Ec 2390: Notes Section 3

Jack Willis∗

November 16, 2015

1 Hsieh and Klenow (2009): Misallocation and Manufacturing TFP

in China and India

• Hsieh and Klenow use micro-data on manufacturing establishments to quantify the potential extent

of misallocation in China and India versus the United States.

• They measure sizable gaps in marginal products of labor and capital across plants within narrowly

defined industries

• What are the factors causing misallocation?

– The McKinsey Global Institute (1998) argues that a key factor behind low productivity in

Brazil’s retail sector is labor-market regulations driving up the cost of labor for supermarkets

relative to informal retailers.

– Political connections & access to subsidized credit

• Goal of the paper: Provide quantitative evidence of resource misallocation on aggregate TFP

– H&K measure how much aggregate manufacturing output in China and India could increase if

capital and labor were reallocated to equalize marginal products within each four-digit sector

to the extent observed in the United States.

∗
These notes follow closely the notes of Frank Schilbach, TA for the class Spring 2011 & 2012.

1
– The United States is a critical benchmark because may be measurement error and factors

omitted from the model (such as adjustment costs and markup variation)

2
• Moving to US efficiency would increase TFP by 30 to 50% in China and 40 to 60% in India.

– Output gains twice as large if capital accumulated in response to aggregate TFP gains.

– In both India and China, larger plants within industries appear to have higher marginal prod-

ucts

– Pattern is much weaker in the United States.

1.1 The model

• Standard model of monopolistic competition with heterogeneous firms (Melitz (2003) without trade)

– Key theoretical result: revenue productivity (the product of physical productivity and a firm’s

output price) should be equated across firms in the absence of distortions.

1.1.1 Final good sector

– Single final good Y is produced by a representative firm in a perfectly competitive final output

market.

– Cobb-Douglas production technology:

S
Y S
X
Y = Ysθs , θs = 1 (1)
s=1 s=1

3
– The final good producer maximizes

S
X S
Y S
X
max P Y − P s Ys = Ysθs − Ps Ys (2)
Ys
s=1 s=1 s=1

and the first order condition associated with this problem is

S
θs Y θs
Ys = Ps ⇒ Ps Ys = θs P Y. (3)
Ys
s=1

QS  θs
Ps
where PS refers to the price of industry output Ys and P = s=1 θs represents the price

of the final good (which is normalized to 1).

4
1.1.2 Intermediate goods sector (s)

– Each intermediate s is produced using Ms further intermediates:

" # σ
X σ−1 σ−1
σ
Ys = Ysi . (4)
Ms

– The demand for these intermediates is given by maximizing

σ
Ms Ms
" #
X X σ−1 σ−1 X
max Ps Ys − Psi Ysi = Ps Ysiσ
− Psi Ysi (5)
Ysi
i=1 Ms i=1

and the associated first order condition is

" # σ −1
σ X σ−1 σ−1 σ − 1 σ−1 −1
Ps Ysi σ Ysi σ = Psi (6)
σ−1 σ
Ms
1
− σ1
⇒ Ps Ysσ Ysi = Psi (7)
1 σ−1
⇒ Ps Ysσ Ysi σ = Psi Ysi (8)

5
• How is Ps determined?

– the cost minimization problem reads as follows:

σ
Ms Ms
! σ−1
X X σ−1
σ
Ps Ys = min Psi Ysi s.t. Ys = Ysi . (9)
Ysi
i=1 i=1

– Rewrite the constraint and set up the Lagrangian

Ms Ms
!
X σ−1 X σ−1
Ls = Psi Ysi + λs Ys σ
− Ysi σ
,
i=1 i=1

where λs is the multiplier on the constraint.

– Taking first-order conditions yields

∂Ls σ − 1 − σ1
= Psi − λs Ysi = 0 (10)
∂Ysi σ
σ − 1 − σ1
⇒ Psi = λs Ysi . (11)
σ

This shows that total costs are given by

Ms Ms s M
X X σ − 1 − σ1 σ−1X σ−1 σ − 1 σ−1
Psi Ysi = λs Ysi Ysi = λs Ysi σ = λs Ys σ (12)
σ σ σ
i=1 i=1 i=1

and that the demand function for Ysi is given by

 σ−1
σ−1 1 σ−1
Ysi σ = λs . (13)
Psi σ

6
– We are now going to solve for the multiplier λs . The demand function (13) implies

σ σ
Ms Ms 
! σ−1 ! σ−1
σ−1 σ 1 σ−1
X σ−1
  X 
σ
Ys = Ysi = λs (14)
σ Psi
i=1 i=1

This implies that the multiplier λs is given by

−1
Ms 
! σ−1
1 σ−1

σ 1 X
λs = Ysσ . (15)
σ−1 Psi
i=1

Plugging this in (12) yields

Ms
X σ − 1 σ−1
Psi Ysi = λs Ys σ (16)
σ
i=1
1
!− σ−1
Ms 
1 σ−1
1 X  σ−1
σ σ
= Ys Ys
Psi
i=1
1
Ms
! 1−σ
X
= Ys Psi 1−σ
i=1

So, finally we arrive at Ps , the price index to buy the composite good Ys :

1
Ms
! 1−σ
X
Ps = Psi1−σ . (17)
i=1

More specifically: if the M individual goods have prices [Psi ]M

i=1 , then one unit of the composite
  σ
PMs σ−1 σ
σ−1
good Ys = i=1 Ysi costs Ps , if each individual variety Ysi is bought in the efficient

proportions. Ps is sometimes referred to as the ideal price index and in many models Ps is

used as the numeraire.

7
1.1.3 Further intermediate goods sector (si)

• Each (further) intermediate is made using Cobb-Douglas function of firm TFP, capital, and labor:

αs 1−αs
Ysi = Asi Ksi Lsi (18)

• There are two distortions:

– markup on the cost of capital τKsi

– tax on the price of the good τY si

• Hence, profits of each intermediate producer are given by

πsi = (1 − τY si )Psi Ysi − wLsi − (1 + τKsi )RKsi (19)

and each intermediate producer i in sector s maximizes:

max (1 − τY si )Psi Ysi − wLsi − (1 + τKsi )RKsi (20)

Lsi ,Ksi
1 σ−1
=(1 − τY si )Ps Ysσ Ysi σ Asi − wLsi − (1 + τKsi )RKsi (21)
1
σ−1
αs 1−αs σ
=(1 − τY si )Ps Ysσ Asi Ksi Lsi Asi − wLsi − (1 + τKsi )RKsi

• Taking first order conditions with respect to Ksi and Lsi yields

Ksi αs
   
1 σ−1 − σ1
(1 − τY si )Ps Ys σ
Ysi Asi (1 − αs ) =w (22)
σ Lsi
Ksi αs −1
   
1 σ−1 − σ1
(1 − τY si )Ps Ysσ
Ysi Asi αs = (1 + τKsi )R (23)
σ Lsi
Ksi αs w
⇒ = (24)
Lsi 1 − αs (1 + τKsi )R

• Plugging the last equation into the first FOC (22) and using the equation for Psi from above (7),

8
we get

   αs
1 σ−1 − σ1 αs w
(1 − τY si )Ps Ys
σ
Ysi Asi (1 − αs ) =w
σ 1 − αs (1 + τKsi )R
 αs  1−αs
σ R w (1 + τKsi )αs
⇒ Psi = (25)
σ − 1 αs 1 − αs Asi (1 − τY si )

This is the standard condition that the firm’s output price is a fixed markup over its marginal cost.

9
• Why does the expression on the RHS of (25 equal marginal cost?

– Using (18) and (24), we can write Lsi as a function of Ysi :

 αs
αs w
Ysi = Asi Lsi L1−α
si
s
(26)
1 − αs (1 + τKsi )R
  αs
αs w
= Asi Lsi (27)
1 − αs (1 + τKsi )R
  αs
1 1 − αs (1 + τKsi )R
⇒ Lsi = Ysi (28)
Asi αs w

– Since wage payments Lsi w are a constant share (1 − α) of total cost, marginal cost is given by

∂Lsi
(1 − α)M Csi = w
∂Ysi
1 − αs (1 + τKsi )R αs
 
1 1
⇒ M Csi = w
(1 − α) Asi αs w
 αs  1−αs
R w (1 + τKsi )αs
= ,
αs 1 − αs Asi (1 − τY si )

which is the expression above.

• Here is another (slightly more elegant) to see this. Since the production function is homogeneous

of degree one, the firm maximizes:

1 σ−1
max Psi Ysi − ACsi Ysi = Ps Ysσ Ysi σ − ACsi Ysi (29)
Ysi

where ACsi is the cost of producing one unit of Ysi . Note that this unit cost does not depend on

the scale of production Ysi .

• The first order condition implies

σ−1 1 −1
ACsi = Ps Ysσ Ysiσ (30)
σ

10
• Again due to homogeneity of degree one in the production function, we have (using (8)):

1 σ−1
σ−1 σ σ σ−1
σ−1 σ Ps Ys Ysi σ Psi Ysi σ−1
1 −1
M Csi = ACsi = Ps Ysσ Ysiσ = = = Psi
σ Ysi Ysi σ
σ
⇒ Psi = M Csi
σ−1

This implies the familiar fixed markup over marginal cost.

11
1.2 Marginal revenue products, TFPR and TFPQ

• Rewriting the two first order conditions (22) and (23), we can se that the marginal revenue products

of labor and capital1 are proportional to the revenue per worker and the revenue-capital ratio,

respectively:

σ − 1 Psi Ysi 1
M RP Lsi = (1 − αs ) =w (31)
σ Lsi 1 − τY si
σ − 1 Psi Ysi 1 + τKsi
M RP Ksi = αs =R (32)
σ Ksi 1 − τY si

Firms equalize the after-tax marginal revenue products of capital and labor. Hence, the before-tax

marginal revenue products are dispersed to the distortions.

• Hsieh and Klenow define their productivity measures as follows2 :

Ysi
T F P Qsi = Asi = αs 1−αs (33)
Ksi Lsi
Psi Ysi
T F P Rsi = Psi Asi = αs 1−αs (34)
Ksi Lsi

While TFPQ captures “physical productivity”, TFPR measures “revenue productivity”.

• This distinction is important since measuring the former requires firm-specific prices, while the latter

can be calculated using sector-specific deflators.

• Furthermore, in Hsieh and Klenow’s model, TFPR does not vary across plants within an industry

unless plants face capital and/or output distortions.

• High plant TFPR sign that the plant confronts barriers that raise the plant’s marginal products of

capital and labor, rendering the plant smaller than optimal.

• More capital and labor should be allocated to plants with higher TFPR to the point at which their

higher output results in a lower price and the exact same TFPR as at smaller plants.
1
The marginal revenue products are given by the products of (physical) marginal products and marginal revenue. Here,
they are just the derivatives of revenue net of taxes, (1 − τY si )Psi Ysi , with respect to Ksi and Lsi , respectively. They tell us
the marginal increase in (net) revenue if we increase labor or capital.
2
Note that there are a couple of typos and algebraic errors in the paper (among them the definitions of TFPQ and TFPR).
You can find a document with corrections here or on the course website.

12
 • Using (31) and (32), we can show that plant TFPR is proportional to a geometric average of the

plant’s marginal revenue products of capital and labor:

σ − 1 Psi Ysi αs σ − 1 Psi Ysi 1−αs

   
αs 1−αs
(M P RKsi ) (M P RLsi ) = αs (1 − αs )
σ Ksi σ Lsi
σ − 1 P Y
si si
= αsαs (1 − αs )1−αs αs 1−αs
| {z σ K Lsi
} | si {z }
sector level constant
T F P Rsi

1−αs αs (1 + τKsi )αs

= | {z R }
w
1 − τY si
sector level constant
(1 + τKsi )αs
⇒ T F P Rsi ∝ (M P RKsi )αs (M P RLsi )1−αs ∝
1 − τY si

• Notice when the τY si and τKsi are all zero, TFPR must be the same for all firms even though TFPQ

obviously differs. This is because the more productive firm faces a lower price exactly in proportion

to its higher productivity.

• Algebra leads to the following expression of GDP3 :

S
Y
θs
Y = TFPs Ksαs L1−α
s
s
(35)
s=1

where

1
"M 
s σ−1 # σ−1
X TFPRs
TFPs = Asi (36)
TFPRsi
i=1

where TFPRs is a geometric average of the average marginal revenue product of capital and labor

in the sector.

3
A sketch of this is given in the correction appendix. Note that equation (16) of the paper is wrong. However, Hsieh and
Klenow use their equation (15) for their empirical estimates, so this does not affect their results.

13
 • Comments:

– If marginal products were equalized across plants, TFP would be

1
Ms
! σ−1
X
As = Aσ−1
si . (37)
i=1

1
"M 
s σ−1 # σ−1
X TFPRs
TFPs = Asi (38)
TFPRsi
i=1

– If σ > 2, more variance of Asi increases TFP. Why4 ?

∗ Holding constant distortions, firms with high Asi have lower marginal cost (see (25)).

∗ Since there is a constant markup over marginal cost, they also charge (proportionately)

lower prices.

∗ Hence, their output will be higher, and they get more weight in the calculation of TFP

than the firms with lower Asi .

∗ The higher the elasticity of substitution σ, the larger will be the output response to price

differences, and hence, the higher TFP for a given distribution of Asi .

– If σ > 2, more variance of TFPRsi decreases TFP. Why?

∗ Holding Asi constant, a firm with higher TFPRsi has higher marginal cost (see (25)) and,

hence, (proportionately) higher prices.

∗ This will induce the firm to produce less than in the absence of distortions.

∗ If TFPRsi and Asi are positively correlated, then the distortions render firms with high

physical productivity (high Asi ) smaller than optimal, which hurts aggregate TFP (since

those firms get less weight). This is the case in Hsieh and Klenow’s data.

∗ Note that the correlation between TFPRsi and Asi matters.

4
Note that xσ−1 is (strictly) concave for σ < 2 and (strictly) convex for σ > 2 such that Jensen’s inequality tells us

1
that an increase in the variance of x (holding the mean constant) increases E[xσ−1 ] (and, hence E[xσ−1 ] σ−1 ) if σ > 2 and
decreases the E[xσ−1 ] if σ < 2. Since Hsieh and Klenow assume σ = 3 in their analysis (and σ = 5 in their robustness check),
the relevant case to consider here is σ > 2.

14
1.3 How can we bring this to the data?

Psi Ysi Ysi

• We can measure TFPRsi = Psi Asi = αs 1−αs from the data, but how do we get Asi = αs 1−αs ?
Ksi Lsi Ksi Lsi

– Issues: We don’t know Psi , so cannot back out Ysi without further assumptions (for more on

this, see De Loecker (2011)5 ).

– Use the fact that (see above)

1 σ−1
Psi Ysi = Ps (Ys ) σ Ysi σ (39)

to write

 1
 −σ σ
1−σ
Ysi = Ps (Ys ) σ (Psi Ysi ) σ−1 . (40)

Hence, we have

σ
(Psi Ysi ) σ−1
Asi = κs αs 1−αs (41)
Ksi Lsi

1
where κs = (Ps Ys )− σ−1 /Ps is a sector-specific constant6 .

– We can ignore κs for any intra-industry reallocation exercise, since it is a constant that does

not affect the variances or covariances of log(Asi ). This is because Var(log(aX)) = Var(log(X))

if a is a constant:

Var(log(aX)) = Var(log(a) + log(X)) = Var(log(a)) + Var(log(X)) = Var(log(X))

5
For example, see Product Differentiation, Multi-Product Firms and Estimating the Impact of Trade Liberalization on
Productivity or Recovering Markups from Production Data
6
Note that the expression for κs below equation (19) in the paper is incorrect.

15
1.4 Results

• Figure I plots the distribution of TFPQ for the latest year in each country: India in 1994, China in

2005, and the United States in 1997. There is more TFPQ dispersion in India than in China, but

this could reflect the different sampling frames (small private plants are under-represented in the

Chinese survey). The US and Indian samples are more comparable.

MISALLOCATION AND TFP IN CHINA AND INDIA 1417

India
0.3

0.2

0.1

0
1/256 1/64 1/16 1/4 1 4

China
0.3

0.2

0.1

0
1/256 1/64 1/16 1/4 1 4

United States
0.3

0.2

0.1

0
1/256 1/64 1/16 1/4 1 4
FIGURE I
Distribution of TFPQ

the United States. Table II provides TFPR dispersion statistics for

a number of country-years. The ratio of 75th to 25th percentiles
of TFPR in the latest year are 2.2 in India, 2.3 in China, and
1.7 in the United States. The ratios of 90th to 10th percentiles of
TFPR are 5.0 in India, 4.9 in China, and 3.3 in the United States.
These numbers are consistent with greater distortions in China
and India than the United States.15

15. Hallward-Driemeier, Iarossi, and Sokoloff (2002) similarly report more

TFP variation across plants in poorer East Asian nations (Indonesia and the
Philippines vs. Thailand, Malaysia, and South Korea).

16
• Figure II plots the distribution of TFPR for the last year in each country. There is clearly more

dispersion of TFPR in India than in the US. Even China, not fully sampling small private estab-

lishments, exhibits notably greater TFPR dispersion than the US.

MISALLOCATION AND TFP IN CHINA AND INDIA 1419

India
0.6

0.4

0.2

0
1/8 1/4 1/2 1 2 4 8

China
0.6

0.4

0.2

0
1/8 1/4 1/2 1 2 4 8

United States
0.6

0.4

0.2

0
1/8 1/4 1/2 1 2 4 8
FIGURE II
Distribution of TFPR

For India and China, Table III gives the cumulative percent-
age of the variance of TFPR (within industry-years) explained by
dummies for ownership (state ownership categories), age (quar-
tiles), size (quartiles), and region (provinces or states). The results
are pooled for all years, and are cumulative in that “age” includes
dummies for both ownership and age, and so on. Ownership is less
important for India (around 0.6% of the variance) than in China
(over 5%). All four sets of dummies together account for less than
5% of the variance of TFPR in India and 10% of the variance of
TFPR in China.

17
1420 QUARTERLY JOURNAL OF ECONOMICS

TABLE III
PERCENT SOURCES OF TFPR VARIATION WITHIN INDUSTRIES

Ownership Age Size Region

India 0.58 1.33 3.85 4.71

China 5.25 6.23 8.44 10.01

Notes. Entries are the cumulative percent of within-industry TFPR variance explained by dummies for
ownership (state ownership categories), age (quartiles), size (quartiles), and region (provinces or states). The
results are cumulative in that “age” includes dummies for both ownership and age, and so on.

Although it does not fit well into our monopolistically compet-

itive framework, it is useful to ask how government-guaranteed
monopoly power might show up in our measures of TFPQ and
TFPR. Plants that charge high markups should evince higher
TFPR levels. If they are also protected from entry of nearby com-
petitors, they may also exhibit high TFPQ levels. Whereas we
frame high TFPR plants as being held back by policy distortions,
such plants may in fact be happily restricting their output. Still,
such variation in TFPR is socially inefficient, and aggregate TFP
would be higher if such plants expanded their output.
We next calculate “efficient” output in each country so we
can compare it with actual output levels. If marginal products
were equalized across plants in a given industry, then industry
Aσsi−1 ) σ18
Ms 1
TFP would be Ās = ( i=1 −1 . For each industry, we calculate

the ratio of actual TFP (15) to this efficient level of TFP, and
then aggregate this ratio across sectors using our Cobb-Douglas
aggregator (1):
 2nd quartile 7.3 5.9 5.3 6.6
3rd quartile 8.5 6.0 5.2 5.4
Bottom quartile 10.5 5.9 4.5 4.2
India 1994 0–50 50–100 100–200 200+

• Gains from
Topreallocation
size quartile (elimination
8.7of distortions):
4.7 if marginal
4.6products were
7.1 equalized across
2nd quartile 10.7 4.6 4.1 5.7
plants in 3rd
a given industry, then industry
quartile 11.4 TFP would
5.0 be 4.0 4.7
Bottom quartile 13.8 3.9 3.3 3.8
1
United States 1997 0–50 Ms 50–100! σ−1 100–200 200+
X
Top size quartile 4.4As = Aσ−1
10.0
si . 6.7 3.9 (42)
2nd quartile 4.4 i=1 9.6 5.8 5.1
3rd quartile 4.5 9.8 5.4 5.4
Bottom quartile 4.7 12.0 4.3 4.1
• For each industry, H&K calculate the ratio of actual TFP to this efficient level of TFP, and then
Notes. In each country-year, plants are put into quartiles based on their actual value-added, with an equal
aggregatenumber
this ratio using the Cobb-Douglas aggregator. Table VI shows the resulting TFP (from
of plants in each quartile. The hypothetically efficient level of each plant’s output is then calculated,
assuming distortions are removed so that TFPR levels are equalized within industries. The entries above show
eliminating distortions)
the percent relative
of plants with to theoutput
efficient/actual US. levels in the four bins 0%–50% (efficient output less than
half actual output), 50%–100%, 100%–200%, and 200%+ (efficient output more than double actual output).
The rows add up to 25%, and the rows and columns together to 100%.

TABLE VI
TFP GAINS FROM EQUALIZING TFPR RELATIVE TO 1997 U.S. GAINS

China 1998 2001 2005

% 50.5 37.0 30.5

India 1987 1991 1994

% 40.2 41.4 59.2

S  Ms Asi
Notes. For each country-year, we calculated Yefficient /Y using Y/Yefficient = s=1 i=1 ( As
TFPRs σ −1 θs /(σ −1) P Y
TFPR
) and TFPRsi ≡ αs si si 1−α .
si Ksi (wsi Lsi ) s
We then took the ratio of Yefficient /Y to the U.S. ratio in 1997, subtracted 1, and multiplied by 100 to
yield the entries above.

In Table VI we report the percent TFP gains in China and

India relative to those in the United States in 1997 (a conserva-
tive point of comparison because U.S. gains are largest in 1997).
For China, hypothetically moving to “U.S. efficiency” might have
boosted TFP by 50% in 1998, 37% in 2001, and 30% in 2005. Com-
pared to the 1997 U.S. benchmark, Chinese allocative efficiency
improved 15% (1.5/1.3) from 1998 to 2005, or 2.0% per year. For

19
 • Figure III plots the “efficient” vs. actual size distribution of plants in the latest year.
1422 QUARTERLY JOURNAL OF ECONOMICS

China
0.25

0.2 Actual

0.15

0.1

0.05 Efficient

64
1/64

512
1/512

8
1/8
India
0.25

0.2

0.15 Actual

0.1

0.05 Efficient

64
1/64

512
1/512

8
1/8

0.25 United States

0.2
Actual
0.15

0.1 Efficient

0.05

0
64
1/64

512
1/512

1/8

FIGURE III
Distribution of Plant Size

Although we expressed the distortions in terms of output

(τY si ) and capital relative to labor (τ Ksi ), in Appendix III, we show
∗
that these are equivalent to a particular combination of labor (τ Lsi )
1.5 Issues ∗
and capital (τ Ksi ) distortions. In Appendix III, we also report that
more efficient (higher TFPQ) plants appear to face bigger distor-
tions
• Measurement error on both capital and labor.

– Extreme outliers

20
 – Classical measurement error? Unlikely, since TFPR and TFPQ are systematically related to
MISALLOCATION
plant ownership AND
categories and plant TFP
exit in IN CHINA
mostly ways in India1427
AND INDIA
reassuring and China.

TABLE VII
TFP BY OWNERSHIP

TFPR TFPQ

China
State −0.415 −0.144
(0.023) (0.090)
Collective 0.114 0.047
(0.010) (0.013)
Foreign −0.129 0.228
(0.024) (0.040)
India
State (central) −0.285 0.717
(0.082) (0.295)
State (local) −0.081 0.425
(0.063) (0.103)
Joint public/private −0.162 0.671
(0.037) (0.085)

Notes. The dependent variable is the deviation of log TFPR or log TFPQ from the industry mean. The
independent variables for China are dummies for state-owned plants, collective-owned plants (plants jointly
owned by local governments and private parties), and foreign-owned plants. The omitted group is domestic
private plants. The independent variables for India are dummies for a plant owned by the central government,
a plant owned by a local government, and a plant jointly owned by the government (either central or local) and
by private individuals. The omitted group is a privately owned plant (both domestic and foreign). Regressions
are weighted least squares with industry value-added shares as weights. Entries are the dummy coefficients,
with standard errors in parentheses. Results are pooled for all years.

surprisingly, collectively owned (part private, part local govern-

ment)
• Industry plants
shares have
are set 11%those
to equal higher TFPR.
in the Foreign-owned
corresponding plants have
US manufacturing industry, because
23% higher TFPQ on average, but 13% lower TFPR. The latter
H&Kcould
presume that the
reflect US isaccess
better comparatively undistorted.
to credit But what treatment
or preferential about distortion
in across (as
export
opposed processing
to within) sectors? zones. Consistent with this interpretation, ex-
porting plants have 46% higher TFPQ but 14% lower TFPR. In the
• HsiehUnited States,
and Klenow show exporters have
that firm exit a similar
is related TFPQ
to TFPR advantage
and TFPQ. But in(50%)
their hypothetical
but display higher rather than lower TFPR (+6% on average).19
experiments, no firms exit or enter due in response to reallocation. How important is endogenous
In India, all types of plants with public involvement exhibit
entry lower
and exitTFPR: 29% lower
for aggregate TFP? for plants owned by the central govern-
ment, 8% lower for those owned by local governments, and 16%
• Whatlower
about for joint public-private
uncertainty and risk? plants. Public involvement also goes
along with 40%–70% higher TFPQ, although this might reflect
monopoly rights that guarantee demand.
We next look at the correlation of TFPR with plant exit. One
would expect true TFPR to be lower for exiters. If TFPR is mea-
sured with more error in China and India, the coefficient from
a regression of plant exit on TFPR 21 should be biased downward.

19. The high TFPQ of exporters could reflect the “demand shock” coming from
accessing foreign markets, rather than just physical productivity.