Transcript ex-ante volatility

Credit and
Productivity
Background material for
DIA 2009
Roadmap
•
•
Stylized Facts
Financial development and TFP
–
–
•
Macroeconomic volatility and TFP
–
–
•
Analytical discussion
Empirical evidence
Analytical discussion
Empirical evidence
• Ex post volatility
• Ex ante volatility
Final remarks
Roadmap
•
•
Stylized Facts
Financial development and TFP
–
–
•
Macroeconomic volatility and TFP
–
–
•
Analytical discussion
Empirical evidence
Analytical discussion
Empirical evidence
• Ex post volatility
• Ex ante volatility
Final remarks
Stylized Facts
Stylized Facts (cont.)
Correlation TFP and FD: LAC=0.80; Asia=0.54
Stylized Facts (cont.)
.01
.005
.04
TFP Growth and Rajan & Zingales - LAC
0
.5
Rajan & Zingales
1.5
Fitted values
0
(mean) gltfp2
1
.02
-.5
TFP Growth
-.005
0
TFP Growth
.015
.02
TFP Growth and Rajan & Zingales - United States
-.02
.04
TFP Growth and Rajan & Zingales - Asian Tigers
0
.5
Rajan & Zingales
0
.02
(mean) gltfp2
-.02
TFP Growth
-.5
-.5
0
.5
Rajan & Zingales
(mean) gltfp2
1
Fitted values
1.5
1
Fitted values
1.5
Stylized Facts (cont)
Volatility of GDP Growth
(relative to industrial countries)
4
Ratio of Standard Deviations of GDP growth rates
3.5
3
2.5
LAC
Asia
2
1.5
1
0.5
0
1970-1979
1980-1989
1990-1999
Period
2000-2006
all sample
Stylized Facts (cont.)
Volatility of TFP Growth
(relative to industrial countries)
3.5
3
2.5
2
LAC
ASIA
1.5
1
0.5
0
1970-1979
1980-1989
1990-1999
2000-2006
Full Sample
Stylized Facts (cont)
Volatility of the Real Effective Exchange Rate
(relative to industrial countries)
Ratio of standard deviations of real exchange rates
2.5
2
1.5
LAC
Asia
1
0.5
0
1980-1989
1990-1999
2000-2006
Period
all sample
Roadmap
•
•
Stylized Facts
Financial development and TFP
–
–
•
Macroeconomic volatility and TFP
–
–
•
Analytical discussion
Empirical evidence
Analytical discussion
Empirical evidence
• Ex post volatility
• Ex ante volatility
Final remarks
Financial Development and TFP
• Literature on financial development and TFP
growth goes as far as Bagehot (1873) and
Schumpeter (1912)
– Financial markets promote efficient capital
reallocation across productive units
• Hsieh and Klenow (2007), Restuccia and
Rogerson (2007), Buera and Shin (2008), Buera
et al (2008) point in a similar direction.
Financial Development and TFP
• Examples of the channels:
– Collateral constraints limit entrepreneurship
– Financial underdevelopment limits the possibility
of entering in highly productive sectors with high
fixed costs
– Credit market imperfections reduce long term
investments (prod. enhancing) vis a vis short term
ones
Financial Development and TFP
• Empirical evidence:
– Cross country is abundant (Beck et al, Levine and
Servos, Rioja and Valev, Acemoglu, Aghion and
Zilibotti, …).
– Sectoral level studies focus more on channels
(FD vs elasticity of investment to GDP, FD and
sensitivity of R&D expenditure to shocks, firm
growth) and less on the final impact (TFP).
Financial Development and TFP
• We add:
– Sectoral data: impact of credit availability on TFP
– Firm level data: Survey data + Colombia country
study
Sector Level Evidence: TFP Estimation
• Unido Dataset: panel 77 countries, 26
manufacturing sectors, annual data 1970-2003.
• Compute series of capital stock using the
perpetual inventory method. (Caselli 2005)
• Assume Cobb-Douglas technologies:

Yi ,c ,t  Ai ,c ,t K i ,c ,t Li ,c ,t

Sector Level Evidence: TFP Estimation (cont.)
• TFP: Regression residual
26
26
i 1
i 1
y i ,, c , t    i k i , c ,t    i l i , c ,t   i   c   i , c ,t
• TFP1: Fixed-cost shares
ltfp1i ,c ,t  y i ,c ,t  0.3k i ,c ,t  0.7li ,c ,t
• TFP2: Industry-specific cost shares (Fleiss 2008,
Bernanke and Gurkaynak, 2001)
ltfp2 i ,c ,t  y i ,c ,t   Ki k i ,c ,t  (1   Ki )li ,c ,t
Sector Level Evidence: TFP Estimation (cont.)
gltfp
gltfp1
gltfp2
ltfp
ltfp1
ltfp2
Growth Rates
gltfp
gltfp1
1
0.997
1
0.996
0.995
Levels
ltfp
ltfp1
1
0.568
1
0.358
0.556
gltfp2
1
ltfp2
1
Notes: ltfp is estimated by OLS, ltfp1 using fixed input shares
for the whole sample, and ltfp2 allows input shares to vary
per industry.
Sector Level Evidence: Methodology
• Estimation Equation (1)
gr _ TFPi ,c ,t  Sharei ,c ,t   RZ i  FDc ,t    i ,t  c ,t   i ,c ,t
• Estimation Equation (2)
gr _ TFPi ,c ,t   1 Sharei ,c ,t  1 RZ i  FDc ,t    2 FDc ,t  X c,t   i ,t   i ,c ,t
Sector Level Evidence: Results
Sector Level Evidence: Results (cont.)
Sector Level Evidence: Results (cont.)
Firm Level Evidence: WBES
• Using the WBES we construct measures of TFP for firms
in 54 developing countries (17 LAC)
• We construct three measures of TFP based on cost
shares and a prod function including labor, capital and
intermediate inputs.
– Cost shares are the same across countries and industries
(TFP)
– Cost shares are the same across countries but differ
across industries (TFPj)
– Cost shares differ across countries and industries (TFPij)
WBES TFP Estimations
All countries
2.70
2.60
2.50
2.40
2.30
2.20
2.10
2.00
1.90
LARGE
MEDIUM
TFP
Latin America
TFPj
SMALL
TFPij
3.70
3.50
3.30
3.10
2.90
2.70
2.50
2.30
2.10
1.90
LARGE
MEDIUM
TFP
TFPj
SMALL
TFPij
WBES: Access to credit
Size
% of WK
Access to financed
credit line with formal
or
lending
overdraft
sources
small >=5 and <=19
medium >=20 and <=99
large >=100
28.84
51.03
61.40
6.85
16.76
25.15
small >=5 and <=19
medium >=20 and <=99
large >=100
59.11
76.12
83.79
16.56
27.78
34.41
% of WK
financed
with trade
credit
All
11.70
12.26
10.60
LAC
15.02
16.35
14.88
% of WK
financed
with other
lending
sources
% of Inv
financed
with bank
lending
% of Inv
financed
with
C.Mkts.
16.29
13.71
9.89
9.16
22.06
30.48
0.68
1.53
1.89
4.25
4.84
3.74
15.23
10.97
10.80
10.54
8.11
5.58
16.04
23.64
33.26
1.00
1.77
1.88
7.95
8.05
7.10
10.62
6.40
3.05
% of Inv % of Inv
financed financed
with trade with other
credit
sources
Marginal impact of access to credit line
1.6
***
1.4
1.2
***
***
1
***
TFP
***
0.8
TFPj
***
**
0.6
**
TFPij
***
0.4
0.2
0
Large
Medium
Small
Regression: TFP(ij) = f(export, size, access, size*access, CI-FE)
Instrument by: past firm growth, share of firms with access in cluster,
share*size
Roadmap
•
•
Stylized Facts
Financial development and TFP
–
–
•
Macroeconomic volatility and TFP
–
–
•
Analytical discussion
Empirical evidence
Analytical discussion
Empirical evidence
• Ex post volatility
• Ex ante volatility
Final remarks
Macroeconomic Volatility and TFP – Some
Related Literature
Crisis, or ex-post volatility I:
•
•
Recent literature suggests a close connection between crisis
and TFP performance:
– Calvo et al (2006): Phoenix miracles: Collapse in TFP
performance
– Fernandez Arias et al (2007): TFP seldom recovers to trend
– Cerra-Saxena (2007): output does not recover to pre-crisis
trend levels
Periods of financial crisis are associated with large RER
depreciation and RER volatility:
– Calvo et al (2004): 63% of large RER depreciations in EMs
associated to Sudden Stops
– Calvo et al (2006): RER volatility (relative price of tradables visà-vis non-tradables) increases with Sudden Stops
Macroeconomic Volatility and TFP: sector level
evidence
Full Sample
(1)
gltfp
RZ * RER_Vol
Industry Share
Observations
R-squared
RZ * RER_Vol
(2)
gltfp1
-0.111**
-0.134***
[0.046]
[0.045]
-0.664***
-0.864***
[0.173]
[0.139]
10316
9840
0.425
0.429
Developing Countries
(1)
(2)
gltfp
gltfp1
(3)
gltfp2
-0.124***
[0.045]
-0.618***
[0.160]
10316
0.425
(3)
gltfp2
-0.110**
-0.127**
-0.123**
[0.049]
[0.049]
[0.048]
Industry Share
-0.648***
-0.837***
-0.600***
[0.183]
[0.153]
[0.170]
Observations
7663
7355
7663
R-squared
0.409
0.417
0.409
Notes: Robust standard errors in brackets *** p<0.01, ** p<0.05, * p<0.1. All
specifications include country-time and industry-time fixed effects.
Macroeconomic Volatility and TFP – Some
Related Literature
Crisis, or ex-post volatility II:
•
Connection between crisis and productivity through credit
markets:
– Caballero et al (1994): The now standard view of recessions: A
cleansing effect
– Barlevy (2003): If credit frictions exist, there could be
“uncleansing” effects
– Efficient but credit-constrained firms with loose connections to
credit markets (or little collateral) could be wiped out, leaving
larger but less efficient incumbents in the market
– This connection between macroeconomic volatility, credit
markets and TFP is what we are working on at the firm level for
the case of Colombia.
Macroeconomic Volatility and TFP – Some
Literature
•
Exposure to frequent crises and large RER fluctuations also raises
ex-ante volatility issues:
–
Calvo (2005): Greater price volatility increases the profitability of
more malleable, less productive technologies
–
Goldberg (2001): Exchange rate volatility affects the share of
foreign direct investment in total investment
–
We take from Calvo the idea that price volatility conspires
against the choice of more productive technologies, and from
Goldberg the idea that volatility affects the composition of
investment, and ask:
–
Can volatility affect the sectoral allocation of investment away
from what TFP differences would indicate?
Ex-ante Volatility – How does volatility introduce
distorsions in investment allocation?
Invijt
Inv jt
•
  1 jt   2it  
Tfpij ,t 1
Tfp j ,t 1
  j ,t 1
Tfpij ,t 1
Tfp j ,t 1
Where
Invijt
Inv jt
: Investment ratio, country j sector i over investment country j
Tfpij ,t 1
: Tfp ratio, country j sector i over Tfp country j.
Tfp j ,t 1
j
 1 jt ,  2it
: Measure of volatility in country j
: Country time and Industry time fixed effects
  ijt
Relationship Between Investment Ratio and
TFP ratio: An Example
Consider the case of a 1 period model in which a firm decides on investment in two
activities with different productivity levels
max   a1 f ( k1 )  pa2 f ( k2 )  rk1  rk2
k1 ,k2
F.O.C.:
a1 f ' (k1 )  r
p a2 f ' ( k 2 )  r
(1)
(2)
Consider f (ki ), homogeneou s of degree" n"; then :
f ' (k1 )k1  n f (k1 )
(3)
f ' (k 2 )k 2  n f (k 2 )
(4)
Relationship Between Investment Ratio and
TFP ratio: An Example
k1
Consider the ratio
k1  k2
(i.e, investment in activity 1
over total investment)
from (1), (2), (3) & (4) :
k1
a1 f (k1 )

, or
k1  k 2 a1 f (k1 )  p a2 f (k 2 )
k1
a1

k1  k 2 a1  p a2 f (k 2 ) f (k1 )
I. Pooled OLS
Dependent Variable: Investment Ratio
Tfp1
Tfp2
(1)
vol1
(2)
vol2
(3)
vol3
(4)
vol4
(5)
vol1
(6)
vol2
(7)
vol3
(8)
vol4
Tfpt-1
0.352
[0.226]
0.119
1.11***
[0.19]
0
0.438*
[0.23]
0.057
1.07***
[0.196]
0
0.173
[0.142]
0.223
0.679***
[0.145]
0
0.227
[0.146]
0.119
0.644***
[0.178]
0
Tfp*volatilityt-1
-1.32
[0.857]
0.123
-3.01***
[0.588]
0
-1.93
[0.858]
0.024
-2.92***
[0.633]
0
-0.382
[0.58]
0.511
-1.8***
[0.453]
0
-0.829
[0.582]
0.155
-1.71***
[0.472]
0
Observations
10430
10766
10239
10766
9991
10317
9807
10317
R-squared
0.512
0.519
0.511
0.518
0.531
0.534
0.53
0.534
Notes: Robust standard errors in brackets *** p<0.01, ** p<0.05, * p<0.1.. All specifications include industry-time and
country time fixed effects. vol1=ave~12, vol2=ave~f, vol3=end~12, vol4=end~f.
legend: b/se/p
II. IV-Panel
Note 1. According to Exogeneity test (C and Hansen J statistic) we can consider Tfp to be
exogenous, in each of the eight models.
Note 2. However, due to perpetual inventory methodology used to construct the capital series
which involves past Investment values, we could think if Investment ratios exhibit enough
persistence, it could influence the Tfp path, that is why we conduct an IV-Panel regression.
Tfp1
Tfp2
(1)
vol1
(2)
vol2
(3)
vol3
(4)
vol4
(5)
vol1
(6)
vol2
(7)
vol3
(8)
vol4
Tfp t-1
0.476**
[0.21]
0.023
1.07***
[0.21]
0
0.511**
[0.235]
0.029
1.08***
[0.204]
0
0.226*
[0.135]
0.095
0.669***
[0.138]
0
0.271*
[0.142]
0.057
0.663***
[0.137]
0
Tfp*volatility t-1
-2.21**
[0.943]
0.019
-2.88***
[0.453]
0
-2.42***
[0.804]
0.003
-2.95***
[0.47]
0
-0.734
[0.594]
0.217
-1.77***
[0.329]
0
-1.1**
[0.52]
0.034
-1.76***
[0.352]
0
Observations
9050
9382
8860
9382
8669
8991
8486
8991
Notes: Robust standard errors in brackets *** p<0.01, ** p<0.05, * p<0.1.. All specifications include industry-time and
country time fixed effects. vol1=ave~12, vol2=ave~f, vol3=end~12, vol4=end~f.
legend: b/se/p
Marginal Effect
(Different Volatility Levels)
Model (4)
Tfp1_end_f
2
0
-2
-4
0
.25
.5
.75
Volatility Value
Marginal Effect
Max
1
Min
1.25
IV. Dynamic Panel
Tfp1
Tfp2
(1)
vol1
(2)
vol2
(3)
vol3
(4)
vol4
(5)
vol1
(6)
vol2
(7)
vol3
(8)
vol4
Investment(-1)
-0.332*
[0.176]
0.059
-0.342*
[0.204]
0.094
-0.346*
[0.181]
0.058
-0.343*
[0.204]
0.094
-0.505***
[0.175]
0.004
-0.444**
[0.193]
0.021
-0.506***
[0.18]
0.005
-0.444**
[0.193]
0.021
Tfp
0.012
[0.062]
0.843
0.443***
[0.148]
0.003
0.008
[0.068]
0.908
0.465***
[0.176]
0.008
0.018
[0.057]
0.757
0.393***
[0.142]
0.006
0.069
[0.063]
0.272
0.434***
[0.160]
0.007
Tfp*volatility
0.599
[0.706]
0.397
-1.198***
[0.439]
0.006
0.263
[0.511]
0.604
-1.341**
[0.57]
0.019
0.431
[0.561]
0.443
-1.077**
[0.425]
0.011
-0.219
[0.440]
0.619
-1.27**
[0.521]
0.015
Observations
10430
10766
10239
10766
9991
10317
9807
10317
Notes: Robust standard errors in brackets *** p<0.01, ** p<0.05, * p<0.1.. All specifications include industry-time and
country time fixed effects. vol1=ave~12, vol2=ave~f, vol3=end~12, vol4=end~f.
legend: b/se/p
Systemic Sudden Stops: Total Factor Productivity in
EM Collapses & the US Great Depression
US Great Depression
Collapses in EM Economies
Collapse
Recovery
140
110
Collapse
125
Recovery
110
135
120
130
108
GDP
125
120
106
TFP
GDP
106
TFP
104
104
115
110
115
TFP
110
105
105
102
102
100
100
1936
1935
t+2
1934
t+1
1933
t
1932
t-1
95
1931
t-2
95
1930
100
1929
100
TFP
GDP
GDP
108
Case study: Colombia
• We use plant level data to estimate TFP and combine
this data with a sectoral level data base to identify
access to finance (firm level panel 1995 – 2005).
• Our general questions refer to:
– Relationship between productivity and crisis/volatility at the
firm level, analyzing the role played by credit constraints.
Does credit access help smooth shock?
– Relationship between entry-exit and firm productivity. Is
there a cleansing effect of crisis/volatility? Does credit play
a role in the way that crisis affect firms with different
productivities?
Case study: Colombia
• We estimate regressions of the sort to estimate
the impact of access to credit on productivity in
general and during crisis:
ait   i   t   o * CCit   it
ait   i   t   o * CCit  1 * CCit * crisis t   it
• We do this at a firm level and at a sectoral level
Case study: Colombia
• To estimate the impact of access to finance on
firm survival:
Pr(eit  1)   s   t   o * CC st   p * ait   2 * CC st * ait   it
• On entry at a sectoral level
eSt   s   t   o * CCst   1 * a St 1   2 * CCst1 * a St   St
eSt   s   t   o * CCst   2 * CCst * crisis t   3 * a St 1   St