Download - Identifying Factor Productivity by Dynamic Panel Data and Control Function Approaches: A Comparative Evaluation for EU Agriculture" (extended version)

Transcript
Page 1: Identifying Factor Productivity by Dynamic Panel Data and Control Function Approaches: A Comparative Evaluation for EU Agriculture" (extended version)

Identifying Factor Productivity by Dynamic Panel Data and Control

Function Approaches: A Comparative Evaluation for EU Agriculture

by Martin Petrick and Mathias Kloss

Mathias Kloss

Economics Cluster Seminar Wageningen UR | 3 October 2013

Page 2: Identifying Factor Productivity by Dynamic Panel Data and Control Function Approaches: A Comparative Evaluation for EU Agriculture" (extended version)

www.iamo.de 2

Page 3: Identifying Factor Productivity by Dynamic Panel Data and Control Function Approaches: A Comparative Evaluation for EU Agriculture" (extended version)

www.iamo.de 3

Page 4: Identifying Factor Productivity by Dynamic Panel Data and Control Function Approaches: A Comparative Evaluation for EU Agriculture" (extended version)

www.iamo.de 4

Stagnating agricultural productivity

growth in Europe

Source: Coelli & Rao 2005, p. 127.

Page 5: Identifying Factor Productivity by Dynamic Panel Data and Control Function Approaches: A Comparative Evaluation for EU Agriculture" (extended version)

www.iamo.de 5

Stagnating agricultural productivity

growth in the EU

Source: Piesse & Thirtle 2010, p. 171.

Page 6: Identifying Factor Productivity by Dynamic Panel Data and Control Function Approaches: A Comparative Evaluation for EU Agriculture" (extended version)

www.iamo.de 6

Outline

β€’ An insight into recent innovations in production function estimation

Comparative evaluation of 2 recently proposed production function estimators

How plausible are these for the case of agriculture?

β€’ Unique and current set of production elasticities for 8 farm-level data sets at the EU country level

β€’ Some evidence on shadow prices

β€’ Conclusions

Page 7: Identifying Factor Productivity by Dynamic Panel Data and Control Function Approaches: A Comparative Evaluation for EU Agriculture" (extended version)

www.iamo.de 7

Two problems of identification

A general production function:

𝑦𝑖𝑑 = 𝑓 𝐴𝑖𝑑 , 𝐿𝑖𝑑 , 𝐾𝑖𝑑, 𝑀𝑖𝑑 + πœ”π‘–π‘‘ + νœ€π‘–π‘‘

with

y Output

A Land

L Labour

K Capital (fixed)

M Materials (Working capital)

πœ” Farm- & time-specific factor(s) known to farmer, unobserved by analyst

νœ€ Independent & identically distributed noise

i, t Farm & time indices

Page 8: Identifying Factor Productivity by Dynamic Panel Data and Control Function Approaches: A Comparative Evaluation for EU Agriculture" (extended version)

www.iamo.de 8

Two problems of identification

Collinearity problem β€’ If variable and intermediate inputs are chosen simultaneously factor use across farms varies only with πœ” (Bond & SΓΆderbom 2005; Ackerberg et al. 2007)

Production elasticities for variable inputs not identified!

Endogeneity problem β€’ πœ” likely correlated with other input choices β€’ Need to take Ο‰ into account in order to identify 𝑓, as πœ”π‘–π‘‘ + νœ€π‘–π‘‘

is not i.i.d No identification of 𝑓 possible if Ο‰ is not taken into account!

Page 9: Identifying Factor Productivity by Dynamic Panel Data and Control Function Approaches: A Comparative Evaluation for EU Agriculture" (extended version)

www.iamo.de 9

Traditional approaches to solve the

identification problems

1. Ordinary Least Squares: forget it. Assume Ο‰ is non-existent.

– Bias: elasticities of flexible inputs too high (capture Ο‰)

2. β€œWithin” (fixed effects): assume we can decompose Ο‰ in

– Assumption plausible?

– Bias: elasticities too low as signal-to-noise is reduced

– Collinearity problem not adressed

time-specific shock

farm-specific fixed effect

remaining farm- and time specific shock

Page 10: Identifying Factor Productivity by Dynamic Panel Data and Control Function Approaches: A Comparative Evaluation for EU Agriculture" (extended version)

www.iamo.de 10

Recent solutions to solve the

identification probems

3. Dynamic panel data modelling

– current (exogenous) variation in input use by lagged adjustment to

past productivity shocks (Arellano & Bond 1991; Blundell & Bond 1998)

β€’ feasible if input modifications s.t. adjustment costs (Bond & SΓΆderbom 2005)

β€’ plausible for many factors (e.g. labour, land or capital ) but less so for intermediate inputs

– one way to allow costly adjustment: πœπ‘–π‘‘ = πœŒπœπ‘–π‘‘βˆ’1 + 𝑒𝑖𝑑, with 𝜌 < 1

– dynamic production function with lagged levels & differences of inputs

as instruments in a GMM framework (Blundell & Bond 2000)

– Bias: hopefully small. Adresses both problems if instruments induce

sufficient exogenous variation

𝜌 autoregressive parameter

𝑒𝑖𝑑 mean zero innovation

Page 11: Identifying Factor Productivity by Dynamic Panel Data and Control Function Approaches: A Comparative Evaluation for EU Agriculture" (extended version)

www.iamo.de 11

Recent solutions to solve the

identification probems

4. Control Function approach

– assume Ο‰ evolves along with observed firm characteristics (Olley/Pakes

1996, Econometrica)

– materials a good control candidate for Ο‰ (Levinsohn & Petrin 2003)

– further assume: (a) M is monotonically increasing in Ο‰ & (b) factor

adjustment in one period

1. Estimate β€œclean” A & L by controlling Ο‰ with M & K

2. Recover M & K from additional timing assumptions

– solves endogeneity problem if control function fully captures Ο‰ β€’ productivity enhancing reaction to shocks less input use violating (a)

β€’ some factors (e.g. soil quality) might evolve slowly violating (b)

– collinearity problem not solved β€’ Solutions by Ackerberg et al. (2006) and Wooldridge (2009)

Page 12: Identifying Factor Productivity by Dynamic Panel Data and Control Function Approaches: A Comparative Evaluation for EU Agriculture" (extended version)

www.iamo.de 12

Data

FADN individual farm-level panel data made available by EC

Field crop farms (TF1) in Denmark, France, Germany East, Germany

West, Italy, Poland, Slovakia & United Kingdom

T=7 (2002-2008) (only 2006-2008 for PL & SK)

Cobb Douglas functional form (Translog examined as well)

Annual fixed effects included via year dummies

Estimation with Stata12 using xtabond2 (Roodman 2009) & levpet

estimator (Petrin et al. 2004)

Page 13: Identifying Factor Productivity by Dynamic Panel Data and Control Function Approaches: A Comparative Evaluation for EU Agriculture" (extended version)

www.iamo.de 13

Cobb Douglas production elasticities

Blundell/Bond

Page 14: Identifying Factor Productivity by Dynamic Panel Data and Control Function Approaches: A Comparative Evaluation for EU Agriculture" (extended version)

www.iamo.de 14

Cobb Douglas production elasticities

LevPet

Page 15: Identifying Factor Productivity by Dynamic Panel Data and Control Function Approaches: A Comparative Evaluation for EU Agriculture" (extended version)

www.iamo.de 15

Elasticity of materials LevPet

Page 16: Identifying Factor Productivity by Dynamic Panel Data and Control Function Approaches: A Comparative Evaluation for EU Agriculture" (extended version)

www.iamo.de 16

Elasticity of materials:

Comparison of estimators (LevPet)

Page 17: Identifying Factor Productivity by Dynamic Panel Data and Control Function Approaches: A Comparative Evaluation for EU Agriculture" (extended version)

www.iamo.de 17

Returns to scale LevPet

Point estimates

Page 18: Identifying Factor Productivity by Dynamic Panel Data and Control Function Approaches: A Comparative Evaluation for EU Agriculture" (extended version)

www.iamo.de 18

Returns to scale LevPet

Not sig. different from 1 displayed as 1

Page 19: Identifying Factor Productivity by Dynamic Panel Data and Control Function Approaches: A Comparative Evaluation for EU Agriculture" (extended version)

www.iamo.de 19

Examining the Translog specification

β€’ OLS: Highly implausible results at sample means

β€’ Within: Interaction terms not sig. in the majority of cases

β€’ BB: No straightforward implementation, as assumption of linear

addivitity of the fixed effects is violated

β€’ LevPet: No straightforward implementation, as M & K are assumed to

be additively separable

Page 20: Identifying Factor Productivity by Dynamic Panel Data and Control Function Approaches: A Comparative Evaluation for EU Agriculture" (extended version)

www.iamo.de 20

-10

0-5

00

50

10

015

020

0

Sha

dow

price o

f w

ork

ing c

apita

l (%

)

DK FR DEE DEW IT PL SK UKexcludes outside values

Shadow interest rate of materials (%):

distributions per country

Page 21: Identifying Factor Productivity by Dynamic Panel Data and Control Function Approaches: A Comparative Evaluation for EU Agriculture" (extended version)

www.iamo.de 21

-10

010

20

30

40

50

60

Sha

dow

wage

(E

UR

/hou

r)

DK FR DEE DEW IT PL SK UKexcludes outside values

Shadow wage (€/h): distributions per country

Page 22: Identifying Factor Productivity by Dynamic Panel Data and Control Function Approaches: A Comparative Evaluation for EU Agriculture" (extended version)

www.iamo.de 22

Conclusions

β€’ Adjustment costs relevant for important inputs in agricultural production – LP and BB identification strategies a priori plausible

β€’ LP plausible results combined with FADN data but is a second-best choice – corrected upward (downward) bias in OLS (Whithin-OLS) regressions

– conceptual problems in identifying flexible factors

β€’ BB only performed well with regard to materials

β€’ Materials most important production factor in EU field crop farming (prod.

elasticity of ~0.7)

β€’ Fixed capital, land and labour usually not scarce

β€’ Shadow price analysis reveals heterogenous picture – Credit market imperfections: Funding constraints (DEE, IT) vs. overutilisation

(DEW, DK)? Effects of financial crisis?

– Low labour remuneration (except DK)

Page 23: Identifying Factor Productivity by Dynamic Panel Data and Control Function Approaches: A Comparative Evaluation for EU Agriculture" (extended version)

www.iamo.de 23

Future research

β€’ Estimated shadow prices as starting point for analysis of

drivers & impacts

β€’ Extension to other production systems (e.g., dairy)

β€’ Examine other identification strategies

β€’ Wooldridge (2009) is a promising candidate

β€’ unifies LP in a single-step efficiency gains

β€’ solves collinearity problem

Page 24: Identifying Factor Productivity by Dynamic Panel Data and Control Function Approaches: A Comparative Evaluation for EU Agriculture" (extended version)

www.iamo.de 24

The END.

Page 25: Identifying Factor Productivity by Dynamic Panel Data and Control Function Approaches: A Comparative Evaluation for EU Agriculture" (extended version)

www.iamo.de 25

Appendix

Page 26: Identifying Factor Productivity by Dynamic Panel Data and Control Function Approaches: A Comparative Evaluation for EU Agriculture" (extended version)

www.iamo.de 26

Blundell/Bond in detail

β€’ Substituting 𝑣𝑖𝑑 = πœŒπ‘£π‘–π‘‘βˆ’1 + 𝑒𝑖𝑑 and πœ”π‘–π‘‘ = 𝛾𝑑 + πœ‚π‘– + 𝑣𝑖𝑑 into the production

function implies the following dynamic production function

𝑦𝑖𝑑 = 𝛼𝑋π‘₯𝑖𝑑 βˆ’ π›Όπ‘‹πœŒπ‘₯π‘–π‘‘βˆ’1 + πœŒπ‘¦π‘–π‘‘βˆ’1 + 𝛾𝑑 βˆ’ πœŒπ›Ύπ‘‘βˆ’1𝑋

+ 1 βˆ’ 𝜌 πœ‚π‘– + νœ€π‘–π‘‘

β€’ Alternatively:

𝑦𝑖𝑑 = πœ‹1𝑋𝑋

π‘₯𝑖𝑑 + πœ‹2𝑋𝑋

π‘₯π‘–π‘‘βˆ’1 + πœ‹3π‘¦π‘–π‘‘βˆ’1 + π›Ύπ‘‘βˆ— + πœ‚π‘–

βˆ— + νœ€π‘–π‘‘βˆ—

subject to the common factor restrictions that πœ‹2𝑋 = βˆ’πœ‹1π‘‹πœ‹3 for all X.

(allows recovery of input elasticities)

β€’ Farm-specific fixed effects removed by FD, allows transmission of πœ” to

subsequent periods

Page 27: Identifying Factor Productivity by Dynamic Panel Data and Control Function Approaches: A Comparative Evaluation for EU Agriculture" (extended version)

www.iamo.de 27

Olley Pakes and Levinsohn/Petrin in

detail

β€’ Log investment (𝑖𝑖𝑑) as an observed characteristic driven by πœ”π‘–π‘‘:

β€’ 𝑖𝑖𝑑 = 𝑖𝑑 πœ”π‘–π‘‘ , π‘˜π‘–π‘‘ and π‘˜π‘–π‘‘ evolves π‘˜π‘–π‘‘+1 = 1 βˆ’ 𝛿 π‘˜π‘–π‘‘ + 𝑖𝑖𝑑, with 𝛿=

depreciation rate

β€’ Given monotonicity we can write πœ”π‘–π‘‘ = β„Žπ‘‘ 𝑖𝑖𝑑 , π‘˜π‘–π‘‘

β€’ Assume: πœ”π‘–π‘‘ = 𝐸 πœ”π‘–π‘‘|πœ”π‘–π‘‘βˆ’1 + πœ‰π‘–π‘‘,

– πœ‰π‘–π‘‘ is an innovation uncorrelated with π‘˜π‘–π‘‘ used to identify capital

coefficient in the second stage

β€’ Idea

1. control for the influence of k and Ο‰

2. recover the true coefficient of k as well as Ο‰ in the second stage

Page 28: Identifying Factor Productivity by Dynamic Panel Data and Control Function Approaches: A Comparative Evaluation for EU Agriculture" (extended version)

www.iamo.de 28

Olley/Pakes and Levinsohn/Petrin

continued

β€’ Plugging πœ”π‘–π‘‘ = β„Žπ‘‘ 𝑖𝑖𝑑 , π‘˜π‘–π‘‘ into production function gives

𝑦𝑖𝑑 = π›Όπ΄π‘Žπ‘–π‘‘ + 𝛼𝐿𝑙𝑖𝑑 + π›Όπ‘€π‘šπ‘–π‘‘ + πœ™π‘‘ 𝑖𝑖𝑑 , π‘˜π‘–π‘‘ + νœ€π‘–π‘‘

β€’ πœ™ is approximated by 2nd and 3rd order polynomials of i and k in the first stage

β€’ Here parameters of variable factors are obtained by OLS

β€’ Second stage:

1. using πœ™π‘‘ and candidate value for 𝛼𝐾, πœ” 𝑖𝑑 is computed for all t

2. Regress πœ” 𝑖𝑑 on its lagged values to obtain a consistent predictor of that part of Ο‰ that is free of the innovation ΞΎ (β€œclean” πœ”π‘–π‘‘)

3. using first stage parameters together with prediction of the β€œclean” πœ”π‘–π‘‘ and 𝐸 π‘˜π‘–π‘‘πœ‰π‘–π‘‘ = 0 consistent estimate of 𝛼𝐾 by minimum distance

Page 29: Identifying Factor Productivity by Dynamic Panel Data and Control Function Approaches: A Comparative Evaluation for EU Agriculture" (extended version)

www.iamo.de 29

Elasticity of materials:

Comparison of estimators (BB)

Page 30: Identifying Factor Productivity by Dynamic Panel Data and Control Function Approaches: A Comparative Evaluation for EU Agriculture" (extended version)

www.iamo.de 30

Comparison of estimators - East German field

crop farms: marginal return on materials

Page 31: Identifying Factor Productivity by Dynamic Panel Data and Control Function Approaches: A Comparative Evaluation for EU Agriculture" (extended version)

www.iamo.de 31

-10

0-9

5-9

0-8

5-8

0-7

5-7

0-6

5-6

0-5

5

Sha

dow

price o

f fixe

d c

apita

l (%

)

DK FR DEE DEW IT PL SK UKexcludes outside values

Shadow interest rate of fixed capital (%):

distributions per country

Page 32: Identifying Factor Productivity by Dynamic Panel Data and Control Function Approaches: A Comparative Evaluation for EU Agriculture" (extended version)

www.iamo.de 32

-.000

50

.000

5

Sha

dow

price o

f la

nd

(E

UR

/ha)

DK FR DEE DEW IT PL SK UKexcludes outside values

Shadow land rent (€/ha): distributions per

country

Page 33: Identifying Factor Productivity by Dynamic Panel Data and Control Function Approaches: A Comparative Evaluation for EU Agriculture" (extended version)

www.iamo.de 33

The Wooldridge-Levinsohn-Petrin

approach

β€’ Unifies the Olley/Pakes and Levinsohn/Petrin procedure within a

IV/GMM framework

– Estimation in a single step

– Analytic standard errors

– Implementation of translog is straightforward

β€’ Suppose for parsimony:

𝑦𝑖𝑑 = 𝛼 + 𝛽1𝑙𝑖𝑑 + 𝛽2π‘˜π‘–π‘‘ +πœ”π‘–π‘‘ + 𝑒𝑖𝑑, and remember

πœ”π‘–π‘‘ = β„Ž π‘˜π‘–π‘‘ , π‘šπ‘–π‘‘ ,

– Now assume:

𝐸 𝑒𝑖𝑑|𝑙𝑖𝑑 , π‘˜π‘–π‘‘ , π‘šπ‘–π‘‘ , 𝑙𝑖,π‘‘βˆ’1, π‘˜π‘–,π‘‘βˆ’1 , π‘šπ‘–,π‘‘βˆ’1, … , 𝑙𝑖1, π‘˜π‘–1 , π‘šπ‘–1 = 0

Page 34: Identifying Factor Productivity by Dynamic Panel Data and Control Function Approaches: A Comparative Evaluation for EU Agriculture" (extended version)

www.iamo.de 34

The Wooldridge-Levinsohn-Petrin

approach

β€’ Again, assume: πœ”π‘–π‘‘ = 𝐸 πœ”π‘–π‘‘|πœ”π‘–π‘‘βˆ’1 + πœ‰π‘–π‘‘ and

𝐸 πœ”π‘–π‘‘|π‘˜π‘–π‘‘ , 𝑙𝑖,π‘‘βˆ’1, π‘˜π‘–,π‘‘βˆ’1 , π‘šπ‘–,π‘‘βˆ’1, … , 𝑙𝑖1, π‘˜π‘–1 , π‘šπ‘–1

= 𝐸 πœ”π‘–π‘‘|πœ”π‘–π‘‘βˆ’1 = 𝑓 πœ”π‘–π‘‘βˆ’1 = 𝑓 β„Ž π‘˜π‘–,π‘‘βˆ’1, π‘šπ‘–,π‘‘βˆ’1,

β€’ Plugging into the production function gives

𝑦𝑖𝑑 = 𝛼 + 𝛽1𝑙𝑖𝑑 + 𝛽2π‘˜π‘–π‘‘ + 𝑓 β„Ž π‘˜π‘–,π‘‘βˆ’1, π‘šπ‘–,π‘‘βˆ’1, + νœ€π‘–π‘‘

where νœ€π‘–π‘‘ = πœ‰π‘–π‘‘ + 𝑒𝑖𝑑.

β€’ Now, we have two equations to identify the parameters

𝑦𝑖𝑑 = 𝛼 + 𝛽1𝑙𝑖𝑑 + 𝛽2π‘˜π‘–π‘‘ + β„Ž π‘˜π‘–π‘‘ , π‘šπ‘–π‘‘ + 𝑒𝑖𝑑

𝑦𝑖𝑑 = 𝛼 + 𝛽1𝑙𝑖𝑑 + 𝛽2π‘˜π‘–π‘‘ + 𝑓 β„Ž π‘˜π‘–,π‘‘βˆ’1, π‘šπ‘–,π‘‘βˆ’1, + νœ€π‘–π‘‘

Page 35: Identifying Factor Productivity by Dynamic Panel Data and Control Function Approaches: A Comparative Evaluation for EU Agriculture" (extended version)

www.iamo.de 35

The Wooldridge-Levinsohn-Petrin

approach

β€’ And

𝐸 𝑒𝑖𝑑|𝑙𝑖𝑑 , π‘˜π‘–π‘‘ , π‘šπ‘–π‘‘ , 𝑙𝑖,π‘‘βˆ’1, π‘˜π‘–,π‘‘βˆ’1 , π‘šπ‘–,π‘‘βˆ’1, … , 𝑙𝑖1, π‘˜π‘–1 , π‘šπ‘–1 = 0

𝐸 νœ€π‘–π‘‘|π‘˜π‘–π‘‘ , 𝑙𝑖,π‘‘βˆ’1, π‘˜π‘–,π‘‘βˆ’1 , π‘šπ‘–,π‘‘βˆ’1, … , 𝑙𝑖1, π‘˜π‘–1 , π‘šπ‘–1 = 0.

β€’ Unknown function β„Ž approximated by low-order polynomial and 𝑓

might be a random walk with drift.

β€’ Estimation:

– Both equations within a GMM framework, or

– Second equation by IV-estimation and instrument for 𝑙 (Petrin and

Levinsohn 2012)