Diapositiva 1

DECOMPOSING DIFFERENCES IN TOTAL
FACTOR PRODUCTIVITY ACROSS FIRM SIZE
Laia Castany, Enrique López-Bazo, Rosina Moreno
COINVEST Conference
Intangible Investments at Macro and Micro Levels and Their
Role in Innovation, Competitiveness and Growth
Lisbon, 18th-19th March 2010
2010
1. Motivation
•
Productivity is an issue of major concern. In the last years, the productivity growth has
slowed down in many advanced economies. [O'Mahoney & van Ark, 2003]
•
The Spanish economy has also suffered a deceleration process since the mid nineties [Gual et
al., 2006; Segura et al, 2006].
•
Spain should increase its competitiveness through efficiency to guarantee sustained growth 
higher investment in technological and human K (Intangible Assets) is required [National
Reform Program for Spain in the Lisbon Agenda]
•
Large firms are more productive [Bartelsman and Doms, 2000; Ruano, 2002]
•
Limitation Spain: predominance of SMEs (Large: 0.1% in Spain vs. 0.2% in EU-15; 20% vs.
35% of employment)
1. Motivation
•
This study contributes to the empirical evidence that the innovative activity and the use of skilled
labor foster firms’ productivity.
•
Firm size is an important source of heterogeneity in productivity across firms.
•
Explores differences in Total Factor Productivity across firm size (TFP-size gap) for the Spanish
manufacturing sector between 1990-2002.
Assessing the contribution of INN & HK (Intangible Assets)
in the TFP-size gap:
Differences in characteristics and returns.
ROADMAP
• Motivation
• Theoretical background
• Empirical specification
• Dataset
• Descriptive analysis
• Estimation
• Decomposing the TFP-size gap
• Conclusions
2. Theoretical background
2.1. Heterogeneity at microeconomic level:
• Empirical evidence on heterogeneity (in productivity, technology, entry-exit patterns, etc.)
across firms with similar characteristics.
• Size as a source of heterogeneity in productivity: large firms are systematically found to be
more productive [Bartelsman & Doms, 2000].
• Large firms have advantages:
– Scale economies effect, scope economies effect, experience effect or organization effect [Audretsch
et al., 1998)].
– At different production levels some technologies would be more appropriate than others.
– Industry effect.
• Small firms are “engines of growth” (importance for the economy as a whole) [Schumpeter;
Audretsch, 2002]:
–
–
–
–
Employment creators.
Innovators.
Entrepreneurship character.
Initiators, catalysts and media for wider technical change.
2. Theoretical background
2.2. Direct and Indirect effect of size (Geroski, 1998):
• Direct effect: Size, as a variable that ceteris paribus improves efficiency.
 size as a regressor, TFP=f(size, …)
• Indirect effect: Size, conditioning the effect of other variables on productivity.
 different coefficients of INN and HK for S&L firms are allowed.
TFPsize=f(HKsize, INNsize, …)
Intuition: Does a large firm get higher returns from an additional investment in HK?
 Large more productive.
 Small & large firms show different patterns of behaviour
2. Theoretical background
2.3. The relation between knowledge capital, TFP and size:
Innovation  productivity
Griliches (1979);
Mairesse & Sassenou (1991);
CDM (1998);
Huergo & Jaumandreu(2004a).
Large innovate more
Schumpeter (1942);
Acs et al (1994);
Huergo & Jaumandreu (2004b)
Large obtain higher returns to innovation
Klepper (1996);
Cohen & Klepper (1996);
Máñez et al. (2006);
Parisi et al (2002).
2. Theoretical background
2.3. The relation between knowledge capital, TFP and size:
Innovation  productivity
Griliches (1979);
Mairesse & Sassenou (1991);
CDM (1998);
Huergo & Jaumandreu(2004a).
Large innovate more
Schumpeter (1942);
Acs et al (1994);
Huergo & Jaumandreu (2004b)
Large obtain higher returns to innovation
Klepper (1996);
Cohen & Klepper (1996);
Máñez et al. (2006);
Parisi et al (2002).
Labour qualification  productivity
Becker (1964)
De la Fuente (2004);
Griliches & REgev (1995);
Haltiwanger et al (1999)
Better educated workers are employed in large
firms
Evans & Leighton (1989);
Zábojník and Bernhardt (2001)
Large firms pay higher wages because they
obtain higher returns from HK
Oosterbeek & van Praag(1995);
El-Attar & López-Bazo (2007)
2. Theoretical background
Two reasons for higher productivity in large firms:
(1)
higher investment in INN and HK (Intangible Assets)
(2)
higher returns from these investments  an additional INN or skilled worker in
a large firm could result in higher returns
Assessing the contribution of INN & HK in the TFP-size gap:
Differences in characteristics and returns.
3. Empirical specification

 


TFP


INN

HK

X'


u
it
0
1
it

1
2
it

1
it
i
it
where:
- TFP is the total factor productivity index in firm i in year t
- INN is a dummy variable that equals 1 when the firm has innovated (process) (*)
- HK is the % of qualified workers (*)
- X (controls):
SIZE is the number of workers
AGE is the number years since the constitution
IND is a set of 20 industrial dummies
YEAR is a time dummy
- μ Firm-specific effect (random effects)
(*) Possible endogeneity between TFP and INN & KH  lagged 1 period
Robustness analysis including: structure of ownership, market competition, economic cycle and regions.
3. Empirical specification
TFP Index: Good, Nadiri and Sickles (1996)
1
ln
TFP

(lnY

lnY
)

(S

S
)(lnX

lnX
)
it
it
t
it
t
it
t
2
t
t
1

(
lnY

lnY
)

(
S

S
)(
lnX

lnX
)

s
s

1 
s
s

1
s
s

1
2
s

2
s

2
where:
• Y = output; X = inputs L, K, M; S = cost-based share of inputs;
• subscripts i, t refer to firm and time period and the bar over the variables denotes
their arithmetic mean.
Permits avoiding problems of endogeneity related to the estimation of the input shares
 It accounts for technological change (lower part of the expression = cumulative change in Y,X
in the reference firms from the initial year to year t ).
 Transitive
 Superlative
 It accounts for market power
 Input shares are specific of every firm

4. Dataset: Encuesta sobre Estrategias Empresariales (ESEE)
ESEE = “Survey on Business Strategies”
- Annual survey, widely used studies on industrial organization in Spain.
- Unbalanced panel between 1990 and 2002.
- Reference population: manufacturing firms with 10 or more employees.
- Representative by size and industry.
- Small firms 10-200 employees; Large >200.
- Cleaning procedure: observations with incomplete data and anomalous
observations according to some criteria (following Ornaghi, 2006).
- After this: 13000 observations for 2100 firms, around 800-1000 observations per
year  TFP index.
- Periods of analysis: 1994 (852 firms); 1998 (968 firms); 2002 (864 firms)
5. Descriptive Analysis
LnTFP
years
Total sample
Small
Large
Test Eq Means
1990
-0.0682
-0.097
-0.0182
6.05***
1994
0.0149
-0.0095
0.0823
6.23***
1998
0.0747
0.0594
0.1314
4.07***
2002
0.1184
0.1054
0.1657
3.04***
• Confirm previous empirical evidence:
- TFP increases almost every year: 1.6% annual
- Slowdown in the 2nd half of 90’ (from 2.3% to 0.8%)
• Large firms are significantly more productive on every year  TFP-size gap.
• Differences/gap reduce, as TFP in small firms grew faster in the 2nd half of 90s.
5. Descriptive Analysis
TFP & INNOVATION
Process innovations (binary variable), but similar results using R&D expenditure.
Confirm previous empirical evidence pointing that:
- L innovate more (50% vs. 30%)  statistically significant differences.
- Innovative (INN) firms are significantly more productive.
•
INN contributes to reduce the TFP-size gap  key element for small firms to achieve higher
TFP levels.
•
INN makes a difference in TFP levels for S, but not for L  relationship between TFP-INN is
conditioned by SIZE.
lnTFP in 2002
Small
Large
Innovative
0.1364
0.1768
Test eq
means
1.45*
Non-innovative
0.0949
0.1553
2.33***
1.93**
0.63
5. Descriptive Analysis
TFP & HUMAN CAPITAL
Education (% of skilled workers).
Confirm previous empirical evidence pointing that:
- L use more HK (12% vs. 8%)  statistically significant differences
- Firms that use more HK are significantly more productive.
•
L are significantly more productive regardless of HK  HK does not seem to help small firms
reducing the gap  a possible explanation: L obtain higher returns from HK
•
The causality relationship between INN & HK – TFP might be questioned; Also, we are not
taking into account other factors explaining the gap; further evidence in what follows.
lnTFP in 2002
Small
Large
High % skilled
0.1444
0.1861
test eq
means
1.64**
Low % skilled
0.072
0.1285
1.84**
3.48***
1.70**
6. Estimation
lnTFP
OLS robust estimation
RE robust estimation
total
small
large
total
small
large
0.0652
0.0445
0.1652
0.0686
0.0726
0.1572
(0.0291)
(0.0352)
(0.0949)
(0.0374)
(0.0442)
(0.1156)
0.0344***
0.0355***
0.0374***
0.0190**
0.0132
0.0379***
(0.0092)
(0.0113)
(0.0158)
(0.0081)
(0.0101)
(0.0134)
0.2112***
0.1793***
0.3028***
0.1475***
0.1203***
0.2051**
(0.0458)
(0.0532)
(0.0897)
(0.0430)
(0.0489)
(0.0891)
0.0085**
0.0115*
0.0025
0.0090**
0.0063
-0.0032
(0.0035)
(0.0063)
(0.0133)
(0.0040)
(0.0070)
(0.0155)
0.0010***
0.0013***
0.0006*
0.0009***
0.0009***
0.0009**
(0.0002)
(0.0003)
(0.0004)
(0.0003)
(0.0004)
(0.0004)
Yes
Yes
Yes
Yes
Yes
Yes
19.89***
14.44***
53.34***
254.02***
185.27***
984.9***
nº observ
2684
2061
623
2684
2061
623
nº firms
-
-
-
1585
1211
415
0.2053
0.1868
0.2887
-
-
-
-
-
-
616.81***
476.95***
70.85***
cons
INN
HK
-
INN & HK  TFP
large firms have a greater
incentive to use KH & INN
-
Controls
SIZE
AGE
SECT
H0: SECt=0
2
R (adj)
H0: REt=0
slightly smaller
coefficients for RE model
-
sensitivity analysis:
including additional vbles
 size effect disappears,
but INN & HK remain
-
7. Decomposing the TFP-size gap
The Oaxaca-Blinder decomposition
 Studying the contribution of INN & HK in explaining the differential in TFP between
S&L firms (analysis in the mean of the distribution).
Oaxaca-Blinder methodology decomposes the contribution of each variable as:
- Differences in characteristics
- Differences in returns
• Departs from two auxiliary regressions for the subsample of S&L firms
• Decomposes the TFP differential in two components (Oaxaca and Ransom, 1994):
 

 




ˆ
ˆ
ˆ


TFP

TFP

X
'

X
'

X
'ˆ

X
'ˆ

L
S
L
S
L
L
S
S


















diff
endowments
diff
in
returns
where: X is the vector of firm endowments, β = returns
7. Decomposing the TFP-size gap
lnTFPL-lnTFPS
1994
1998
2002
0.0794
0.0766
0.0603
Total
Charact
97.7%
Returns
2.3%
Charact
104.3%
Returns
-4.3%
Charact
108.6%
Returns
-8.6%
INN
HK
11.4%
9.3%
0.5%
15%
9.5%
11.5%
0.3%
18.2%
13.1%
12.7%
0.4%
23.9%
INN & HK
20.6%
15.5%
21%
18.5%
25.8%
24.2%
• As a whole: firm’s characteristics explain almost completely the TFP differential
7. Decomposing the TFP-size gap
lnTFPL-lnTFPS
1994
1998
2002
0.0794
0.0766
0.0603
Total
Charact
97.7%
Returns
2.3%
Charact
104.3%
Returns
-4.3%
Charact
108.6%
Returns
-8.6%
INN
HK
11.4%
9.3%
0.5%
15%
9.5%
11.5%
0.3%
18.2%
13.1%
12.7%
0.4%
23.9%
INN & HK
20.6%
15.5%
21%
18.5%
25.8%
24.2%
• As a whole: firm’s characteristics explain almost completely the TFP differential
• INN & HK: explain quite an important part of the differential
– INN, modest contribution (10%, characteristics)
– HK, larger contribution (16-36%, endowment 1/3 + returns 2/3) & increasing
• Similar results for decomposition based on RE estimation  although smaller contribution of
INN & HK
8. Conclusions
1.
Large firms are more productive  TFP-size gap
1.
Investment in intangible assets contributes to increase firms’ productivity.
1.
Differences in HK & INN together explain quite a large part of the TFP-size gap.
1.
Increasing the levels of INN in small firms could improve their TFP
1.
But: an effort to increase HK requires higher returns to HK in small firms so that TFP
increases (small firms have less incentives to use HK)
1.
National Reform Program (Spain should increase its human capital levels)  Increasing
skilled workers in SMEs would only have a positive impact on productivity if returns
increased  Otherwise the effort would have a limited impact (in SMEs & total economy).
DECOMPOSING DIFFERENCES IN TOTAL
FACTOR PRODUCTIVITY ACROSS FIRM SIZE
Laia Castany, Enrique López-Bazo, Rosina Moreno
COINVEST Conference
Intangible Investments at Macro and Micro Levels and Their
Role in Innovation, Competitiveness and Growth
Lisbon, 18th-19th March 2010
2010