measuring micro and macro productivity dynamics
Post on 01-Feb-2016
40 Views
Preview:
DESCRIPTION
TRANSCRIPT
Measuring Micro and Measuring Micro and Macro Productivity Macro Productivity
DynamicsDynamics
Remarks By John HaltiwangerRemarks By John Haltiwanger
OverviewOverview
Objectives:Objectives: Measure productivity at micro and macro levels in Measure productivity at micro and macro levels in
consistent fashionconsistent fashion Reallocation dynamics at micro level important for Reallocation dynamics at micro level important for
aggregateaggregate Track firm dynamics including entry and exit to Track firm dynamics including entry and exit to
measure and analyze importance of static and measure and analyze importance of static and dynamic allocative efficiency vs. “within firm dynamic allocative efficiency vs. “within firm productivity” growthproductivity” growth
Challenges:Challenges: TFP difficult to measureTFP difficult to measure Firm dynamics difficult to measureFirm dynamics difficult to measure
Methodological IssuesMethodological Issues
CAPITAL
MATERIALS
LABOROUTPUTTFP
materialslabor
materials
labor
ln1
ln
lnlnln
Challenges:1. Difficult to measure outputs and inputs especially capital a. Capital
b. Prices especially at micro level 2. Estimating factor elasticities:
a. OLS b. Cost Shares c. Proxy methods (Olley-Pakes; Levinsohn-Petrin) d. IV
Mixed Messages on Mixed Messages on Challenges…Challenges…
Labor productivity and TFP often highly Labor productivity and TFP often highly correlatedcorrelated More important to get representative sample and More important to get representative sample and
firm dynamics than TFP?firm dynamics than TFP? Capital measurement difficult and potentially Capital measurement difficult and potentially
important for policies/institutionsimportant for policies/institutions Price/output measurement difficult given Price/output measurement difficult given
data limitations and very important for data limitations and very important for positive and normative implicationspositive and normative implications
Table III.1 (B) Definitions of MeasuresTable III.1 (B) Definitions of Measures
Estimation methodology:Estimation methodology:
11 Net book value of total capital, Share of total outputNet book value of total capital, Share of total output
22 Net book value of machinery and equipment, Share of total outputNet book value of machinery and equipment, Share of total output
33 Replacement value of total capital, Share of total outputReplacement value of total capital, Share of total output
44 Replacement value of machinery and equipment, Share of total outputReplacement value of machinery and equipment, Share of total output
55 Net book value of total capital, Share of total costNet book value of total capital, Share of total cost
66 Net book value of machinery and equipment, Share of total costNet book value of machinery and equipment, Share of total cost
77 Replacement value of total capital, Share of total costReplacement value of total capital, Share of total cost
88 Replacement value of machinery and equipment, Share of total costReplacement value of machinery and equipment, Share of total cost
99 Net book value of total capital, OLSNet book value of total capital, OLS
1010 Net book value of machinery and equipment, OLSNet book value of machinery and equipment, OLS
1111 Replacement value of total capital, OLSReplacement value of total capital, OLS
1212 Replacement value of machinery and equipment, OLSReplacement value of machinery and equipment, OLS
Source: Haltiwanger and Schweiger (2004)
Table III.2 Average Pairwise Correlation of TFP Measures Using Alternative Estimation and Table III.2 Average Pairwise Correlation of TFP Measures Using Alternative Estimation and Measurement MethodsMeasurement Methods
r(lnprod,lntfp10p)r(lnprod,lntfp10p) 0.6834*0.6834*
r(lntfp1p,lntfp10p)r(lntfp1p,lntfp10p) 0.8093*0.8093*
r(lntfp2p,lntfp10p)r(lntfp2p,lntfp10p) 0.8679*0.8679*
r(lntfp3p,lntfp10p)r(lntfp3p,lntfp10p) 0.7889*0.7889*
r(lntfp4p,lntfp10p)r(lntfp4p,lntfp10p) 0.7995*0.7995*
r(lntfp5p,lntfp10p)r(lntfp5p,lntfp10p) 0.9826*0.9826*
r(lntfp6p,lntfp10p)r(lntfp6p,lntfp10p) 0.9897*0.9897*
r(lntfp7p,lntfp10p)r(lntfp7p,lntfp10p) 0.9784*0.9784*
r(lntfp8p,lntfp10p)r(lntfp8p,lntfp10p) 0.8599*0.8599*
r(lntfp9p,lntfp10p)r(lntfp9p,lntfp10p) 0.9951*0.9951*
r(lntfp11p,lntfp10p)r(lntfp11p,lntfp10p) 0.9718*0.9718*
r(lntfp12p,lntfp10p)r(lntfp12p,lntfp10p) 0.9784*0.9784*
Note: Reported correlations are based upon pooled data. See Table III.1 (B) for definitions of Note: Reported correlations are based upon pooled data. See Table III.1 (B) for definitions of measures. Lntfp is log TFP so using Table III.1 (B) we can see that ntfp10p is the log TFP using measures. Lntfp is log TFP so using Table III.1 (B) we can see that ntfp10p is the log TFP using pooled production function estimated via OLS using net book value of machinery and equipment.pooled production function estimated via OLS using net book value of machinery and equipment.
Source: Haltiwanger and Schweiger (2004)
PRODUCTIVITY SHOCKSPRODUCTIVITY SHOCKS• TFP - residual from KLEM production function:TFP - residual from KLEM production function:
where the production function is estimated in where the production function is estimated in logs:logs:
Inputs may be correlated w/ productivity, so use Inputs may be correlated w/ productivity, so use IVs:IVs:
(a)(a) Output of downstream producers.Output of downstream producers.
(b)(b) Regional governments expenditures.Regional governments expenditures.
(c)(c) Energy and materials prices.Energy and materials prices.
• Estimated factor elasticities by sector using cost-Estimated factor elasticities by sector using cost-share approach and results are robust (e.g., share approach and results are robust (e.g., correlation between two measures is 0.88 and correlation between two measures is 0.88 and moments very similar).moments very similar).
jtjtjtjtjtjtjt VMEHLKY )(
jtjtjtjtjtjtjt MEHLKYTFP logˆlogˆ)log(logˆlogˆlog
Source: Eslava et al. (2005)
TABLE 2: PRODUCTION FUNCTIONTABLE 2: PRODUCTION FUNCTION
TABLE 3: DIFFERENT TFP MEASURESTABLE 3: DIFFERENT TFP MEASURES
TABLE 6: DETERMINANTS OF EXIT PROBABILITY TABLE 6: DETERMINANTS OF EXIT PROBABILITY
Specification:Specification: [1][1] [2][2] [3][3] [4][4] [5][5] [6][6] [7][7]
Revenue TFPRevenue TFP -0.062-0.0620.0110.011
Physical TFPPhysical TFP -0.031-0.0310.0100.010
-0.059-0.0590.0120.012
-0.028-0.0280.0090.009
PricesPrices -0.034-0.0340.0140.014
-0.078-0.0780.0170.017
Demand ShockDemand Shock -0.038-0.0380.0020.002
-0.038-0.0380.0020.002
U.S. Exit Selection Results
Source: Foster, Haltiwanger and Syverson (2005)
Aggregate productivity and allocationAggregate productivity and allocation
Olley and Pakes (1996) static decomposition:Olley and Pakes (1996) static decomposition:
where: N: # of firms in a sector; where: N: # of firms in a sector; The first term is the unweighted average of firm-level productivity, The first term is the unweighted average of firm-level productivity, The second term reflects allocation of resources: do firms with higher The second term reflects allocation of resources: do firms with higher
productivity have greater market share.productivity have greater market share. Requires representative cross sectional samples but does not require accurate Requires representative cross sectional samples but does not require accurate
longitudinal linkageslongitudinal linkages Cannot quantify directly importance of entry and exitCannot quantify directly importance of entry and exit
By construction, cross term takes out country effects so abstracts from some By construction, cross term takes out country effects so abstracts from some aspects of measurement erroraspects of measurement error
i i
tittitittt PppNP ))(()/1(__
The cross-sectional efficiency of the The cross-sectional efficiency of the
allocation of activityallocation of activity
0.0
0.2
0.4
0.6
0.8
Data for Hungary, Indonesia and Romania use Three-Year Differencing.Excluding Brazil and Venezuela.
Five-Year Differencing, Real Gross Output, Manufacturing
The Gap Between Weighted and Un-WeightedLabor Productivity, 1990s
Source: Bartelsman, Haltiwanger and Scarpetta (2005)
The cross-sectional efficiency of the The cross-sectional efficiency of the
allocation of activityallocation of activity
0.0
0.2
0.4
0.6
0.8
Five-Year Differencing, Real Gross Output, Manufacturing.Data for Hungary and Romania use Three-Year Differencing.
in Transition Economies over the 1990s
The Evolution of the Gap Between Weightedand Un-Weighted Labor Productivity
Source: Bartelsman, Haltiwanger and Scarpetta (2005)
Olley-Pakes Decomposition for Colombian Manufacturing
0.80
0.90
1.00
1.10
1.20
1.30
1.40
1.50
1.60
1.70
1.80
82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98
Year
0.00
0.10
0.20
0.30
0.40
0.50
0.60
0.70
0.80
0.90
1.00
Aggregate (Weighted) Simple Average Cross-term
Source: Eslava et al. (2005)
Dynamic decomposition of productivity growthDynamic decomposition of productivity growth
Advantages:Advantages: Explicit measure of contribution of entry and exitExplicit measure of contribution of entry and exit
Disadvantages:Disadvantages: Horizon dependentHorizon dependent Experimentation, learning and selection may make especially dynamic Experimentation, learning and selection may make especially dynamic
economies have a low contribution of net entry at short horizons but high economies have a low contribution of net entry at short horizons but high contribution at long horizonscontribution at long horizons
)()(
)(
ktkitXi
kitktitNi
it
Ciititkt
Cikititit
Cikit
t
PpPp
pPppP
Within-firm productivity growth Within-firm productivity growth dominates at five-year horizons in Mfgdominates at five-year horizons in Mfg
-0.5
0.0
0.5
1.0
1.5
Argentina: 1995-2001. Chile: 1985-1999. Colombia: 1987-1998. Estonia: 2000-2001.Finland: 2000-2002. France: 1990-1995. West Germany: 2000-2002. Korea: 1988 & 1993.Latvia: 2001-2002. Netherlands: 1992-2001. Portugal: 1991-1994. Slovenia: 1997-2001.Taiwan: 1986, 1991 & 1996. UK: 2000-2001. USA: 1992 & 1997.Excluding Brazil and Venezuela.
Labor Productivity - Five-Year Differencing, Real Gross OutputFHK Decomposition Shares - Manufacturing
Within Between Cross
Entry Exit Firm Turnover(i)
Source: Bartelsman, Haltiwanger and Scarpetta (2005)
Contribution of Net Entry to Productivity Growth (10-year horizon)
-1
-0.5
0
0.5
1
1.5
2
Retail-All DeptStores
GenMerch
Mfg-All
Shar
e Continuing Estabs
Net Entry
Source: Foster, Haltiwanger and Krizan (2005)
VariableVariable
Weighted RegressionWeighted Regression
Exit DummyExit Dummy Entry DummyEntry Dummy
Revenue TFPRevenue TFP -0.0340-0.03400.00490.0049
0.04480.04480.00550.0055
Physical TFPPhysical TFP -0.0305-0.03050.00580.0058
0.09990.09990.00640.0064
PricePrice -0.0035-0.00350.00400.0040
-0.0551-0.05510.00450.0045
Demand ShockDemand Shock -0.6364-0.63640.02930.0293
-0.0927-0.09270.03260.0326
Differences Between continuing, entering and exiting establishments in U.S. Manufacturing
Source: Foster, Haltiwanger and Syverson (2005)
TotalTotalGrowthGrowth
Components of Decomposition (FHK/BHC)Components of Decomposition (FHK/BHC)
WithinWithin BetweenBetween CrossCross EntryEntry ExitExit Net EntryNet Entry
RevenueRevenue 5.095.09 3.383.38 -0.51-0.51 1.321.32 0.710.71 0.190.19 0.900.90
PhysicalPhysical 5.095.09 3.453.45 -0.40-0.40 0.700.70 1.221.22 0.120.12 1.341.34
Components of Decomposition (GR)Components of Decomposition (GR)
WithinWithin BetweenBetween EntryEntry ExitExit Net EntryNet Entry
RevenueRevenue 5.095.09 4.044.04 0.130.13 0.540.54 0.370.37 0.910.91
PhysicalPhysical 5.095.09 3.803.80 -0.07-0.07 1.041.04 0.310.31 1.351.35
Sensitivity of U.S. Manufacturing Decomposition to Prices
Source: Foster, Haltiwanger and Syverson (2005)
Guidance?Guidance?
Micro/macro links critical Micro/macro links critical Large fraction of level of productivity associated with Large fraction of level of productivity associated with
allocative efficiencyallocative efficiency Important contribution of changes in allocative efficiency Important contribution of changes in allocative efficiency
and/or reallocation/net entry components for productivity and/or reallocation/net entry components for productivity growthgrowth
Measurement of TFP challenge in general and Measurement of TFP challenge in general and especially at micro levelespecially at micro level
Greater weight on representative sample with Greater weight on representative sample with micro/macro links than TFP vs. Labor productivity?micro/macro links than TFP vs. Labor productivity? But measurement matters:But measurement matters:
Prices and market structure matter and are often missing Prices and market structure matter and are often missing piece in micro/macro linkpiece in micro/macro link
top related