P A L K A N S A A J I E N T U T K I M U S L A I T O S • T Y Ö P A P E R E I T A
LABOUR INSTITUTE FOR ECONOMIC RESEARCH • DISCUSSION PAPERS
effects of low-
We are grateful to Matz Dahlberg, Tomi Kyyrä and seminar participants at the Athens University of
Economics and Business, the Helsinki Centre for Economic Research, the CESifo Area Conference
on Labour Economics, the University of Jyväskylä, and the 2009 EALE Conference (Amsterdam), the
Nordic Summer Institute in Labour Economics (Aarhus), and the Nordic Seminar for Public Economics
(Uppsala, Sweden) for useful comments. Kari Eerola provided excellent research assistance and Ossi
Korkeamäki helped with calculations involving the FLEED data. Financial support from the Finnish
Employees’ Foundation is gratefully acknowledged.
*Labour Institute for Economic Research
**Corresponding Author, Address: Department of Economics, 33014 University of Tampere,
Finland. email: email@example.com
***Government Institute for Economic Research, IZA and IFAU
Low-wage subsidies are often proposed as a solution to the unemployment problem among the
low skilled. Yet the empirical evidence on the effects of low-wage subsidies is surprisingly
scarce. This paper examines the employment effects of a Finnish payroll tax subsidy scheme,
which is targeted at the employers of older, full-time, low-wage workers. The system’s clear
eligibility criteria open up an opportunity for a reliable estimation of the causal impacts of the
subsidy, using the difference-in-difference-in-differences approach. Our results indicate that the
subsidy system had no effects on the employment rate. However, it appears to have increased
the probability of part-time workers obtaining full-time employment.
Key words: low-wage subsidies, employment, social security contributions.
JEL Classification: H24, J23, J68.
Usein esitetty ratkaisu vähän koulutetun työvoiman työttömyysongelmaan ovat
matalapalkkatuet, mutta niitä arvioivaa empiiristä tutkimusta ei ole paljon. Tässä työssä
tutkitaan suomalaisen työnantajille suunnatun matalapalkkatuen työllisyysvaikutuksia. Tuen
selkeä kohdentuminen (siihen ovat olleet oikeutettuja täysaikaisia, yli 54-vuotiaita
matalapalkkaisia palkanneet) tarjoaa hyvän mahdollisuuden arvioida tukijärjestelmän
vaikutuksia luotettavalla tilastotieteellisellä tavalla. Tulostemme mukaan tukijärjestelmä ei
kasvattanut kohderyhmän kokonaistyöllisyyttä. Sen sijaan osa osa-aikaisista työntekijöistä
siirtyi täysaikaisiksi tukijärjestelmän ansiosta.
Asiasanat: matalapalkkatuki, työllisyys, sosiaaliturvamaksut
One way to reduce unemployment among the low skilled is to cut labour taxation for low-
income workers. This raises the question about the general effectiveness of targeted tax cuts
and also the more nuanced question about what the best way is to implement these cuts. Phelps
(1994, 1997) and Dreze and Malinvaud (1994), for instance, argue that a subsidy given to the
employers of low-skilled workers could be more effective in increasing the demand for those
workers than a reduction in the taxes paid by the employees. The reason is that if wages are
rigid downwards, the subsidy reduces labour costs and therefore increases labour demand more
than a reduction in the labour income tax paid by the employees themselves.1
Low-wage subsidies, in the form of targeted cuts to employers’ social security contributions,
have also been implemented in practice, for example in Belgium and in France. In some other
countries, notably Germany, there is a debate on whether low-wage subsidies should be
introduced.2 Kramarz and Philippon (2001) and Crepon and Desplatz (2002) provide
econometric evaluations of the French system.3 While these papers represent high-quality
empirical work on the subject, some elements of the French system render its evaluation a
complicated task. In particular, minimum wages have been changed simultaneously with the
subsidy system. The French system also provides benefits to all low-wage workers, which
makes it hard to find a proper comparison group that could be used to predict what would have
happened to low-wage employment without the subsidy scheme.
Low-wage subsidies have also attracted a large amount of theoretical analysis, based on search-
theoretic frameworks (see e.g. Chéron, Hairault and Langon (2008) and Brown, Merkl and
Snower (2007) and the references there). This strand of literature often involves simulation
analyses with empirically plausible parameter values or structural econometric work. Another
sizable literature deals with hiring subsidies (see Katz 1996 for a survey and Bell, Blundell and
van Reenen (1999) or Kangasharju (2007) for more recent evidence). The evidence on the
effects of these temporary subsidies is, however, not necessarily suitable for assessing the
effectiveness of permanent low-wage subsidies.
1 Edlin and Phelps (2009) argue, in addition, that low-wage subsidies might be an especially effective way to
increase demand in a recession.
2 See, for instance, Knabe and Schöb (2008), which builds on the theoretical analysis in Knabe et al. (2006).
3 A large amount of other relevant empirical work exists. This literature is reviewed in more detail in Section 2.
On balance, it seems fair to say that empirical evidence on the effectiveness of low-wage
subsidies is still scarce. This is in marked contrast to the evidence of the targeted tax cuts for
employees. This evidence builds on the experience from in-work benefit systems implemented
in the US and the UK. The Working Families Tax Credit (WFTC) in the UK and the Earned
Income Tax Credit in the US appear to have, indeed, been successful in increasing employment
among the target groups. The design of these systems, involving a clear division between the
target group and non-eligible persons, has opened up an opportunity for reliable econometric
work that has been able to isolate the causal impacts of these policies (see Eissa and Liebman
(1996) and Eissa and Hoynes (2004) for the US case and Blundell 2006 for the UK evidence).
Even more compelling is experimental evidence from the Self-Sufficiency Project, which
provides earnings subsidies for welfare leavers in Canada. (See e.g. Michalopoulos et al. (2002)
and Card and Hyslop (2005).)
In addition, because of large differences in labour market institutions, it is not entirely clear that
a scheme that works in the Anglo-Saxon countries is directly applicable to countries, say, in
Continental Europe or in the Nordic states. The difference in the nominal recipient of the
subsidy (the employee in the case of the EITC and the WFTC, the employer in the original
Dreze-Malinvaud and Phelps proposals) can also imply that the success of low-wage subsidy
schemes cannot be evaluated based on the evidence on in-work benefits.
The purpose of this paper is to offer new evidence on the causal effects of low-wage subsidies
by examining the impacts of a highly targeted low-wage subsidy experiment that started in
Finland in 2006. Finland is a good case for analysing the effectiveness of payroll tax subsidies.
Union contracts have led to a relatively narrow wage distribution which could have contributed
to the gap in the unemployment rates between low-skilled and high-skilled workers that is
among the largest in Europe, according to the Eurostat Labour Force Survey.
The design of the Finnish low-wage subsidy scheme makes evaluating its impacts relatively
straightforward. In order to be eligible for the subsidy, workers must be over 54 years of age,
earn a salary between 900 and 2,000 euros per month and work full time. This means that we
can find several comparison groups for the targeted workers, thus allowing the difference-in-
difference-in-differences (DDD) approach. We can simultaneously control for any permanent
differences across the eligible and ineligible groups and take into account time-varying
differences in labour demand for different skill groups. The latter, particularly, may be quite
important if skill-biased technical change or globalisation changes the relative productivity of
different workers. In this paper we use this setup to study the impacts of the low-wage subsidy
scheme on wage rates, hours worked and employment (mainly via retention rates), offering a
full analysis of the incidence and the employment effects of the system.
The paper proceeds as follows. Section 2 reviews earlier relevant empirical work.4 Section 3
explains the Finnish subsidy system in more detail. The actual empirical analysis in the paper
has many phases and utilises several different data sets. Therefore it may be useful to set out a
fairly detailed plan of the empirical content of this paper.
In section 4, we look at the trends in the employment rates of workers in different age groups
using data from the Finnish Labour Force Survey. The main purpose is to examine the overall
employment impacts of the subsidy system.
In section 5, we use register-data of the unemployed and examine in detail whether the subsidy
system affected re-employment rates. The main interest is in finding out whether there was a
change in re-employment rate of older unemployed workers, and in particular in re-
employment rate of those older unemployed with low pre-unemployment income level or low
education level. This section therefore investigates whether the subsidy system led to increased
entry to the workforce of those unemployed who were likely to benefit from the subsidy
Section 6, representing the bulk of our analysis, builds on a detailed data set that covers all
workers of employers that are members of the Finnish Employers’ Confederation. The benefit
of these data is that they contain detailed information about the working hours and monthly
wages of the workers that are needed to strictly target the analysis to the affected workers. This
enables us to build a clearly defined treatment group and corresponding control groups (in
other words, the DDD analysis). This data is then used to examine the exit rates of workers
employed by these firms. In addition, we estimate the impacts of the subsidy system on the
working hours and wages of those workers who keep their jobs.
Finally, section 7 discusses results from a number of extensions to the analysis above. Section 8
4 For the sake of space, we do not cover the theoretical literature on payroll tax subsidies here. Brown et al.
(2007) contains an extensive list of theoretical work in the area.
2. EARLIER EMPIRICAL WORK
The best-known scheme of targeted social security cuts has been implemented in France, where
payroll taxes were first reduced in March 1994 by 5.4% for workers earning no more than 1.1
times the minimum wage and 2.7 per cent for workers in the range between 1.1 and 1.2 times
the minimum wage. The subsidy was increased dramatically in September 1995 and its range
was extended in October 1996. At the end of 1996, the subsidy was 18% at the minimum wage
and decreased linearly thereafter, hitting zero at 1.33 times the minimum wage.5
The employment effects of the French payroll tax subsidy scheme have been evaluated by
Kramarz and Philippon (2001) and Crepon and Desplatz (2002). Given that the French
subsidy is universal, both studies suffer from a lack of a natural comparison group. Kramarz
and Philippon base their evaluation on household survey data and examine the effects of
changes in the minimum labour costs - hence capturing the effects of both the minimum wage
changes and the changes in payroll tax subsidies at the minimum wage level. By comparing
workers affected by the minimum wage increases with workers just above the new minimum
wage, they show that increases in labour costs increase transitions to non-employment.
However, their analysis regarding the effects of a decrease in the labour costs due to an
increase in the payroll tax subsidy reveals no significant employment effects. The authors
measure this as an increase in minimum-wage workers coming from non-employment.
Crepon and Desplatz (2002) perform their analysis using firm-level employment as the key
dependent variable. They calculate the ex-ante change in labour costs due to the payroll tax
subsidies, using payroll tax parameters and the composition of the firm’s labour force before
the introduction of the payroll tax changes. They find that employment in firms that received
larger subsidies grew more than employment in firms that employed fewer low-wage workers
and hence received fewer subsidies. The authors interpret this as strong evidence for the
employment effects of low-wage subsidies. Since the outcome variable is total employment,
the authors cannot discover whether the increase in employment occurs in the targeted low-
wage group or whether the increase in employment is due to an increase in high-wage workers.
5 The Finnish and French subsidy systems are roughly similar in magnitude, but the Finnish subsidy is phased
out more slowly and hence has an impact on labour costs at much higher wage levels. Another important
difference is naturally that the French subsidy affects all low-wage workers, while the Finnish subsidy is targeted
at older workers.
Targeted payroll tax subsidies have also existed in the Netherlands and in Belgium. We are not
aware of the econometric evaluations of the Dutch system, but Bovenberg et al. (2000) evaluate
its effects using a simulation model calibrated to Dutch data. Their conclusion is that the most
effective way of reducing unemployment is the introduction of in-work benefits, though the
simulation results between the benefits paid to the low-wage workers or to the employers of
these workers are roughly similar. Goos and Konings (2007) evaluate the effects of changes
occurring in the ‘Maribel subsidies’ system in Belgium in the late 1990s using firm-level data.
These subsidies reduced the payroll taxes paid on manual workers. Even though the subsidy
was not specifically targeted to the employers of low-wage workers, its lump-sum structure
reduced the payroll taxes for the low-wage workers more than for the other groups. Goos and
Konings find that the subsidy had significant effects on employment.
The subsidies discussed above involve a decrease in the payroll taxes of the subsidized
workers. Gruber (1994) analyses the effect of a reverse experiment, increasing the costs of
hiring certain groups by increasing mandatory employer contributions. He examines the effects
of forcing employers to purchase health insurance that includes maternity benefits, a change
mainly affecting young women. He shows that the costs of these group-specific mandates are
mainly borne by workers in terms of lower wages and that the additional costs have little
effects on employment.
While permanent non-categorical subsidies to all employers of the low-wage workers are rather
rare, there is a large literature evaluating the effects of temporary subsidies to employers who
hire long-term unemployed persons or workers with disabilities. Many of these programs have
been evaluated using randomised trials. In his comprehensive survey of the US programs Katz
(1996) concludes that wage subsidies have been effective in improving the earnings and
employment of disadvantaged groups, at least when combined with training elements. More
recent evidence is available from Britain, where the so-called ‘New Deal’ system has led to
modest improvements in the productivity of the target group (Bell et al. 1999), and Finland
where temporary subsidies for the unemployed who find work have been found to be effective
(Kangasharju 2007)6 Finally, related literature looks at the impacts of regional employment
subsidies (see e.g. Korkeamäki and Uusitalo (2009) and Bennmarker et al. (2009).
6 See also Gesine (2009) that focuses on the wage effects of hiring subsidies. His paper also includes an
extensive survey of the empirical work on hiring subsidies. Since temporary hiring subsidies are often one part
As discussed in the introduction, an alternative to employer-based subsidies is to target the
subsidy to employees. This is the way in which the Earned Income Tax Credit in the US and
the Working Families Tax Credit in the UK are designed. Despite the different nominal
recipients, there are also similarities that make the results from the evaluation of these subsidy
schemes relevant for the Finnish case. All these schemes share the property that the subsidy is
targeted to the low-wage workers and that the subsidy gradually decreases after earnings
increase above some threshold level. All these schemes are also intended to be permanent
subsidies for the low-wage workers instead of temporary subsidies for the newly hired.
Importantly for the evaluation, all these schemes also have other eligibility criteria in addition
to low earnings. This allows us to compare wage and employment changes after the
introduction or expansion of the subsidy in the eligible group and in some comparison group
that is in a reasonably similar position in the labour market. Using this strategy, Eissa and
Liebman (1996) and Blundell (2006) compare the changes in labour supply between single
mothers and (ineligible) single women without children. Both of these studies find substantial
effects on the labour supply.
3. THE FINNISH EMPLOYER LOW-WAGE SUBSIDY SCHEME
Since January 1st 2006 Finnish employers have been eligible for a wage subsidy if they employ
a low-wage worker that is over 54 years old. The subsidy-scheme is temporary and will be in
force until December 2010.7 The subsidy depends on the wage level and may be up to 16 per
cent of the gross wage or 13 per cent of the total pre-reform labour costs including payroll
The subsidy covers full-time workers who are employed at least 140 hours per month and
whose wage is between 900 and 2,000 euros per month. The subsidy equals 44 per cent of the
part of the monthly wages that exceeds 900 euros. The maximum subsidy per employee is 220
euros a month. The amount of the subsidy is reduced by 55 per cent of the monthly wages
exceeding 1,600 euros.
of active labour market policies, our paper is also related to the literature of evaluating these policies. For one
survey, see Calmfors et al (2004). Recent evidence is available e.g. in Adda et al (2007) and Sianesi (2008).
The wage subsidy is paid to the employer and can be seen as simply a reduction in the payroll
tax rate for the firms that employ old low-wage workers. In 2006 the average payroll tax rate
was 20.9 per cent of the gross wage. The tax is levied on all wages. The revenues are mainly
used for funding the employee’s pension system and the sickness insurance. The tax rate is
slightly higher for larger and more capital-intensive firms. For large firms, pension payments
also vary according to the age structure of the employees and according to the disability and
unemployment pensions granted to former employees. Still, even for large firms the payroll tax
is a proportional tax on all wages paid.
The 2006 tax subsidy created a system where the payroll tax rate is a decreasing function of
monthly wages when wages are between 900 and 1,400 euros. The payroll-tax rate is at its
minimum (5.2%) when the monthly wage equals 1,400€. When wages are between 1,400 and
1,600 euros, firms get the maximum subsidy of 220 euros. The subsidy is gradually reduced
when wages increase above 1,600 euros so that the subsidy reaches zero when the monthly
wage equals 2,000 euros. In this phase-out range the payroll taxes are strongly progressive with
tax rates increasing from 7.2 % at the wage level of 1,600 €/month to roughly 21% at the wage
level of 2,000 €/month. Figure 1 illustrates the effects of the low-wage subsidy by plotting the
reduction in the payroll tax due to the system and the corresponding average payroll tax rate.
Finland has no minimum wage laws. However, union contracts also cover non-union workers
so that in practice the lowest legal wages are set in union contracts for about 95 per cent of the
workers. In union contracts, the lowest wages vary across sectors, regions and tasks. In typical
low-wage sector contracts, the lowest full-time wages were around 1,300 in 2006. (1,280
€/month in the retail trade, 1,301 €/month in hotels and restaurants and 1,249 €/month for
security guards). In comparison one could note that according to Statistics Finland the average
wage for full-time workers was around 2,500 euros in 2006. The subsidy is therefore targeted at
workers that are well below the average wage, with the maximum subsidy paid to those whose
wage is close to the minimum wage.
7 Whether the system will be continued will depend on its effectiveness during the experimental period. Since
the experiment will be quite long, it is reasonable to believe that firms can react to it and therefore the evaluation
of this system will also reveal relevant information for a truly permanent scheme.