Learning-by-doing, learning-by-exporting, and productivity: evidence - PowerPoint PPT Presentation

Learning-by-doing, learning-by-exporting, and productivity: evidence from Colombia Ana M. Fernandes Alberto E. Isgut The World Bank Wesleyan University LACEA Conference Paris, October 27-29, 2005

Does participation in export markets increase plant-level productivity? � Appealing idea but inconclusive empirical evidence so far � What is the direction of causality? � Learning-by-exporting hypothesis: Export participation => Improvement in plant-level productivity � Or self-selection hypothesis: Higher plant productivity => Export participation?

What have other researchers found? � Method 1: Difference-in-differences specifications 1 ( ) ∆ ≡ − = β + β + β + ε ln Y ln Y ln Y D ' Z iT iT i 0 0 1 i 0 2 i 0 it T Sample partitioned into a treatment group and a control group, e.g. entrants into export markets vs. nonexporters Y iT measure of plant performance, e.g. TFP D i0 =1 if plant i belongs to treatment group Z i0 controls, e.g. plant size, industry, year

Study Country Estim. method LBE? Bernard and Wagner (1997) Germany OLS No Bernard and Jensen (1999) USA OLS No Aw et al. (2000) Korea, Taiwan OLS Some (Taiwan) Isgut (2001) Colombia OLS No Delgado et al. (2002) Spain Nonparametric Some (young) Castellani (2002) Italy OLS (export Yes intensity) Hallward-Driemeier et al. Indonesia, Korea, OLS, c-s (born Yes (except (2002) Malaysia, exporters) Korea) Philippines, Thailand Fafchamps et al. (2002) Morocco IV, c-s (time of No 1 st exports) Baldwin and Gu (2003) Canada OLS Yes Girma et al. (2004) UK Matched samples Yes Hahn (2004) Korea OLS Yes Arnold and Hussinger (2004) Germany Matched samples No Alvarez and Lopez (2004) Chile OLS No De Loecker (2005) Slovenia Matched samples Yes

� Method 2: Dynamic specifications = β + β + β + β + ω + ε ln Y D ln Y ' X − − it 0 1 it 1 2 it 1 2 it it it Y it plant performance measure D it-1 = 1 if plant i exported in t -1 X it controls (e.g., industry, year) ? it plant-specific effect unobserved by the econometrician

Study Country Estim. method LBE? Clerides et al. (1998) Colombia, Morocco FIML w/export No participation equation GMM Some (Morocco) Kraay (1999) China IV (lagged export Yes intensity) Bigsten et al. (2004) Cameroon, Ghana, Kenya, FIML w/ export No Zimbabwe participation equation FIML nonparametric Yes errors Blalock and Gertler (2004) Indonesia Various (current Yes exports) Van Biesebroeck (2005) Cameroon, Cote d’Ivoire, SYS-GMM Yes Ghana, Kenya, Tanzania, FIML w/ export Yes Zambia, Zimbabwe participation equation Olley and Pakes (1996) Yes

What previous studies have in common? • Most measure export involvement using an export participation dummy variable � But shouldn’t learning-by-exporting depend on the extent of exposure to export markets? • Most use the whole sample of manufacturing plants � But are all manufacturing plants equally likely to learn?

What is learning by exporting? • Define learning-by-exporting drawing on Arrow (RES, 1962) characterization of learning-by-doing : � “Learning is the product of experience. Learning can only take place through the attempt to solve a problem and therefore only takes place during activity” (p. 155). � “Learning associated with repetition of essentially the same problem is subject to sharply diminishing returns… To have steadily increasing performance, then, implies that the stimulus situations must themselves be steadily evolving rather than merely repeating” (pp. 155-6).

What do firms learn from exporting? • To adopt stringent technical standards demanded by sophisticated consumers • To use new, more efficient equipment that might need to be introduced for export production • To meet orders in a timely fashion and ensure the quality of their products

Implications for measurement � If learning is the product of experience, then learning-by-exporting should depend on measures of export experience rather than export participation � If learning is subject to sharply diminishing returns, then learning-by-exporting should not be observed in established exporters => focus on young exporters � Firms might learn both by exporting and “by doing” => we control for learning-by-doing effects in our estimation of learning-by-exporting effects

Experience measures • Output experience: o Plant age − t 1 ∑ = YE Y / Y o Index of cumulative production up to t -1: τ it i i , FY i τ = FY i o In the regressions we enter these variables in reciprocal form • Export experience: o Number of years plant exported up to t -1 − t 1 ∑ = EE E / E o Index of cumulative exports up to t -1: τ it i i , FE i τ = FE i

Empirical specification • Production function: ( ) β β β = β Y L M K exp ' Q A l m k it it it it q it it ( ) = β + β + ω + ε A exp YE EE it YE it EE it it it L it labor M it materials K it capital Q it vector of factor quality measures, including skill intensity S it , wage premium W it , and capital vintage V it A it total factor productivity

Estimation • Two-step approach 1) Olley-Pakes and Levinsohn-Petrin assumptions: a. ? it follows a Markov process b. Manager chooses variable inputs based on the expected value of ? it and the other state variable, capital 2) To control for differences in input quality, we add an additional state variable, capital vintage, and two additional inputs, skill ratio and wage premium 3) For now we assume that the choice of inputs is uncorrelated with output and export experience

Estimation (Cont’d) 4) In the first step we estimate using Levinsohn and Petrin ( ) ( ) β β β = β ω + ε Y L M K exp ' Q exp l m k it it it it q it it it = ω + ε ˆ ˆ 5) Our measure of TFP is ˆ a it it it 6) In the second step we estimate = β 0 + β + β + + ˆ a YE EE f u it YE it EE it i it -Notice that we add a fixed plant effect

Estimation (Cont’d) 7) Because ? it follows a Markov process, the error term u it is likely to be autocorrelated. We model it as = ρ + ν ρ ≤ ν σ u u , 1 , ~ i . i . d .( 0 , ) − ν it it 1 it it If ? ? 0, the estimating equation can be written as ( ) ( ) = − ρ β + ρ + β − ρ a ˆ 1 a ˆ YE YE − − it 0 it 1 YE it it 1 ( ) ( ) + β − ρ + − ρ + ν EE EE 1 f − EE it it 1 i it Taking first differences to get rid of the fixed effect we get ∆ = ρ ∆ + β ∆ − ρβ ∆ ˆ ˆ a a YE YE − − it it 1 YE it YE it 1 + β ∆ − ρβ ∆ + ∆ ν EE EE − EE it EE it 1 it

Estimation (Cont’d) 8) As we know from the literature on dynamic panel data ∆ ∆ ν ≠ models, ˆ . Also, using the cumulative Cov ( a , ) 0 − it 1 it ∆ = ∆ production index as output experience, , YE Y 1 / Y − it it i , FY i which is also correlated with the error term. 9) According to the self-selection into export markets hypothesis, past and current favorable productivity shocks can facilitate a firm’s access to export markets. We assume, following Kraay (1999) that future productivity shocks are uncorrelated with export it ν = > participation: . Therefore, the Cov ( E , ) 0 , s 0 + it s first differences of our measures of export experience are correlated with the error term.

Estimation (Cont’d) 10) Conclusion: we need to use an IV method. In the paper we use Blundell and Bond’s (1998) system- GMM estimator. 11) Notice that if ? = 1, the model simplifies to ∆ ˆ = β ∆ + β ∆ + ν a YE EE it YE it EE it it In this case we do not need to instrument for the differences in experience variables because they are transformations of lagged output and exports, which are uncorrelated with the error term at time t .

Data • Plant-level data from manufacturing census of Colombia 1981-1991 (DANE) • Output and intermediates deflated by industry-level price indexes, accounting for exports and imported materials • Main sample includes plants born in or after 1981 o No export data before 1981 o Exclude plants with three or less consecutive observations, in industries with less than 100 observations, first year of each plant, and outliers o 15,457 plant-year observations (3,091 plants)

Effect of export experience on plant TFP: System-GMM estimation – Assumption: ? ? 0 1 2 0.092 ** Number of years plant exported (0.037) -0.097 ** Lagged number of years plant exported (0.043) Cumulative export index 0.0013 (0.0018) Lagged cumulative export index -0.0010 (0.0024) 0.995 *** 0.999 *** Lagged productivity (0.008) (0.006) Sample of plants with 8 or more observations; robust standard errors in parentheses; regressions include year dummies

o Our estimates of ? are not significantly different from 1, with p -values of 0.25 and 0.47. ˆ a o Unit root tests on confirm this result. it o Therefore, we feel we are justified to estimate the model with the simpler first difference specification. o We show results for different subsamples o In all cases we control for current exports by adding a current exports dummy (in first differences) to the estimating equation.