NEW EXPORTER DYNAMICS

• Author: Jonathan L. Willis,Kim J. Ruhl
• DOI: http://doi.org/10.1111/iere.12232
• Date: 01 August 2017
• Published date: 01 August 2017

INTERNATIONAL ECONOMIC REVIEW

Vol. 58, No. 3, August 2017

NEW EXPORTER DYNAMICS∗

BYKIM J. RUHL AND JONATHAN L. WILLIS1

Pennsylvania State University, U.S.A; Federal Reserve Bank of Kansas City, U.S.A.

We document that new exporters initially export small amounts, grow gradually, and are most likely to exit

the export market in their first few years. We find that the standard sunk-cost model cannot replicate these new

exporter dynamics: New exporters grow too large too quickly and live too long. In a modified sunk-cost model

that can account for these facts, the entry costs needed to match the data are three times smaller than in the

sunk-cost model. Dynamic models with richer plant-level heterogeneity are needed.

1. INTRODUCTION

Models in which heterogeneous plants face fixed export entry costs have become important

tools for studying international trade patterns, measuring trade frictions, and evaluating policy.2

These models, which are typically focused on steady-state analysis, have been successful in

replicating several key features of the plant-level data in the cross section, such as the low

export participation rate and the fact that exporters are larger than nonexporters.

Although the steady-state properties of this class of models have provided insight into the

plant’s export decision, the presence of a sunk entry cost makes the dynamic content of these

models extremely rich.3The key innovation in the dynamic models is that plants enter and

exit the foreign market in response to changes in relative prices and productivity. In seminal

work, Baldwin (1988), Baldwin and Krugman (1989), and Dixit (1989) demonstrated that the

hysteresis implied by these models is important for understanding the effects of large and

small exchange rate movements. More recent work has incorporated this type of industrial

structure into stochastic dynamic models disciplined by increasingly accessible plant-level data.

For example, Ghironi and Melitz (2005) and Alessandria and Choi (2007) use these types of

models to study how the inclusion of plants’ exporting decisions affects real exchange rate and

net export dynamics over business cycles, and Ruhl (2008) demonstrates how export entry can

produce asymmetric responses to temporary and permanent changes in expected export profits.

Das et al. (2007) estimate a dynamic structural model of export entry and exit and use it to study

the impact of trade policy. These models have typically focused on the aggregate implications

of export entry and exit.

In this article, we ask whether sunk-cost models can reproduce the dynamics of new exporters.

We begin by using plant-level data on Colombian manufacturers to document two key properties

of new exporter behavior. First, we show that new exporters begin by exporting small amounts

∗Manuscript received July 2014; revised March 2016.

1We would like to thank George Alessandria, Costas Arkolakis, Roc Armenter, Tim Kehoe, and two anonymous

referees for helpful comments. We thank Mark Roberts and Jim Tybout for access to the Colombian plant data.

This work was undertaken with the support of the National Science Foundation under grant SES-0536970. The views

expressed herein are solely those of the authors and do not necessarily reflect the views of the Federal Reserve Bank

of Kansas City.

Please address correspondence to: Kim J. Ruhl, Department of Economics, Pennsylvania State University, 615 Kern

Building, University Park, PA 16802. E-mail: kjr42@psu.edu.

2Notable papers in this expansive literature include Melitz (2003), Helpman et al. (2008), and Eaton et al. (2011).

3In this class of models, a plant pays a fixed entry cost to enter the export market (f0) and a fixed cost in each period

thereafter to continue exporting (f1). The difference between the two (f0−f1) is the part of the entry cost that is

irrecoverable—the part of the entry cost that is sunk.

703

C

(2017) by the Economics Department of the University of Pennsylvania and the Osaka University Institute of Social

and Economic Research Association

704 RUHL AND WILLIS

Ratio of exports to total sales

0 1 2 3 4

0.00

0.03

0.06

0.09

0.12

0.15

0.18

years since export entry

export−total sales ratio

Continuing exporter average (data)

New exporter average (data)

Conditional survival rate

0 1 2 3 4

0.50

0.60

0.70

0.80

0.90

1.00

years since export entry

conditional survival rate

New exporters (data)

All exporters (data)

(a) (b)

FIGURE 1

NEW EXPORTER DYNAMICS [COLOR FIGURE CAN BE VIEWED AT WILEYONLINELIBRARY.COM]

relative to total production and—conditional on continuing in the export market—gradually

expand their export volumes over several years. This can be seen in Figure 1(a), in which we

plot the average export to total sales ratio (the export–sales ratio) of the new exporters in our

sample. Second, we show that, compared to plants that have exported for several years, new

exporters are more likely to exit the export market. In Figure 1(b), we plot the share of plants

that export a+1 years after entry, conditional on having exported for ayears. It takes three

years for a new exporter’s survival rate to level off. Our two findings have been subsequently

confirmed in other data: Kohn et al. (2016), for example, document similar patterns among

Chilean manufacturers.

Next, we construct a dynamic stochastic model of a plant’s export decision and calibrate it to

the cross-sectional facts that are commonly used to parameterize fixed entry cost models. We

find that the model cannot—either qualitatively or quantitatively—reproduce the behavior of

new exporters that we observe in the data. In the model, as opposed to the data, an exporter

sells the most in the export market (relative to its total sales) in its first periods of exporting,

and there is no gradual growth in export sales in subsequent years. The model is also unable

to generate exit rates that decline with the number of years that a plant has exported. In the

model, a plant is least likely to exit the foreign market in its first few years of exporting.

The failure of the standard sunk-cost model to capture the dynamics of new exporters is

important. The present value of exporting—the key determinant of export entry—is fundamen-

tal to understanding both the plant-level and the aggregate responses to trade policy and to

measuring export barriers, such as the costs of export entry. The present value of exporting is

determined by both how long a plant expects to stay in the export market and when export

profits are realized. By generating new exporters that live too long and export too much too

soon, the model significantly overestimates the value of exporting. At the plant level, this im-

plies that the estimated export entry costs will be overstated, which we quantify in Section 6.

Export entry costs fall by a factor of three when we augment the sunk-cost model so that export

volumes grow as slowly as they do in the data.4Although the estimated entry cost falls when we

account for new exporter dynamics, this does not mean that sunk-costs are no longer important.

We still estimate that a significant portion of the entry cost is sunk.

Why does the sunk-cost model fail to capture the dynamic properties of new exporters? With

fixed entry costs and mean reverting shocks to productivity and the real exchange rate, the

4In the model with stochastic entry cost, 70% of entering plants pay no entry cost and 30% pay an entry cost that is

61% as large as the one in the standard model. See Subsection 6.2 for details.