Time, uncertainty and trade flows

• Date: 01 September 2020
• Author: Mauro Boffa,Ben Shepherd,Jean‐François Arvis,José Ansón,Matthias Helble
• Published date: 01 September 2020
• DOI: http://doi.org/10.1111/twec.12942

World Econ. 2020;43:2375–2392. wileyonlinelibrary.com/journal/twec

|

2375

© 2020 John Wiley & Sons Ltd

Received: 15 February 2017

|

Revised: 9 August 2017

|

Accepted: 27 January 2020

DOI: 10.1111/twec.12942

ORIGINAL ARTICLE

Time, uncertainty and trade flows

JoséAnsón1

|

Jean-FrançoisArvis2

|

MauroBoffa3

|

MatthiasHelble4

|

BenShepherd5

1UPIDO AG, Bern, Switzerland

2World Bank, Washington, DC, USA

3Universal Postal Union, Bern, Switzerland

4Asian Development Bank, Manila, Philippines

5Developing Trade Consultants, New York, NY, USA

KEYWORDS

international trade, logistics, postal delivery, trade time, uncertainty

1

|

INTRODUCTION

Trade costs are a major determinant of the observed pattern of trade and production across countries.

They consist of many factors, and the best estimates available suggest that total trade costs remain high

even though they have fallen substantially in recent years (Anderson & Van Wincoop, 2004; Arvis,

Duval, Shepherd, Utoktham, & Raj, 2016). Applied international trade work based on the gravity

model typically uses a set of proxies for bilateral trade costs, with geographical distance playing a

central role as an indicator of transport costs. However, distance is a very rudimentary measure, as it

neglects transshipments and geographical obstacles. One main element of trade costs that has already

been identified in the literature is time (Evans & Harrigan, 2005; Hummels, 2001). For example,

Djankov, Freund, and Pham (2010) use Doing Business data on the time taken to export—moving

goods from the producer's factory to the dock—to show that an extra day's delay can decrease bilateral

trade by around 1%.

Another salient feature of the current world economy is the extension of global value chains (GVCs)

to sectors and countries that have not traditionally been included in them. GVCs are now apparent,

at least in rudimentary form, in many parts of the world. Lead firms, often based in developed coun-

tries but sometimes also in major emerging markets, coordinate complex networks of suppliers that

provide the goods and services needed to produce a final product that is then shipped to a destination

market. The GVC business model relies on the swift movement of intermediate goods across borders,

something that occurs numerous times during the production process. Even more important, however,

is the way in which GVCs deal with risk. Clearly, delay is a major risk for the just-in-time production

methods used by these business models. Basic intuition suggests that maintaining inventories would

be a rational response. But doing so is costly, and the management techniques favoured by GVCs try to

2376

|

ANSÓN et Al.

keep inventory carrying costs to an absolute minimum. It is therefore vital that delivery of goods not

only be speedy, but that the time required be certain. Indeed, businesses often report anecdotally that

they can manage delay by building it into their production model, but uncertainty creates risk, which

in turn increases costs and decreases their competitiveness. In the GVC context, lead firms will tend

to source goods from markets that can provide a narrow window of delivery times. Countries where

time delays are highly uncertain will have difficulty joining GVCs and leveraging international trade

for their development.

The importance of uncertainty as a source of costs for businesses involved in GVCs is reflected at

a policy level. For example, the 21-member Asia Pacific Economic Cooperation (APEC) has adopted

a Supply Chain Connectivity Framework Action Plan, in which economies committed to reduce the

time, cost and uncertainty associated with supply chain transactions by 10% by 2015. Given the prom-

inence of the issue, it is notable that the applied international trade literature has not yet dealt with it.

The key contributions on the role of time in determining bilateral trade flows—Djankov et al. (2010),

and Hummels and Schaur (2013)—have focused on the central tendency of the time distribution,

rather than its second moment, which would capture uncertainty.

Against this background, we build on and extend the previous literature in two main ways. First,

we use a large Universal Postal Union (UPU) database of parcel deliveries to derive new data on

bilateral international transport times for 161 exporting countries and 165 importing countries. Our

data are based on individual transaction records tracked using parcel scans and reported to the UPU

by national postal authorities. As a result, our measures of time are richer than the consensus esti-

mates from trade specialists used by Djankov et al. (2010) and the schedule data used by Hummels

and Schaur (2013).

Second, we exploit the fact that the original database is recorded at the level of individual postal

transactions to calculate bilateral measures of time uncertainty on all trade routes. As a proxy for

uncertainty, we use the standard deviation of recorded international transit times at the transaction

level. This indicator captures the dispersion of transport times around their central tendency and

can reasonably be said to be one, although not the only, proxy for the uncertainty facing firms when

they make shipments. This is the first time that such a measure has been calculated and brought

into contact with international trade data, and the core novelty of our paper lies in our ability to

draw conclusions on the effect of time-based uncertainty on bilateral trade, in addition to average

or median time.

To estimate the effects of international transit time and uncertainty on bilateral trade flows, we use

a standard theory-based gravity model. The variables of interest vary at the bilateral level, so we can

control for multilateral resistance and economic size using fixed effects. We estimate by Poisson pseu-

do-maximum likelihood to deal with heteroscedasticity and zero trade observations. The model fits

the data well and provides statistically significant results on the coefficients of interest, even with the

inclusion of standard gravity model controls including distance, historical and geographical factors,

and membership of a regional trade agreement.

The paper proceeds as follows. The next section discusses our data set, focusing on the inter-

national transport time data. Section 3 presents our empirical model, implements it and discusses

results. The final section concludes and discusses possible policy implications of our findings.

2

|

DATA AND PRELIMINARY ANALYSIS

The data set for this paper consists of standard bilateral trade data and gravity controls, along with

exploitation of a new source on international transport times based on parcel delivery data from