investment and institutions stijn claessens, kenichi ueda, and yishay yafeh international monetary...
TRANSCRIPT
Investment and Institutions
Stijn Claessens, Kenichi Ueda, and Yishay YafehInternational Monetary Fund and University of Amsterdam, International Monetary Fund, andHebrew University
16th Dubrovnik Economic Conference, June 24, 2010
The views expressed in this paper are those of the authors and should not be attributed to the International Monetary Fund, its Executive Board or its management.
2
MotivationMotivation
Capital is not always allocated efficiently (Hsieh and
Klenow, 2009; Abiad, Oomes, and Ueda, 2008)
TFP is the most important factor for growth.
Does the institutional environment affect the cross-
country differences in investment efficiency, and, if so,
which institutions and how?
How should we judge the cross-country differences in
investment efficiency?
3
Our ConjectureOur Conjecture
Tobin’s Q is a measure of investment
efficiency.
Tobin’s Q should be 1 in a perfect world.
If not, it should approach 1 over time.
This adjustment may be slower in countries
with worse institutions.
4
Main ResultsMain Results
Good institutions can affect the investment
efficiency through two channels:
Lower required return allows a larger adjustment
in Q (from above) as the required capital gain (given
current profits) is less.
Less financial frictions create less divergence of
Tobin’s Q from its steady state to begin with,
implying slower adjustment in Q.
5
Main ResultsMain Results
Thus, Implication of efficient investment on overall
adjustment speed of Q is theoretically unclear.
We estimate the institutional effects through each of
two channels separately using a canonical model of
investment.
Regression results support beneficial effects of good
corporate governance, operating through both
channels, and high general institutional quality,
operating through financial friction channel.
6
Literature: Finance-Efficiency LinkLiterature: Finance-Efficiency Link
Efficiency of allocating capital across sectors has
been estimated using a variety of measures
GDP growth or TFP growth (e.g., Beck, Loayza, and
Levine, 2000; De Nicolo, Laeven, and Ueda, 2008)
Industry growth (e.g., Rajan and Zingales, 1998, and
Wurgler, 2000).
Dispersion of firm-level productivity (e.g., Hsieh and
Klenow, 2009, and Abiad, Oomes, and Ueda, 2008).
7
Literature: Tobin’s QLiterature: Tobin’s Q
Tobin's Q measures efficiency in the use of capital (Tobin, 1969)
Speed of investment adjustment relates to Q (Mussa, 1977).
Because of adjustment costs, investment and Tobin’s Q relate in
non-linear ways (Abel and Eberly, 1994) without financial
frictions.
With financial frictions, the sign of cash-flow sensitivity of
investment becomes difficult to predict (Gomes, 2000).
Moreover, market imperfections and varying discount factors
affect movements in Q (Abel and Eberly, 2008).
(Measurement error issues are discussed later.)
8
Literature: Measures of SystemsLiterature: Measures of Systems
Many measures have been developed/collected for
institutional and financial development (e.g.,
Demirguc-Kunt and Levine, 2001, Morck et al, 2000,
La Porta et al., 2008).
These allow comparisons of financial and governance
systems around the world.
9
ModelModel
Develop a canonical model of Tobin’s Q with both adjustment cost of investment and financial frictions (Abel and Blanchard (1986), Abel and Eberly (1994, 2008), Gomes (2000), and Hennessy et al. (2007)).
10
TimingTiming
Given K– and revealed current productivity ε at the
beginning of the current period:
Investment I is determined
Adjustment costs ϕ are wasted on investment
New capital K is formed and usable immediately
Using K, goods are produced with productivity ε
Fees λ paid for external financing (over-the-period)
11
ModelModel
( , ) ( , )t t t t t t tK f K L w L
1(1 ) tt tIK K
1 1, , ( , ) (1 ) ( , )t t t t tt t t t tB K K I K K K K ò
max , ) ( , ; , , )
( , ; ,
1( ; ) (
1
) ( ;, )
KI K X W
B K X
V
K
r
V
K K
EW
12
TimingTiming
First-order condition
Envelope condition
Combined together and use
1 1 1 1 11 2 2 E V
1 1 1(1 )(1 )r V
1Q V
1 1 2 2(1 (1 ))E Q r Q
13
Adjustment Speed of QAdjustment Speed of Q
Adjustment in Q depends on required return and frictions
Lower required return allows a larger adjustment in Q
(from above) as the required capital gain (given current
profits) is less.
Less financial frictions creates less divergence of Tobin’s Q
from its steady state to begin with, implying slower
adjustment in Q.
1 2 21) (1(1 )()
)(
E Qr
Q
Q
Q
14
Empirical Method: Minimize Forecast ErrorsEmpirical Method: Minimize Forecast Errors
Assume adjustment costs/financial frictions are functions of real
environment and firm characteristics X and institutional factors W:
We observe realized value of Q instead of expectation E[Q].
One-period-ahead forecast errors are not serially correlated.
Minimize mean squared errors: OLS is unbiased and consistent.
|Q E Q
1 2
1
2 3
2 3
1
[ | ]
( ) )* ( *
( * ) ( * ) .
E Q
Q Q Q
countryfe X W
X W
X Z W ZZ
1 1 1 2 2where Z
15
Estimation: Parameterization of CostsEstimation: Parameterization of Costs
1 2
2
3( , , )2
b Bb B b K K
KB K
2
32 2 2
a I
Ka
32
2
1( , , )2
aa I a KK
KK
II
11 3
Bb b
K
3
2
2 2 2
b Bb
K
16
Assumptions on Parameters and VariablesAssumptions on Parameters and Variables
Each coefficient (a2, a3, b1, b2, and b3) of
adjustment cost and financial friction functions
are assumed to be linear functions of real
environment and characteristics (X) and
institutional factors (W).
Make similar assumption for the required return
r, which constitute a coefficient c on lagged Q.
c(X, W)
17
Equation to be EstimatedEquation to be Estimated
, ,
, , , , , 1
, , , , ,
2 , , , , ,
2
, , , , , ,3 , ,
, , , 1, , , ,
1 1, , , ,
1, , , , ,, , , , , ,
,
,
,
2
,
,
,
,
1,
2
i j k t i
j k t k
j k t k i j k t
j k t k i j k t
j k t k i j k t
i j k t i j k tj k t k i j k t
i
j k t
i
j k t i j
j k t
i jt
j
tk
k
Q
a Q
b
b
B
X W
X W
X W
X W
X WB
bA A
c X
,
, ,3 , ,
, ,
, , ,
2
, ,
,
,,
, , ,
k t
i j kj k t
i j
i j k t
i j k t
i j k t
k t
IXc
A
18
Firm Level Data
Firm level data: Worldscope, 1990-2007, 48
countries.
Before tax income / After tax income
Capital investment / Capex + security investment
RZ external finance / Increase in debt + equity fin.
19
Macroeconomic and Firm Characteristics (X)
Firm Age: based on Founded Date.
Industry: SIC 2 digit.
Interest rate: Real short-term government rate
Inflation: CPI inflation
Macroeconomic Volatility
Standard deviation of real GDP growth 1995-2006
Coefficient of variation of the exchange rate
Standard deviation of inflation
20
Institutions (W)
Corporate Governance (Shareholder Protection)
Anti-director index (La Porta, et al., 1998; Spamann, 2009)
Self-dealing index (Djankov, et al., 2008)
CGQ index (De Nicolo, Laeven and Ueda, 2008)
Creditor Rights
Strength of Legal Rights (for creditors/borrowers) (Doing
Business)
Creditor Rights (Djankov, McLeish, and Shleifer, 2007)
Efficiency in Bankruptcy (World Economic Forum, 2004)
21
Institutions (W)
General Institutional Quality
Property rights (La Porta et al. 1998)
Rule of law (Kraay and Kaufman, 2000)
Trust in people (World Values Survey, 1990-93)
Competitiveness (of the product market)
Trade barriers (World Economic Forum, 2007)
Degree of new entry in business (World Development
Indicators, 2008)
Number of listed firms in the population (WDI)
22
Institutions (W)
Financial Market Development
Stock market-capitalization-to-GDP ratio
(International Financial Statistics)
Private credit (stocks + debt) to GDP ratio
(International Financial Statistics)
Lack of foreign ownership restrictions (World
Economic Forum)
23
Benchmark All TogetherRequired Return
(-) Fin. Friction Coeff. Ext. Fin.
Fin. Friction Coeff. Capital
(-) Fin. Friction Curvature
[1] [2] [3] [4]
Institutional Factors
Corporate Governance -0.0433 -0.0028 0.0200 0.0230
[-2.403]** [-1.778]* [2.639]*** [1.167]
Creditor Rights -0.0099 -0.0042 -0.0102 0.0399
[-0.454] [-1.119] [-1.673]* [1.148]
Institution -0.0007 0.0091 0.0639 -0.2282
[-0.016] [0.734] [3.683]*** [-1.750]*
Competitiveness 0.0772 0.0003 -0.0071 -0.0950
[1.864]* [0.045] [-0.423] [-0.858]
Financial Markets 0.0001 0.0000 0.0001 -0.0004
[0.357] [-0.167] [0.414] [-0.508]
24
One-by-One: Almost the same resultsa -b1 b2 -b3
Required Return (-) Fin. Friction Coeff. Ext. Fin.
Fin. Friction Coeff. Capital
(-) Fin. Friction Curvature
[1] [2] [3] [4]
Corporate Governance -0.0494 -0.0037 0.0222 0.0335
[-2.665]*** [-1.603] [2.964]*** [1.443]
Creditor Rights -0.0184 -0.0039 0.0077 0.0002
[-1.144] [-1.587] [1.340] [0.010]
Institution -0.0632 -0.0062 0.0535 -0.0794
[-1.534] [-0.893] [3.299]*** [-1.187]
Competitiveness 0.0858 0.0041 -0.0264 -0.0775
[2.154]** [0.737] [-1.814]* [-0.965]
Financial Market -0.0003 -0.0001 0.0002 0.0009
[-0.920] [-1.782]* [1.494] [1.684]*
25
Findings on Required Return
Only corporate governance significantly lowers the required return (in many specifications)
One std dev change in anti-director rights (1.3), mean Q goes down by 0.2 for average firm with Q=3.
Product market competition increases required return (though not so often)
Firm age very slightly increase required return.(As a firm becomes older and bigger, its returns comove more with market portfolio, reducing insurance premium.)
Other factors do not have robust effects.
26
Findings on Internal Fin Frictions (2)-(4)
Slope and curvature of costs associated with the size of
external finance is little affected by any institutional
factors.
Better general institutional quality sometimes worsen the
curvature but not robust.
27
Findings on Internal Fin Frictions (2)-(4)
Extra costs that small firms need to pay (small-firm
premium) are less in country with better corporate
governance and general institutional quality (column 3).
One std dev improvements in CG lowers the premium by
about 3 cents per dollar asset.
One std dev improvements in institutional quality lowers
premium by about 4 cents per dollar assets.
Other factors do not have robust effects on
financial frictions.
28
Real Adj. Cost of Investment
Real adjustment cost of investment are not affected by X.
How about institutional factors W?
Entrenchment of managers under private information (Myers and Majluf
(1984) and workers’ sabotage (Parente and Prescott, 2000)
We find:
Better corporate governance and general institutional quality reduce
technological/managerial diseconomy of scale.
But somewhat offset by increased curvature.
Without coeff on investment, overall effect is unidentified.
Other institutional factors do not have significant effects.
All effects on financial frictions and required returns are unchanged.
29
Real Adj. Cost of Investment – CG lowers
Required Return (-) Fin. Friction Coeff. Ext. Fin.
Fin. Friction Coeff. Capital
(-) Fin. Friction Curvature
Inv. Adj. Cost Coeff. Capital
(-) Inv. Adj. Cost Curvature
[1] [2] [3] [4] [5] [6]
Institutional Factors
Corporate Governance -0.0424** -0.0027* 0.0249*** 0.0220 -0.1738*** -0.9204**
[-2.346] [-1.759] [3.193] [1.117] [-3.300] [-2.060]
Creditor Rights -0.0102 -0.0042 -0.0100 0.0411 0.0324 -0.0422
[-0.465] [-1.142] [-1.571] [1.187] [0.503] [-0.185]
Institution 0.0010 0.0094 0.0638*** -0.2332* -0.2343* -0.1380
[0.023] [0.761] [3.584] [-1.786] [-1.663] [-0.238]
Competitiveness 0.0782* 0.0005 -0.0013 -0.1008 0.1335 -0.9498
[1.885] [0.076] [-0.074] [-0.903] [0.852] [-1.633]
Financial Markets 0.0001 0.0000 0.0000 -0.0003 0.0029 0.0063
[0.356] [-0.158] [0.252] [-0.498] [1.291] [0.586]
30
Measurement Error IssuesMeasurement Error Issues
Stock Price Movements may not always reflect
fundamental values:
Abel and Blanchard (1986), Blanchard, Rhee, and
Summers (1994), Phillippon (2009) – need long time
series
Accounting Issues:
Difference between marginal and average Q Hayashi’s
(1982) assumptions make them the same; we also allow
for industry and age specific effects.
Blanchard, Rhee, and Summers (1994): market valuation
in debt and the replacement cost of capital—need long T.
31
Measurement Error IssuesMeasurement Error Issues
Different Timing:
The productivity shock may not be revealed at the end of
last period. If so, we observe E[Q –] instead of Q –.
The problem is similar to the case with noisy stock price.
These three measurement errors may be big or small.
Need to test.
32
Testing Measurement ErrorsTesting Measurement Errors
With measurement errors, the OLS errors are serially
correlated (even if measurement errors are not).
Cannot reject Ho (zero autocorrelation) in OLS errors.
* * 'OLS OLSu X W ' ' '( * ) '( * )OLS OLS OLSE u u E v v E v X v E v W v
33
IV EstimationIV Estimation
Measurement errors turns out small, if any, relative
to one-period-ahead forecast errors.
Still, to check robustness, we run IV estimation.
Note that, if any, there is little autocorrelation in
measurement errors. Large swings in stock prices is
likely to dominate the other sources.
We use twice-lagged Q as an instruments for lagged
Q (and fitted cross-terms as IV for cross-terms).
TSLS results are similar to the OLS-FE estimation.
34
IV estimation – size and curvature effects Required Return
(-) Fin. Friction Coeff. Ext. Fin.
Fin. Friction Coeff. Capital
(-) Fin. Friction Curvature
[1] [2] [3] [4]
Institutional Factors
Corporate Governance -0.0209 -0.0022 0.0164 0.0361
[-1.373] [-1.473] [2.321]** [2.202]**
Creditor Rights -0.0120 -0.0009 -0.0040 0.0221
[-0.558] [-0.205] [-0.623] [0.530]
Institution -0.0113 0.0022 0.0578 -0.2333
[-0.254] [0.129] [3.440]*** [-1.291]
Competitiveness 0.0811 0.0015 -0.0215 -0.1098
[1.968]** [0.162] [-1.501] [-0.782]
Financial Markets 0.0001 0.0000 0.0001 -0.0015
[0.286] [0.713] [0.390] [-1.789]*
35
ConclusionConclusion
We have investigated how institutional environment
affect investment efficiency.
Good corporate governance and general institutional
quality, though less robust, are the main driving forces
to lower financial frictions, in particular, the small-firm
premium.
Also, better corporate governance lowers the required
return in many specifications.
36
ConclusionConclusion
Why is corporate governance, not creditor rights,
important? Our interpretation:
At the margin, the cost of equity finance determines the cost
of borrowing.
Cost of external finance implicitly measures investors’ fear
on mismanagement of injected cash.
37
AppendixAppendix
Suppose the firm-level shock following cdf F can be
decomposed into the aggregate shock following cdf G,
industry-specific shock following cdf H, and idiosyncratic
shock. Three components are assumed to be orthogonal each
other.
Firm managers can figure out overall shock ε when they make
investment decision, but cannot know the size of each
component.
38
AppendixAppendix
max , ; , ) , ) ( , ; , , )
( , ; , , ) ( )
max , ; , ) ,
( ; , ,
) ( , ; , , )
( , ; , , ) ( )
ma
)
( (
; , ,
( (
; , ,
1(
1 [ ],x , ) ( , ; , , )
( ;
,
, ,
K
K
K
X W I K X W
B K X W V K dF dGdH
X W dGdH I K X W
B K X W V K d
V K X W
m K
X W
m K
X W
K
F
I K X Wr E X W
B K X
; ,, ) ( ) .,X WW V K dF