wages. Journal of Economic
Literature 33.4:1950–65.
Kerr, Clark. 1977. Labor Markets and Wage Determination. Berkeley: University
of California Press.
Krueger, Alan, and Lawrence Summers. 1988. Ef?ciency Wages and the Inter-
industry Wage Structure. Econometrica 56.2:259–93.
Lees Smith, H. B. 1907. Economic Theory and Proposals for a Legal Minimum
Wage. Economic Journal 17.68:504–12.
Leonard, Thomas. 1997. The Reason of Rules in the Intellectual Economy: The
Economics of Science and the Science of Economics. Ph.D. diss., The George
Washington University.
Lester, Richard. 1946a. Shortcomings of Marginal Analysis for Wage-Employment
Problems. American Economic Review 36.1:62–82.
———. 1946b. Wage Diversity and Its Theoretical Implications. Review of Eco-
nomics and Statistics 28.3:152–59.
———. 1947. Marginalism, Minimum Wages, and Labor Markets. American Eco-
nomic Review 37.1:135–48.
Longe, F. D. 1894. The Coal Strike and a “Minimum” Wage. The Economic Journal
4.13:25–34.
Machlup, Fritz. 1946. Marginal Analysis and Empirical Research. American Eco-
nomic Review 36.4:519–54.
———. 1947. Rejoinder to an Antimarginalist. American Economic Review
37.1:148–54.
The Minimum-Wage Controversy 143
Marshall, Alfred. 1897. The Old Generation of Economists and the New. Quarterly
Journal of Economics 11.2:115–35.
Mill, John Stuart. 1848. Principles of Political Economy with Some of Their Appli-
cations to Social Philosophy. Boston: Little & Brown.
Miller, Merton. 1996. Commentary on the Minimum Wage. Wall Street Journal, 25
April, A20.
Millis, H. A. 1914. Some Aspects of the Minimum Wage. Journal of Political
Economy 22.2:132–55.
Mincer, Jacob. 1976. Unemployment Effects of Minimum Wages. Journal of Polit-
ical Economy 84.4, part 2: 87–104.
Musgrave, Alan. 1981. “Unreal Assumptions” in Economic Theory: The F-Twist
Untwisted. Kyklos 34.3:377–87.
National Industrial Conference Board. 1927. Minimum Wage Legislation in Mas-
sachusetts. New York: NICB.
Neumark, David, Mark Schweitzer, and William Wascher. 1998. The Effects of
Minimum Wages on the Distribution of Family Incomes: A Non-Parametric
Analysis. NBER Working Paper no. 6535. Cambridge, Mass.: NBER.
Neumark, David, and William Wascher. 1995. The Effect of New Jersey’s Mini-
mum Wage Increase on Fast-Food Employment: A Reevaluation Using Payroll
Records. NBER Working Paper no. 5224. Cambridge, Mass.: NBER.
———. 1996. Is the Time-Series Evidence on Minimum Wage Effects Contami-
nated by Publication Bias? NBER Working Paper no. 5631. Cambridge, Mass.:
NBER.
Obenauer, Marie, and Bertha von der Nienburg. 1915. Effect of Minimum-Wage De-
terminations in Oregon. Bulletin of the U.S. Bureau of Labor Statistics. Number
176. Washington, D.C.: U.S. Government Printing Of?ce.
Osterman, Paul. 1995. Comment for review symposium on Myth and Measurement:
The New Economics of the Minimum Wage, by David Card and Alan Krueger.
Industrial and Labor Relations Review 48.4:839–42.
Peterson, John. 1957. Employment Effects of MinimumWages, 1938–1950. Journal
of Political Economy 65.5:412–30.
Pigou, A. C. 1922. Trade Boards and the Cave Committee. The Economic Journal
32.127:315–24.
Prasch, Robert. 1996. In Defense of the Minimum Wage. Journal of Economic
Issues 30.2:391–97.
Report of the Minimum Wage Study Commission. 1981. Washington, D.C.: U.S.
House of Representatives.
Reynolds, Lloyd, and C. Taft. 1956. The Evolution of the Wage Structure. New
Haven, Conn.: Yale University Press.
Robinson, Joan. [1933] 1950. The Economics of Imperfect Competition. London:
Macmillan.
Rottenberg, Simon, ed. 1981. The Economics of Legal MinimumWages. Washington,
D.C.: American Enterprise Institute.
144 Thomas C. Leonard
Sidgwick, Henry. 1886. Economic Socialism. Contemporary Review 50 (Novem-
ber): 620–31.
Smith, Ralph, and Bruce Vavrichek. 1987. The Minimum Wage: Its Relation to
Incomes and Poverty. Monthly Labor Review 110.6:24–30.
Stigler, George. 1946. The Economics of Minimum Wage Legislation. American
Economic Review 36.3:358–65.
———. 1947. Professor Lester and the Marginalists. American Economic Review
37.1:154–57.
———. 1982. The Economist as Preacher and Other Essays. Chicago: University
of Chicago Press.
Webb, Sidney. 1912. The Economic Theory of a Legal Minimum Wage. Journal of
Political Economy 20.10:973–98.
Webb, Sidney, and Beatrice Webb. [1897] 1920. Industrial Democracy. London:
Longmans, Green.
Welch, Finis. 1995. Comment for review symposium on Myth and Measurement:
The New Economics of the Minimum Wage, by David Card and Alan Krueger.
Industrial and Labor Relations Review 48.4:842–49.
Whaples, Robert. 1996. Is There a Consensus among American Labor Economists?
Journal of Labor Research 17.4:725–34.
doc_615295575.pdf
The modern minimum-wage controversy derives from a recent literature in empirical labor economics, which argues, foremost among several claims, that moderate increases in mandated minimum wages do not lead to adverse employment outcomes for low-wage workers.
The Very Idea of Applying Economics:
The Modern Minimum-Wage Controversy
and Its Antecedents
Thomas C. Leonard
Disagreements among economists rarely breach the academy’s walls.
The recent minimum-wage controversy is a signi?cant exception. At
?rst glance, all the sound and fury is a puzzle.
1
Minimum-wage effects,
at least for current U.S. magnitudes, are small potatoes.
2
Of several
more important policy concerns—entitlement reform, health insurance,
CPI calculation, for example—none has generated anything like the
vituperation that has characterized the minimum-wage controversy. But
if the minimum-wage controversy is not especially important to the
economy, it is very important to economics, and, thereby, to the status
of economics as a policy science.
Correspondence may be addressed to Professor Thomas C. Leonard, Department of Econom-
ics, Fisher Hall, Princeton University, Princeton, NJ 08544; e-mail: [email protected]. I
am grateful to Jeff Biddle, Bob Goldfarb, an anonymous referee, and participants in the 1999
HOPE conference on the history of applied economics for helpful comments on an earlier
draft of this material.
1. In the wake of Card and Krueger 1995, one Nobel laureate intoned, “I tremble for my
profession” (Miller 1996). Another made reference to “camp-following whores” (Buchanan
1996). A third eminence called Card and Krueger’s main interlocutors, David Neumark and
William Wascher, “heroes” (Barro 1996).
2. Even a 27 percent increase in the U.S. minimum wage (from $3.35 to $4.25, in 1990 and
1991) affected only about 8 percent of the labor force, and these workers were concentrated
in the retail trade and food service sectors. Ignoring employment effects, Card and Krueger
estimate that $5.5 billion was transferred to low-wage workers, only 0.2 percent of total U.S.
annual earnings (1995, 277). Card and Krueger estimate that if ?rms had passed along all of
the labor cost increases to consumers, prices would have increased (once and for all) only
about 0.3 percent (393).
118 Thomas C. Leonard
The reason, I want to argue, is that the core of modern economics—
neoclassical price theory—is seen to be at stake. In particular, minimum-
wage research has come to be seen as a test of the applicability of
neoclassical price theory to the determination of wages and employment.
The modern minimum-wage controversy is not just a technical quarrel
over the sign and magnitude of wage-elasticity coef?cients; it is the
latest chapter in a longstanding methodological dispute over whether
and in what domains neoclassical price theory can be said to properly
apply. One does not make opposition to minimum wages a disciplinary
litmus test, unless there is something this important at stake.
3
The modern minimum-wage controversy engages economists be-
cause it brings to light methodological issues ordinarily submerged in
the course of normal science. I present a history of the economics of min-
imum wages designed to illuminate two such issues: (1) does prevailing
theory explain and predict the determination of wages and employment,
and (2) by what means do economists decide when explanation and
prediction are successful?
The history—really three vignettes—is designed to show, ?rst, that
the current debate is a modern-dress version of an old family argument
in economics. Second, it provides some evidence for my claim that the
minimum-wage controversy is not merely a substantive disagreement
among labor specialists, but also a recurring touchstone for method-
ological debate about the application of economics.
What Is the Modern Minimum-Wage
Controversy About?
The modern minimum-wage controversy derives from a recent literature
in empirical labor economics, which argues, foremost among several
claims, that moderate increases in mandated minimumwages do not lead
to adverse employment outcomes for low-wage workers.
4
Many of these
?ndings are collected or summarized in David Card and Alan Krueger’s
3. George Stigler said in 1976 that “one evidence of professional integrity of the economist
is the fact that it is not possible to enlist good economists to defend . . . minimum wage laws”
(Stigler 1982, 60).
4. Card and Krueger say that their “strongest and most important ?nding” is the absence of
disemployment effects from moderate increases in minimum wages (1995, 387). The quali?er
“moderate” is important. No economist believes that ?xing a wage ?oor at $100 per hour
would not create disemployment. Card and Krueger also recognize that the intensity of the
minimum-wage debate is out of proportion to its economic importance (1995, 395).
The Minimum-Wage Controversy 119
in?uential 1995 book, Myth and Measurement: The New Economics of
the Minimum Wage (hereafter cited as Myth).
The current controversy arises because “the new economics of the
minimum wage” is at odds with neoclassical price theory, which pre-
dicts disemployment, and with a generation of time-series econometric
research, which consistently ?nds evidence of disemployment effects.
Myth’s ?ndings are also at odds with what most U.S. economists, es-
pecially labor economists, believe (Whaples 1996). Card and Krueger
understand this; indeed, they expressly conceive of their project as a
refutation of neoclassical price theory’s predictions and of the econo-
metric evidence adduced in support (1995, 396–97).
Following the contemporary emphasis, I focus on employment and
welfare effects: do minimum-wage increases disemploy low-wage
workers, and do minimum-wage increases, on balance, help low-
wage workers? Later I take up the less emphasized but essential matter
of distribution: To what extent do workers affected by minimum-wage
legislation live in poor or near-poor households?
5
The Neoclassical Theory of
Minimum-Wage Effects
The neoclassical theory is familiar. Mandated minimum wages are
treated as any other price ?oor. If the mandated wage exceeds the
market-clearing wage, then, given a downward-sloping labor demand
curve, ?rms reduce the quantity of labor demanded. The magnitude of
the reduction depends upon the wage increase and the wage elasticity
of labor demand, but the direction of change is unambiguous. Welfare
effects also depend upon the wage elasticity of demand. Some workers
will receive the higher wage and are better off. Other workers, those
whose product is worth less than the new minimum, will be laid off, or
will work fewer hours. If the quantity of labor refers to employment,
then the wage gains of those who keep their jobs must be traded off
against the wage losses of those who lose their jobs.
6
5. Poor and near-poor families are those with incomes at or below 1.5 times the U.S.
poverty level.
6. Suppose, for example, that the wage elasticity of demand is ?0.2 and that the minimum-
wage increase is 10 percent. This implies that 2 percent of workers (if L is employment) will
lose their jobs, and that 98 percent will get a 10 percent raise, meaning that affected low-
wage workers collectively realize income gains of 7.8 percent [(0.98)(0.1) ?(.02)(1.00)]. By
implication, minimum wages will help the class of affected workers when the wage elasticity
120 Thomas C. Leonard
Disemployment is a logically necessary result if one assumes that
(1) ?rms maximize pro?ts and (2) the low-skilled labor market is com-
petitive, that is, that ?rms have no monopsony power (Kennan 1995).
Neoclassical theory can accommodate a weakening of assumption 2.
If labor markets are monopsonistic, ?rms exercise their market power
and pay a wage lower than workers’ marginal revenue product (MRP
L
).
Monopsonistic ?rms offer higher wages to attract newhires, which leads
to costlier labor expenditures (more pay to the inframarginal workers)
and inef?ciently low employment.
In principle, then, a judiciously chosen minimum wage (set between
monopsony wage and MRP
L
) can increase employment and ef?ciency.
In practice, however, labor economists traditionally regard monopsony
as an empirical curiosum (Brown 1988, 134 n. 1).
7
Individual low-wage
?rms typically employ a small percentage of the low-wage work force
(Brown, Gilroy, and Kohen 1982, 489), so monopsony power is seen as
limited in practice.
8
Neoclassical economics is less able to accommodate a weakening
of the pro?t-maximization assumption, which is more fundamental.
With few exceptions (see, for example, Prasch 1996), contemporary
economists maintain the pro?t-maximization assumption, generally re-
verting to monopsony models derived from ef?ciency-wage explana-
tions.
If some ?rms pay more than the going wage, and pro?t maximization
is assumed, then the only remaining explanation is that some ?rms ?nd it
pro?table to pay above the going rate, which de?nes an ef?ciency-wage
explanation (Krueger and Summers 1988). Ef?ciency-wage accounts
propose that wages can exceed MRP
L
when (1) higher wages bring
forth greater worker productivity, a reversal of the traditional causality,
is less elastic than ?1.0. But there are two important caveats. First, not all affected workers
get a 10 percent raise; those between the old and new minima get a smaller increase. Second,
wage elasticities generally derive from studies that calculate employment effects for entire
demographic groups, not just for the minimum-wage workers within those groups. Thus,
an elasticity estimate such as ?0.2 measures the effect of minimum-wage increases on the
employment of some group (say, teenagers), which includes higher-wage (hence unaffected)
workers, and thereby underestimates the employment effects on the lowest-wage workers.
(These points are due to Neumark, Schweitzer, and Wascher 1998, 3–4).
7. Robert Whaples ?nds that “six out of seven labor economists believe that ‘few American
workers are employed in monopsonistic labor markets’” (1996, 726).
8. Even with monopsony power, the range of employment-increasing wage increases is
limited. A minimum set above the MRP
L
at the monopsonist’s pro?t-maximizing output also
results in disemployment.
The Minimum-Wage Controversy 121
(2) there is an informational asymmetry such that ?rms cannot readily
monitor worker effort, and (3) ?rms have high recruitment and turnover
costs (the last is emphasized by Card and Krueger 1995, 373–79). Since
pro?t maximization and perfect competition entail disemployment with
higher minima, it is not surprising that, historically, both assumptions
recur in arguments against and for the minimum wage.
History of Minimum-Wage
Legislation and Its Economics
The idea of administered wages is ancient. Wage payments to laborers
were conventional and noncontractural long before the industrial revo-
lution and the concomitant rise of freely contracting labor. But the idea
of administered wages as a legal means of improving the lot of the urban
working poor has its origins in nineteenth-century reform movements,
especially in the United Kingdom. Minimum wages were connected to
mid-Victorian-era trade union movements and were an important item
on the Fabian socialist agenda to “democratize” unregulated industries
and improve working conditions (Webb and Webb [1897] 1920).
9
The ?rst laws setting wage ?oors, as against a ?xed wage or an output-
price peg, have their origin in Australasia. The District Conciliation
Boards of NewZealand, established in 1894 principally for arbitration of
labor disputes, had statutory authority to set minimum wages (Douglas
1919, 701). The ?rst independent legal agencies designed speci?cally
for setting minimum wages were established by the 1896 Factory and
Shops Act, enacted by the state of Victoria to ?x minimum wages in
“sweated” industries (701).
10
Britain passed the Trade Boards Act of
1909, which likewise set minimum wages in several industries deemed
to be “sweating.”
9. Aminimumwage means only that workers cannot legally accept a wage belowit. Higher
wages are legal. How and in what settings the minimum is to be ?xed is more ambiguous, as
the compelling but vague term living wage suggests. See Pigou 1922.
10. De?nitions of “sweating” vary. Reporting on the 1888–90 investigation of “sweated
industries” by a select committee in the British House of Lords, Beatrice Webb offers the
following de?nition: “Lord Derby and his colleagues decided that sweating was no particular
method of remuneration, no particular form of industrial organization, but certain conditions
of employment—viz., unusually low rates of wages, excessive hours of labor, and unsani-
tary workplaces. When we get any one of these conditions in an exaggerated and extreme
form, . . . then we may say, that the labor is sweated, and that the unfortunates are working
under the sweating system” (cited in Holcombe 1910, 574).
122 Thomas C. Leonard
The late-nineteenth-century minimum-wage movement in the United
States was originally voluntarist in spirit. Organizations like the Na-
tional Consumers League hoped, by information provision and moral
suasion, to persuade consumers to avoid ?rms that paid especially low
wages to workers—a minimum wage was seen as a moral rather than
a legal standard (Kelley 1912). State legislation overtook their efforts.
Massachusetts (1912) was the ?rst, followed a year later by Wisconsin,
Minnesota, Washington, Oregon, California, Colorado, Nebraska, and
Utah (see Millis 1914).
Some laws empowered supervisory boards to determine minimum
wages; others directly ?xed minima by statute. In all cases, the states’
laws, unlike their British Commonwealth precursors, applied to women
and minors only. Until the advent of a federal minimum in 1938, men
were deemed outside such paternal protection, in light of the U.S. Con-
stitution’s explicit protection of the right to free contract.
11
Stretching back into the classical era is the question of whether a
general theory—that of free markets and how they determine price and
quantity—can be applied to matters of wage and employment deter-
mination. In other words, is it the case, as Adam Smith said, that the
market for men is like the market for goods, so that there exists a “labor
market” in which wages are determined as prices are in other markets?
At least as early as John Stuart Mill, we see a further application of
this application: what does the general “free market” model, as applied
to the workers and employers’ behavior, tell us about the effects of a
particular policy measure, the wage ?oor?
Mill’s 1848 political economy text has a recognizably modern ar-
gument on employment effects. Ignoring the peculiarities of Mill’s
wages-fund doctrine, which he had yet to abandon, the outline of the
neoclassical disemployment argument is present 150 years ago. Said
Mill:
If law or opinion succeeds in ?xing wages above this [competitive]
rate, some laborers are kept out of employment; as it is not the
intentions of the philanthropists that these should starve, they must
be provided for. (1848, 432)
Minimum wages will disemploy some workers, says Mill, and some-
thing must be done to help the newly disemployed. But, whatever is done
11. For a brief history of the dramatic political events leading up to the Fair Labor Standards
Act (FLSA), see Grossman 1978.
The Minimum-Wage Controversy 123
to help—public assistance, for example—will create a disincentive to
work.
Mill points to the now familiar ef?ciency cost of all interventions—
the tension between alleviating poverty and providing incentives to
work. If laid-off workers receive a guaranteed income, then there is
no incentive to work unless the minimum wage is well above this level.
But if the minimum is so set, previously viable jobs are eliminated,
which is wasteful (Kennan 1995).
The disemployment argument is in place, as is a familiar theme of
British political economy: unintended policy consequences. The sundry
social reform movements of nineteenth-century Britain generally re-
garded Mill’s late classical economics as a kind of vulgar laissez-faire,
a mere rati?cation of the raw excesses of Victorian industrial capitalism.
But Mill was an ardent social reformer. He quarreled not with the goal
of meliorating adverse free-market outcomes, but with the ef?cacy of
policy instruments such as a legal minimum.
Henry Sidgwick (1886) shared Mill’s skepticism regarding the net
bene?ts of government interventions, especially utopian socialist
schemes (see Hutchison 1953, 53–54). And Sidgwick was at pains to
distinguish British political economy from more truly laissez-faire ap-
proaches, such as Claude-Frédéric Bastiat’s. British political economy
does not argue that low-wage workers get what they deserve, but that
well-intentioned interventions may do more harm than good. Says Sidg-
wick:
You can make it illegal to employ a man under a certain rate of
wages, but you cannot secure his employment at that rate, unless
the community will undertake to provide for an inde?nite number of
claimants work remunerated at more than its market value; in which
case its action will tend to remove, to a continuously increasing extent,
the ordinary motives to vigorous and ef?cient labor. (1886, 631)
In an 1897 article by Alfred Marshall, the disemployment argu-
ment is also quite clear: “[The academic economist] must for instance
analyse the methods which people are tempted to take for securing a
high minimum wage . . . and must show which of them will have indi-
rect effects that will cause workingmen as a whole a loss greater than
the bene?t. . . . ordinary people do not see that the means most com-
monly advocated . . . would impoverish all” (1897,128). Here, in an ad-
dress to the inaugural meeting of the Cambridge Economics Club (29
124 Thomas C. Leonard
October 1896), we also ?nd Marshall elevating ef?ciency as the badge
of the academic economist. The economist must ?ght conventional wis-
dom on minimum wages, says Marshall, not because one disagrees with
the goal of alleviating poverty among the working poor, but because
minimum wages do more harm than good (1897, 129).
The Marshallian theory of minimum wages ?nds its most thorough
and lucid exposition in a 1907 essay by H. B. Lees Smith; it is a re-
markable paper written in partial response to minimum-wage arguments
in Sidney and Beatrice Webb’s Industrial Democracy ([1897] 1920). I
present Lees Smith’s argument truncated and translated into modern
argot, but it bears reading in the original.
At the outset, Lees Smith assumes that productivity is independent of
cost: “we may assume for the present that the ef?ciency of both wage-
earners and employers is the same before and after the enactment of the
minimum” (1907, 504). He notes that statutory coverage may be partial
or total, taking partial coverage ?rst.
If the minimum is actually binding, higher labor costs must be funded
fromreduced expenditures on other inputs, higher output prices, or lower
pro?ts (504). Take reduced expenditures ?rst. If other inputs such as
skilled labor are substitutes, and they generally are, then a binding min-
imum will increase, not decrease, their prices and expenditures thereon,
which eliminates other input costs as a means of funding higher costs
of low-skilled labor (505).
If, alternatively, output prices are increased to fund higher low-skilled
labor costs, then output demand will fall, unless such demand is perfectly
inelastic. And, if ?rms and workers are already optimizing, the reduction
in output quantity demanded will entail an inef?cient reallocation of
inputs to other uncovered sectors (506). Thus, present in Lees Smith is
the law of demand and the assumption of pro?t maximization:
A rise in the price of the commodity will lead to a decrease in the
demand for it. This means that a certain amount of the labor and
capital hitherto employed in the production of the commodity must
seek other occupations. But this is likely to involve both the laborers
and the employers in a loss. Employers and laborers were presumably
occupied where, with such choices as they had, they thought they
would do the best for themselves. . . . It is no kindness to the workers
in a trade to merely turn them out. (506; emphasis added)
The Minimum-Wage Controversy 125
Finally, increased wages may be funded out of pro?ts. But if pro?ts
are diminished by higher labor costs, capital will migrate to more re-
munerative sectors. This will occur unless the covered sector’s output
market is so uncompetitive that supranormal pro?ts can fund the higher
labor costs, a claim nowhere made by minimum-wage proponents.
Thus, each possible alternative entails low-skilled labor being dis-
employed. With universal coverage, the same outcome obtains, only
low-skilled labor becomes unemployed or leaves the labor force. Con-
cludes Lees Smith: the disemployment verdict of “pure theory” is clear.
Lees Smith fully anticipates the neoclassical argument, as exempli?ed
by Stigler 1946.
I do not wish to suggest that political economists were uniformly
opposed to minimum wages. Not only is it anachronistic to assume
that economics was, around 1900, a uniform or even professionalized
discipline, but, all along the way, there were those who denied that
late-classical and then Marshallian models applied to labor markets.
F. D. Longe (1894), who was one of the ?rst to challenge Mill’s wages-
fund doctrine, dissented, as did a number of American Institutional-
ists, notably John R. Commons, who helped draft the ?rst Wisconsin
minimum-wage statute.
Perhaps the most conspicuous proponents of minimum wages in this
era are Beatrice and Sidney Webb. Sidney Webb (1912) offers an espe-
cially useful counterpoint to Lees Smith 1907.
Webb argues that, for the least productive workers, wages are not de-
termined by labor productivity. He says, “The wages of the lowest grade
of labor are ?xed, not by ‘worth’ in any sense—not even the possible
‘value in exchange’—of the individual laborer, but (as we nowadays
sadly concede) largely by the urgent necessities of the ‘marginal’ man,
or, rather, the ‘marginal’ woman” (1912, 990). By implication, either
low-wage ?rms are not pro?t maximizing, or they ?nd it pro?table,
along ef?ciency-wage lines, to hire workers whose productivity is very
low.
Webb makes recourse to both claims. At one point he says that “a
Legal Minimum positively increases the productivity of the nation’s
industry, by ensuring the surplus of unemployed workmen shall be ex-
clusively the less ef?cient workmen . . . [which] is plainly not the case
under ‘free competition’” (979). For this to be true, ?rms cannot be, ex
ante, pro?t maximizing, so Webb is offering, in rough form, an early
version of the “shock-therapy” argument.
126 Thomas C. Leonard
Arguing in the alternative, Webb also makes an ef?ciency-wage ar-
gument. “To put it plumply,” he says, “if the employers paid more, the
labor would be worth more” (990). Why might higher productivity fol-
low higher wages? At one point Webb argues that affected ?rms will
be obliged to innovate—“[the minimum wage] acts as a constant incen-
tive to the further improvement of the manual laborers, the machinery
and the organizing ability used in industry”—implicitly assuming, as
against Lees Smith, that factor substitution would not, on balance, work
against low-skilled workers (985). Webb also argues that if the minimum
were to increase wages above subsistence, laborers will be physically
stronger and thus more productive. This latter claim requires that ?rms
had failed to perceive that a wage above subsistence would increase
pro?tability.
What we observe, then, is that Webb takes issue with the key ideas in
Lees Smith’s account, not least pro?t maximization and perfect compe-
tition. Webb also argues, at least implicitly, against the law of demand.
Where Lees Smith insists that higher output prices will entail lower
output quantity demanded, Webb suggests that “the community [con-
sumers] would willingly pay more for it, and yet consume as much, or
nearly of much of it as it now does” (990).
I do not wish to suggest that Webb and Lees Smith would have de-
scribed their disagreement solely in terms of pro?t maximization and la-
bor market structure. The hope, rather, is that this sketch suf?ces to show
that the leading contemporary issue in minimum-wage economics—em-
ployment effects—is already being debated in familiar terms around the
turn of the century. The disemployment account is recognizable in Mill
150 years ago, and, in the sense of considering the labor-market condi-
tions under which disemployment does and does not obtain, is almost
fully formed 90 to 100 years ago.
What is missing from the minimum-wage arguments of this era is
empirical evidence. The evidence for the successful application of eco-
nomics to labor markets is entirely theoretical—that is, based on appeal
to reason and logic. Clearly, U.S. data became available only follow-
ing the states’ minimum-wage legislation. But what is striking through
much of the twentieth century is the extent to which the application of
economics to labor markets remains almost entirely theoretical. George
Stigler’s landmark paper, for example, presents only a few stylized facts
for evidence, admitting that “no precise estimate of the effects of the
minimum wage upon aggregate employment is possible” (1946, 361).
The Minimum-Wage Controversy 127
The mid-twentieth-century debate recapitulates many of the old ar-
guments about whether the general theory of competitive markets, now
full-blown neoclassical price theory, is applicable to labor markets. But,
here our second theme emerges: the role of empirical evidence as the
arbiter of which theory (if any) is indeed applicable. The marginalist
controversy and the economics of minimum wages come together in the
person of Richard Lester.
Minimum Wages and the Marginalist Controversy
Richard Lester, the labor economist to whom Card and Krueger dedicate
their book, argued that ?rms do not always maximize pro?ts and that
labor markets generally are not competitive. Lester’s conclusion was
based on surveys of manufacturing CEOs who did not know marginal
costs and revenues, and on wage surveys, which found a wide vari-
ance in wages paid to production workers of roughly equal productivity
(Lester 1946a, 1946b). Lester read ignorance of marginal magnitudes
as implying that ?rms do not maximize pro?ts; and he read the unex-
plained wage variation as suggesting that ?rms do not take wages as
given.
Lester was not the ?rst to argue that ?rms do not maximize pro?ts,
12
but his work generated a volley of neoclassical responses, an exchange
which has come to be known as the marginalist controversy. Some re-
sponses were directed at Lester’s challenges explicitly (Machlup 1946,
1947; Stigler 1947) and others more indirectly (Alchian 1950; Friedman
1953).
13
It was Lester’s investigation of labor markets, and his strong
conclusions, that gave rise to a withering exchange on the nature of eco-
nomic theory and on the role of empirical evidence in testing theories
and their implications.
That labor research was the occasion for the methodological dispute
we ordinarily associate with the marginalist controversy is not coin-
cidence. Lester and other institutionally minded labor economists of
the mid-century era—especially John Dunlop, Clark Kerr, and Lloyd
12. R. L. Hall and C. J. Hitch (1939,18) interpret their interviews of businessmen as
arguing that ?rms do not equate marginal revenues and marginal costs, but use rules of thumb
for marking up average costs.
13. Lester’s rejoinder to Machlup 1946 (itself a reply to Lester 1946a) and reply to Stigler
1946, the neoclassical locus classicus, are found in Lester 1947.
128 Thomas C. Leonard
Reynolds (DKLR)—practiced a kind of labor economics that differed
signi?cantly from the neoclassical labor economics that later predom-
inated (Freeman 1989, 317–42). The DKLR method was built around
institutional expertise, detailed case studies, and surveys of management
and workers. They surely resented what they regarded as neoclassical
theory-tropism: relative indifference to empirical work, and hostility to
survey questionnaires in particular (Machlup 1946).
Related, and perhaps more important, DKLRrejected the neoclassical
model’s assumptions of pro?t maximizing and perfect competition, at
least for empirical purposes.
The labor market is not a bourse; wage rates are typically quoted
prices. (Dunlop 1944, cited in Freeman 1989, 330)
The labor market is by nature an imperfect instrument. . . . all kinds
of wage distortions exist even under non-union conditions. (Reynolds
and Taft 1956, 168, cited in Freeman 1989, 319)
Reasoning about labor markets as though they were commodity mar-
kets seems to be an important explanation for erroneous conclusions
on such matters as minimum wages. (Lester 1947, 146, responding
to Stigler 1946)
A [competitive market–determined wage structure] is a useful norm
for theoretical speculations but an unstable departure point for em-
pirical studies. (Kerr 1977, 46, cited in Freeman 1989, 318)
In short, DKLR disagreed with the neoclassical presumption that labor
markets, especially the determination of wages and employment, operate
no differently than do goods markets.
The controversy following Lester’s challenge had methodological
consequences that reverberate still. Fritz Machlup’s 1946 response to
Lester 1946a was important less for its argument that Lester’s survey
methodology was sloppy, than for its methodological claim that pro?t
maximization is not a descriptive claim but a useful ?ction. Managers
may not know marginal magnitudes, nor need they be calculating on
all conceivable margins. But their habits and routines can be thought of
as learned strategies that themselves derive from a goal of pro?t maxi-
mization. Methodologically, it is as if ?rms maximize pro?ts (Machlup
1946).
The Minimum-Wage Controversy 129
Alchian (1950) added a natural selection story to the gap that Machlup
opened between intent and effect. Whether or not ?rms intend to maxi-
mize pro?ts, Alchian argued, a Darwinian business environment ensures
that surviving ?rms will be maximizing pro?ts, as underperforming
?rms are weeded out by competition. It is a competitive output market
structure that creates pro?t maximizing behavior, albeit as a by-product
rather than as the necessary goal of ?rms.
Milton Friedman’s famous paper (1953) borrowed elements from
both Machlup 1946 and Alchian 1950. With Machlup, Friedman em-
braced the as if method, which he in?ated into a grander claim: theo-
retical assumptions need not be true. While it is clearly unrealistic to
assume that billiard players solve systems of differential equations when
making shots, it is useful for prediction, says Friedman, to pretend that
they do. Friedman goes so far as to make “unrealistic” (read: false) as-
sumptions into a theoretical virtue: “The more signi?cant the theory, the
more unrealistic the assumptions” (1953, 14).
From Alchian, Friedman carries over the idea that an agent’s intent is
unimportant; economics should be seen not as a theory of motivational
states but of market outcomes. Friedman concedes that ?rms need not
intend to maximize pro?ts, but boldly argues that this does not matter
methodologically. Theories are to be judged not by the truth or falsity
of their assumptions, but, rather, by the success of their predictions. The
assumption that ?rms consciously optimize is false, he concedes, but
the prediction that only pro?t-maximizing ?rms will survive in the long
run is empirically accurate.
Methodologically, the realism-of-assumptions debate is something of
a muddle. Our focus is on the role of empirical results in adjudicating
among competing theories. Much turns on whether one sees assump-
tions as restricting the applicability of the theory—a domain assump-
tion—or whether one sees assumptions’ truth or falsity as negligible for
theoretical purposes, such as prediction (Musgrave 1981).
If we regard the assumption of competitive labor markets as a domain
assumption, then the disemployment prediction obtains only when the
assumption is true, that is, only when labor markets are competitive.
Stigler’s reply to Lester 1946a at one point reverts to just this domain-
assumption interpretation (1947, 155).
14
Alternatively, one can predict
14. Joan Robinson, parenthetically commenting on the minimum-wage debate, connects
the matter of assumptions to the Marshallian-era version of the controversy: “This [monopsony]
analysis may throw light on the dispute between Marshall and Mr. and Mrs. Webb upon the
130 Thomas C. Leonard
disemployment regardless of whether the competitive-market assump-
tion is true or not, a negligibility assumption. If the power of ?rms to
set wages below MRP
L
is negligible, then the falsity of the assumption
will not much matter because, ceteris paribus, the disemployment pre-
diction will still be borne out. This is Friedman’s interpretation of how
assumptions work in economics.
Both readings have important implications for application. Since
Stigler agrees that assumptions constrain the applicability of theory,
then it is clearly important that assumptions not be too restrictive; oth-
erwise, there is nothing to apply to. But Stigler, of course, does not
really attend to the question of whether maximization and competition
are overly restrictive assumptions. As we have seen, his own contribu-
tion offers only limited empirical evidence. And he is quite dismissive
of Lester’s attempt to empirically test the key assumptions.
In any event, it was Friedman’s reading that carried the day:
economists need not worry about the truth or falsity of assumptions; only
predictive accuracy matters. By implication, empirical work should be
con?ned to testing the predictions of theory.
15
Testing of assumptions,
of the kind that Lester has undertaken, can be dismissed as methodolog-
ically beyond the pale.
Although Friedman prevailed, his methodological injunction—em-
pirically test theoretical predictions—went mostly unheeded. With few
exceptions (e.g., Peterson 1957), the relative lack of mainstream empir-
ical work on minimum-wage effects persists well into the 1970s. Robert
Goldfarb’s survey (1974) of the minimum-wage literature ?nds that the
empirical support for disemployment effects, especially for adults (ages
twenty and up), who have generally composed the majority of low-wage
workers, is quite limited. With the exception of teenage employment, it
is not hyperbole to suggest that, in the Nixon era, the basis for profes-
sional priors was essentially unchanged from that of three generations
earlier.
‘marginal productivity theory’ of wages (see [Marshall’s] Principles, 705). It seems to have
arisen because Mr. and Mrs. Webb failed to realized the implications of the assumptions
of perfect competition, while Marshall failed to recognize the extreme unreality of those
assumptions” (Robinson [1933] 1950, 300–301 n. 2).
15. I do not take up Friedman’s implicit claim that theoretical components are separable,
that is, that there is a meaningful distinction between theoretical assumptions and theoretical
implications.
The Minimum-Wage Controversy 131
Empirical Evidence on Employment
Effects of Minimum Wages
The empirical literature grew rapidly with the expansion of time-series
econometric techniques. Roughly twenty-?ve papers are reviewed in
Charles Brown, Curtis Gilroy, and Andrew Kohen’s (1982) survey.
Goldfarb (1981) reports empirical progress, especially with regard to
distribution effects.
Most empirical papers in this period make use of a regression with
something like the following speci?cation:
Y = f (MW, D, X
1
. . . X
n
),
where Y is labor force status (e.g., labor force participation ratio,
employment-population ratio, unemployment rate), MW is some mea-
sure of the minimum wage,
16
D is a business cycle variable to account
for changes in aggregate economic activity, and X
i
is an exogenous
variable controlling for labor supply, school enrollment, participation
in the armed forces, and so on (Brown, Gilroy, and Kohen 1982, 497).
Some speci?cations add a time trend as well (see Mincer 1976). There
are different functional forms to the speci?cations, but nearly all studies
use Current Population Survey (CPS) data.
Brown, Gilroy, and Kohen summarize the literature as follows:
teenage employment-rate elasticities are on the order of ?0.1 to ?0.3,
but “the lower half of that range is to be preferred.” For young adults
(ages 20–24), the effect of the minimum wage on employment is “neg-
ative and smaller” than that for teens. But Brown, Gilroy, and Kohen
are careful to say that “uncertainty about the effects on [older] adults
is a serious gap in the literature . . . ,” echoing Goldfarb 1974 (1982,
524). These conclusions generally comport with contemporaneous sur-
veys and collections (Rottenberg 1981; Report of the Minimum Wage
Study Commission 1981; Eccles and Freeman 1982).
The uncertainty surrounding adult employment effects notwithstand-
ing, the disemployment prediction seems comfortably consolidated by
the 1980s (Alston, Kearl, and Vaughan 1992). There may have been
16. “The key variable, minimum wage, has generally been measured by the ratio of the
nominal legal minimum wage to average hourly earnings weighted by coverage, as devised
by Kaitz (1970). Ratios of minimum wage rates to average hourly earnings are calculated for
each industry, weighted by the proportion of workers covered. These are combined into an
index in which the weight for each industry ration is the number of persons employed in the
industry as a percentage of total employment” (Brown, Gilroy, and Kohen 1982, 499).
132 Thomas C. Leonard
quibbles about the exact magnitude of disemployment effects, but none
about whether they existed (Deere, Murphy, and Welch 1995, 232).
Indeed, there was very little empirical work on minimum wages under-
taken during the mid-eighties. What changed?
First, the few new time-series estimates which did incorporate data
from the eighties seemed to suggest that employment elasticities with
respect to minimumwages were becoming more inelastic. Second, while
the nominal value of the federal minimum wage stayed at $3.35 per
hour between 1981 and 1990, several states raised their minima, which
created a research opportunity for cross-sectional studies. Finally, the
increase in income inequality during the 1980s spurred political action
for redistributive measures. In the nineties, minimum-wage research
revived, and with dramatic consequences.
The “New” Economics of Minimum Wages
and Its Critics
I emphasize two lines of attack that Card and Krueger mount on the
neoclassical account. First, they present cross-sectional evidence from
four studies—most famously from the New Jersey–Pennsylvania (NJ-
PA) fast food experiment—which ?nds that the employment elasticity
of the minimum wage is positive, or at least nonnegative. Of Myth’s
seven analyses of employment effects, one shows zero effect and six
show positive effects, but only two of these are signi?cantly different
from 0 at the 0.05 level (1995, 388).
Second, and less well known, Card and Krueger review the estab-
lished econometric literature and judge it altogether inadequate. They
conclude that previous ?ndings are unreliable, tainted by publication
bias and speci?cation search. Paul Osterman calls this aspect of Myth “a
damning indictment of how labor economics has been practiced over the
past three decades” (1995, 839). Card and Krueger offer their own empir-
ical approach—“natural experiment” and “model-free empiricism”—as
evidentially superior to earlier econometric studies.
Card and Krueger’s Evaluation
of Previous Empirical Work
Card and Krueger (CK) review earlier empirical work on minimum-
wage effects and judge it inadequate and tendentious. CK replicate and
The Minimum-Wage Controversy 133
extend some earlier work, ?nding that employment elasticities attenuate
when more recent years are included. A meta-analysis of previous time-
series estimates concludes that earlier work is corrupted by publication
bias and speci?cation search.
CK argue that time-series estimates, over time, involve larger and
larger samples, which, ceteris paribus, should increase t -statistics. If
the additional data in newer studies are independent of the older data,
“then a doubling of the sample size should result in an increase in the
absolute value of the t -ratio of about 40 percent” (1995, 187).
17
If the t -
statistics do not increase with sample size, then, CK suggest, something
is ?shy. Using the time-series studies reviewed in Brown, Gilroy, and
Kohen 1982, CK ?nd that the t -statistics are, in fact, declining in sample
size (188).
18
CK also conduct a related meta-analysis that compares the relation-
ship between the estimated coef?cients of employment elasticity with
their respective standard errors, for the same set of time-series studies.
If the estimated coef?cients (b) are truly unbiased, CK argue, then there
should be no systematic relationship between them and their respec-
tive standard errors (se). If, on the other hand, editors will only publish
studies with robust test statistics (t > 2, say), then this publication bias
should be re?ected in the meta-analysis, given that t = b/se. Plotting
“implicit” standard errors against the coef?cients, CK ?nd a rather good
?t (for b = 2se), with only two or three exceptions, suggesting a problem
with independence (1995, 191).
From this meta-evidence, CK offer three possible explanations. One
explanation is that structural change made the employment effects of
minimum wages decline over time. But structural change calls into
question the validity of time-series work to begin with. The alternative
17. Card and Krueger note the lack-of-independence problem that occurs with time-series
data more generally, but point out that previous authors themselves assume independence
following adjustments for serial correlation in errors and similar procedures.
18. Actually, they plot the absolute value of the t -statistic against the square root of
the degrees of freedom, or (n ? k)
1/2
. The rationale is explained by Kennan 1995. The
t -statistic in a regression with n observations on k explanatory variables satis?es the equa-
tion t
2
(n ? k) = r
2
/(1 ? r
2
), where t is the t -statistic of a regressor and r is the par-
tial correlation coef?cient between this regressor and the dependent variable. As the sam-
ple size increases, r converges to its population value, assuming stationarity of the partial
correlation, so t must grow at the same rate as the square root of n ? k (Kennan 1995,
1963).
134 Thomas C. Leonard
explanations are speci?cation search and publication bias. CK argue
that the early literature was contaminated by a preference (on the part
of data-mining researchers and biased editors) for statistically signi?-
cant results that comported with the disemployment predictions of the
neoclassical model.
The econometric speci?cations which became standard were chosen,
CK suggest, because they produced the desired negative sign and got
the t -statistics above 2. Later researchers tended to reuse these speci?-
cations. Because true statistical signi?cance was overstated in the earlier
research, however, more recent work (presumably not so susceptible to
temptation) discovered weaker effects (1995, 192–93). Three of Finis
Welch’s studies are singled out by CK as examples of this econometric
malpractice.
Natural Experiments and “Model-Free”
Empiricism
Myth is notable for its emphasis on natural experiments (NE) and, more
generally, the idea of “model-free” empirical work. CK take the NJ-
PA study as their most important result, precisely because they believe
that its setting is closest to the ideal of the randomized controlled ex-
periments of the “hard” sciences. They imply that an NE, even if an
imperfect substitute for a randomized controlled experiment, is eviden-
tially superior to traditional econometric techniques.
The New Jersey–Pennsylvania (NJ-PA) study compares employment
responses in the fast-food market in New Jersey and eastern Pennsylva-
nia to increases in the minimum wage (New Jersey raised its state min-
imum from $4.25 to $5.05, while Pennsylvania’s remained at $4.25).
The “experiment” compares the employment effects on affected work-
ers against the Pennsylvania “control” group. If all differences between
New Jersey and Pennsylvania remain unchanged before and after the
minimum-wage increase, then any employment (E
it
) “difference in
differences” (DD) can be attributed to the minimum-wage effects, as
follows:
DD = (E
NJ2
? E
PA2
) ? (E
NJ1
? E
PA1
),
where period 1 is February 1992 (before the new minimum takes effect
in April 1992) and period 2 is November 1992. Neoclassical theory
predicts that, ceteris paribus, employment in New Jersey should fall
The Minimum-Wage Controversy 135
relative to employment in Pennsylvania, that is, that DDwill be negative.
CK, however, ?nd it to be positive.
19
CK sometimes make extravagant claims regarding the signi?cance
of this ?nding, even as they are careful to acknowledge the relative
weaknesses of NEs. First, NEs do not randomly assign subjects into
treatment and control groups, which is essential for attributing any dif-
ferential outcomes to the minimum wage exclusively. Minimum-wage
legislation is the outcome of a political process, and states that enact
minimum wages may differ from those that do not; they may have more
robust economic growth, for example (1995, 23). Second, NEs are not
“blind.” Subjects know when they are (or are not) being treated. Third,
NEs require ceteris paribus. Any current differences between treatment
and control groups must persist in the absence of the minimum-wage
intervention (e.g., both groups should exhibit identical rates of growth)
(23).
CK are con?dent that these departures from classical experiments are
small relative to the advantage NEs provide over traditional econometric
approaches. The NE approach is “model free” in that its results do not
hinge on a particular econometric speci?cation or, therefore, upon the
theoretical assumptions that underwrite a given speci?cation. And, at
least in principle, NEs can reduce the incentive for speci?cation search,
the temptation to ?t the model to the data.
Replies to Card and Krueger
The replies to Myth are too numerous to document fully, a measure of
its in?uence. Professional critics have focused less on the nature than
on the execution of CK’s empirical methods, especially in the famous
NJ-PA study.
The most serious replies are as follows: (1) the survey is poorly de-
signed as re?ected in the noisy data that result (Kennan 1995; Welch
1995); (2) the study is not a proper NE, particularly with respect to the
control group (Hamermesh 1995; Kennan 1995; Welch 1995); (3) re-
sults are short run only (Hamermesh 1995); (4) employment increases
should mean higher output and lower output prices, but CK ?nd evi-
dence of higher output prices (Brown 1995); and, most conspicuously,
19. Employment is measured in full-time equivalents (FTEs), where a part-time worker is
counted as 0.5 FTE. The results: 2.7 = [21.0 ? 21.2] ? [20.4 ? 23.3].
136 Thomas C. Leonard
(5) results cannot be replicated and are contradicted when superior data
are employed (Neumark and Wascher 1995).
20
Low-wage employers may well have already adjusted to the New
Jersey minimum-wage increase before it took effect in April 1992, given
that the statute passed in 1990. Or their adjustment may have occurred
after the second survey. And the replication by David Neumark and
William Wascher (1995) does contradict CK’s original work.
21
And
CK’s most important result—that employment in New Jersey increases
relative to employment in Pennsylvania—obtains principally because
employment fell in Pennsylvania, the control. It is, John Kennan argues,
“like having a drug trial in which the drug has no effect but the placebo
makes people sick” (1995, 1958).
Conclusion: The Very Idea of Applying Economics
The “new” economics of the minimum wage is not really new at all.
Although Card and Krueger do not mention it, their method of natu-
ral experimentation has direct empirical antecedents in a little-known
statistical literature developed in the 1910s and 1920s. There are large
repositories of “before and after” employment data collected by the U.S.
Bureau of Labor Statistics and by various state agencies.
22
And Card and
Krueger’s empiricism and their monopsony explanation have acknowl-
edged af?nities with the DKLR approach. But two things do distinguish
the contemporary controversy: (1) the substantive emphasis on employ-
ment effects and (2) the methodological emphasis on empirical work as
the arbiter of successful theoretical application.
As the debate over minimum wages increasingly stresses its role as a
test of the applicability of price theory to labor markets, there is a con-
comitant emphasis on employment effects. Reactions by nonspecialists
to Myth are almost entirely in this vein. Said one Nobel laureate:
20. A more detailed account is available in Leonard 1997, 268–73.
21. Card and Krueger 1998 replies to Neumark and Wascher 1995.
22. Kennan 1995 instances a remarkably detailed study of minimum-wage effects in Ore-
gon (Obenauer and von der Nienburg 1915), which begins, “Does the establishment of a legal
minimum wage for women do more harm than good to the women it is designed to assist?” and
which presents employment and wage data for tens of hundreds of affected girls and women.
The BLS published eleven bulletins on minimum-wage issues prior to Obenauer and von der
Nienburg 1915, and there are detailed data in numerous state publications, such as National
Industrial Conference Board 1927.
The Minimum-Wage Controversy 137
Just as no physicist would claim that “water runs uphill,” no self-
respecting economist would claim that increases in the minimum
wage increase employment. Such a claim, if seriously advanced, be-
comes equivalent to a denial that there is even minimal scienti?c
content in economics, and that, in consequence, economists can do
nothing but write as advocates for ideological interests. (Buchanan
1996)
Labor specialists are less intemperate but equally preoccupied with dis-
employment. Myth examines many consequences of minimum wages,
but CK themselves make central their own ?ndings and their critique of
previous econometric work on employment effects.
The emphasis on employment has crowded out a crucial, and answer-
able, policy question: Do minimum wages reduce poverty? A desire to
help the working poor, of course, lies at the root of minimum-wage
policy. An obviously relevant question is, Are affected workers poor?
(Stigler 1946).
23
The surprising answer is, generally, no. The reasons for
a positive but weak relationship between minimumwages and household
income are well understood and well documented (Gramlich 1976).
The relationship is weak when low-wage workers are members of
nonpoor households (teens, say), or when households have nonwage
income, or when households have no workers at all. Low-wage workers
are more likely to be in poor families than are all workers, but the great
majority of low-wage workers are not poor. Ralph Smith and Bruce
Vavrichek (1987) found that 75 percent of affected workers in 1985 were
neither poor nor near-poor. CK’s own distribution analysis ?nds that
only 30 percent of affected workers live in poor and near-poor families
(1995, 280). In other words, over two-thirds of income redistributed by
minimum-wage increases—even assuming zero disemployment—will
go to individuals in families neither poor nor near-poor.
Myth concludes, “We believe that our ?ndings suggest a reorientation
of policy discussions away from the ef?ciency aspects of the minimum
wage and toward distributional issues” (393). But, at the same time,
CK are obliged to say that the minimum wage is a “blunt instrument”
for redistributing income to poor families and ?nd that “the effect of
23. Not all economists accept alleviation of poverty as the goal of minimum wages. Some
argue that ?attening the income distribution is worthwhile for its own sake, and others sug-
gest that progressive income redistribution stimulates aggregate demand, owing to a higher
propensity to consume in lower-income households (Prasch 1996).
138 Thomas C. Leonard
the minimum wage on the overall poverty rate of adults is statistically
undetectable” (1995, 285, 280; see also Burkhauser, Couch, and Witten-
berg 1996).
24
Even if one believes that minimum-wage increases do not
disemploy low-wage workers, or is prepared to suffer disemployment as
a cost of redistribution, it is signi?cant that minimum wages do so little
for the goals of poverty reduction or progressive income redistribution.
The evidence on distribution—where the data are comparatively reli-
able and uncontroversial—has been completely overshadowed by the
emphasis on whether the disemployment prediction obtains.
The methodological emphasis on empirical work as the arbiter of suc-
cessful theoretical application is also new to the contemporary period.
Friedman (1953), we have seen, made testing of theoretical predictions
the paramount goal of economic science, but only in the contemporary
controversy does empirical work actually assume this role. CK are ex-
plicit: they regard their work not merely as four papers contributing to
a large empirical literature, but as “a simple and direct test of the kind
of theoretical reasoning that economists routinely apply to other, more
complicated phenomena” (1995, 397).
Some are prepared to renounce Myth on a priori grounds. Evidence
that says “water runs uphill” is to be rejected on its face. But labor
is the quintessentially empirical ?eld in economics. Labor economists
take pride in their willingness to submit theoretical claims to the tribunal
of empirical evidence. In other corners of economics, such as macro-
economics, says Richard Freeman, “the evidentiary base is suf?ciently
diffuse to allow economists with different priors to reach different con-
clusions—to make a lawyer’s case as it were” (1995, 831). But, Free-
man argues, not so in minimum-wage research, where changes occur
in discrete exogenous jumps, and evidence permits far less scope for
“disparate interpretation” (831).
Nearly all labor specialists in the controversy, critics and supporters
alike, take for granted (at least nominally) that attempts to apply theory
must be tested by empirical evidence. CK’s valedictory to Myth says that
“although many economists may disagree with our interpretation of the
?ndings in this book, we hope that they will at least agree on the value of
testing the implications of standard economic theory, and on the validity
of our empirical approach” (1995, 399). Their hope has been met. I can
24. A minimum wage can help only the employed. Card and Krueger (1995, 305) estimate
that about two-thirds of adults in poor families do not work at all. Burkhauser, Couch, and
Wittenberg (1996, 551) estimate that fully one-quarter of poor families have no working adults.
The Minimum-Wage Controversy 139
locate no serious objection, even among CK’s strongest critics, to the
idea of empirical evidence as testing theory or to the evidential virtues
of natural experiments. But a methodological ambivalence persists. If
all agree that the data should be decisive, no one agrees on when the
data are decisive.
I conjecture that the ambivalence regarding empirical testing has two
sources. One is methodological. To answer the key substantive ques-
tion—do Myth’s results refute price theory’s disemployment predic-
tion?—one must know the answer to a prior methodological question:
what evidence constitutes a refutation? If Myth really offers “a simple
and direct test” of neoclassical price theory, then one must know what
counts as a simple and direct test. Without methodological argument,
still less consensus, there are no ready means for evaluating CK’s claim.
The other source of ambivalence derives from a residual unease with
the robustness of the empirical evidence on disemployment. The evi-
dence for adults, we have seen, was never as strong as that for teenage
workers. Furthermore, CK’s most contentious claim—that the entire
empirical literature before circa 1990 is contaminated by econometric
malpractice—has gone virtually unrebutted. The silence is fairly deaf-
ening. The only paper to directly take issue with CK’s extraordinary
indictment is Neumark and Wascher 1996.
25
Finally, good evidence on minimum-wage effects is hard to obtain,
even given the relatively advantageous NE setting, as Kennan 1995
points out. Even ignoring legal noncompliance and survey measurement
error, there are thorny problems. First, employers can respond to wage
increases by reducing nonpecuniary compensation—health insurance,
training, pleasant work environment, scheduling ?exibility, discounts
on goods and services. This simply rearranges the composition of the
worker’s pay packet and should not affect employment (Kennan 1995).
Second, employers might reduce hours without reducing employment,
so measuring employment will miss changes in labor demand. Third,
affected workers already have highly volatile employment patterns. The
noise in the employment data makes it dif?cult to disentangle changes
25. Neumark and Wascher (1996) examine Myth’s meta-analysis of the time-series litera-
ture and argue that CK’s results are better explained by structural change than by publication
bias leading to speci?cation search. Parameter values changed, they say, producing spuri-
ous evidence of publication bias. Neumark and Wascher do not consider the implications of
structural change for the validity of time-series evidence.
140 Thomas C. Leonard
caused by statutory minima from those caused by confounding fac-
tors—the proverbial needle-in-the-haystack problem (Kennan 1995).
All of these empirical challenges are well known to labor researchers.
But they are troubling. Methodologically, the belief that empirical ev-
idence “decides” when theory applies obviously has less force when
the evidence tends to be equivocal. Substantively, if the evidence for
minimum-wage effects is equivocal—when, according to its producers,
it offers the best of empirical opportunities—then what does this imply
about the quality of evidence in economics more generally?
Herein lies a basic disciplinary tension. Modern labor economics is
empirical. Indeed, in some respects, Myth can be read as an empiricist
manifesto—go forth and test, and if the data are inconvenient for the
theory, so much the worse for theory. But modern labor economics is
also, unlike its mid-century variant, ?rmly neoclassical (Freeman 1989,
318). Price theory, like its predecessors, has the goal of producing a
general theory of economic behavior, which places the determination of
wages and employment comfortably within its purview—markets are
markets are markets. And a general theory that does not apply in one
domain is not a general theory. These twin disciplinary ambitions—to
be an empirical policy science, and to have a general theory of economic
behavior—can clearly come into con?ict, especially when, as with min-
imum wages, compelling evidence meets core theoretical belief.
The economist’s dilemma is as follows: do I reject the evidence or
reject the application of theory? Most respondents to Myth have chosen
the ?rst option—some on principle, but most on grounds that the new
evidence is too fragile to justify abandoning core doctrine. Still others
argue that more evidence is required. Kennan says:
Myth and Measurement’s lasting contribution may well be to show
that we just don’t know how many jobs would be lost if the minimum
wage were increased to $5.15, and that we are unlikely to ?nd out
using more sophisticated methods of inference on the existing body
of data. What is needed is more sophisticated data. (1995, 1964)
More sophisticated data are surely desirable. But the con?icting ambi-
tions of economics will continue to manifest themselves in recurring
controversy, at least until the data are so sophisticated that economists
routinely answer methodological questions such as, What evidence
would convince you that your theory is refuted?
The Minimum-Wage Controversy 141
References
Alchian, Armen. 1950. Uncertainty, Evolution, and Economic Theory. Journal of
Political Economy 58.3:211–21.
Alston, Richard, J. R. Kearl, and Michael Vaughan. 1992. Is There a Consensus
among Economists in the 1990s? American Economic Review Papers and Pro-
ceedings 82.2:203–9.
Barro, Robert. 1996. Higher Minimum Wage, Higher Dropout Rate. Wall Street
Journal, 11 January, A14.
Brown, Charles. 1988. Minimum Wage Laws: Are They Overrated? Journal of
Economic Perspectives 2.3:133–45.
———. 1995. Comment for review symposium on Myth and Measurement: The
New Economics of the Minimum Wage, by David Card and Alan Krueger. Indus-
trial and Labor Relations Review 48.4:828–30.
Brown, Charles, Curtis Gilroy, and Andrew Kohen. 1982. The Effect of the Mini-
mum Wage upon Employment and Unemployment. Journal of Economic Liter-
ature 20.2:487–528.
Buchanan, James. 1996. Commentary on the Minimum Wage. Wall Street Journal,
25 April, A20.
Burkhauser, Richard, Kenneth Couch, and David Wittenberg. 1996. “Who Gets
What” from Minimum Wage Hikes: A Re-estimation of Card and Krueger’s
Distributional Analysis in Myth and Measurement: The New Economics of the
Minimum Wage. Industrial and Labor Relations Review 49.3:547–52.
Card, David, and Alan Krueger. 1995. Myth and Measurement: The New Economics
of the Minimum Wage. Princeton, N.J.: Princeton University Press.
———. 1998. AReanalysis of the Effect of the NewJersey MinimumWage Increase
on the Fast-Food Industry with Representative Payroll Data. NBER Working
Paper no. 6386. Cambridge, Mass.: NBER.
Deere, Donald, Kevin Murphy, and Finis Welch. 1995. Employment and the
1990–1991 Minimum-Wage Hike. American Economic Review Papers and Pro-
ceedings 85.2:232–43.
Douglas, Dorothy. 1919. American Minimum Wage Laws at Work. American Eco-
nomic Review 9.4:701–32.
Dunlop, John. 1944. Wage Determination under Trade Unions. New York: Augustus
M. Kelley.
Eccles, Mary, and Richard Freeman. 1982. What! Another Minimum Wage Study?
American Economic Review Papers and Proceedings 72.2:226–32.
Freeman, Richard. 1989. Labor Markets in Action. Cambridge: Harvard University
Press.
———. 1995. Comment for review symposium on Myth and Measurement: The
New Economics of the Minimum Wage, by David Card and Alan Krueger. Indus-
trial and Labor Relations Review 48.4:830–34.
Friedman, Milton. 1953. The Methodology of Positive Economics. In Essays on
Positive Economics. Chicago: University of Chicago Press.
142 Thomas C. Leonard
Goldfarb, Robert. 1974. The Policy Content of Quantitative Minimum Wage Re-
search. Industrial Relations Research Association Proceedings: 27:261–68.
———. 1981. The Context of Recent Research. In The Economics of Legal Mini-
mum Wages, edited by Simon Rottenberg. Washington, D.C.: American Enter-
prise Institute.
Gramlich, Edward. 1976. Impact of MinimumWages on Other Wages, Employment,
and Family Incomes. Brookings Papers on Economic Activity, no. 2:409–51.
Grossman, Jonathan. 1978. Fair Labor Standards Act of 1938: Maximum Struggle
for a Minimum Wage. Monthly Labor Review 101.6:22–30.
Hall, R. L., and C. J. Hitch. 1939. Price Theory and Business Behavior. Oxford
Economic Papers, no. 2:13–45.
Hamermesh, Daniel. 1995. Comment for review symposium on Myth and Mea-
surement: The New Economics of the Minimum Wage, by David Card and Alan
Krueger. Industrial and Labor Relations Review 48.4:835–38.
Holcombe, A. 1910. The British Minimum Wages Act of 1909. Quarterly Journal
of Economics 24.3:574–77.
Hutchison, Terrence. 1953. A Review of Economic Doctrines 1870–1929. Oxford:
Clarendon Press.
Kelley, Florence. 1912. Minimum-Wage Laws. Journal of Political Economy
20.10:999–1010.
Kennan, John. 1995. The Elusive Effects of Minimum Wages. Journal of Economic
Literature 33.4:1950–65.
Kerr, Clark. 1977. Labor Markets and Wage Determination. Berkeley: University
of California Press.
Krueger, Alan, and Lawrence Summers. 1988. Ef?ciency Wages and the Inter-
industry Wage Structure. Econometrica 56.2:259–93.
Lees Smith, H. B. 1907. Economic Theory and Proposals for a Legal Minimum
Wage. Economic Journal 17.68:504–12.
Leonard, Thomas. 1997. The Reason of Rules in the Intellectual Economy: The
Economics of Science and the Science of Economics. Ph.D. diss., The George
Washington University.
Lester, Richard. 1946a. Shortcomings of Marginal Analysis for Wage-Employment
Problems. American Economic Review 36.1:62–82.
———. 1946b. Wage Diversity and Its Theoretical Implications. Review of Eco-
nomics and Statistics 28.3:152–59.
———. 1947. Marginalism, Minimum Wages, and Labor Markets. American Eco-
nomic Review 37.1:135–48.
Longe, F. D. 1894. The Coal Strike and a “Minimum” Wage. The Economic Journal
4.13:25–34.
Machlup, Fritz. 1946. Marginal Analysis and Empirical Research. American Eco-
nomic Review 36.4:519–54.
———. 1947. Rejoinder to an Antimarginalist. American Economic Review
37.1:148–54.
The Minimum-Wage Controversy 143
Marshall, Alfred. 1897. The Old Generation of Economists and the New. Quarterly
Journal of Economics 11.2:115–35.
Mill, John Stuart. 1848. Principles of Political Economy with Some of Their Appli-
cations to Social Philosophy. Boston: Little & Brown.
Miller, Merton. 1996. Commentary on the Minimum Wage. Wall Street Journal, 25
April, A20.
Millis, H. A. 1914. Some Aspects of the Minimum Wage. Journal of Political
Economy 22.2:132–55.
Mincer, Jacob. 1976. Unemployment Effects of Minimum Wages. Journal of Polit-
ical Economy 84.4, part 2: 87–104.
Musgrave, Alan. 1981. “Unreal Assumptions” in Economic Theory: The F-Twist
Untwisted. Kyklos 34.3:377–87.
National Industrial Conference Board. 1927. Minimum Wage Legislation in Mas-
sachusetts. New York: NICB.
Neumark, David, Mark Schweitzer, and William Wascher. 1998. The Effects of
Minimum Wages on the Distribution of Family Incomes: A Non-Parametric
Analysis. NBER Working Paper no. 6535. Cambridge, Mass.: NBER.
Neumark, David, and William Wascher. 1995. The Effect of New Jersey’s Mini-
mum Wage Increase on Fast-Food Employment: A Reevaluation Using Payroll
Records. NBER Working Paper no. 5224. Cambridge, Mass.: NBER.
———. 1996. Is the Time-Series Evidence on Minimum Wage Effects Contami-
nated by Publication Bias? NBER Working Paper no. 5631. Cambridge, Mass.:
NBER.
Obenauer, Marie, and Bertha von der Nienburg. 1915. Effect of Minimum-Wage De-
terminations in Oregon. Bulletin of the U.S. Bureau of Labor Statistics. Number
176. Washington, D.C.: U.S. Government Printing Of?ce.
Osterman, Paul. 1995. Comment for review symposium on Myth and Measurement:
The New Economics of the Minimum Wage, by David Card and Alan Krueger.
Industrial and Labor Relations Review 48.4:839–42.
Peterson, John. 1957. Employment Effects of MinimumWages, 1938–1950. Journal
of Political Economy 65.5:412–30.
Pigou, A. C. 1922. Trade Boards and the Cave Committee. The Economic Journal
32.127:315–24.
Prasch, Robert. 1996. In Defense of the Minimum Wage. Journal of Economic
Issues 30.2:391–97.
Report of the Minimum Wage Study Commission. 1981. Washington, D.C.: U.S.
House of Representatives.
Reynolds, Lloyd, and C. Taft. 1956. The Evolution of the Wage Structure. New
Haven, Conn.: Yale University Press.
Robinson, Joan. [1933] 1950. The Economics of Imperfect Competition. London:
Macmillan.
Rottenberg, Simon, ed. 1981. The Economics of Legal MinimumWages. Washington,
D.C.: American Enterprise Institute.
144 Thomas C. Leonard
Sidgwick, Henry. 1886. Economic Socialism. Contemporary Review 50 (Novem-
ber): 620–31.
Smith, Ralph, and Bruce Vavrichek. 1987. The Minimum Wage: Its Relation to
Incomes and Poverty. Monthly Labor Review 110.6:24–30.
Stigler, George. 1946. The Economics of Minimum Wage Legislation. American
Economic Review 36.3:358–65.
———. 1947. Professor Lester and the Marginalists. American Economic Review
37.1:154–57.
———. 1982. The Economist as Preacher and Other Essays. Chicago: University
of Chicago Press.
Webb, Sidney. 1912. The Economic Theory of a Legal Minimum Wage. Journal of
Political Economy 20.10:973–98.
Webb, Sidney, and Beatrice Webb. [1897] 1920. Industrial Democracy. London:
Longmans, Green.
Welch, Finis. 1995. Comment for review symposium on Myth and Measurement:
The New Economics of the Minimum Wage, by David Card and Alan Krueger.
Industrial and Labor Relations Review 48.4:842–49.
Whaples, Robert. 1996. Is There a Consensus among American Labor Economists?
Journal of Labor Research 17.4:725–34.
doc_615295575.pdf