The impact of employment subsidies on job creation: nonexperimental evidence using firm level data

1
The impact of employment subsidies on job
creation non-experimental evidence using firm
level data
preliminary

• Maarten Goos and Joep Konings (KUL)

2
motivation (1)

• With the joining of China, India and the
  ex-Soviet bloc, there has been a great doubling
  of the global workforce and consequently a fall
  in the global capital-labor ratio especially for
  low-skilled manual workers.
• Recent technological change driven by
  computerization is biased against low-skilled
  manual workers.
• Both facts put pressure on low-skilled manual
  workers either in terms of their employment
  prospects or in terms of their wage.

3
motivation (2)

• Whether wage dispersion or unemployment
  temporarily increases during this transition
  period depends on the degree of wage flexibility
  in the low-skilled labor market.
• The literature suggests that wage flexibility
  will provide a smoother longer-run transition of
  low-skilled manual workers into other industries
  or skill-sets but that real wages are rigid in
  European countries.
• Could it be that subsidizing low-skilled manual
  employment is an effective instrument to increase
  wage flexibility and maintain low-skilled manual
  employment at least in the short-run?

4
motivation (3)
5
motivation (4)

• This paper will use a panel of Belgian firms to
  examine the impact of lump-sum employment
  subsidies for full-time manual workers known as
  the Maribel subsidies
• Given that the Maribel subsidies introduced
  progressivity in payroll taxes, we will first
  provide a framework to analyze its impact under
  different assumptions about the nature of labor
  and product markets
• Second, given that Maribel subsidies were not
  applied across the board and given that we have
  information about the actual subsidy received by
  each firm, a set a straightforward
  non-experimental evaluation techniques can be
  used to analyze the impact of Maribel subsidies
  on full-time manual employment.

6
presentation outline

• a framework to understand the impact of Maribel
  subsidies on job creation
• the history of Maribel
• data
• empirical analysis
• conclusions

7
a framework (1)

• A Maribel subsidy is a per-period lump-sum
  subsidy paid to the employer for each full-time
  manual worker employed at that firm. A framework
  is needed to understand the possible impact of
  Maribel subsidies on full-time manual employment.
  
• This framework must account for the existence of
  a proportional payroll tax, wage bargaining
  between unions and firms and the fact that each
  of many firms has some product market power.
• It will be argued that under realistic
  assumptions about labor and product markets,
  Maribel subsidies are expected to increase
  full-time manual employment.

8
a framework (2)

• Assume the pre-tax and post-tax wage are given
  by
• (1)
• with t the proportional marginal payroll tax and
  the lump-sum employment subsidy.
• (2) and
• (3) and

9
a framework (3)

• when the firm has no product and no labor
  market power
• A group of homogeneous firms maximizes profits
• (4) with p possibly
  different from the economy wide aggregate price
  index normalized to unity and such that
  and
  .
• (5) which for given p
  implicitly defines the unconditional demand for
  labor function.

10
a framework (4)

• when the firm has no product and no labor
  market power
• Unconditional labor demand is given by
• (6) with
• Uncompensated labor supply is given by
• (7) with
• Setting (6) (7) solves for employment and
  wages
• (8)

11
a framework (5)

• when the firm has no product and no labor
  market power
• Totally differentiating (8) wrt the pre-tax wage
  and the employment subsidy gives
• (9)
• (10) and
• Maribel subsidies are expected to increase
  full-time manual employment if labor and product
  markets are perfectly competitive

12
a framework (6)

• wage-bargaining when the firm has no product
  market power
• Assume unions and firms bargain over the pre-tax
  wage and firms then choose the level of
  employment
• (12)
• with .

13
a framework (7)

• wage-bargaining when the firm has no product
  market power
• It must be true in equilibrium that (13)
• The impact of employment subsidies on the pre-tax
  wage is
• (14)

14
a framework (8)

• wage-bargaining when the firm has no product
  market power
• Remember from (9) and (10) that
• and
• An increase in employment subsidies decreases
  employment if
• is larger than 1/1t. An interesting
  question is under what conditions employment
  subsidies lead to job destruction rather than job
  creation (as was true for perfectly competitive
  labor markets).

15
a framework (9)

• wage-bargaining when the firm has no product
  market power
• A necessary condition for (14) to be larger than
  1/1t is that
• Given the concavity of the Nash-bargain in
  pre-tax wages, this implies employment subsidies
  decrease employment only if
• (16)

16
a framework (10)

• wage-bargaining when the firm has no product
  market power
• For Maribel subsidies to decrease full-time
  manual employment, the labor demand function must
  be sufficiently concave and/or, in absolute
  value, the elasticity of labor demand must be
  sufficiently small relative to the elasticity of
  profits.
• This excludes a wide range of production
  technologies including the commonly assumed case
  of isoelastic production functions.

17
a framework (11)

• wage-bargaining when the firm has no product
  market power

E.g. Cobb-Douglas technology in the single
factor case
18
a framework (12)

• wage-bargaining when the firm has no product
  market power
• Therefore, for a wide range of production
  functions
• In sum, assuming wage bargaining, Maribel
  subsidies are expected to increase full-time
  manual employment under standard assumptions
  about production technologies.

19
a framework (13)

• wage-bargaining when the firm has product
  market power
• So far we have looked at conditions imposed on
  production technologies for employment subsidies
  to increase employment assuming labor markets are
  imperfectly competitive.
• But there imperfect competition on output markets
  too in the data examined below. In particular, a
  common assumption is that of monopolistic
  competition assuming that each takes the actions
  of other firms as given.

20
a framework (14)

• wage-bargaining when the firm has product
  market power
• The Nash-bargain is now given by
• (18)
• where is implicitly given by

21
a framework (15)

• wage-bargaining when the firm has product
  market power
• In line with the case where the firm is a price
  taker, employment subsidies decrease employment
  only if
• (19)
• This excludes a wide range of production
  technologies including the commonly assumed case
  of isoelastic production technologies and
  isoelastic product demand.

22
a framework (16)

• wage-bargaining when the firm has product
  market power
• E.g. Cobb-Douglas production technologies and
  isoelastic product demand in the single factor
  case

23
a framework (17)

• wage-bargaining when the firm has product
  market power
• In sum, under realistic assumptions about
  production technologies and product demand,
  Maribel subsidies are expected to increase
  full-time manual employment

24
presentation outline

• a framework
• the history of Maribel
• data
• empirical analysis
• conclusions

25
the history of Maribel (1)
Table 1 The history of Maribel
26
the history of Maribel (2)
Table 2 Maribel II/III
27
the history of Maribel (3)
Table 3 Maribel IV
28
presentation outline

• a framework
• the history of Maribel
• data
• empirical analysis
• conclusions

29
data (1)

• BELFIRST a panel of company accounts for the
  period 1995-1999
• Header and balance sheet firm identifier,
  industry classification, average annual
  employment at the firm
• Social balance sheet total full-time, full-time
  manual, part-time manual and full-time non-manual
  employment on 31/12, annual hours worked by
  full-time workers, average annual full-time
  employment, average annual part-time employment
  and annual Maribel subsidy received by the firm

30
data (2)
Table 4 Comparing BELFIRST
31
data (3)
Table 5 Subsidized and non-subsidized firms in
BELFIRST
32
data (4)
Table 6 Firm level Maribel subsidies in BELFIRST
33
data (5)
Figure 2 Actual and predicted total subsidies
for Maribel II/III in 1996
34
data (6)
Figure 2 (cont.) Actual and predicted total
subsidies for Maribel IV in 1998 given that
Xgt0.66
35
presentation outline

• a framework
• the history of Maribel
• data
• empirical analysis
• conclusions

36
empirical analysis (1)

• The aim of this section is to derive
  treatment-on-the-treated effects of Maribel
  subsidies on full-time manual employment.
• Since Maribel has been an uncontrolled
  experiment, one has to make choices about how to
  construct counterfactuals.
• This section uses the different possibilities
  with the data at hand applying DID, matching and
  IV estimators
• In line with the theory presented above, it will
  be argued that Maribel subsidies have increased
  full-time manual employment indeed.

37
empirical analysis (2)

• Difference-in-differences estimates compare the
  within-firm variation in employment between firms
  switching Maribel participation and firms not
  switching Maribel participation.
• Combining years 1996-1999 allows us to apply the
  fixed-effects estimator to the following
  functional form
• (21)
• Allowing for an arbitrarily different time trend
  for firms having ever received Maribel subsidy
  does not change the difference-in-differences
  estimates

38
empirical analysis (3)
Table 7 Difference-in-differences estimates of
the employment impact of Maribel subsidies
39
empirical analysis (4)

• the estimated coefficients are in line with the
  prediction that under standard assumptions about
  labor and product markets, Maribel subsidies
  increase full-time manual employment
• The estimated effects are relatively large
  suggesting that Maribel subsidies created at
  least between 56 000 (0.060x20.66x22827x2 for
  1999) and 85 000 (0.080x19.04x27644x2 for 1998)
  jobs or about 2 percent of economy wide
  employment.
• But this is not to say that these effects would
  be accurate estimates if the government were to
  double the number of firms entitled to receive
  benefits since these are treatment-on-
  the-treated effects.

40
empirical analysis (5)
Table 8 The impact of Maribel subsidies on
alternative employment measures
41
empirical analysis (6)

• Fixed-effects estimates reported above allow for
  unobserved time persistent firm characteristics
  such as time averaged firm size to be correlated
  with Maribel participation.
• But if, for example, smaller firms are likely to
  grow faster, the estimated difference-in-differenc
  es presented above will be downward biased (given
  that smaller firms are less likely to receive
  Maribel subsidies).
• A solution to this problem is to draw a sample
  from the group of non-receivers that better
  resembles the treatment group for example in
  terms of beginning-of-year full-time manual
  employment.

42
empirical analysis (7)
Table 9 Matching estimates of the employment
impact of Maribel subsidies
43
empirical analysis (8)

• The analysis so far has used whether or not firms
  received Maribel subsidies as the explanatory
  variable (variation in Maribel subsidies at the
  extensive margin).
• But receiving firms also differ in the amount of
  per-worker subsidy received in any given year
  (variation in Maribel subsidies at the intensive
  margin).
• However, using the amount of Maribel subsidy as
  an explanatory variable introduces measurement
  error as was argued above. This will bias the
  estimated impact towards 0.
• One possible solution to this problem is to apply
  a 2SLS estimator using the legal information as
  an instrument.

44
empirical analysis (9)

• Consider the following second-stage equation of
  interest
• (22)
• To deal with measurement error in S and the
  possible endogeniety between S and N, consider
  the following first-stage
• (23)
• where all right-hand-side variables are taken in
  1997.

45
empirical analysis (10)
Table 10 2SLS estimates using the group of
firms receiving Maribel subsidies in 1998
Notes All regressions include a lagged dependent
variable.
46
empirical analysis (11)

• Under realistic assumptions about the nature of
  labor and product markets, Maribel subsidies are
  expected to increase full-time manual employment.
• At the extensive margin, it was shown that firms
  receiving subsidies created more jobs compared to
  firms non-receivers using difference-in-difference
  s and matching estimators indeed.
• Also using variation at the intensive margin
  showed that among receivers the creation of
  low-skilled jobs is bigger the bigger is the
  subsidy.

47
empirical analysis (12)

• Note that under standard assumptions about
  production technologies and product demand, the
  impact of Maribel subsidies is positive
  independent of who has the bargaining power
• It can also be shown that the impact of Maribel
  subsidies is expected to be bigger the more
  elastic is product and therefore labor demand.

48
empirical analysis (13)
Figure 3 The employment impact of Maribel
subsidies and union-bargaining power in
manufacturing industries

49
empirical analysis (14)
Figure 4 The employment impact of Maribel
subsidies and import penetration in manufacturing
industries

50
conclusions (1)

• The question we started from was whether or not
  payroll tax subsidies are likely to increase wage
  flexibility and therefore employment?
• We developed a framework to show that lump-sum
  employment subsidies are expected to increase
  employment under realistic assumptions about
  labor and product markets.
• We used different sources of variation in firm
  level panel data exploited by different
  estimators to show that Maribel subsidies
  increased low-skilled manual employment by at
  least about 2 percent of the working population.

51
conclusions (2)

• But this is not to say that employment subsidies
  for low-skilled workers are a policy necessarily
  sustainable in the long-run.
• New cohorts of low-skilled workers should be
  educated or trained to do different jobs than
  they do today. Fine-tune our educational system
  to the needs of jobs in those sectors in which
  advanced countries aim to maintain a comparative
  advantage.
• Invest in RD to generate exports in selected
  high-tech industries in which advanced countries
  aim to maintain a comparative advantage and
  provide the skills needed to do these jobs.