Science and Technology Division

1
Science and Technology Division
Effects of Innovation on Employment in Latin
America the microeconomic evidence Comparative
results Gustavo Crespi and Ezequiel Tacsir
9th GLOBELICS Conference. Buenos Aires,
Argentina November15th, 2011
2
http//www.iadb.org The Inter-American
Development Bank Discussion Papers and
Presentations are documents prepared by both Bank
and non-Bank personnel as supporting materials
for events and are often produced on an expedited
publication schedule without formal editing or
review. The information and opinions presented in
these publications are entirely those of the
author(s), and no endorsement by the
Inter-American Development Bank, its Board of
Executive Directors, or the countries they
represent is expressed or implied. This
presentation may be freely reproduced.
3
Employment and Innovation in LAC

• The potential for inclusive growth in the region
  depends on its capacity to generate good quality
  employment.
• There are high expectations by regional Policy
  Makers on the potential of STI to trigger
  inclusive growth.
• However, we know very little on the transmission
  mechanisms and dynamics between innovation and
  job performance.
• Results will help us to better understand a very
  complex phenomena (very little studied in the
  region) and to generate inputs for the design of
  public policies that maximize the employment
  generation impacts of innovation and at the same
  time mitigate its negative consequences.

4
Motivation

• Our own work on the link between innovation and
  productivity research shows that the lack of
  innovation negatively affects productivity
  levels. How this affects labor market outcomes?
• However, we still lack knowledge on the relation
  between employment and innovation. Evidence from
  more developed countries or regions might not be
  applicable. Need to provide evidence for
  policy-making.
• In Latin America the production structure is
  strongly dominated by small and medium
  enterprises (SMEs). Indeed, SMEs innovation is
  strongly dominated by informal search routines
  and learning from already available knowledge and
  technologies.
• Importance of assessing the effects on employment
  (level and composition). It is not only quantity
  what matters, but also quality.

5
Focus

• This paper focuses on the (short term) link
  between innovation and employment. Based on
  national country studies performed by different
  research teams.
• Coverage Comparative research in four countries
  Argentina, Chile, Costa Rica and Uruguay.
• Novel features (as a project) Emphasis on SMEs,
  sectoral differences (low/high tech sectors).
  Extensions of the project during this session and
  following session. To further increase
  comparability and homogeneity among countries
  possibility to use Enterprise Surveys for LAC.

6
Research questions

• How different types of innovations (product,
  process, organizational) create or displace
  employment.
• (2) How different types of business innovation
  strategies (make or buy) influence the capacity
  of innovation to generate or destroy employment.
• (3) To analyze how (1) and (2) might render
  differential impacts across firms of different
  size and sectors?
• (4) To assess how different types of innovations
  and innovation strategies affect the quality of
  employment measured by types of skills.

7
Employment effects of innovation A Conceptual
framework

• The relationship between innovation and
  employment at the firm level is not
  straightforward. It is usually believed that
• Process innovation destroys jobs by substituting
  labor
• Product innovation creates jobs by an increase
  in demand
• But
• If process innovation increase productivity firms
  can reduce prices and increase the demand for
  their products
• If the firm with a new product gains market
  power, it can be the case that it is more
  profitable to increase price, reduce the quantity
  sold , and employment accordingly.
• Harrison, Jamandreu, Mairesse, and Peters (2008)
  (HJMP, from now onwards) provide an empirical
  framework to address the issue at the firm level.

8
Employment effects of innovation A Conceptual
framework
Displacement Compensation
Process Innovation Productivity effect (-) less labor for a given output Price effect () cost reduction, passed on to price, expands demand
Product Innovation Productivity differences of the new product (- or ) Demand enlargement effect ()
Depends of firms behavior
Innovation activities
Depends on competition
9
Methodology

• Data Research is based on innovation surveys
  (collected in several countries in the region,
  normally repeated cross-sections using similar
  methodologies). Need to match with other surveys
  (i.e., annual manufacturing surveys) or register
  data (social security).
• Argentina Second Innovation Survey (1998-2001),
• Chile 4 waves of National survey of Innovation
  (EIT) (1995, 1998, 2001, 2007), and National
  Annual Manufacturing Survey (ENIA) (1993-2007)
• Costa Rica Innovation survey for Costa Rica from
  2006 to 2007. Matched for performance with
  Manufacturing survey and social security records
• Uruguay 4 waves of (Manufacturing) Innovation
  Surveys (MIS) (1998-2000, 2001-2003, 2004-2006
  and 2007-2009) Annual Economic Activity Surveys
  (EAS) for the period 1998-2007. Both surveys have
  the same sample and statistical framework.

10
Effects on employment quantity

• the research setting assumes that a firm can
  introduce (i) product or process innovations and
  (ii) produce new and old products.
• 2 types of products are distinguished existing
  products and the production of new products. The
  change in employment is then decomposed into the
  part due to the increased efficiency in
  production of old products (related to process
  organizational innovations) and the part due to
  the introduction of new products (product
  innovations)
• HMJP (2008) suggests the following regression to
  estimate the effects of innovation on employment
  
• Where l is total employment growth, g1 is the
  nominal growth in sales of old products, g2 is
  the nominal growth in sales of new products
  (product innovations) and d captures the
  introduction of process innovations in the
  production of old products. If process
  innovations in the production of old products
  displace employment, a1lt 0. If product
  innovations create employment, ßgt0. ß also
  captures the relative efficiency in the
  production of old/new products (if lt1 new are
  more efficient than old).

11
Foundations of the empirical model

• Multiproduct production function (two products,
  CRS, etc)
• Production function
• K capital
• L labour
• M material
• ?jt technological efficiency
• ?it unobserved idiosynchratic productivity
  shocks
• ? unobserved firm-specific productivity shocks
  

12
Foundations of the empirical model

• Cost function
• Assumption of constant returns to scale
• Conditional labour demand for old products
• Conditional labour demand for new products
• Simplifying assumption
  
• Given that the production of new products at the
  beginning of the period is nil ( ) so
  we can approximate the employment growth
  decomposition as follows

13
Foundations of the empirical model

• Decomposition of employment growth into growth of
  employment due to production of the old products
  and the new products
• Leads to

14
Identification strategy

• If v is correlated to innovation, OLS estimates
  are inconsistent. Productivity is omitted and
  therefore v could be correlated to innovation.
  Innovations are the result of investment
  decisions (such as RD) and those decisions
  depend on the firms productivity. If
  productivity is in the error term (because it is
  an omitted variable) the error term will be
  correlated (hence endogeneity problem).
• Productivity can be thought as wit wi uit
• If the correlation is through time invariant
  characteristics wi (e.g. managerial
  capabilities), endogeneity is less important
  because the equation is in first differences
• If the correlation is through the time varying
  part of productivity (productivity shock), uit,
  timing in the relation innovation-productivity
  becomes crucial. If investments decisions are
  taken in advance innovations variables wont be
  related with the error term (OLS is consistent).
  If not, innovation outputs might become
  endogenous.

15
Identification strategy

• Another source of endogeneity is the presence of
  measurement error in g1 and g2.
• Ideally, we would use growth in real production
  but we only observe nominal output. Hence, the
  growth in prices (of both old and new) are left
  in the error term. Correlation between growth in
  prices and g2 can create an attenuation bias in
  the estimation of ß
• Hence (and since we generally do not have firm
  level prices), we use industry price indexes and
  a proxy for the growth in prices of old products.
  We use IV correlated with real growth in the
  production of new products but uncorrelated with
  its nominal growth.
• IV methods are used to correct these issues.
  Variety of instruments (but preferred is the
  increased range of products). 2 conditions
  partially correlated with product innovation but
  not correlated with the error term.

16
Descriptive statistics
17
Descriptive statistics (small firms)
18
Employment effects of types of innovation (OLS)
19
Employment effects of types of innovation (IV)
20
Employment effects by types of sectors (IV)
21
Implications

• Results show that the introduction of new
  products is associated with employment growth at
  the firm level, with similar efficiency between
  the production of old and new products. Hence, no
  evidence that product innovation displace
  employment (due to increased efficiency) being
  prevalent the creation effect of employment.
• Only in Uruguay (for all firms and in the high
  tech sectors), process innovations present
  displacement effects.
• In both Chile and Costa Rica, the compensation
  effects due to new products imply employment
  growth even when the replacement of old products
  is taken into account.

22
Robustness checks

• d as endogenous similar coefficients but with a
  general loss in precision.
• Overidentification by including further
  instruments weak instruments are not a concern.
• Allowing for change in the slope of product
  innovation if these innovations were introduced
  together with process innovation (interaction
  term between g2 and a dummy) generally, there is
  evidence that the positive impact on employment
  growth is weaker when products and process are
  introduced jointly.

23
Employment growth decomposition

• We decompose the employment growth observed in
  each country (and type of firm) over four
  different components. Using our preferred
  specification, we can write employment growth for
  each firm in the following way
• 1st component measures the change in its
  employment attributable to the (industry
  specific) productivity trend in production of old
  products
• 2nd component estimates the change in employment
  associated with the gross productivity effect of
  process innovation in the production of old
  products
• 3rd component corresponds to the employment
  change associated with output growth of old
  products for firms that do not introduce new
  products
• 4th component gives the net contribution of
  product innovation (i.e., contribution after
  allowing for any substitution of new products for
  old products).
• The last term is a zero-mean residual component.

24
Employment growth decomposition
25
Effects on employment quality two approaches

• Based on equation (1), we can split the growth
  rate of employment in both skilled (ls) and
  unskilled workers (lus). Therefore, we can study
  the impact of both process and product innovation
  on skilled and unskilled labor growth
• Once again we use instrumental variables as
  discussed before in order to address the
  identification problem related to correlation
  between d and g2 and the error term.
• In addition, and as robustness check, we also
  follow different variations of Berman, Bound and
  Griliches (1994) to estimate the relative demand
  of skilled labor

26
Descriptive statistics skill composition
27
Effects on skills
28
Effects on skill composition

• In Argentina there is weak evidence suggest that
  product innovations are more skill intensive (one
  side test. Similar for small firms), the opposite
  for small firms in Uruguay. In all other
  countries, there is no evidence of skill bias on
  product innovation.
• In Chile we find a negative effect of process
  innovation on unskilled employment, except in low
  tech sector. In Costa Rica and Uruguay there is
  evidence of displacement effect in the case of
  unskilled for the whole sample, but not for small
  firms. No biases from process in the case of
  Argentina.

29
Conclusions

• Relationship between innovation and employment is
  complex. Innovation could trigger different
  effects at different levels of aggregation and
  the relation depends on the transmission
  mechanisms.
• Idiosyncratic nature of innovation in the region
  means that the recent evidence in developed
  countries cannot be simply extrapolated. This
  project allow to fill the knowledge gap on the
  effects of innovation on employment (both
  quantity and quality).
• The evidence presented supports the idea that the
  negative performance in the labor market tend to
  be related to the lack of product innovators
  rather than to the introduction of innovations
  (either process or product).
• Considering that innovation positively affect
  employment generation, these results provide
  support for the current emphasis to the promotion
  of firm-level innovation.

30
(No Transcript)