^

+*
*rf
r
^F SUMMARY

Opponents of minimum wages argue that they hurt jobs in Europe;

supporters say that they combat exploitation and help the poor. We try to sort
out myth from reality. Differing policy prescriptions reflect different views of
how the labour market actually works. Contrary to popular wisdom, it is as
easy to make a theoretical case against minimum wages as for them.
Evidence, not theory, is what is needed now.
Relative to average wages, minimum wages have not risen in Europe over
the last 30 years; they caused higher unemployment only if they prevented
a necessary fall in the wages of the low paid. Second, minimum wages for
young workers are often a lower proportion of average earnings in Europe
than in the USA. Third, we find no general evidence that minimum wages
reduced employment, except perhaps for young workers. The (good or bad)
effect of minimum wages has been exaggerated.

—Juan Dolado, Francis Kramarz, Stephen Machin, Alan Manning,

David Margolis and Corn Teulings

Economic Policy October 1996 Printed in Great Britain

© CEPR, CES, MSH, 1996.
r
MINIMUM WAGES 319

The economic impact of

minimum wages in Europe

Juan Dolado, Francis Kramarz, Stephen Machin,

Alan Manning, David Margolis and Coen Teulings
Bank of Spain; CREST, INSEE; University College, London and Centre for Economic
Performance, LSE; Centre for Economic Performance, LSE; University of Montreal; University
of Amsterdam

1. INTRODUCTION
The minimum wage is a controversial policy. Its proponents typically argue that it
helps prevent the worst excesses of exploitation in the labour market, and brings the
living standards of the lowest paid up to some minimum acceptable standard; its
opponents argue that it may price low-skill workers out of jobs, and that many of
diose typically affected by the minimum wage do not live in poor households. The
economics of the minimum wage has recendy been the subject of renewed interest
for two reasons.
First, the minimum wage is widely seen as one of the labour market institutions
causing high and persistent unemployment in Europe. As unemployment rose up
die policy agenda, interest in the minimum wage revived. The recent OECD Jobs
Study (OECD, 1994) recommended, in an uncharacteristically forthright statement,
that the member countries should

Reassess the role of statutory minimum wages as an instrument to achieve redistributive

goals and switch to more direct instruments. If it is judged desirable to maintain a legal

We would like to thank Peter Burgess, Ana Rute Cardoso, Richard Dickens, Susan Harkness, Claudio Lucifora,
Costas Kanellopoulos, Joep Konings, Steve Mcintosh, Erik Strojer Madsen, Marco Manacorda, Sally Marullo,
Christoph Schmidt, Henri Sneessens, Erienne Wasmer and Fredrik Wulfsberg for help with this paper, and Jim Sesil
in particular for his research assistance. We would also like to thank the Economic Policy Panel and especially David
Begg, Michael Keen and Gilles Saint-Paul for their comments.
320 JUAN DOLADO ETAL

minimum wage as part of an anti-poverty strategy, consider minimising its adverse

employment effects, including:

• Indexing it to prices, rather than average earnings

• Ensuring sufficient differentiation in wage rates by age and region to prevent die
minimum wage from harming employment prospects for young people or low-
productivity regions.

Second, the cosy consensus among economists about die employment effect of die
minimum wage - summarized in the survey of Brown et al. (1982) as 'a 10%
increase in the minimum wage reduces teenage employment by 1 to 3 % ' , diough
with no ^identifiable effect on the adult labour market - has been challenged by
new research. Card and Krueger (1995) examine evidence on the employment
effects of die minimum wage in the USA using a number of different data sources
and methods, and conclude that

recent minimum-wage increases have not had the negative employment effects predicted
... Some of die new evidence points towards a positive effect of die minimum wage on
employment; most shows no effect at all. Moreover, a reanalysis of previous minimum-
wage studies finds little support for die prediction diat minimum wages reduce
employment.

These conclusions have not gone unchallenged (see, for example, the reviews of
Card and Krueger in the Industrial and Labor Relations Review, July 1995; papers on die
minimum wage in the American Economic Review, May 1995; Neumark and Wascher,
1995;' and Schmitt, 1996). The final verdict is not yet in.
We aim to assess the effect of minimum wages in Europe, cutting through die
emotive positions of proponents and opponents to offer a more balanced assessment
of its impact in different countries; there is no substitute for a careful study country
by country. However, one paper cannot consider all European countries. We start
with a brief description of the minimum wage systems currendy in operation in a
range of countries, tiien focus on four: France, the Netherlands, Spain and the UK.
Each has some intrinsic interest, but their experiences have broader lessons: France
is often cited as an object lesson in the adverse effects of very high levels of die
minimum wage, die Nedierlands had a large cut in youdi minimum wages in die
early 1980s, Spain had a very large rise in die youth minimum wage in 1990, while
the UK abolished all minimum wages (except in agriculture) in 1993.
We start with a brief description of die minimum wage systems in operation in die
countries of western Europe and assess their impact on the wage distribution. We
then consider die dieory about the minimum wage, arguing that many common
perceptions about die minimum wage are at best half-truths. We try to paint a more

1
Neumark and Wascher received a lot of publicity, but their continued reluctance to make their data available makes
it hard to evaluate their contribution.
MINIMUM WAGES 321

balanced picture. Our discussion is informal, but its claims are backed up by a
formal model sketched in a Technical Appendix. Our four country studies focus
mainly on the crucial issue of the employment effects of minimum wages. Finally, we
offer policy conclusions.

2. MINIMUM WAGES IN EUROPE

The systems of minimum wages in operation in most European countries are
summarized in Table 1. There are five main types of system. First, a statutory
minimum can be set by government (possibly in consultation widi employers and
unions), as in France, Spain, the Netherlands, Portugal and Luxembourg. Second,
as in Belgium, Greece and Denmark, a national minimum wage can be set as part of
national collective bargains. Third, different minimum wages can be determined in
sectoral collective agreements (generally extended to employers who were not party
to the original agreements), as in Germany, Italy, Austria and, to some extent,
Switzerland. Fourth, as in Sweden, Norway and Finland, collective agreements can
cover effectively everybody and generally contain minimum rates without any
formal provision for extension of these rates to non-signatory employers. Finally, as
in Ireland and the U K (prior to 1993), minimum wages can be set in selected low-
paying industries.
There is also considerable variation in how many different minima are set: most
countries have variation by age; some have variation across region, occupation,
industry, job tenure, size of firm, and even marital status and number of
dependants.
One would like some measure of the importance of the minimum wage in die
wage determination process. The real level of die minimum wage may be inappro-
priate in comparisons across time or between countries with differing productivity: we
expect die effect of minimum wages to depend on dieir level relative to labour
productivity. The commonest measure is therefore die minimum wage as a fraction
of average earnings, known in die USA as die Kaitz index (often weighted to allow
for the fact diat not all workers are covered by minimum wage laws). The Kaitz
index is analogous to die most widely accepted measure of die impact of die wage
floor provided by welfare benefits: namely, the replacement ratio.
Table 1 presents estimates of die Kaitz index for as many countries as possible
(die USA is included as a comparison). Minimum wages in most European countries
are about 5 0 - 7 0 % of average earnings compared to about 33% in die USA. In
countries witii a number of different minimum wages, diere are obviously difficulties
in computing a single measure of die Kaitz index. Measures are dierefore a general
indication of minimum wages rather dian precise estimates. This is a particularly
serious problem in countries like Germany and Italy, which have a very large
number of minimum rates diat may be very different. In countries witii a single
statutory minimum, die effective Kaitz index will be much higher for less skilled
Table 1. Minimum wage systems in Europe and the USA
Country Determination Variation by: Kaitz % of workers Youth minimum as % of adult Replacement
index at or near minimum ratio
(year) minimum

Austria Legally binding collective Industry, region, dependants, 0.62 4% Embodied in industry agreements 0.5
agreement at industry level age, job tenure (1993)

Belgium Negotiated by unions and Age, job tenure 0.60 4% Small reduction for <23 (lower 0.6
employers as part of national (1992) rates for short job tenure)
agreement

Denmark Negotiated as part of collective Industry, age 0.54 6% 40% (<18) 0.9
agreements (1994)

Finland Negotiated as part of collective Age, occupation, industry, region 0.52 0.63
agreements (1993)

France Set by government constrained by Age, training 0.50 11% 80% (age 16); 90% (age 17) + 0.57
formula (1993) schemes + 3 0 - 7 5 % for trainees

Germany Part of collective agreements, Age, qualifications, trainee status, 0.55 Embodied in industry agreements 0.63
then extended region (1991)

Greece Part of national collective Manual/non-manual, job tenure, 0.62 20% Lower rates for short job tenure
agreement marital status, qualifications (1995)

Ireland Joint Labour Committees in 16 Age, industry, region, occupation, 0.55 Varies; 6 3 % (<18); 8 1 % (<21) in 0.37
low-paying industries job tenure (1993) hotels
Italy Extension of collective Age, industry, job tenure (0.71 Embodied in industry agreements 0.20
agreements (1991)

Luxembourg Statutory minimum wage Age, skill, family characteristics 0.56 11% 70% (<21)

Netherlands Statutory minimum wage Age 0.55 3.2% 34.5% (age 16) rising to 84% 0.70
(1993) (age 22)

Norway Negotiated as part of collective Industry, age, job tenure, job 0.64 0.65
agreements (1993)

Portugal Statutory minimum wage Age, trainee status, industry 0.45 8% 75% (<18) 0.65
(1993)

Spain Statutory minimum wage Age, homeworkers, casual workers 0.32 6.5% 66% (<18) 0.70
(1994)

Sweden Negotiated as part of collective Age, industry, job tenure, 0.52 0% 85% (<24) 0.80
agreements occupation (1992)

Switzerland Industry-level collective Age, industry 0.52 Embodied in industry agreements 0.70
agreements (coverage 30%) (1993)

United Pre-1993 set by Wages Councils Age, industry (more complex 0.40 0% (<21) (1986-93) 0.38
Kingdom by industry; now only pre-1986) (1993)
agriculture

United States Federal minimum wage (higher Limited youth sub-minimum 0.39 4% No reduction 0.50
in some states) (1993)

Note: We are grateful to Ana Rute Cardoso, Costas Kanellopoulos, Claudio Lucifora, Joep Konigs and Erik Madsen for help in compiling this table.
Sources: Principally die European Industrial Relations Review and ILO Yearbook ofLabour Statistics. Further details available from the authors.
324 JUAN DOLADO ETAL

than for more skilled workers because the numerator in the index is the national
minimum. But in countries where there are many different minimum rates, this is
not necessarily the case because the minimum rate is generally lower for less skilled
workers and the gap in the minimum rates can be very large. In both Italy and
Germany, die highest minimum rate is over diree times the lowest minimum rate,
so diat altfiough die average Kaitz index for Italy looks extremely high (75%), die
Kaitz index for low-paid workers is only slighdy higher.
Given a higher minimum wage, relative to average earnings, in Europe than die
USA, minimum wages might cause job losses in Europe even if mey do not in die
USA. It is tempting to blame high European unemployment at least in part on high
minimum wages. But this may be mistaken. Many commentators feel diat the effect
of minimum wages is strongest in the youth labour market. The USA has little
provision for the payment of lower minimum wages to young workers (the evidence
suggests that die restrictive youth sub-minimum is not widely used), but most
European countries have extensive variation in die minimum by age (see Table 1,
noting diat a number of countries, especially Belgium and Greece, vary die
minimum wage by job tenure, which has the effect of leading to lower minimum
wages for young workers). Given mis variation in the use of die youm minimum, die
variation in die Kaitz index for teenagers across countries differs from diat for
adults, die US level (85%) being much higher in 1992 dian in many European
countries (e.g. Spain 50%, Sweden 80%).
Akhough die Kaitz index is the most widely used measure of die impact of die
minimum wage, concerns are expressed about its use as a measure of die impact of
minimum wages. First, a rise in die minimum wage will generally affect die average
wage; in the extreme case of a corresponding induced percentage increase in die
average wage, die Kaitz index would be unchanged. Typically, of course, die induced
effect on die wages of higher-paid workers will be smaller, so die Kaitz index will rise.
Bom dieory (see die Technical Appendix) and die available evidence suggest diat diis
is die case (see Grossman (1983) for die USA; Bazen and Martin (1991) and Bazen
etd. (1995) for France; and Dickens et al. (1994a) for die UK): a 10% rise in the
minimum wage raises die average wage by at most 3%, akhough effects on die average
wage in die youdi market are, of course, larger. If European labour market institutions
make it easier to transmit rises in die minimum wage to further up die wage structure,
it is dangerous to compare Kaitz indices across continents or even perhaps across
countries; it is safer to use it to analyse changes over time in a given country.
Second, the Kaitz index only measures the impact of the minimum wage on die
structure of earnings (minimum relative to average). It cannot direcdy measure
whether wage costs as a whole are too high in Europe because of labour market
institutions such as the minimum wage. However, it has recently become more
fashionable to argue that European labour market institutions interfere witii die
structure of wages in a way which leads to poor labour market performance (see, for
example, OECD, 1994). The Kaitz index is well suited to analysing tiiis argument.
MINIMUM WAGES 325

Third, the Kaitz indices in Table 1 take no account of taxes paid by employers
and workers which may be important in understanding the problems of European
labour markets. One could compute a producer Kaitz index (the ratio of the cost of
minimum wage labour to the cost of 'average' labour for an employer) or a
consumer Kaitz index (the real take-home wage of a minimum wage worker relative
to the 'average' worker). These differ from the Kaitz indices presented here only to
the extent that the tax system is not proportional. As taxes on workers are generally
progressive, die Kaitz index used here will generally understate die consumer Kaitz
index. But employer payroll taxes are regressive in many countries, in which case
our measure of the Kaitz index will understate the producer Kaitz index.
If higher employer payroll taxes are especially bad for employment in the
presence of minimum wages, can die Kaitz index pick this up? For example,
concern is often expressed in France that die level of employer payroll taxes rose
from 27% in 1950 to 44% in 1992. Workers paid above die minimum can pardy
absorb an increase in payroll taxes dirough accepting lower wages than diey would
odierwise have done, mitigating the rise in labour costs for employers. Workers paid
die minimum wage cannot. But men the Kaitz index rises because its denominator has
fallen, so die index is perfecdy capable of detecting such effects.
Finally, measures of the ratio of the minimum to average earnings may
misrepresent die effect of minimum wages where odier institutions - benefit systems
and collective bargaining - provide odier floors to wages. Another common
measure of the impact of minimum wages is the fraction of workers paid at or close
to die minimum wage. In many sectors widi a minimum wage, there is a noticeable
spike in die wage distribution at die minimum wage; the larger tiiis spike, die more
likely the impact of the minimum wage on die wage distribution. The Kaitz index
and the spike may give different impressions of the importance of minimum wages.
In Sweden, the Kaitz index is higher tiian in die USA, but other institutions
compress die wage distribution so strongly diat nobody actually receives die
minimum (Ostros, 1994)2. Some estimates of die spike are also presented in Table 1:
diey tend to be in the region 5-10%. In order to give some idea of the comparison
of minimum wages and benefits as a floor to wages, we also include a measure of
replacement ratios.
Table 1 refers only to the minimum wage at a point in time, but its importance
may change over time. Figure 1 presents the Kaitz index for a number of
countries. Although minimum wages have risen substantially in all of our
countries, tins has been broadly in line with the increase in average earnings: there
are few dramatic changes in the Kaitz index in European countries. It is hard to
argue from this that aggressive increases in minimum wages caused the spectacular

2
The model of the Technical Appendix shows that the relationship between the Kaitz index and the spike would be
expected to depend on the dispersion of earnings.
326 JUAN DOLADO ETAL

1993 1975
Year

Figure 1. Kaitz index i n s e l e c t e d countries

Sources: Available from the authors.

o Spike 1 a Spike 2

20 -

15

10

5 -

0 -

1954 1960 1965 1970 1975 1980 1985 1990

Figure 2. T h e evolution o f the spike i n France

Notes: Spike 1, taken from the annual ACEMO series reported in Combault (1995), measures the
proportion of workers at end-June earning between the old and new SMIC. Spike 2 is an estimate of the
proportion of workers in die Enquete Empkn who report earnings in the band which contains the SMIC
multiplied by their usual hours of work (actual earnings are not reported).
MINIMUM WAGES 327

rise in European unemployment. But failure to decrease the index in response to

changed market circumstances might still have had an effect. Powerful forces such
as globalization and skill-biased technieal change in all OECD countries may
require widening the wage distribution (see OECD (1994) for a forceful statement
of this view).
Although minimum wages may be a constant fraction of average earnings, they
may be a rising fraction of the earnings of those workers affected. If so, we would
expect to see evidence of a rising proportion of workers paid at or near the
minimum, or a more general compression of wages at the bottom of the distribution.
Unfortunately, it is hard to get estimates of changes in the spike for most countries.
We have two estimates for France (Figure 2). Both show some rise in the spike in the
1980s, consistent with the view that the minimum wage now has a bigger impact
than in the 1970s (akhough the highest levels of the spike were in the 1950s, which
also had the highest Kaitz index). But given the ongoing debate about whether
trends against the less skilled are the main explanation of rising unemployment (see
Nickell and Bell, 1995; Card et al. 1995), we conclude that there is no strong
evidence that minimum wages are a more serious constraint on the European
economies than 30 years ago.

3. THE ECONOMICS OF THE MINIMUM WAGE

The conventional wisdom about the implications of standard economic dieory can
be encapsulated in five propositions:

• A minimum wage cannot increase employment and generally reduces it.

• Its adverse employment effects are largest in a small open economy where
international competiveness is most significant.
• Lower tax rates or higher subsidies are a better way to improve both the
employment prospects and the incomes of the low paid.
• Young workers are most affected.
• Minimum wage earners do not usually come from the poorest households, so
minimum wages do little to alleviate poverty.

Is this picture accurate? Our discussion will be largely informal, but it is based on
a formal model of the labour market in the Technical Appendix. Firms are
heterogeneous, so that different firms are affected in different ways by the
minimum wage.

3.1. Economic theory predicts that minimum wages must reduce

employment
Predictions of economic theories are almost always sensitive to assumptions. We are
surprised by an unconditional claim like the one made above and sceptical uiat
328 JUAN DOLADO ETAL

anyone actually believes it. Yet it pervades analysis of die minimum wage. Consider
the OECDJobs Study (OECD, 1994). On page 49 it states:

A number of full-time workers in Canada are estimated to have received hourly wages
below the average minimum in 1987, but mere is a clear jump in the number of young
persons once die minimum wage was passed. While diis suggests diat legislation has
impeded employment for young people in Canada, diere is little evidence of any jumps in
die wage distribution in the United States in 1987. This in turn would call into doubt any
negative employment effects of die minimum wage in diat year

and on the following page:

The link between minimum wages and youdi wages in France and Canada ... does tend
to indicate die potential for a negative impact on employment of statutory wage floors.

This argument is based on the familiar competitive labour market model in which
firms have to pay the market wage and workers get paid their marginal product, as
in Figure 3a. If the labour market clears in the absence of a minimum wage, a
binding minimum wage (Wt) must reduce employment from jV0 to JV,. The
employment loss from a given minimum wage will be larger the more elastic is the
demand for labour.

Real
wage

0 N, W0 Employment

Figure 3a. M i n i m u m w a g e s and e m p l o y m e n t i n a competitive labour m a r k e t

Real
wage

IV,

Wa
Wm

0 Ni Wm W0 Employment

Figure 3 b . A monopsonistic labour m a r k e t

MINIMUM WAGES 329

However, embedding analysis of the effect of a minimum wage within a labour

market assumed to be perfectly competitive is not the only possibility; it is less well
known mat within other contexts a minimum wage may raise employment. Labour
market textbooks show how mis occurs under monopsony, where there is a single
buyer of labour. The employer can choose what wage to pay, a higher wage bringing
forth a higher supply of workers. A profit-maximizing employer will choose the level of
employment where the marginal product of an extra worker is equal to the marginal
cost of an extra worker, which exceeds the wage: attracting the extra worker requires
raising wages also for existing workers.3 Hence each worker gets paid a wage below his
or her marginal product: it is this gap which allows the minimum wage to raise the
wage of workers without entirely removing their profitability to firms.
Consider the monopsonistic labour market in Figure 3b. In the absence of a
minimum wage, wage and employment will be at Wm and JVm. A binding minimum
wage at Wx makes me marginal cost of labour to the firm flat until it hits die labour
supply curve (hiring extra workers no longer requires raising the wage). Previously,
the employer chose low wages and suffered the consequence of low employment; but
die low-wage option is no longer allowed and so the employer might as well take on
the extra workers forthcoming at the minimum wage. However, there is a limit to
how high the minimum wage can be set without hurting employment. Employment
is maximized when the minimum wage is set at the level mat would prevail in a
competitive labour market (where the labour supply and MRPL curves cross).
Attempting to raise die minimum wage still furmer will result in employment losses
(relative to the maximum attainable). The extent to which minimum wages can be
raised witiiout destroying jobs depends on bom the elasticity of the labour demand
curve and die elasticity of the labour supply curve facing die firm, with minimum
wages being more effective die more elastic is the labour demand curve
Of course, die level of wages that would prevail in a competitive market will differ
across segments of the labour market; it might be lower for young workers and in
some regions. A single national minimum wage is men an extremely blunt policy
instrument, being set too low in some markets (employment could be raised by
having a higher minimum) and too high in otiier markets (employment is reduced).
If mere are forces leading to wider dispersion in wages, mis problem will worsen
over time and will act as a constraint on the minimum wage. It is better to vary the
minimum wage across labour markets where the competitive wage is expected to
differ, notably by age or region. It is also possible that any binding minimum wage
reduces employment even diough all labour markets are monopsonistic. This could
occur where the labour supply to die market as a whole is inelastic, but where mat
to me individual firm is more elastic (see die Technical Appendix for more details).

3
Unless a different wage can be paid to each and every worker. While employers can partially discriminate across
workers, perfect discrimination is implausible.
330 JUANDOLADO£T.li.

Having explained monopsony, most labour economics textbooks then argue that
few employers are a single purchaser of labour and hence monopsony is extremely
rare. This conclusion is seriously misleading. The important features of monopsony
will be reproduced in any situation where firms have some discretion over die wages
diey pay. For example, it is plausible to tfiink mat the higher the wage that a firm
pays, die easier it finds it to recruit and retain workers. Put diis way, it is perfect
competition that is die extreme and implausible outcome, widi its assumption mat
an employer which cuts wages by even a fraction of an ecu will find diat all its
workers immediately leave. Nobody would support such an extreme view; and
anything else is monopsony. The important question is the extent of monopsony
power: is perfect competition a tolerable approximation to reality?4 The key point is
diat economic theory has no unambiguous prediction about the employment effects of
minimum wages. Empirical research is required.
There are otiier meoretical arguments why a minimum wage might raise
employment. Calvo and Wellisz (1979) argue that, as a firm grows in size, it
becomes more expensive to monitor workers and hence the average cost of labour
rises (mis makes these models akin to monopsony models), an idea recendy revived
by Rebitzer and Taylor (1995). In efficiency wage models more generally, where die
productivity of labour depends on die wage paid, the minimum wage may raise
employment (Manning (1995) provides a general condition). One can also make a
case for a minimum wage even if it reduces employment for some workers. Cahuc
and Michel (1996) present a model in which a binding minimum wage will price
some workers out of the market unless tiiey raise their productivity to die level of die
minimum by education and training; the effect of die minimum wage is mat some
workers undertake training tiiey would otherwise not have done. This is an outcome
that is socially desirable because tiieir model is set up so mat there is some
underinvestment in training. One should note diat it has also been argued that die
minimum wage discourages training, and that models can be constructed in which
tiiis is the case. The conclusion should be that a perfecdy literate economic
argument can be made for or against a minimum wage. Theory alone will not
resolve the debate: evidence is what is needed. Where and how should one look?
We provide a brief description of die main methods diat have been used togemer
widi an assessment of tiieir respective strengms and weaknesses. We will argue diat
diere is a danger in routinely applying US methodology to Europe: diere are good
reasons why die approaches are less likely to be successful on diis side of die Adantic.
One approach is simply to look at die correlation of employment changes with
minimum wage changes, controlling for other relevant factors. This approach

4
Is there other evidence for monopsony power in labour markets? Proponents argue that the existence of equilibrium
wage dispersion, the employer-size wage effect, discrimination, the earnings losses suffered by displaced workers, and
the general training paid for by employers are all explained most plausibly by labour market frictions which are the
source of monopsony power. Opponents claim that other explanations can be found for these empirical findings.
MINIMUM WAGES 331

typically exploits variation in minimum wages (or more commonly the Kaitz index)
over time and/or across industries and regions. The approach has a very simple
interpretation (the Technical Appendix provides some justification), but also some
potential weaknesses. First, it is difficult to control fully for other supply and demand
factors that affect employment, and die Kaitz index will pick up the effect of omitted
supply and demand factors (die denominator in the Kaitz index is die average
wage, which will be affected by demand and supply shocks). So it is important to
establish mat die variation in the Kaitz index is driven primarily by variation in the
minimum wage. The approach has also been criticized for failing to recognize die
endogeneity of die minimum wage (e.g. some have argued that the minimum wage
is increased faster in good times). The seriousness of mis problem is a matter of
debate: minimum wages may be set a considerable time before diey take effect, and
one might question die economic sophistication of the process by which diey are
often set.
A second general approach (fashionable in recent years) is to exploit 'natural
experiments': episodes in which minimum wages were raised. Most notable is die
study by Card and Krueger (1994) of die rise in the minimum wage from $4.25 to
$5.05 in New Jersey in April 1992 on employment in fast-food restaurants. As a
control group, unaffected by minimum wage changes but facing other relevant
shocks, Card and Krueger used the experience of restaurants in neighbouring parts
of Pennsylvania. This type of methodology is easier to apply in the USA (and also
Canada): quoting Judge Louis Brandeis, 'it is one of die happy incidents of die
federal system mat a single courageous state may, if its citizens choose, serve as a
laboratory and try novel social and economic experiments widiout risk to die rest of
the country' (Card and Krueger, 1995). It is much less easy in Europe, where
minimum wages are generally a national minimum.
But mere are odier control groups to use. For example, a rise in die minimum
wage should have more effect on the employment of low-wage groups man high-
wage groups: what might be called die 'differential' impact approach. The groups
might be regions (as in Card (1992a, 1992b) and die studies of France and Spain
below), firms (as in Card and Krueger, 1994), individuals (as in Currie and Fallick,
1994), segments of the labour market as defined by age and education (as in the
study of France below), industries (as in die study of die abolition of minimum
wages in the U K below) or occupations (as in the study of die Netherlands below).
This approach has intuitive appeal, but again a number of weaknesses. First, it is
only as good as die control group used. For example, Card and Krueger (1995)
criticize Currie and Fallick for ascribing all die difference in die risk ofjob loss between
low- and high-wage individuals to the effect of the minimum wage. Much me same
relationship between risk of job loss and wages exists for workers not covered by the
minimum wage. As we show in die Technical Appendix, mere are potential dangers in
using high-wage agents as control groups for low-wage agents when we have no
explanation as to why a high wage was chosen in die first place. Second, die approach
332 JUANDOLADO£7M£

is most likely to produce reliable and precise results when there is a large rise in the
minimum wage, so that differences in the experience of the treatment and control
groups are most plausibly due to the effect of the minimum wage. Again the USA is a
more fruitful field for this approach because it has infrequent but large changes in
minimum wages. European countries in which minimum wages are upgraded
annually in line with average wages offer fewer examples of large changes in minimum
wages.5 Third, this approach typically compares employment shortly before and
shortly after changes in the minimum wages. This may not be such a problem in the
USA, where labour turnover rates are high and firing costs are low; but it might be
more of a problem in Europe, where lower turnover rates and firing costs mean that
employers are more likely to adjust employment in advance of predictable minimum
wage changes, and the adjustment process may be more drawn out. Looking at
employment changes over a longer period reduces this problem, but makes it more
difficult to ascribe observed changes to changes only in minimum wages.
A third approach (to which the OECD Jobs Study (OECD, 1994) devotes most
discussion) was developed by Meyer and Wise (1983a, 1983b) and subsequently
used by, among others, Albaek and Madsen (1987, 1989) and van Soest (1989,
1994). It attempts to infer the employment effect of minimum wages using only data
on the cross-sectional distribution of wages, claiming to estimate the effect of
minimum wages without having any variation in the minimum at all. But some very
strong assumptions are necessary to do this: a particular distribution of wages in the
absence of the minimum (e.g. log-normal), and an assumption that workers earning
above a certain level are unaffected by the minimum. One then estimates die
parameters of the wage distribution from those earning above the minimum, infers
how many people should be earning at or below the minimum, and compares this
figure with the number of people actually there. The theoretical foundation for this
approach is very weak (it is totally inappropriate in the model of the Technical
Appendix) and there is a serious danger that the results are very sensitive to the
precise assumptions made about the distribution of wages and the way it is affected
by minimum wages. For die UK, Dickens et al. (1994b) conclude diat these
problems are so serious as to make this approach unusable.

3.2. Minimum wages are inappropriate for small open economies

Many discussions in Europe emphasize the need to encourage wage moderation to
preserve international competitiveness. It is argued that a single economy that raises
its minimum wage will raise its wages relative to competitors with adverse effects on
traded goods industries.

* We study some of these large changes below. Others that we discovered but have not examined are Greece in 1982,
Denmark in 1977 and Switzerland in 1995.
Table 2. Distribution of workers by sector (% of workers)
Country Category of Agriculture Mining Clothing Other Electricity, Construction Trade, Transport Finance, Personal
worker and leather manufacturing gas, water restaurants, insurance, services
hotels etc.

France All 0.4 2.1 20.4 1.0 7.7 18.1 6.8 10.1 33.2
(1989) Min. wage 0.0 6.2 10.3 0.0 5.2 25.7 2.9 4.12 45.5
Netherlands All 4.8 0.2 22.8 0.9 9.4 16.1 8.1 11.0 26.5
(1993) Min. wage 1.7 0.0 8.3 0.1 2.4 43.9 2.8 19.5 21.4
Portugal All 0.8 5.5 23.4 1.2 9.8 19.4 5.8 5.4 28.7
(1989) Min. wage 0.4 4.2 213.0 0.0 7.0 26.5 0.9 1.1 38.6
Spain All 21.3 1.4 9.0 67.3
(1994) Min. wage 9.8 0.1 8.8 81.3
United All 1.1 0.4 2.0 16.5 0.9 4.6 20.4 6.7 13.2 32.2
Kingdom Min. wage 2.1 0.1 41.0 6.8 0.1 1.9 41.4 4.3 9.6 25.5
(1994)

Notes: In the UK, minimum wage workers are defined as those wim an hourly wage below £3.50 in April 1995 prices. In the Netherlands, the minimum wage is defined ai
those paid within 5% of the statutory minimum. For Spain, Portugal and France, data are official estimates.
Sources: ILO Yearbook ofLabour Statistics; Combault (1995); supplemented by data for particular countries (details available from die authors).
334 JUAN DOLADO ETAL

This argument makes sense if the labour market is competitive (as in Figure 3a),
for then a given rise in the minimum wage will have a larger negative effect on
employment the more elastic (flatter) is the labour demand curve, as is most likely
with severe international competition. But we also saw diat the employment-raising
potential of the minimum wage in monopsonistic labour markets is larger if the
labour demand curve becomes more elastic. With a large international market,
increased employment and increased output can be achieved with litde reduction in
price. The marginal revenue product of labour does not fall as fast as employment
rises. Again, the effect of minimum wages is very different in competitive and
monopsonistic labour markets.
Anodier important point is mat die low paid in European countries are typically not
concentrated in traded goods industries. Table 2 presents information on minimum
wage workers by sector for our four countries (plus Portugal, a lower-wage country)
and (as a comparison) me sectoral distribution of employment for all workers. Low-
wage workers are generally underrepresented in manufacturing: typically doming and
learner (and also furniture making) is the only manufacturing sector wim many low-
paid workers. The comparative advantage of developed economies tends to be in
relatively skill-intensive industries. However, the importance of tourism, which may be
price sensitive, may mean mat the wages of workers in many service industries are
important for the competitive position of die Spanish and Portuguese economies.

3.3. Lower labour taxes or subsidies are a better way to help the low paid
One of the main aims of the minimum wage is to improve the living standards of me
low paid. The minimum wage is not the only way in which this might be done.
Those who want to argue diat a minimum wage is desirable need to demonstrate
that it is a more effective policy tool than alternatives. Perhaps the most prominent
alternative is to reduce the taxes or even subsidize the employment of me low paid.
A lower tax rate on labour will lead to a rise in employment and living standards
for the affected workers in both competitive and monopsonistic labour markets. For
a given producer wage, labour supply increases. But reduced labour taxes will
reduce tax revenue, so they must be paid for in some way. The usual proposal is mat
the taxes on the low paid should be reduced and those on die higher paid increased:
a more progressive tax system. In a competitive labour market, mis might be a
better policy than a minimum wage because it raises the living standards of die low
paid while raising their employment at the same time 6 . The reason for the difference
is that the subsidy received by the low paid is paid by different groups: the employers
of the low paid in the case of a minimum wage and higher-paid workers in the case
of a reduced tax burden.

6
This policy might still reduce incentives to invest in the acquisition of skills.
MINIMUM WAGES 335

But matters differ if the labour market is monopsonistic. Making die tax system
more progressive makes the effective supply of labour facing die monopsonistic firm
more inelastic (steeper): it would change from NS^ to JVS2 in Figure 4. This raises
die marginal cost of labour from MCLX to MCL^ so that the wage paid by the
employer and employment fall. Effectively, one is providing subsidies for employers
to cut wages. This effect can be so large mat the living standards of the low paid
actually fall when dieir labour is subsidized (an effect that is akin to the overshifting
of taxes tiiat can occur in oligopolistic product markets). The minimum wage in this
case is a much better policy because making die labour supply curve facing die
employer flatter is a move towards efficiency. Again, policy recommendations
depend on one's view of die way in which labour markets work.

3.4. Most minimum wage workers are young

A high proportion of both research and policy discussion on the effects of die
minimum wage focuses on die young. From this, one might conclude mat young
workers are die main group affected by the minimum wage. But while it is certainly
true that young workers are more likely to be low paid than the average worker, it is
not necessarily true diat young workers make up die bulk of die low paid. Young
workers typically make up a small proportion of die total workforce. Table 3
presents some information for our four countries on the proportion of low-paid
workers who fall into a number of categories. Young workers used to be a higher
proportion of die low paid, but dieir importance has declined widi rises in school
enrolment and women's labour market participation. For example, in the UK, the
proportion of workers in the bottom decile of the hourly wage distribution who are
teenagers fell from 40% in die mid-1970s to 18% in the early 1990s. The focus on
die effect of minimum wages on the youth labour market leads to a misleading
impression of die effect on the labour market as a whole. It may be, for example,
tiiat the minimum wage does reduce employment for young workers, but that it
increases employment of women and diat the latter effect dominates (as suggested

Real . MCL2
wage /MCL,
NS,

_NS,

W2 •' | N .
MRPL
r
0 W2 W, Employment

Figure 4. Labour subsidies for the low paid in a monopsonistic labour market
336 JUAN DOLADO ETAL

Table 3 . Characteristics o f the low paid

France Netherlands Spain United Kingdom

(1990) (1985) (1990) (1994)

Proportion female 0.54 0.81 0.58 0.66

Proportion age <21 0.09 0.01 0.41 0.29
Proportion part time 0.11 0.71 0.37 0.52

Sources: French computations from Enquete Emploi; Netherlands computations from 1985 Wage Survey,
Spanish computations from Vaktes (1992), Labour Force Survey and Tax Returns; UK computations from
Labour Force Survey.

below in evidence for Spain). Unfortunately, some of this paper only reinforces this
emphasis on the youth labour market, since the obsession with the youth labour
market means that most of the policy changes have concerned youth minimum
wages and economists need such changes to study die effects of minimum wages. We
know very little about die effect of minimum wages on womenfc employment
although there is evidence from some countries mat quite large increases in womens
pay achieved dirough equal pay legislation had no noticeable adverse effects on dieir
employment (see Manning (1996) for a study of the U K Equal Pay Act). It is
possible that the effect of minimum wages might be similar. More research is
needed.

3.5. Most minimum wage workers are not in poor households so a

minimum wage does little to alleviate poverty
There is an important element of truth in this. The main cause of poverty in most
European countries is unemployment, and die minimum wage does nodiing to raise
die living standards of those households in which nobody is in work. But it is
unreasonable to expect the minimum wage to have such an effect. It might be more
reasonable to focus on diose households which do have someone in work.
Table 4 presents information on the position of minimum wage workers in the
household income distribution for the countries diat we study. For every individual
in a household with someone in work, we assign a level of equivalized income based
on total household income adjusted for the number of adults and children in diat
household. We dien order individuals by mis equivalized income level and ask what
proportion of workers in each decile are minimum wage workers, yielding a picture
of where minimum wage workers are in the household income distribution.
As we move up die household earnings distribution, the proportion of minimum
wage workers falls. Low-paid workers do tend to be in poorer households. However,
mere are two ways in which die minimum wage may not be a very effective anti-
poverty tool. First, low-paid workers may not all be in poor households (altiiough we
know litde about die distribution of income widiin households). Second, many poor
households may contain no minimum wage worker. In all our countries, die picture
MINIMUM WAGES 337

Table 4 . T h e relationship b e t w e e n m i n i m u m w a g e s and income distribution

Decile France Nedierlands Spain United Kingdom

% %of % %of % %of % %of

affected affected affected affected affected affected affected affected

All 7.2 100.0 11.5 100.0 6.6 100.0 14.5 100.0

1 10.1 13.2 35.0 30.4 23.5 35.5 41.7 28.7
2 11.3 15.8 21.0 18.2 12.2 18.4 19.1 13.2
3 13.1 17.7 24.0 20.9 9.6 14.5 18.2 12.5
4 8.4 11.4 16.0 13.9 7.4 11.2 12.2 8.4
5 6.6 9.1 8.0 6.9 6.2 9.4 11.8 8.1
6 7.4 10.7 5.0 4.3 3.2 4.8 10.3 7.1
7 7.1 9.8 2.0 1.7 2.1 3.2 12.1 8.3
8 5.1 7.4 2.0 1.7 1.3 2.0 8.0 5.5
9 2.2 4.1 1.0 0.9 0.6 0.9 5.4 3.7
10 0.6 0.8 1.0 0.9 0.1 0.2 6.9 4.7

Notes: For each country, the first column is the proportion of individuals in each decile of the equivalized
household income distribution who are minimum wage earners; the second column is the fraction of
individuals who are affected by the minimum wage who are in each decile. For the Netherlands it is the
adult rate that is used for all workers, which is why the incidence is high. Households without any
worker are excluded from these computations. These figures make no allowance for taxes and benefits
that may be very important in practice. See Sutherland (1995) for an analysis of the UK, Nolan (1993)
for Ireland, for analyses including the tax/benefit system.
Sources: Spain and UK as in Table 3; France from the Enquete Actifs Financiers (1990); the Netherlands
from a micro model used by the Ministry of Social Affairs.

is similar: minimum wage workers tend to be in low-income households, but a

considerable proportion are quite a long way up the household income distribution.

4. THE EFFECT OF MINIMUM WAGES ON EMPLOYMENT

In this section we present more detailed studies of four European countries. In each
we provide an account of particular episodes, but we are not necessarily in a position
to provide an overview of me effect of the minimum wage system as a whole because
most of our countries have essentially had the same system throughout our study
period.

4.1. France
France first instituted minimum wages in 1950, when a law was passed establishing
the SMIG (salaire minimum interprofessionel garanti, or guaranteed minimum earnings).
The SMIG was raised in line with prices when the inflation rate exceeded 5% per
year (reduced to 2% per year in 1957), but was not indexed to earnings growth. The
consequence (see Figure 1) was that the Kaitz index fell from a very high level in
1950 to a very low level by the late 1960s. A major revision of the minimum wage
law took place in 1970, converting the SMIG to the SMIC (salaire minimum
interprofessionel de croissance, or minimum wage for growth), revised each July to
338 JUAN DOLADO ETAL

increase in line with prices and by no less than 50% of the amount by which the real
TSH (taux de salaire koraire ouvrier, or hourly blue-collar wage) increased in the
previous year. Neitfier minimum wage law ruled out increases above die required
minima; nor did they prevent the government from increasing the minimum at
dates odier dian 1 July. The largest discretionary increases in minimum wages were
in 1968, the early 1970s and (more modesdy) the early 1980s (see Figure 1). From
1988 to 1993 the SMIC evolved at the same rate as the TSH. In 1993 and 1994 the
SMIC increased by die bare minimum. Newly elected President Chirac raised the
SMIC by 4% in July 1995.
The SMIC is not the only form of minimum wages in France. French law allows
minimum wages to be set by collective bargaining agreements that apply only if diey
exceed the statutory minimum wage. At the beginning of die 1980s, many of die
minimum rates in diese agreements were above the SMIC, but since then sectoral
minima have grown less rapidly than die SMIC: by 1990 60% of sectoral agree-
ments had some point on dieir pay scales below the SMIC, so one reason why the
proportion of workers paid the SMIC was higher in the 1980s (see Figure 2) may be
a reduction in sectoral minima radier man a more generous SMIC.
There have also been important changes in recent years in die operation of die
SMIC in response to growing concern about the possible adverse effect of minimum
wages on the employment of vulnerable groups: notably those aged less than 25, but
also some for the long-term unemployed. These schemes offer subsidies to employers
(often in the form of exemptions from payroll taxes) and also often reduce the wage
received by die worker. For example, the apprenticeship and qualification contracts
(the largest in terms of numbers on die schemes) allow a 16-year-old to be paid 25%
of die SMIC, rising to 78% for a 22-5-year-old in die final year of the contract. The
growth in the number of workers on these schemes has been dramatic, from about
8000 in 1976 to about 900 000 in 1993 with die largest rise after 1986 (Gelot, 1995).
The effect on wages paid is noticeable. The proportion of workers paid below die
SMIC rose from 1% in 1981 to 4% in 1989. And the OECD reported that die
average wage for French teenagers in 1987 was actually below die SMIC. This should
be taken not as evidence diat die SMIC is set at a very high rate, but as an
indication diat the SMIC is not the effective minimum wage for tiiese workers.
Let us now consider evidence of the SMIC's effect on employment and unemploy-
ment in France. Bazen and Martin (1991) used a time-series approach to conclude
diat me minimum wage had a modest adverse effect on youtii employment, but
dieir mediodology assumes both die validity of the competitive model and diat
minimum wages can affect employment only through an effect on die average wage.
However, diey concluded diat their estimates might not be robust, a conclusion also
reached by Benhayoun (1994) in his study of the effect of the minimum wage on
youdi employment. Abowd et al. (1995) compare France and die USA, claiming to
find evidence of modest employment reductions associated with the SMIC. What
can we say about the employment effects?
MINIMUM WAGES 339

A commonly heard argument is that the SMIC was raised substantially in the
early part of the 1980s and is important in explaining high and persistent
unemployment in France. Does the experience of the 1980s support this view? Both
the Kaitz index and the estimates of the spike (see Figures 1 and 2) suggest that the
SMIC became more important from 1981 to 1985 (although the size of the increase
was quite modest), since when its importance has declined slighdy. In looking for
employment effects of the SMIC in the 1980s, we therefore divided the 1980s into
two periods: 1981-5 when Mitterrand (as president) raised the SMIC relative to
average earnings, and 1985-9 when Chirac (as prime minister) did not. Our
intention is to estimate the effect of the SMIC by comparing the labour market
fortunes of parts of the labour market strongly influenced by the SMIC with those
parts that are not. So we divide the labour market into six age groups, eight
education groups and the two sexes, making 96 segments in total. As our measure
of the impact of the SMIC on each segment, we use the fraction of workers paid a
wage near the SMIC in 1985 7 . As we are also interested in the effect of the
schemes, we also compute the proportion of workers in each segment paid below
the SMIC in 1985, which we expect to pick up the groups particularly affected by
the schemes. We then look at how labour market outcomes are related to these two
variables. The results are presented in Table 5, with separate estimates for men and
women.
The first row of each part of the table shows the results of a regression of the
change in the real log hourly wage on the proportion of workers paid the SMIC and
the proportion below it. This regression was run for the whole period and for
1981-5 and 1985-9 separately. From column (3) one can see that real wage growth
in die period 1981-5 is positively related to the proportion of workers paid close to
the SMIC, but that the estimated coefficient is not significandy different from zero.
There is litde evidence that the rise in the minimum wage under Mitterrand led to
larger wage increases for low-paid groups. One could argue that the relative wages
of die lower paid would have fallen further in die absence of die rise in the
minimum. If so, we would expect to see a very strong fall in the relative wages of the
low paid from 1986 to 1989, but the coefficient in column (5) shows diat this is not
die case. The coefficient on the proportion of workers paid below the SMIC is also
inconsistent widi what one would expect, with stronger negative wage growth in
1981-5 (when the schemes were relatively small scale) man in 1986-9 (when they
were on a much larger scale). So the changes in the structure of wages do not
provide much evidence mat the Mitterrand period was one in which minimum
wages were increased very strongly, while the Chirac period was one in which
minimum wages did not increase. This is consistent widi other studies documenting

As in Figure 2, 'near the SMIC is denned by having monthly earnings in the band which includes the SMIC
multiplied by usual monthly hours. We compute the proportions using 1985 data, when schemes had become large
and data more reliable.
340 JUAN DOLADO ETAL

T a b l e 5 . C h a n g es i n t h e F r e n c h l a b o u r m a r k e t , 1981-9
Change in: 1981 - 9 1981 - 5 1985 - 9

Proportion Proportion Proportion Proportion Proportion Proportion

near below near below near below
SMIC SMIC SMIC SMIC SMIC SMIC
(1) (2) (3) (4) (5) (6)

Men
Log hourly wage 0.89 -2.08 0.36 -1.17 0.33 -0.15
(1.66) (1.94) (1.09) (1.78) (0.90) (0.21)
Proportion 0.96 2.74 1.26 3.08 0.32 -0.43
near SMIC (1.87) (2.07) (3.38) (3.18) (0.11) (1.89)
Proportion 1.01 -0.64 1.33 0.13 0.24 -0.48
below SMIC (5.83) (1.42) (7.81) (0.29) (3.61) (3.56)
Employment -0.47 0.93 -0.12 -0.62 -0.06 0.90
rate (1.32) (1.32) (0.45) (1.13) (0.25) (2.00)
Unemployment 0.89 -1.63 0.28 0.68 0.42 -1.78
rate (3.94) (3.50) (2.01) (2.45) (2.04) (4.26)

Women
Log hourly wage 0.43 -1.04 0.18 -0.38 0.07 0.01
(1.09) (1.43) (0.71) (0.81) (0.34) (0.03)
Proportion 0.49 -0.31 0.49 0.12 0.39 -0.91
near SMIC (1.73) (0.26) (1.83) (0.11) (4.46) (5.70)
Proportion 0.71 -1.30 0.76 -1.22 0.24 -0.34
below SMIC (4.28) (1.90) (4.72) (1.86) (0.06) (3.30)
Employment -0.52 0.79 0.44 -1.10 -0.61 1.15
rate (2.04) (1.69) (2.11) (2.86) (2.94) (3.01)
Unemployment 1.27 -1.54 0.00 0.97 0.85 -1.51
rate (6.54) (4.29) (0.05) (3.41) (4.89) (4.75)

Notes: The numbers reported (^-statistics in parentheses) are coefficients from a regression of the variable
in the left column on the proportion of workers paid near the SMIC and the proportion paid below it in
1985; cells are across 8 education groups and 6 age groups. Each regression thus has 48 observations.
Data from the Enquete Emploi.

the stability of the French wage structure in diis period (Card et ai, 1995): one
possible explanation is the declining importance of me sectoral minima in the 1980s.
The second and third rows present estimates of regressions in which the change in
the proportion paid near the SMIC (the second row) and below the SMIC (the third
row) are regressed on the proportion paid near and below the SMIC in 1985. For men
the second row supports the hypothesis that the growth in the proportion paid near the
SMIC grew more rapidly in 1981-5 than in 1985-9, but this is not true for women.
There is also little evidence here for the view diat the proportion of workers paid below
the SMIC grew more strongly in the second period than in the first period. Again these
regressions do not provide strong support for the perception of France in the 1980s as
a country in which die minimum wage was being strengmened very rapidly.
Now let us turn to employment outcomes. The fourth and fifth rows of Table 5
MINIMUM WAGES 341

present regressions similar to those presented above, but widi employment rates (the
fourth row) and unemployment rates (the fifth row) as the dependent variable. The
employment and unemployment effects almost always go in opposite directions; our
discussion is confined to unemployment effects. For men there is evidence that the
rise in die unemployment rate in the Mitterrand period was larger for groups with a
high proportion of workers paid at or below the SMIC; in the Chirac period, die
change in the unemployment rate was positively related to the proportion of workers
paid near the SMIC, but negatively related to the proportion paid below die SMIC.
These estimates are consistent with the view that the SMIC was responsible for some
of the rise in unemployment in die 1980s, an effect diat was mitigated in die second
half of die decade by the expansion of the schemes allowing workers to be paid
below the SMIC.
Before accepting this conclusion, one must recognize mat tiiese results are
consistent with an alternative hypomesis. The years 1981-5 were a recession and
die years 1985-9 were a weak recovery. The unemployment rates of low-paid
workers, particularly die young, rise more in recession but fall more in booms. As
low-paid workers are more likely than others to be paid die SMIC, the results in
Table 5 may be the normal response to recession and recovery.
The results reported in Table 5 are not robust to the inclusion of dummy variables
for die age and education of die workers. The behaviour of wages is arguably more
consistent witii this story man witii the minimum wage story, since die young are die
main group of workers paid below die SMIC and die wage growtii of workers paid
below the SMIC was particularly strong in the period 1986-9. There seems to be no
way, using diese data, to distinguish die two hypotheses using information from die
1980s alone. It is also possible that die employment and unemployment effects of die
minimum are understated: adjustment of employment is very slow in France
(because of high firing costs), which, combined with die fact tiiat minimum wage
increases are largely predictable, means it is likely mat some employment adjustment
takes place before rises in die minimum wage.
The main problem in interpreting die experiences of the 1980s arises because
changes in die minimum wage in mat period were relatively modest (much smaller
man die episodic large increases in die US federal minimum) and because die time
periods we consider are relatively short. But looking at Figures 1 and 2, we can see
diat diere had been much bigger changes in the minimum wage in 1968 and
1973-4. We might hope that there is a better chance of estimating die minimum
wage effects from these episodes. Regressions similar to those reported above
indicate that the segments of die labour market with large proportions of workers
near die SMIC had larger rises in dieir unemployment rates in die 1970s, but die
same problem of interpretation arises because this was a period of a general rise in
unemployment. So we take a different approach here. Because the SMIC is a single
national minimum and mere is considerable regional wage dispersion, we expect big
rises in die SMIC to have larger effects on the economy in low-wage regions. This is
342 JUAN DOLADO ETAL

the methodology used by Card (1992a) in his study of the rise in the federal
minimum wage in the USA. For this purpose we use data on wages and employment
by dipartement for the period 1967-92 (the beginning of the period marks the end of
regional differences in minimum wages).
In a year with a large rise in the minimum wage, we expect larger wage rises in
the regions which initially had very low wages and hence a reduction in the
dispersion of wages across regions. There is evidence that increases in the SMIC
have led to reductions in regional wage inequality. Figure 5 plots the change over
time in the log of the Kaitz index and the variance across departements of log earnings
for male blue-collar workers for the period 1967-92. The sharp decline in regional
wage dispersion until the early 1980s is very clear, since when it has been approxi-
mately constant. This is the mirror image of the behaviour of the Kaitz index, which
rose until the early 1980s and then was approximately constant. The relationship
appears to be strong even when one looks at correlations of the change in the
variance with the change in the log Kaitz index (the estimate of the slope has a t-
statisticof4.9).
The same information is presented differently in Table 6. In the first column we
present the results of a regression of the change in wages from the late 1960s to the
mid-1980s (the period of the rise in the Kaitz index) on the initial level of wages. As
can be seen, wage growth was much faster in regions diat initially had low wages,

o Kaitz index A Regional wage dispersion

i i i i i i i
0.55 - 0.024296

0.5 -

0.45-

0.4- 0.005095
1 i 1 1 i i i
1965 1970 1975 1980 1985 1990 1995
Year

Figure 5. Regional w a g e dispersion and the Kaitz index in France, 1 9 6 7 - 9 2

Notes: The right-hand axis is the variance in the log of hourly blue-collar wages across 95 departements.
The left-hand axis is die Kaitz index as used in Figure 1.
MINIMUM WAGES 343

Table 6. Wage and e m p l o y m e n t growth across French regions, 1 9 6 7 - 9 2

Late 1960s- Mid-1980s- Late 1960s- Mid-1980s-

mid-1990s 1990s mid-1980s early 1990s

Wage growth Wage growth Employment Employment

Dependent growth growth
variable (1) (2) (3) (4)

Coefficient on -0.50 -0.10 -0.38 0.10

initial wage
(f-statistic) (23.2) (3.97) (3.20) (1.14)
No. of observations 95 95 95 95
R2 0.85 0.14 0.10 0.00

Notes: We also estimated regressions with extra controls: the average share of employment in agriculture,
manufacturing, services and construction, for each departement over the 1980s (the only period for
which we have such). The results on wages were identical. The results for employment differ slighdy,
with the coefficient on the initial wage now insignificantly different from zero in the first period (but still
negative) and significandy positive in the second period. High-wage regions still did relatively worse in
the period when the minimum wage increased.
Source: Data from the annual 'Les Salaires dans l'industrie, le commerce et les services', based on die
Decimation Annuelk de Salaires.

confirming that this was a period of strongly declining regional wage inequality. We
also ran the same regression using the change in wages from the mid-1980s to the
early 1990s, a period in which the change in the Kaitz index was much more
modest. There is still a relationship between wage growth and initial wages, but it is
much weaker. This is consistent with the rise in the SMIC in the first period leading
to a reduction in regional wage inequality.
Now consider the effects on employment growth. If minimum wages reduce
employment, we expect a positive relationship between employment growth and the
initial level of wages when minimum wages were raised strongly relative; if minimum
wages raise employment, we expect a negative relationship. Column (3) presents the
results of a regression of the change in employment from die late 1960s to the mid-
1980s on the initial level of wages. Employment growth is significantly higher in
departements that initially had low wages. This is not what we would expect if
minimum wages were destroying jobs in this period: this evidence suggests, if
anydiing, that the rise in the SMIC over this period actually led to higher employ-
ment. A similar regression run over the period from the mid-1980s to the early 1990s
(column (4)) shows no such relationship in this period, suggesting that this is not
because of some long-term trend. So low-wage regions did relatively well in the
period 1967-85, a period when minimum wages were raised very dramatically.
In conclusion, French evidence suggests that the substantial rise in the SMIC to
the mid-1980s had no adverse effect on employment. One can make a case that, by
the mid-1980s, the SMIC had reached levels that were causing employment losses,
but other explanations can be offered for the deterioration in the labour market
position of the low paid in France.
344 JUAN DOLADO ETAL

4.2. Netherlands
National minimum wage regulations have existed in die Nedierlands since the
Second World War, but the current system dates from an Act of 1968. Initially only
workers aged 24 or over were covered, but diis was changed to 23 or older in 1970,
and youdi minimum wages (albeit at a lower rate) were introduced in 1974. The
minimum wage for youdis is expressed as a fraction of the full rate, with the fraction
ranging from 30% for 15-year-olds to 85% for 22-year-olds, and diere being a
separate minimum wage for each year of age. Initially the minimum wage was
indexed to average private sector wage growth, but this indexation was stopped in
1982. In 1984 die minimum wage was actually reduced by 3 % in nominal terms and
it has remained fixed in nominal terms since that date. The effect of this change can
be seen in the Kaitz index in Figure 1. The Kaitz index rose from 1964 to 1981,
since when it has fallen back to slighdy below where it was in 1964, although it
remains at a fairly high level by international standards in die 1990s. Looking at
wage distributions suggests that the proportion of workers paid the minimum has
fallen over the 1980s, with substantial numbers left in relatively few sectors like retail
and cleaning.
There are two other complications worth noting. First, something like 75% of
workers are covered by sectoral wage bargains between employers and unions (what
are known as CAOs), which also set minimum wages. These are typically higher
man the legal minimum, with the gap being larger for younger workers. This means
mat die gap between the youdi and adult minimum may not be as large as the legal
minimum might suggest. But most of the CAOs followed die legal minimum in
reducing the youth minimum in 1981 and 1984 and nominal minimum wages in
1984 (although die adult rates were reduced less often than die youdi rates). As a
result, die time-series and cross-sectional variation in minimum wages is not really
affected by the role of the CAOs.
Second, it should also be noted diat unemployment benefits are direcdy linked to
die minimum wage, so mat all die changes in die minimum wage implied changes
in benefit levels as well. This obviously makes it difficult (if not impossible) to
disentangle die effects of benefit changes from minimum wages, even though tiiese
effects might be different. The Nemerlands is a country in which there has been a
relatively large amount of research into die employment effects of minimum wages.
Perhaps more tiian in any other country diis research has made use of 'structural'
models of the labour market. So van Soest (1989, 1994) estimates a variant of the
Meyer and Wise (1983a, 1983b) model and Koning et al. (1994) estimate a fully
specified search model of the labour market. Bodi studies claim to find very large
negative employment effects of minimum wages, but do so using data in which there
is effectively no variation in minimum wages. In bodi cases die methods used to
identify die effect of minimum wages on employment are rather dubious. In the case
of van Soest (1994) 'identification of the minimum wage effect relies on model
MINIMUM WAGES 345

assumptions such as linearity and normality of error terms' (p. 106) in the author's
own words. As there is no strong a priori reason to believe in these assumptions, and
there are no specification or robustness tests, one should be a little uncomfortable
with the results. And Koning et al. (1994) choose to estimate a variant of the search
model in which the minimum wage is constrained to reduce employment (so the
only issue is the size of the effect) and then seemingly estimate the size of the effect by
looking at the rate at which exit rates from unemployment decline with unemploy-
ment duration, ignoring die large amount of research into other possible reasons for
this phenomenon. 8
Here we take a simpler approach to analysis of the employment effects of the
minimum wage, but one that has the considerable virtue of using variation in
minimum wages to look at employment effects. The biggest change in the operation
of the system was a change in the youth minimum rates relative to adult rates in
1981 and 1983. The fraction of the adult minimum received by 20-year-olds went
from 77.5% to 61.5%, and by 16-year-olds from 47.5% to 34.5% (Mot and
Teulings, 1990). This had the expected effect on wages. While wages for diose aged
23 or more increased by 9% from 1980 to 1984, the wage for diose aged less than
23 actually fell, with die largest falls for the youngest workers (the wage of 17-year-
olds fell by 14%). This is consistent widi evidence mat many young workers were
receiving the minimum wage in 1979 and strongly suggests mat minimum wages or
benefits or both were important constraints on the wage distribution in this period.
Now consider the employment effects. In the whole economy, the share ofyoudi
employment fell from 19.4% in 1979 to 16.4% in 1985, perhaps partly because of
shifts in demand against younger workers (during a severe recession) or because of a
fall in their share in the working-age population; outcomes alone cannot be used to
infer the effect of the minimum wage in isolation.
To try to avoid this problem, one might mink of using the idea of Card (1992a)
that this change should have a bigger impact in some areas of the economy man
others, as we did in the French study above. Unfortunately, regional wage differentials
are very small in me Netherlands, so that we cannot use this exact mediodology.
Radier we consider die effect on different occupations. Table 7 presents information
on the nine occupations which are the most intensive users of youth labour (die share
of employment of 17-22-year-olds is only 6% in die excluded occupations). These
occupations differ in die level of wages. Unfortunately we have no data on wages, but
we do have information on die educational attainment of die workers in diese
occupations, a good proxy of low-wage sectors. The occupations witii a relatively high
proportion of less educated workers are kitchen and household staff, agricultural
labour, unskilled industrial labour and leatiier and textile workers.

In fact, if one altered their structural model so that minimum wages could only raise employment, then one could
use their estimate of die employment loss as the employment gain associated with die minimum wage.
346 JUANDOLADO£7MZ.

Table 7. Youth employment in selected occupations in t h e N e t h e r l a n d s ,

1979-5

Occupation Basic Lower Intermediate Youth employment share

education education education
(%) (%) (%) 1979 1981 1983 1985

Auxiliary medical 11 24 65 50 47 43 37
personnel

Secretaries, typists 5 36 59 36 35 26 24

Office staff 5 33 62 25 24 20 18
assistants

Shop assistants 16 50 34 41 40 38 37

Kitchen and 38 44 18 26 25 25 24
household
personnel

Agricultural 30 42 28 21 22 22 25
labour

Unskilled industrial 39 42 18 14 14 14 14
labour

Skilled industrial 14 43 43 17 17 14 14
labour

Leather and 37 44 19 24 24 23 26
textile workers

Whole 18 35 48 19 20 17 16
economy

Notes: Youth employment refers to workers aged between 17 and 22 years. All employment figures are in
full-time equivalents. Individuals with higher and university education are excluded, since they are
usually unaffected by the minimum wage.
Source: Labour Market Survey, 1979-85.

Did the 1981 and 1983 changes lead to changes in youth employment in these
sectors relative to the rest of the economy? Since any observed changes might be
caused by sectoral shocks, we consider changes in the share of youth employment
(those aged 17-22) in total employment (measured in full-time equivalents) in low-
wage occupations relative to the whole economy. The employment shares of young
workers in these low-wage occupations over the period 1979-85 are presented in
the final columns of Table 7. While the youth share of total employment in the
whole economy fell over this period, that for kitchen and household staff went from
26% to 24%, for agricultural labour from 2 1 % to 25%, for unskilled industrial
labour from 14% to 14% and for leather and textile workers from 24% to 26%.
This is some indication that the cut in youth minimum wages did increase the
relative employment of young workers relative to the economy as a whole in
occupations most likely to be affected. However, these differences are, at best, on die
margins of statistical significance. Also, if work within an occupation is relatively
MINIMUM WAGES 347

homogeneous, we would expect to see large substitution of adult for youdi labour in
the low-wage occupations: mere is little evidence for this. So evidence from the
Netherlands about the adverse employment effects of a minimum wage is scarcely
compelling.

4.3. Spain
The current system of minimum wages in Spain, the salario minimo interprofesional
(SMI), was introduced in 1963, replacing an earlier system in which minimum
wages varied by region and age. It is a statutory minimum wage set by the
government in consultation with trade unions and employer organizations, and its
stated purpose (according to die Workers' Charter) is to protect wage earners and
ensure 'a guarantee of their purchasing power and participation in die economic
development of the nation'.
It currendy sets one rate for workers aged 18 or over and one rate for those aged
16-17 (akhough prior to 1990 it had set different rates for 16-year-olds) with special
rates for homeworkers and casual/temporary workers. The change in youdi rates in
1990 produced a very large rise in the minimum wages for teenagers, with an 83%
increase for 16-year-olds and a 15% increase for 17-year-olds. We will study diis
episode in more detail below. At the start of 1994, apprenticeship contracts were
introduced for workers under the age of 25. These contracts must be between six
months and diree years in lengdi and allow the employer to pay 70% of the SMI in
the first year, 80% in the second and 90% in the diird. Unemployment and health
benefits are not included and they can be used by firms widi fewer man 25
employees. But it is too soon for us to say what the effect of these schemes will be.
As in the Nedierlands, the level of a number of welfare benefits is linked to the
minimum wage. For example, unemployment assistance (paid after entidement to
unemployment insurance has been exhausted) is generally 75% of die SMI (and
100/125% for unemployed heads of households aged 45 or more with 2/2 +
dependent relatives). Again this makes it very difficult to disentangle die effects of die
level of benefits from the effects of die level of minimum wages perse.
As in die odier countries, a major difficulty widi assessing die impact of die
minimum wage on die Spanish economy as a whole is diat diere is a national
minimum diat has varied very litde in relation to average earnings (see Figure 1). It is
obviously radier difficult to argue mat die minimum wage is responsible for die high
level of Spanish unemployment, since die Kaitz index is currendy at a lower level
man it has ever been in relation to average earnings. Any attempt to evaluate the
overall effect is hampered by die lack of variation in die variable we wish to study, and
any conclusions must be treated widi caution. What we might hope to be able to do is
to exploit not just die variation in the minimum wage over time, but also the fact diat
we would expect a given minimum wage to have more effect on low-wage industries.
So we have investigated die relationship between minimum wages and employment
348 JUAN DOLADO ETAL

using a panel of six sectors over the period 1967-94. In column (1) of Tables 8a and
8b we present a simple regression relating the change in employment to the change in
the Kaitz index for the period 1967-94, controlling also for the rate of growth of
sectoral output, the rate of growth of aggregate output and the rate of capacity
utilization (to try to control for cyclical factors). The equations include sector-specific
fixed effects and are estimated by instrumental variables (see die notes to Tables 8a
and 8b). The Sargan instrument validity test does not reject the overidentifying
restrictions in any instance. We do tiiis exercise for both total employment and youth
employment, the latter defined in terms of workers aged 16-19, since a finer
definition of employment by age is not available for the whole sample.
As can be seen, diere is a significandy positive relationship between die change in
employment and the change in die Kaitz index for all workers, but a negative one
for young workers (although only significant at die 10% level). This suggests diat
minimum wages tend to reduce the employment of young workers, but to raise die
employment of odier workers by more. There are two concerns one might have
about these results. First, tiiey might be driven by die omission of relevant aggregate
time-varying factors, a fear heightened by die fact that it is conventional to include
dummy variables for each year in panel regressions of diis type. But in tiiis case, one
can reject die hypothesis that diese additional year effects are significant when tiiey

Table 8a. The relationship between minimum wages and total employment in
Spain, 1967-94 (dependent variable: change in log employment)
Data Panel Aggregate Energy Chemicals Engineering Other Construction Services
and manufacturing
water
(1) (2) (3) (4) (5) (6) (7) (8)

Change in 0.082 0.077 0.044 0.063 0.056 0.134 0.136

logKaitz (0.019) (0.045) (0.024) (0.021) (0.031) (0.033) (0.042)
index
Change in 0.296 0.329 0.332 0.273 0.416 0.832 0.312 0.193
sectoral (0.062) (0.148) (0.154) (0.136) (0.168) (0.251) (0.134) (0.092)
output
Lagged 0.338 0.329 0.273 0.169 0.263 0.331 0.352 0.293
change in (0.156) (0.148) (0.151) (0.082) (0.104) (0.096) (0.111) (0.164)
employment
Sargan 14.8 7.8 4.6 5.2 6.7 3.9 7.7 8.2
test (11) (6) (6) (6) (6) (6) (6) (6)
Serial 0.8 1.3 -1.2 1.1 0.8 0.6 -0.3 -0.5
correlation

Notes: Equations also contain the following controls: fixed effects, change in log hours, change in log
aggregate output and the aggregate capacity utilization rate. The equation in column (1) also contains
year effects when testing for their joint significance. The estimation method is IV. The instruments used
are the controls listed above, sectoral fixed effects, and lags of the real minimum wage, the real average
wage, the change in output, the change in employment, the change in hours and the change in the
sectoral price deflator. When testing for year effects, they are also included. In column (1) two lags are
used. Hcteroscedastic-consistent standard errors are reported in parentheses. Sargan test for the validity
of the overidentifying restrictions. Degrees of freedom are in parentheses. The serial correlation is a
N(0.1) statistic for the presence of first-order serial correlation in the residuals.
Source: ContalilidadNational Sectorial (various issues), and Garcia etal. (1994).

A
MINIMUM WAGES 349

Table 8b. The relationship between minimum wages and youth employment
(16-19) in Spain, 1967-94 (dependent variable: change in log employment)
Data Panel Aggregate Energy Chemicals Engineering Other Construction Services
and manufacturing
water
(1) (2) (3) (4) (5) (6) (7) (8)

Change in -0.154 -0.096 0.036 -0.042 -0.096 -0.164 -0.216 -0.194

log Kaitz (0.093) (0.073) (0.031) (0.028) (0.083) (0.081) (0.112) (0.106)
index
Change in 0.448 0.612 0.492 0.536 0.443 0.962 0.446 0.282
sectoral (0.122) (0.246) (0.163) (0.212) (0.203) (0.272) (0.176) (0.100)
output
Lagged 0.232 0.279 0.362 0.202 0.193 0.431 0.331 -0.036
change in (0.086) (0.128) (0.154) (0.122) (0.077) (0.182) (0.126) (0.056)
employment
No. of 168 28 28 28 28 28 28 28
observations
Sargan 11.4 9.8 8.3 6.4 5.9 6.3 8.7 9.3
test (11) (6) (6) (6) (6) (6) (6) (6)
Serial -1.2 1.6 -1.3 -0.8 1.8 1.6 0.3 -0.4
correlation

Notes: As for Table 8a

are added both to the regressors and to the instruments, so that the specification
presented is a parsimonious representation of the data. A second fear is that the
results are driven by the cross-section variation in the Kaitz index, which is itself
driven by cross-section variation in industry wages rather than the minimum wage.
It would then be very dangerous to interpret die results as saying anything about the
effect of die minimum wage. But the final six columns of Tables 8a and 8b show that
essentially the same correlations between changes in employment and changes in die
Kaitz index apply widiin sectors as well as between mem. The fact mat estimated
effects of the Kaitz index are largest in the sectors where the minimum wage is most
important (sectors (4), (5) and (6)) gives some reason to believe mat we are picking
up effects of me minimum wage in diis regression. So, while one must exercise some
caution in interpreting the results, since there is relatively little variation in
minimum wages in Spain, which increases the danger mat the results are caused by
some other factor, mere does seem to be evidence that the minimum wage increases
total employment while reducing mat of youtfis.
However, mere is one episode in recent Spanish history in which mere has been a
large change in minimum wages, and this episode deserves detailed study. This is
the change in 1990 when the minimum wage for those aged 16 and under was
raised to the level for workers aged 17. Figure 6 shows mat diis change caused a very
dramatic rise in the youdi Kaitz index. How was die employment rate of young
workers affected relative to mat of older workers? Since we expect the minimum
wage to have the largest impact on employment in die low-wage regions, we plot the
change in employment rates over die period 1990-4 (after die change) against the
350 JUAN DOLADO ETAL

o kad A k16
D k17

L_, ! ( ! ! ! j ,_

1960 1965 1970 1975 1980 1985 1990 1995

Year

Figure 6. Kaitz i n d i c e s for different age groups i n Spain

Notes: kad is the adult Kaitz index; kl 7 is the Kaitz index for 17-year-olds; and kl6 is die Kaitz index for
those aged less than 17.
Sources: As in Figure 2.

fraction of workers in the region who were low paid in 1989 (before the change).
Figure 7a shows what has happened to die employment rates of die youngest
workers (aged 16-19) and Figure 7b shows what has happened to slighdy older
workers (diose aged 20-4).
Figure 7a suggests diat mis rise in the minimum wage reduced die employment of
youdis, while Figure 7b suggests diat the employment of slighdy older workers
increased. These results remain the same when one estimates regressions controlling
for other factors, such as the overall employment rate for adult workers (to control
for cyclical factors). The most plausible explanation for the employment effect on
older workers is substitution of employers away from younger workers who had
become relatively more expensive. If the substitution effect is die dominant one in
this episode, one might also wonder if it can explain die time-series results reported
earlier. There are two reasons why this is unlikely to be the case. First, it is very
unlikely diat total employment would rise if a rise in die minimum wage reduced
teenage employment widi a substitution effect for older workers: yet that is what die
time-series results of Table 8 suggest. Second, we regressed the changes in the
employment rate of die 20-64 age group on die adult Kaitz index, the changes in
die employment rate of workers aged 16-19 and controls for region and a trend. If
substitution were all diat mattered, we would expect to see no effect from die Kaitz
MINIMUM WAGES 351

-5 -

-T" 1
15 20 25
% low paid in 1989

Figure 7a. Changes in teenage e m p l o y m e n t rates i n Spanish regions, 1 9 9 0 - 4

Notes: The solid line is the regression line from a regression of the change in the teenage employment
rate (those aged 16-19) from 1990 to 1994 on the proportion low paid in 1989.

-5 H

I
-10

-15
I
15 20 25
% low paid in 1989

Figure 7b. Changes in e m p l o y m e n t rates o f slightly older workers in Spanish regions,

1990-4
Notes: The solid line is the regression line from a regression of the change in the employment rate of
those aged 2 0 - 4 from 1990 to 1994 on the proportion low paid in 1989.
352 JUAN DOLADO ETAL

index and a strong negative effect from the teenage employment rate. The
coefficient on the teenage employment rate is -0.37 (standard error 0.12),
suggesting that there is some substitution, but the coefficient on the Kaitz index is
0.09 (standard error 0.047), suggesting that the monopsony effect is still present. In
conclusion, the evidence suggests that minimum wages reduced die employment of
teenagers, but raised that of odier workers: a conclusion in line widi the time-series
regressions reported earlier.
It is natural to want to try to reconcile diese results widi those of Card (1992a),
who found no effect, or even a weakly positive effect, of die rise in die federal
minimum wage in die USA in die early 1990s on teenage employment; and widi
diose from France reported earlier. A first hypodiesis might be that the level of die
minimum wage is different in the two countries, the youth minimum being at a
much higher level in Spain. If this is die case, it does not show up in die Kaitz index
because die rise in die Spanish youdi minimum led to a rise in the Kaitz index from
slighdy over 0.3 to slighdy over 0.5, whereas die change in die USA was from 0.75
to 0.85. One could argue that it does show up in die spike, since approximately 25%
of US teenagers receive die minimum wage, whereas the figure is closer to 40% in
Spain. It is also possible diat, whereas die minimum wage is virtually die only wage
floor for teenage workers in die USA, die minimum wage has an important
influence on higher wage floors in Spain, so that die Kaitz index understates die
importance of die minimum. Some evidence supportive of diis view comes from die
survey 'Collective Bargaining in Large Firms' for 1985 and 1994. The ratio of die
wage of labourers (die lowest skill category) to die wage of workers aged 16 and 17
fell from 1.83 in 1985 to 1.31 in 1994 (and was even lower in some sectors). As these
negotiated wages are generally above the minimum, mis suggests diat die rise in die
relative wages of youdis was not confined to workers paid the minimum wage. The
rise in the youdi minimum wage may have had a considerable impact on die wages
of young workers paid above die minimum.
To conclude, we do have evidence from Spain diat rising minimum wages in die
early 1990s have reduced youdi employment. But die evidence also suggests a rise in
total employment. One should not conclude that die minimum wage must
necessarily be bad for die labour market.

4.4. United Kingdom

The Wages Councils were established by Winston Churchill in 1909 to protect die
pay of workers in the 'sweated' trades. They set minimum wage rates in a number of
different industries. Over die years, the number of industries covered first increased
(to a peak of about 60 covered sectors in the early 1960s), dien decreased, and by
die early 1990s die 26 remaining Wages Councils set minimum wages for
approximately 2.5 million workers in low-paid sectors (mosdy in hotels and catering,
retail, clodiing manufacture and hairdressing, but also in a number of very small
MINIMUM WAGES 353

industries). If we define the low paid as those earning less than £3.50 per hour, then
35-40% of these workers were in Wages Council sectors. Each Wages Council
consisted of an equal number of employer and worker representatives, plus a
maximum of three independent members (nominated by the government of the day)
who had the casting vote if an agreement was not reached. Until the 1986 Wages
Act, the Councils generally set a myriad of minimum wages differentiated by age,
occupation and region, but after 1986 they set only a single rate for those aged 21 or
over and the minimum wage for younger workers was abolished. The 1993 Trade
Union Reform and Employment Rights Act abolished the remaining 26 Councils.
Since 1993 there have been no minimum wages in operation except in agriculture,
where, for the moment, the Agricultural Wages Boards still exist.
As the rates set by the Wages Councils varied both across industries and over
time, it is natural to use a panel of industries to investigate the effect of minimum
wages on employment. We have done this elsewhere (Machin and Manning, 1994;
Dickens et ai, 1994a). The results can be most succincdy summarized widi a
diagram. Figure 8 plots die change in employment against the change in the Kaitz
index for die period 1975-92. Each dot represents an observation for a particular
Wages Council in a particular year. There is a positive relationship between tfiese
two variables, suggesting that industries and years in which minimum wages rose
fast relative to average wages were associated widi faster man average growth in
employment. A simple picture cannot control for a host of factors that might explain
uiese results, but our regression analysis suggests that the finding is surprisingly
robust.9 We found similar results when we looked at agriculture alone (Dickens
rial., 1995).
The recent experience of the U K also offers an opportunity to study the effect of a
dramatic change in die minimum wage system: namely, die abolition of die Wages
Councils (except in agriculture) in August 1993. One might expect sudden, and
possibly large, falls in wages and to be able to investigate the effect on employment.
Such dramatic falls in wages never happened. Figure 9a shows the distribution of
wages in retail (die largest Wages Council) in April 1993, immediately prior to
abolition. The minimum wage is at the spike in die wage distribution: its influence
can be clearly seen. Figure 9b shows what die distribution of wages looked like in
retail in April 1994, six months after abolition. The most striking feature is mat the
spike at die old minimum, altiiough reduced, is still noticeable.
There are a number of reasons why die minimum continues to exert an influence
on die wage distribution even after abolition. First, any worker employed prior to 30
August 1993 has a right to die same terms and conditions of employment after diat
date if he or she remains widi die same employer. Any attempt to cut wages by die

9
We attempted to control for potential endogeneity of die minimum wage, differences in employment growth rates by
industry, and differences in short- and long-run effects, and we used alternative measures of employment. Our results
were always the same.
354 JUAN DOLADO ETAL.

0.321584 - o
o

o o
o o
0
o o o o»_ o 0
- 0 0 °o O
*o o o
**> ^00»°<fe° ^ ? o ° o
o o * £ • * > *• °»° ° _ - - -
0
0
o % » o « 0 0 9 < _ o o o o
oojfg ""ve r* ° oo o
o
» o °°° °o0
°o
o °*
o O
o
o
°
0
0

0
0.385993 - o
1 1 1 1 1
-0.127583 0.338803
Itough-ltough(-l)
Change in log Kaitz index

Figure 8. Changes i n e m p l o y m e n t and the Kaitz index i n UK Wages Council industries,

1975-92
Notes: Each observation represents a change in log employment and a change in the log Kaitz index in a
particular year in a particular Wages Council where we distinguish between men and women. The solid
line is the fitted regression line.

employer could be argued by the worker to be constructive dismissal, for which

redress could be sought under the 1986 Wages Act.10 Effectively, the minimum wage
is still in force for existing workers and it is only the wages of new employees that the
employer can lower. Even this freedom is curtailed by limits on die extent to which
employers are prepared to pay different wages to workers doing the same job, and
by considerable resistance to nominal wage cuts. Reasons for these practices are not
fully understood, but they are a powerful influence on wage formation.
The minimum wage rates set by the Wages Councils prior to abolition will
probably continue to exert an influence on the bottom of the wage distribution for
some time to come, but dieir importance is reduced as inflation and real-wage
growth cause mem to fall in relation to average earnings. Some employers also cut
wages for new workers. Table 9 shows that wage growth in die old Wages Council
industries was slower in die period 1993-4, aldiough me difference is modest. And a
survey of vacancies by die Low Pay Network found mat a significant proportion of
vacancies were at pay rates below die old minimum rates (akhough again a
significant proportion are still at the old minimum rates).

Many low-paid workers are unaware of their rights or are in no position to enforce them; some existing workers may
also experience wage cuts.
MINIMUM WAGES 355

(a) April 1993 (after abolition)

•S 4 •

2.5 5.0 7.5 10.0 12.5 15.0 17.5

Hourly wage (£)

(b) April 1994 (after abolition)

7r

7.5 10.0 12.5 15.0 17.5

Hourly wage (£)

Figure 9. The distribution of wages in the UK retail sector

Source: New Earnings Survey.

With only modest immediate effects on wages, we are unlikely to discover

significant effects on employment. Table 10 presents information on the share of
employment in the Wages Council industries; there is no evidence of dramatic
changes in employment. But just as we might expect to see the first effect of abolition
on wages in the wages being offered to new workers, so we might expect to see the
first effect in employment on hiring and exit rates from the Wages Council sectors.
Table 11 presents information for the quarters immediately before and after
abolition. There is no noticeable change in the behaviour of the Wages Council
sector relative to the rest of the economy.
There is good reason to believe that the abolition of the Wages Councils will
reduce wages in the long run, but effects will take a long time to work through,
making it hard to identify any effects on employment. But this evidence from the
period of operation of the Wages Councils provides no evidence mat their activities
were associated with job losses.
356 JUANDOLADO£7ML.

Table 9. Average w a g e s in Wages Council industries, April 1991-April 1994

Industry Average hourly earnings {£), in April Annual % change during:

1991 1992 1993 1994 91/92 92/93 93/94

Clothing 4.11 4.39 4.59 4.83 6.6 4.7 5.2

Catering 4.04 4.20 4.30 4.42 3.9 2.3 2.7
Hairdressing 3.33 3.59 3.86 3.91 7.6 7.6 1.4
Retail 4.75 5.04 5.30 5.35 6.2 5.0 1.0
All Wages 4.59 4.87 5.07 5.16 6.9 4.1 1.7
Councils (WC)
All non-WC 6.85 7.35 7.66 7.89 7.3 4.3 2.9
industries
(SIC 0-9)
All non-WC 6.93 7.45 7.76 8.0 7.4 4.2 2.9
service inds
(SIC 6-9)

Notes: 'All Wages Councils' includes Clothing Manufacture, Retail Food, Retail Non-Food, Unlicensed
Place of Refreshment, Licensed Residential and Restaurant, Licensed Non-Residential, Hairdressing,
Boot and Shoe Repair, Laundry, Hat Cap and Millinery, Linen and Cotton Handkerchief, Made Up
Textiles, Rope, Twine and Net, Fur Goods, Toy Manufacture and Soft Drinks Manufacture.
Source: New Earnings Survey Micro Data.

Table 10. Employees in e m p l o y m e n t in Wages Council industries (thousands),

Great Britain

June-Aug. Dec. 1992- June-Aug. Dec. 1993- Sept.-Nov.

1992 Feb. 1993 1993 Feb. 1994 1994

All WC industries 2420 2374 2436 2421 2461

All other inds 18466 18183 18370 18185 18481
WC employment 11.6 11.6 11.7 11.8 11.8
as % of total

Notes: Abolition of the Wages Councils took place at the end of August 1993.
Source: Quarterly Labour Force Survey Micro Data.

Table 11. Hiring and exit rates in Wages Council and other industries

Wages Council Non-Wages Council Differences

Hiring rate Exit rate Hiring rate Exit rate Hiring rate Exit rate

June -Aug. 1992 7.30 6.05 4.51 4.47 2.79 1.58

Dec. -Feb. 1993 5.54 6.91 3.15 3.17 2.39 3.76
June--Aug. 1993 7.29 5.96 4.40 3.55 2.89 2.41
Mar. - M a y 1994 7.65 5.81 3.77 3.24 3.88 2.57
Sep.- -Nov. 1994 7.85 7.07 4.42 3.89 3.43 3.18

Notes: The hiring rate is the fraction of workers in the sector (WC or non-WC) who were not in it in the
previous quarter. The exit rate is the fraction of workers in a sector who had left it by the following
quarter. The Wages Councils were abolished in August 1993.
Source: As in Table 10.
MINIMUM WAGES 357

5. CONCLUSIONS
Much policy discussion about European labour markets suggests that abolition of
the minimum wage is a key issue. We hope to have painted a more balanced view.
The importance of minimum wages has probably been exaggerated.
The theoretical effect of minimum wages on employment depends on details of
the labour market in which it is imposed. The effect can in theory go either way.
Detailed evidence is needed before judgements can be drawn. The common
perception that the theoretical case is clear cut is simply incorrect. From our
empirical analysis we draw four main conclusions.
In most European countries, there has been little change in minimum wages relative
to average earnings over the past 30 years. It is hard to argue that the minimum wage
has played a large part in the rise in European unemployment over mis period.
While minimum wages do seem higher in relation to average earnings in Europe
than in the USA for adults, the opposite is often true for young workers: there is
extensive variation in the minimum wage by age in Europe and very limited
variation in the USA.
Although much policy discussion surrounding the minimum wage focuses on the
effects on the youth labour market, these workers make up a rather small and
declining proportion of minimum wage workers. Increasingly, minimum wage
workers are women, temporary and part-time workers.
The evidence on the employment effects of minimum wages is very mixed. We
have found evidence that higher minimum wages reduced employment in some
cases (particularly for young workers) and raised it in others (particularly for total
employment). But it is surprisingly hard to find strong evidence of any adverse
employment effects of minimum wages in situations where many commentators are
firmly convinced (rightly or wrongly) that the job losses exist: France in the 1980s is
a good example of this. As the employment effects of minimum wages on the youdi
labour market do seem to be worse, diere is a strong case for having a lower
minimum wage for young workers.
We should emphasize that none of our results suggests that the effects (good or
bad) on the economy of current levels of minimum wages are particularly large. The
presence or absence of a minimum wage will not be the difference between
economic success and failure.

Discussion

Gilles Saint-Paul
CERAS, DELTA and CEPR
As we all know, economics is strongly dependent upon fads and fashions. A
currendy fashionable view, which started across the Adantic, and is exemplified by
358 JUAN DOLADO ETAL

the work of David Card and Allan Krueger, is that the minimum wage increases
employment. The idea is that, if monopsony power is prevalent, firms are artificially
reducing the wage below the marginal product of labour, so that a minimum wage,
rather than reducing labour demand, would simply cut into their monopsony mark-
up; at the same time labour supply increases, so that the net employment effect is
positive. This idea is backed by some evidence: for example, on the fast-food
industry in New Jersey and Pennsylvania.
Given that minimum wages are typically much more binding in Europe than in
the USA (see the authors' Table 1), and are a key suspect for high unemployment in
Europe, it is essential for policy purposes to examine whether die conclusions of
Card and Krueger are valid in Europe. At the same time, the variety of minimum
wage legislations in Europe offers a useful laboratory in order to assess the impact of
various levels of minimum wages on employment. The paper by Dolado et al. is
therefore a very welcome contribution to the literature.
How seriously should we take this argument that the minimum wage is good, at
least over some range, for employment? Several remarks are in order. First, it is an
argument about employment, not unemployment. It tells us mat the minimum wage
will increase labour supply and that, when it is introduced, firms will be willing, at
the margin, to absorb this excess labour supply because of a wedge between the
marginal product of labour and the wage. To know whether unemployment will rise
or fall, we need a model of unemployment. Monopsony per se does not generate
involuntary unemployment; everybody is able to find a job at the current wage. In
the pure monopsony model, employment rises but the economy is and stays at full
employment until the minimum wage equals the Walrasian equilibrium one. To get
unemployment we must introduce a distinction between die wage formation
schedule and the Walrasian labour supply curve. My intuition tells me that the net
effect on unemployment depends on the relative steepness of these two schedules. All
mis is important, since part of the evidence which is presented has to do with
unemployment rather than employment.
Second, analysis is mosdy a partial equilibrium, at firm level. We know that a firm
facing an upward-sloping supply curve will increase employment; we do not know
what happens in the economy as a whole. In particular, the aggregate labour supply
curve may be much steeper than the one faced by an individual firm, since the latter
reflects the possibility for workers to move across regions and/or sectors. So in the
aggregate, increases in the minimum wage would show up in higher wages with little
or no effect on employment.
Third, the scope for increasing employment by means of an increase in the
minimum wage depends on the extent of the individual firms' monopsony power.
The fact that we notice a spike at the minimum wage in the distribution of income is
itself evidence in favour of monopsony; if productivity were smoomly distributed
across workers and if these were paid their marginal product, minimum wages
would truncate the distribution of income but not create a mass of people at the
MINIMUM WAGES 359

minimum wage. On the other hand, there are many sectors in which I find it hard
to believe in monopsony. This particularly applies to the original Card and Krueger
finding about the fast-food industry in New Jersey. As far as I know, this is a very
densely populated state; furthermore, the US fast-food industry is extremely
competitive; there are many competing firms having the same location (shopping
malls, food halls), offering highly substitutable products (pizza, hamburgers,
Chinese food, etc.) and all employing the same sort of low-skilled labour. This
sounds to me like one of the worst candidates for monopsony (and monopoly) one
might think of. So it is important to think of other explanations. One possibility is
die Rebitzer and Taylor (1995) paper, which is based on efficiency wage considera-
tions and has predictions quite similar to monopsony models. Another argument can
be made more in line with the European context. Assume that mere are two types of
worker, permanent and temporary. Assume that permanent workers have a high
firing cost and are paid above the minimum wage, while temporary ones are paid
the minimum wage and have no firing cost. Under some circumstances, the firm will
hire both permanent and temporary workers. A rise in the minimum wage will
induce it to substitute permanent for temporary workers. Now, because permanent
workers have a high firing cost, in a slump the firm will fire only temporary workers.
Because it has more permanent workers following the rise in the minimum wage,
total employment may well increase due to the contribution of those firms which are
in a slump. Thus, die minimum wage acts as a 'shadow firing cost'; increasing it is
like increasing firing costs, which we know may be favourable for aggregate
employment by preventing firms from firing.
Fourth, there is a great deal of market imperfection elsewhere in the economy, in
particular on the labour supply side; this calls for maximum, rather than minimum
wages. We do not see many proposals going in that direction. For any existing
institution, it is possible to think of a distortion that makes this institution have
beneficial effects. But this is different from saying that this institution is the most
adequate, or even was intended, to cure the inefficiency.
If minimum wages increased employment, politicians would have surely found
out, as raising minimum wages would certainly be a popular and uncontroversial
measure. Yet in France, when the minimum wage is increased, die government
often offsets its impact on unskilled labour costs by reducing social security
contributions. Evidence on US congressional voting suggests that states witii a
higher proportion of youth in their workforce are actually more likely to oppose
increases in the minimum wage (Uri and Mixon, 1980); that would be strange if
higher minimum wages were to raise both the income and employment level of the
young. Now, it may just be the case that the voice of orthodox economists is so loud
that mey are perversely preventing a political free lunch from being eaten ...
Turning now to the empirical evidence, a key, but unavoidable, problem with die
authors' approach is that it is impossible to design a control group for which die
minimum wage evolved differendy, since we are dealing here widi national
360 JUAN DOLADO ETAL

minimum wages applying to everybody. Therefore, die explanatory variable has to

be some proxy like the proportion of workers paid the minimum wage or the Kaitz
index, and all the cross-sectional variation in such proxy will be driven by factors
other than the minimum wage. This raises many simultaneity problems. For
example, in Table 5, die coefficient of the change in employment (or unemploy-
ment) rate on die proportion paid near die SMIC may be biased, since any source
of unemployment odier dian die minimum wage will push wages down, dius
increasing die proportion of workers paid near die SMIC. Similarly, in Table 6,
which leads to opposite conclusions, a positive employment impact of minimum
wages is inferred from die evidence of higher employment growdi in regions widi
initially lower wages. The argument is, again, diat minimum wages are more biting
in such regions. But, again, it is hard to believe diat die cross-sectional variance in
employment growdi is entirely driven by changes in die national minimum wages.
Any model widi partial adjustment of employment and region-specific wage shocks
will predict a negative correlation between initial wages and subsequent employment
growdi. A similar argument may apply to Figure 9: a positive regression coefficient
of employment on die Kaitz index may simply point to a negative correlation
between employment and average wages, i.e. we are simply recovering a labour
demand curve. This will prevail if cross-section variations in wages are driven by
differences in wage pressure (e.g. different unionization rates).

Michael Keen
University of Essex
The economic impact of minimum wage legislation is a fearsomely contentious
topic, both academically and politically. This paper does great service bodi to
outsiders like me and, I have no doubt, to those working in die area, in providing a
level-headed, painstaking and dioughtful European perspective on some of die
principal issues at stake. In tiiese brief comments I seek only to raise some of die
wider policy questions diat the paper prompts but does not dwell on.
The central focus of die paper is on die employment effects of die minimum wage.
This is right and proper, given die policy priority diat ought to be attached to
unemployment in Europe and, of course, recent empirical and tiieoretical
developments suggesting diat these effects are not as straightforwardly bad as had
been widely believed. Nevertheless, employment itself is typically not an appropriate
maximand for policy design. The minimum wage, as die paper makes clear, is best
diought of as just one potential weapon in a wider attack on poverty (and/or
inequality). Understanding die employment effects of die minimum wage is an
important first step in understanding its potential value as a component of poverty
alleviation strategies; but it is only a first step. Moreover, policy must be made before
diat first step can be taken with full confidence. Tony Blair, leader of die British
Labour Party and apparendy prime minister in waiting, will at some point
(probably) have to specify an exact figure for the minimum wage to whicIThe is
MINIMUM WAGES 361

committed, and might reasonably look to the profession for some guidance as to die
level at which it should be set and, more broadly, on the role of the minimum wage
in a well-designed poverty alleviation programme. It is not clear that we can yet
provide as much guidance on tiiese matters as he might have hoped. Indeed, Card
and Krueger (1995, p. 393) tiiemselves call for 'a reorientation of policy discussions
away from the efficiency aspects of the minimum wage and towards distributional
issues'. Several audiors have indeed provided elements of such a reorientation,
notably Freeman (1996), Guesnerie and Roberts (1987), Marceau and Boadway
(1994) and Saint-Paul (1994). But it remains the case diat the minimum wage has
been only imperfecdy integrated into die analysis of poverty alleviation.
To see some of the issues mat may arise in moving towards such an integration,
many of which have indeed been noted by die audiors just mentioned, it may help
to start widi a very simple example. Suppose that die object of policy is to maximize
a social welfare function of me familiar form

where yh is the income of individual h and higher values of die parameter a ~»0
correspond to increases in inequality aversion: a = 0 is utilitarian, a + °° is die
Rawlsian maximin case." Suppose too that there are only two kinds of person: those
who find work at die minimum wage w and diose who instead take unemployment
benefit at die rate b (which we treat as fixed). Take the proportion of die population
in die former group to be e(w), tfiis being diought of as the reduced form of some
unspecified labour market model. In the competitive case, e(w) is decreasing; in
odiers - textbook monopsony, efficiency wages, matching - it may be increasing at
first, but will surely be decreasing eventually. Choosing die minimum wage to
maximize social welfare gives die necessary condition

{l-o)w1-a + E{bi-°-wl-°) =0 (1)

where E& -exwfe is die elasticity of employment with respect to die minimum
wage.
This provides an immediate reminder of an obvious but important point: at die
optimum, E>0. That is, even if monopsony elements of die kind emphasized by
Dolado et al. lead employment to increase witii die minimum wage over some range,
mat minimum should be pushed to die point at which further increases reduce
employment. This in turn has one implication diat may have some empirical
relevance: if me minimum wage is set optimally, disturbances to it must give an
observed negative relationship between employment and die minimum wage. In
richer models, of course, this may cease to be true. Nevertheless, die example

The logarithmic case a — 1 is excluded for brevity.

362 JUANDOLADO£T^i.

suggests that there may be a sense in which the cards are stacked against observing a
positive effect of the minimum wage on employment even if there is such an effect
over some range.
Solving (1) gives
1/1 -a
w
(2)
b E+o-\
As one would expect, the optimal minimum wage increases with die level of
unemployment benefit and decreases widi the elasticity of labour demand.
Intuitively, the impact of inequality aversion is less obvious. In so far as die
minimum wage is usually rationalized in terms of distributional concerns - is
diought of, broadly, as a 'leftist' tool - one might have expected the optimal
minimum wage to increase with a. But in so far as it diminishes employment, of
course, the minimum wage has an adverse impact on the between-group inequality
of working and non-working poor. Indeed, and somewhat surprisingly, diis turns out
to be the dominant effect in the present simple setting: it can be shown from (2) that
(for given E) an increase in inequality aversion a unambiguously reduces die
optimal minimum wage. Once seen, the reason is obvious: as inequality aversion
increases, so the plight of die very poorest — tiiose on benefit - becomes die
overriding concern, and die object of policy becomes simply to maximize the
number of employed. Thus the minimum wage may not, after all, be as 'left' a
measure as commonly diought.
The framework just used is, of course, flagrandy stylized. Perhaps most obviously,
it ignores botii those employed at above the minimum wage and die revenue cost of
financing unemployment benefits. The implications of die latter may also be more
complex than at first seems, for die cosdiness of raising public funds to finance
benefits for poverty relief might in itself seem to be an argument for a minimum
wage. To the extent that die minimum wage is a tax on the distributionally relatively
undeserving, the marginal social cost of financing transfers to workers by this route
may be less tiian that of explicit transfers financed by taxes that may be both more
distorting and less well targeted distributionally: may be less, diat is, than die
marginal cost of public funds (MCPF). But die framework here emphasizes that diis
natural line of thought may be incorrect, for in this setting one would expect a
higher MCPF to lead to a lower minimum wage rather than a higher, because a
higher MCPF means a higher cost to society of financing any given level of
expenditure on unemployment benefit. In a richer model, this effect would need to
be weighed against any saving that a minimum wage might bring in the cost of
financing in-work benefits. The trade-off between die minimum wage and in-work
benefits, witii potentially divergent effects on both work incentives and tax revenues,
may be a critical concern in designing die fully optimal minimum wage.
As the authors righdy emphasize, establishing the case for a minimum wage
requires one to be persuaded tiiat there is no better way of achieving die purpose in
MINIMUM WAGES 363

hand. In this context they highlight the potential importance of labour market
structure, and indeed this will clearly have a crucial role in integrating die minimum
wage into broader poverty alleviation strategies. Another set of considerations likely
to be important may also deserve some emphasis: the nature of, and any constraints
on, the tax instruments available to government. It is natural, after all, to conceive
of the problem of putting together a poverty alleviation strategy in terms of the
somewhat ill-defined principle of targeting: the most effective policy is likely to be
that which uses instruments most directly related to the quantities of concern. Take,
for example, the argument that a minimum wage may help to overcome under-
investment in skill acquisition. Intuition suggests that a better policy response would
be to act on that underinvestment directly, perhaps by an explicit subsidy. The
advantage of such a targeted policy is that it may avoid unintended distortions: there
may, for instance, be some who could never get their skills up to the level needed for
them to warrant the minimum wage, and whom such a minimum would conse-
quently condemn to unemployment. Thus one needs to know why a more precisely
targeted intervention is not possible. Restrictions on tax instruments available to the
government - a high MCPF inhibiting subsidies, for example, or problems
associated with monitoring the uses to which educational loans are put - are likely
to have an important role to play in justifying adoption of a minimum wage. They
may even be indispensable.
Several examples of this general point come to mind. One is provided by
Guesnerie and Roberts (1987). Imagine a world of competitive labour markets, and
suppose first that the only redistributive instrument available is a linear income tax.
There may then be a case for a binding minimum wage, essentially as a means by
which the low paid can exploit their collective monopoly power. If, however, the
government is able to deploy fully non-linear taxation, the case for a minimum wage
vanishes: the government is able to bring about the same exploitation of monopoly
power by acting on labour supply decisions through the tax system, with no need to
ration labour supply.
As a further instance of the same general principle, it may be that the minimum
wage has advantages over other forms of income support in terms of take-up. In the
UK, for example, take-up of the main benefit to the working poor is very low (at
around 60% of those entitled), a fact often attributed to stigma and/or hassle
connected with claiming the benefit. No such disinclination to take-up is likely to be
associated with the minimum wage. Thus it may be that the disadvantages
associated with leakage of benefit of minimum wage to the low paid who do not
belong to the poor households, a concern emphasized by the authors, may be offset
by its greater reach. On the other hand, the same high take-up is just as likely to
apply to other instruments that are administered through firms rather than people,
such as a marginal employment subsidy. It may then be that the complexities of
administering such schemes, and the possibilities for abuse, again leave something to
be said for the minimum wage.
364 JUAN DOLADO ETAL

One particular set of constraints on the instruments likely to be available are those
implied by the continuing process of economic integration within Europe. While the
paper touches on the possible implications of this for die impact of minimum wage
rules, the obvious policy question lies outside its remit: does closer integration imply
a case for co-ordinating minimum wage policies to avoid these minima being driven
to excessively low levels?
The paper emphasizes the effects of integration in product markets. There may,
however, be effects through other markets. The most obvious such effects are those
through labour mobility. Views on the actual and prospective importance of labour
movements in Europe vary rather widely, so I leave this aside. Effects through
capital markets also deserve consideration. Moving from a situation in which capital
is immobile internationally to one in which it is perfectly mobile flattens the labour
demand curve. The intuition here is simply that mobility enables capital to escape
any burden of the minimum wage, so that an increase in the wage rate has a more
dramatically negative impact on employment. One might then expect capital
mobility to lower the minimum wage that is optimal from the perspective of any one
country; and hence, perhaps, to strengdien the case for co-ordination of policy
across countries. But there may be further effects through product markets pointing
in the opposite direction. With monopsony elements, for example, a minimum wage
may become a device for committing to high output; and thus, for the kind of reason
familiar from the strategic trade literature, non-cooperative behaviour might lead to
excessively high minimum wages. The prospect of closer integration may also affect
the political equilibrium from which the minimum wage emerges: as capital
becomes more mobile, so capital owners become more able to escape the burden of
a minimum wage by instead investing abroad; they might then be inclined to oppose
such minima less fiercely. Again, closer integration might for this reason conceivably
lead to an increase in minimum wages rather than a reduction.
The minimum wage, it seems clear, will remain a contentious topic. We shall
need more such well-balanced analyses as this.

General discussion
The regressions and, in particular, the use of the Kaitz index in the regressions
generated considerable discussion. Gene Grossman argued that regressing total
employment on the Kaitz index was not very meaningful, and that the estimated
coefficient was hard to interpret. He reasoned that the index - the ratio of the
minimum to the average wage - could conceivably explain the ratio of skilled to
unskilled workers in the workforce, but not the total level of employment. Its use in
cross-section analysis was even more problematic, since it implicitly assumed that
production functions are the same across industries. Mathias Dewatripont observed
that trade unions often bargain over both the minimum wage and the average wage.
In such a circumstance, which was usual in Belgium, it would be appropriate to
MINIMUM WAGES 365

include both as separate variables in the regression, radier man using merely their
ratio. For David Begg the issue boiled down to whether changes in employment
reflected a shift in the labour demand curve or movements along it. Alan Manning
defended the use of die index. He argued that die alternative, to use die absolute
level of minimum wages instead of die Kaitz index, would not be sensible because
minimum wages vary over time widi changes in productivity. Not correcting for this
variation would bias die results in a serious way. The Kaitz index corrects for
variations in productivity by deflating die minimum wage by die average wage; here
the deflator is to be viewed only as a measure of average productivity.
The issue of how minimum wages are set in Europe drew various comments.
Frederick van der Ploeg suggested mat Greece and Spain, where wages are much
below die European average, be included to give more balance to me enquiry. He
also pointed out mat mere are two minimum wages to consider, since die legal
minimum wage is quite different from die agreed minimum wage mat obtains under
collective bargaining. In the Nedierlands, die gap between die two was as high as
10%. Axel Weber added tiiat, similarly, in Germany mere was a distinction between
contract wages and minimum wages. Further, given tiiat die proportion of workers
widi minimum wage protection depends on die degree of unionization in a country,
it would be useful to include unionization as a variable in die regressions. Gene
Grossman argued mat, to die extent mat minimum wages in die UK were set
sectorally by Wages Councils, there must be some political economy explanation for
die differences in minimum wages across industries. This could make the evolution
of minimum wages endogenous to industry evolution. Tryphon Kollintzas added
mat in Greece the effect of minimum wages on employment levels was hard to
decipher, primarily because there were otiier rigidities in die labour market, most
notably die high firing cost. Christian Dustmann took issue widi die claim mat die
recent abolition of minimum wages in 1993-4 had not affected wages in die UK.
He argued that die period under study was much too short to support such a strong
claim, and tiiat examining the new entrants' wages might reveal a more pronounced
effect. Rafael Repullo felt diat labour economics tended to have excessively
simplistic models of wage setting, despite its frequent claim diat die labour market
needs special models.
Georges de Menil was not convinced diat diere is an inconsistency between die
relative stability of the Kaitz index and the growing level of unemployment. In a
stochastic world it is the relative rigidity of the Kaitz index that proves costly. In
order to appreciate the true cost of minimum wages, it would be more instructive
to examine particular episodes where die minimum wage was changed. Naturally,
given the downward stickiness of wages, diere might be more information in
episodes where minimum wages were increased rather than decreased. He
wondered if it was somehow possible to retain die legislated minimum wages at
their current levels and yet allow workers to work for a lower wage if diey were so
willing. Frederick van der Ploeg thought this was unlikely to succeed. Given diat
366 JUAN DOLADO ETAL

legislated minimum wages are closely tied to the level of unemployment benefit,
workers have little incentive to work for a lower wage. For him the principal policy
issue was: how do we reduce the gap between the legislated minimum wage and
the agreed minimum wage? He felt that wage demands could be moderated by
offering tax concessions around the minimum wage, and tax credits to firms for
training. Alessandra Casella thought that the excessive concern with the pure
employment consequences of minimum wages ignored the potential impact of
minimum wages on other variables, such as poverty, health and even long-term
growth.

TECHNICAL APPENDIX
Our theoretical framework is based on a simplified version of the model in Dickens
et al. (1994a).12 Assume firm i has a log marginal revenue product of labour curve
given by:

mrpli = a, - rjitj (Al)

where n is log employment and a is a shock to the MRPL which reflects demand or
productivity shocks. We assume that these shock are normally distributed across
firms with mean ju and variance a2. If the labour market were perfecdy competitive,
die elasticity of labour demand would be (1/?/). Labour supply to the firm is given
by:

w{ = 6w+ erii + b (A2)

where w{ is the log wage in firm i and w is the average log wage. If the labour
market is perfecdy competitive then £ = 0, but if e > 0 then the market is, to some
extent, monopsonistic. The dependence of labour supply on the average wage gives
us a simple way of modelling the spillover effect of the minimum wage on wages in
omer firms: it implies that the elasticity of the labour supply curve facing the firm is
more elastic than the labour supply curve in the economy as a whole. If 6 = 1 then
aggregate labour supply is completely inelastic; b is a labour supply shock which, for
simplicity, we assume is constant across firms.
First, consider the equilibrium when there are no minimum wages. Each firm
chooses the level of employment where the log MRPL equals the log marginal cost
of labour, which from equation (A2) is given by:

mcli = h\{\ + e) + wi (A3)

We assume that all firms face the same labour supply curve and that economy-wide averages are averages across
firms of the logs of the relevant variables (this keeps the mathematics simple). One must be a little careful in using this
version of the model to analyse the competitive case, since our assumptions imply that there would be no wage
dispersion in this rase. See Dickens etal. (1994a).
MINIMUM WAGES 367

Equating (Al) and (A3) gives employment in firm i (for a given average wage) as:

[fl,-ln(l+£)]-(ga;+*)
n,- = s 72(a>, a,-, 0) (A4)
Tj + E

and from (A2) the wage is:

e [ a , - l n ( l + e ) ] + i7(Oa;+6)
w,- = = v(w, ah b) (A5)
t] + E
Equations (A4) and (A5) are easy to understand. Revenue shocks, a, have a positive
effect on employment, while supply shocks, b, have a negative effect. In contrast,
both a and b are positively related to wages, although, as we would expect, a has an
effect only to the extent that the labour market is not perfectly competitive (where
e>0).
One can solve the model by then taking expectations of (A5) and using the
assumption that E(w{) = w. We can derive:
„ e[/i-\n{l+e)] + rib
w = (Ab)
e + ri(\-d)
and the average log employment across firms is:

„- = (l-d)[n-ln(l+e)]-b (A?)
e + J7O - 6>)
Now consider what happens if a minimum wage of w is introduced. This will
have different effects on different firms according to their different productivity
shocks. Some firms will pay above the minimum, so that equations (A4) and (A5)
will continue to be relevant. Note that if 0 * 0 , the change in w caused by the
minimum wage will mean that the set of firms initially paying above w will not be
the same as the ones now paying above w , and that although these firms pay
above the minimum, they are still affected by it. A firm will be in this regime if
the desired wage as given by equation (A5) is above w , i.e. if a, > a where a
satisfies:
* rj[6w+b]+e[a*-ln{l+e)]
w = (A8)
rj + e

For firms with a, ^ a , we will have, using equation (A5):

£ -if «4J t )fc sfc

w.(ai) = (a,- - a ) + v(w, a , b) = (a,- - a ) + w (A9)

r/ + e rj + E
All firms with a,: *£ a* will pay the minimum wage w . The fraction of firms paying
the minimum wage (i.e. die spike) will be given by 0(z ) where z ={a -f*)/cr.
368 JUAN DOLADO ETAL

Putting this information together we have diat the expected wage is given by:

w =Q(z*)w* + [1 - *(«*)](»* + {-' + -!- Eifli-a* | a, 5 a*) (A10)

= w* + _i£_ ( r V [ l -«(**)] +^(**»

Equation (A 10) can be solved for the average wage in terms of the minimum wage.
Although mere is generally no explicit analytical solution for w in terms of the
minimum, we can still derive several useful conclusions.
First, the effect of an increase in the minimum wage on the average wage is given
by:

)w <J>*
< 1 (All)
dw* 90(1-0*)
E + 1J

Hence an increase in the minimum wage leads to a less man proportional increase
in me average wage, so that the Kaitz index will rise. This is the basis for the claim
in the paper that we would expect rises in the minimum wage always to be
associated with rises in the Kaitz index. 13 One can also see that a given rise in the
minimum wage has a larger effect on the average wage, the larger is the spillover
effect, 0.
Second, as (o> - w) is the log of the Kaitz index and <&(z ) is a measure of die
spike, one can mink of equation (A 10) as giving a relationship between the Kaitz
index and the spike. It shows that, other things being equal, a rise in the Kaitz index
will be associated with a rise in the spike. But equation (A 10) also shows mat the
relationship between the Kaitz index and the spike is influenced by the parameters
of the economy, so that, for example, an increase in a will, for a given spike, lead to
a reduction in the Kaitz index.
Now let us turn to employment. In die unconstrained firms we will have
employment given by equation (A4), which can be written as:

Ufa) = n(w, a*,b)+ — ( a , - a *) (Al2)

£ + ?/

Now consider firms with a,; < a , i.e. those which pay the minimum wage.
Employment in these firms will be determined by the minimum of labour demand

This result is based on the assumption that the wage distribution with a minimum wage is a censored distribution. If
one were convinced that the minimum wage destroyed all jobs that would have paid below the minimum, it would be
more appropriate to use a truncated distribution. One can then show that the elasticity of the average wage with
respect to the minimum wage is less than 1 if (1 - F[w)) is log-concave, where Fis the distribution function of wages.
This condition is satisfied for most wage distributions used.
MINIMUM WAGES 369

and supply. Labour supply will be given by n{w, a , b), since the firm at the
boundary between these two regions will be on its labour supply curve. Employment
will be supply determined if the marginal revenue product of labour supply exceeds
the minimum wage, i.e. if:

a , - - (w* - dw-b)>w* (A13)

e
which can be rearranged to yield:

a,2*a*-ln(l + £) = a (A14)

Hence, firms for whom a - l n ( l + e)« a , < a pay the minimum wage and have
employment equal to supply n(w, a , b). Firms for whom a - ln(l + e) > a, will have
employment demand determined, in which case, using equation (Al), employment
can be written as:

n(a,) = n(w, a , b)+ — (a,- - a ) (A15)

Combining equations (A 15) and (A 12), we have that average employment is given
by:

n = £(»,-) = n{w, a*, b) + (1 - *(**)). E{a{ - a*\ a, 5* a*)

Tj + E

+ - *(*)£(«,- - a | ai « a)
V

= n(w, a*, b) + — ( r - / ( l - <&(<:*)) + 0(«*)) - - (z*(z) + 4>{z)) (A16)

ij + E rj

Now, using equation (A4), we have that:

+ a — [i az
n(w, a , b) = n(w, fi, b) + = n(w, fi, b) +
rj + e ri + e
d(w — wm) az
= n(wm,p,b)--± '- + -=- (A17)
Tf + E T] + £

And, using equations (A6) and (A8), we can derive:

/ -\ MZ* -{y + e){w* -w)

[w-w ) = (Alo)
E + t}(l-d)
Combining equations (A16), (A 17) and (A 18) we have that:

g(1
„ _ „- = " g ) L*i>(z*) + <t>(z*)) - - (z®tt) + <t>(z))\ (A19)
e + rti-fl) \ n )
370 JUAN DOLADO ETAL.

We can use this relationship to explain and evaluate the different approaches to
estimating the employment effects of the minimum wage.

A1. Time-series and panel studies

The idea here is to estimate an equation like (A 19) including measures of demand and
supply shocks to control for nm and then the Kaitz index to pick up die minimum wage
effects. Note that equation (A19) shows that, given die parameters of the economy, die
Kaitz index or the spike is a sufficient statistic for the employment loss or gain
attributable to die minimum wage independent of demand and supply shocks, so the
Kaitz index is a sensible measure to use of die impact of me minimum wage.

A2. The 'natural experiment' approach

The idea here is to take two markets, one with a big change in the minimum wage
and one without, but which might be expected to have similar demand and supply
shocks. Taking the difference in difference between mese markets in equation (A 19)
men allows an estimate of me effect of minimum wages.

A3. The 'differential impact' approach

Here, the idea is to look at the relationship between employment changes and the
impact of minimum wages across agents in a single market. There are potential
problems with mis. The problem is that the model predicts a non-monotonic
relationship between employment change and the impact of the minimum wage.
To see this, consider the effect of introducing a minimum wage. Assume, for
simplicity, that 0 = 0. The line A7V gives the relationship between employment and

0 a' a

Figure Al. The impact of minimum wages in our theoretical model

MINIMUM WAGES 371

a, before the introduction of a minimum wage. Now consider the imposition of a

minimum wage. Employment in the firms above a will be unaffected, those
immediately below a will have constant employment and those with sufficiently
low a, will have employment falling. The differential impact approach examines the
slope of the relationship between the gap between the two lines (the employment
change) and the gap between the initial wage and the minimum wage. This is not
the same as the total employment change, which is the integral between the two
employment lines. Figure Al needs modification if 6 is not zero, since then j\CV"itself
moves down with the minimum wage, but the basic ideas remain the same.
However, it is unclear how serious this problem is because no evidence of such a
non-monotonic relationship has ever been presented.

A4. The Meyer-Wise approach

The model presented above can be used to illustrate the problems associated with
the Meyer-Wise approach. As shown in Dickens et al. (1994b), this approach
estimates the wage distribution truncated by the minimum. The proportion of
workers who are predicted to have wages below the minimum is then compared
with the actual proportion paid the minimum. In the model presented here,
equation (A9) can be used to show that for those paid above the minimum wage the
distribution of wages is a truncation of the normal distribution given by:

( * E(U - a )
w* + — '-,
e + rj
ea

(e + ri) J
\
(A20)

where the truncation is at w . Assuming that the researcher estimates the correct
distribution, the proportion of firms which should be paying wages at or below the
minimum is given by <&({a - ft)/o). This is the actual proportion paying below the
minimum, so the methodology of Meyer-Wise would induce the researcher to
conclude that there are no employment effects - a conclusion that is generally
wrong.

REFERENCES
Abowd, J. M., F. Kramarz, T. Lemieux and D. N. Margolis (1995). 'Minimum Wages and Youth
Unemployment in the United States', presented at the CIRANO Summer Workshop on Unemployment in
the 1980s,July.
Albaek, K. and E.S. Madsen (1987). 'Wage Distributions and Employment Effects of the Danish Minimum
Wage', in R. Lund and P. Pederson (eds.), Studies in Unemployment, Samfundslitteratur.
(1989). 'Den solidariske lonpolitik og beskaefitgelsen af lavtlonnede', Nationalokonomisk Tidsskrift.
Bazen, S. and J. P. Martin (1991). 'L'impact du salaire minimum sur les salaires et l'emploi des jeunes et des
adultes en France', Revue Economique de I'OCDE.
Bazen, S., G. Benhayoun and N. Skourias (1995). 'The Impact of Minimum Wages in France: A Regional
Approach', IERSO, Universite de Bordeaux IV.
Benhayoun, G. (1994). 'The Impact of Minimum Wages on Youth Employment in France Revisited',
International Journal ofManpower.
Brown, C., C. Gilroy and A. Kohen (1982). 'The Effect of the Minimum Wage on Employment and
Unemployment', Journal of Economic Literature.
372 JUAN DOLADO ETAL

Cahuc, P. and J. Michel (1996). 'Minimum Wage, Unemployment and Growth', European Economic Review,
forthcoming.
Calvo, G. and S. Wellisz (1979). 'Hierarchy, Ability and Income Distribution', Journal of Political Economy.
Card, D. (1992a). 'Using Regional Variations in Wages to Measure die Effects of the Federal Minimum
Wage', Industrial and Labor Relations Review.
(1992b). 'Do Minimum Wages Reduce Employment? A Case Study of California, 1987-89', Industrial
and Labor Relations Review.
Card, D. and A. Krueger (1994). 'Minimum Wages and Employment: A Case Study of me Fast Food Industry
in NewJersey and Pennsylvania', American Economic Review.
(1995). Myth and Measurement: The New Economics of the Minimum Wage, Princeton University Press,
Princeton, NJ.
Card, D., F. Kramarz and T. Lemieux (1995). 'Changes in die Relative Structure of Wages and Employment:
A Comparison of the United States, Canada and France', unpublished, Princeton University.
Combault, P. (1995). 'Les Salaires au Smic enjuillet 1990', INSEE.
Currie, J. and B. Fallick (1994). 'The Minimum Wage and the Employment of Youth: Evidence from the
NLSY', UCLA Working Paper.
Dickens, R., S. Machin and A. Manning (1994a). 'The Effects of Minimum Wages on Employment: Theory
and Evidence from Britain', LSE CEP Discussion Paper.
(1994b). 'Estimating die Effect of Minimum Wages on Employment from die Distribution of Wages: A
Critical View', LSE CEP Discussion Paper.
Dickens, R., S. Machin, A. Manning, D. Metcalf, J. Wadsworth and S. Woodland (1995) 'Minimum Wages
and UK Agriculture', Journal ofAgricultural Economics.
Freeman, R.B. (1996). 'The Minimum Wage as a Redistributive Tool', Economic Journal.
Garcia, S., F. Goerhch and V. Orts (1994). 'Macro Magnitudes Basicas a Nivel Sectorial de la Industria
Espanola: Series Historicas', Economia Industrial, 27.
Gelot, O. (1995). 'Youdi Employment Programs in France', paper presented at NBER, 1995.
Grossman, J.B. (1983). 'The Impact of die Minimum Wage on Odier Wages', Journal ofHuman Resources.
Guesnerie, R. and K. Roberts (1987). 'Minimum Wage Legislation as a Second Best Policy', European Economic
Review.
Koning, P., G. Ridder and G. van den Berg (1994). 'Structural and Frictional Unemployment in an
Equilibrium Search Model widi Heterogeneous Agents', Free University of Amsterdam Research
Memorandum 1994-52.
Machin, S. and A. Manning (1994). 'Minimum Wages, Wage Dispersion and Employment: Evidence from
the UK Wages Councils', Industrial and Labor Relations Review.
Manning, Alan (1995). 'How Do We Know that Real Wages are Too High?', Quarterly Journal ofEconomics.
(1996). 'The Equal Pay Act as an Experiment to Test Theories of die Labour Market', Economical,
fordicoming.
Marceau, N. and R. Boadway (1994). 'Minimum Wage Legislation and Unemployment Insurance as
Instruments for Redistribution', Scandinavian Journal ofEconomics.
Meyer, R. and D. Wise (1983a). 'The Effects of die Minimum Wage on die Employment and Earnings of
Youdi', Journal of Labor Economics.
(1983b). 'Discontinuous Distributions and Missing Persons: The Minimum Wage and Unemployed
Youdi', Econometrica.
Mot, E. and C. Teulings (1990). "The Employment and Wage Effects of Minimum Wage Reductions for
Young People in die Nedierlands', unpublished, University of Amsterdam.
(1995). 'The Effect of New Jerseys Minimum Wage Increase on Fast-Food Employment: A Re-
evaluation using Payroll Records', NBER Working Paper No. 5224.
NickeU, S. and B. Bell (1995). "The Collapse in the Demand for the Unskilled Across the OECD', Oxford
Review ofEconomic Policy.
Nolan, B. (1993). 'Low Pay in Ireland', ESRI General Research Paper No. 159, Dublin.
OECD (1994). Jobs Study, Paris.
Ostros, T. (1994). 'Do Minimum Wages Matter? The Case of Swedish Mining and Manufacturing',
unpublished, Uppsala University.
Rebitzer, J. and L. Taylor (1995). "The Consequences of Minimum Wage Laws: Some New Theoretical
Ideas', Journal of Public Economics.
Saint-Paul, G. (1994). 'Do Labor Market Rigidities Fulfill Distributional Objectives?' IMF StaffPapers.
Schmitt, J. (1996). 'The Minimum Wage and Job Loss: Opponents of Wage Hike Find No Effect', Economic
Policy Institute Briefing Paper, Washington, DC.
Sudierland, H. (1995). 'Minimum Wage Benefits', New Economy.
Uri, N. D. and M.J. Mixon (1980). 'An Economic Analysis of die Determinants of Minimum Wage Voting
Behavior', Journal ofLaw and Economics.
Valdes, T. (1992). 'Diseflo del Panel de Declarantes de IRPF y Patrimonios', in Las Rentas de las Famitiasy su
Tributacion, Institute) de Estudio Fiscales.
van Soest, A. (1989). 'Minimum Wage Rates and Unemployment in die Nedierlands', De Economist.
(1994). 'Youdi Minimum Wage Rates: The Dutch Experience', International Journal ofManpower.