Skip to main content
Log in

Is agglomeration a free lunch for new exporters? Evidence from Chile

  • Original Paper
  • Published:
The Annals of Regional Science Aims and scope Submit manuscript


The extant empirical studies do not wholly support the popular belief held by policy-makers on the positive influence of geographical proximity of incumbent exporters on the export decision of a non-exporter. To reconcile the mixed evidence, this paper examines the nonlinear relationship between the agglomeration of exporters and export decisions. In brief, the results show that the former positively affects the latter. Furthermore, the squared term of agglomeration calculated only by the number of skilled workers is negatively and significantly associated with export decisions, which can be explained by congestion costs in a local labor market for the skilled worker. As a result, these findings suggest an inverted U-shaped relationship between the agglomeration of exporters and the probability of being an exporter.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Similar content being viewed by others


  1. The entry costs for exporting are related to gathering information on foreign consumers’ preference, establishing a distribution system in foreign countries, improving infrastructure necessary for distributing the products abroad, paying the search costs to identify local bankers, networking, adopting the product to new standards, and so on.

  2. The mixed evidence may be attributed to the definition of the local export spillovers (restricted to multinational firms or including all exporters), the level of data disaggregation, and the different estimators used (Koening et al. 2010).

  3. This paper employs the Levinsohn and Petrin (2003) method, which uses intermediate input, such as electricity, as the instrument variable in order to address the endogeneity problem. In order to estimate the production function, this study implements a STATA command (i.e., levpet) introduced by Petrin et al. (2004).

  4. However, multiple plants have the same identification number. These plants are excluded as their percentage in the sample is only around 0.5 %.

  5. Although Angrist and Pischke (2009, p. 196) dealt with peer effects in labor economics, the way to compute peer effects is the same as that for the agglomeration effect. Therefore, their suggestion is applied: “the best shot a causal investigation of peer effects focuses on variation in ex ante peer characteristics, that is, some measure of peer quality that predates the outcome variable and is therefore unaffected by common shocks.” Koenig (2009) and Bernard and Jensen (2004) also employed this approach to avoid the endogeneity problem.

  6. A fixed effects (FE) model is preferable for its assumption that unobserved heterogeneity is random as well as the non-specification of its distribution. For the dynamic FE logit model, Honore and Kyriazidou (2000) proposed a semi-parametric estimator, which is, however, extremely data-demanding. Additionally, as the dynamic FE logit model does not assume the distribution of unobserved heterogeneity, the marginal effects cannot be computed to quantify the research interest (Wooldridge 2005). Thus, the dynamic FE logit estimator is not used.

  7. Stewart’s (2006) STATA code is used, which suggests a shortcut implementation of Heckman’s estimator, i.e., rebprob. Despite Stewart’s shortcut implementation, the Heckman estimator still requires considerable computing time.

  8. According to Stewart (2006), Wooldridge assumes normality of the conditional distribution of unobserved heterogeneity, given the initial observation of the dependent variable, whereas Heckman assumes bivariate normality for the unobserved heterogeneity and the initial observation of the dependent variable.

  9. The Wooldridge’s method needs a set of covariates in order to use Mundlak’s suggestion. The set of covariates is constructed by using the logarithms of TFP, of plant size, of real wage per worker, and of agglomeration of exporting plants. However, the logarithms of real export subsidies and of foreign ownership are excluded due to the presence of zero values.

  10. Heckman’s methods need a set of covariates in order to approximate the conditional density of initial observations, given a vector of the unobserved heterogeneity. The set of covariates is constructed by using the log of plant-level TFP.

  11. This can be explained by the fact that exporting plants can pay higher wages to skilled workers than non-exporting plants can because of the higher productivity of exporting plants (Bernard and Jensen 1999; Kandilov 2009). In particular, exporting plants can pay their skilled workers a higher wage to prevent them from moving to a local competitor, i.e., a non-exporting plant. This is similar to foreign affiliates behavior (Fosfuri et al. 2001; Combes and Duranton 2006).


  • Akay A (2009) The Wooldridge method for the initial values problem is simple: what about performance? IZA discussion paper no. 3943

  • Accetturo A, Bugamelli M, Lamorgese A (2013) Skill upgrading and exports. Econ Lett 121:417–420

    Article  Google Scholar 

  • Andersson M, Weiss JF (2012) External trade and internal geography: local export spillovers by industry characteristics and firm size. Spat Econ Anal 7(4):421–446

    Article  Google Scholar 

  • Angrist JD, Pischke J-S (2009) Mostly harmless econometrics. Princeton University Press, Princeton

    Google Scholar 

  • Atiken B, Hanson GH, Harrison AE (1997) Spillovers, foreign investment, and export behavior. J Int Econ 43(1):103–132

    Article  Google Scholar 

  • Barrios S, Gorg H, Strobl E (2003) Explaining firms’ export behavior: R&D. Spillovers and the destination market. Oxf Bul Econ Stat 65(4):476–496

    Google Scholar 

  • Balsvik R (2011) Is labor mobility a channel for spillovers from multnationals? Evidence from Norwegian manufacturing. Rev Econ Stat 93(1):285–297

    Article  Google Scholar 

  • Bernard AB, Jensen JB (1999) Exceptional exporter performance: cause, effect, or both? J Int Econ 47(1):1–25

    Article  Google Scholar 

  • Bernard AB, Bradford Jensen J (2004) Why some firms export. Rev Econ Stat 86(2):561–569

    Article  Google Scholar 

  • Breschi S, Lissoni F (2001) Knowledge spillovers and local innovation system: a critical survey. Ind Corp Change 10(4):975–1005

    Article  Google Scholar 

  • Castellani D, Zanfei A (2007) Multinational companies and productivity spillovers: is there a specification error? Appl Econ Lett 14(14):1047–1051

    Article  Google Scholar 

  • Choquetter E, Meinen P (2015) Export spillovers: opening the black box. World Econ 38(12):1912–1946

  • Clerides SK, Lach S, Tybout JR (1998) Is learning by exporting important? Micro-dynamic evidence from Colombia, Mexico, and Morocco. Q J Econ 113(3):903–947

    Article  Google Scholar 

  • Combes PP, Duranton G (2006) Labour pooling, labour poaching, and spatial clustering. Reg Sci Urban Econ 36(1):1–28

    Article  Google Scholar 

  • Farole T, Winkler D (2013) Firm location and the determinants of exporting in low- and middle income countries. J Econ Geogr 14(2):395–420

    Article  Google Scholar 

  • Fosfuri A, Motta M, Ronde T (2001) Foreign direct investment and spillovers through workers’ mobility. J Int Econ 53(1):205–222

    Article  Google Scholar 

  • Greenaway D, Kneller R (2008) Exporting. Productivity and agglomeration. Eur Econ Rev 52(5):919–939

    Article  Google Scholar 

  • Harasztosi P (2015) Export spillovers in Hungary. Empirical economics, pp 1–30

  • Heckman JJ (1981) The incidental parameters problem and the problem of initial conditions in estimating a discrete time-discrete data stochastic process. In: Manski CF, McFadden D (eds) Structural analysis of discrete data with econometric applications. MIT Press, New York, pp 114–178

    Google Scholar 

  • Honore B, Kyriazidou E (2000) Panel data discrete choice models with lagged dependent variables. Econometrica 68(4):839–874

    Article  Google Scholar 

  • Karpaty P, Kneller R (2011) Demonstration or congestion? Export spillovers in Sweden. Rev World Econ 147(1):109–130

    Article  Google Scholar 

  • Kandilov IT (2009) Do exporters pay higher wages? Plant-level evidence from an export refund policy in Chile. World Bank Econ Rev 23(2):269–294

    Article  Google Scholar 

  • Koenig P (2009) Agglomeration and the export decisions of French firms. J Urban Econ 66(3):186–195

    Article  Google Scholar 

  • Koening P, Mayneris F, Poncet S (2010) Local export spillovers in France. Eur Econ Rev 54(4):622–641

    Article  Google Scholar 

  • Levinsohn J, Petrin A (2003) Estimating production functions using inputs to control for unobservables. Rev Econ Stud 70(2):317–341

    Article  Google Scholar 

  • Masso J, Roigas K, Vather P (2014) Foreign market experience, learning by hiring and firm export performance. CELSI discussion paper no. 26

  • Melitz M (2003) The impact of trade on intra-industry reallocations and aggregate industry productivity. Econometrica 71(6):1695–1725

    Article  Google Scholar 

  • Mion G, Opromolla L (2014) Managers’ mobility, trade performance, and wages. J Int Econ 94:85–101

    Article  Google Scholar 

  • Molina D, Muendler M (2013) Preparing to export. NBER working papers #18962

  • Mundlak Y (1978) On the pooling of time series and cross section data. Econometrica 46(1):69–85

    Article  Google Scholar 

  • Petrin A, Poi B, Levinsohn J (2004) Production function estimation in stata using inputs to control for unobservables. Stata J 4(2):113–123

    Google Scholar 

  • Saito H, Gopinath M (2009) Plants’ self-selection, agglomeration economies and regional productivity in Chile. J Econ Geogr 9(4):539–558

    Article  Google Scholar 

  • Stewart MB (2006) -Redprob- a Stata program for the Heckman estimator of the random effects dynamic probit model. Mimeo, University of Warwick

  • Verhoogen EA (2008) Trade, quality upgrading, and wage inequality in the Mexican manufacturing sector. Quart J Econ 123:489–530

    Article  Google Scholar 

  • Wooldridge JM (2005) Simple solutions to the initial conditions problem in dynamic. Nonlinear panel data models with unobserved heterogeneity. J Appl Econ 20(1):39–54

    Article  Google Scholar 

Download references


I would like to give thanks to Robert McNown, Jin-Hyuk Kim, anonymous referee and the Editor Martin Andersson for valuable comments and suggestions. I am also grateful to Luis Castro for kindly sharing the Chilean plant-level data which comes from the Chilean National Statistics Institute (INE). This work was supported by the Soongsil University Research Fund of 2015.

Author information

Authors and Affiliations


Corresponding author

Correspondence to Youngho Kang.

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Kang, Y. Is agglomeration a free lunch for new exporters? Evidence from Chile. Ann Reg Sci 57, 195–222 (2016).

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI:

JEL Classification