Pitfalls in the use of foreign direct investment statistics

• Published date: 01 October 2019
• Author: Frank Barry,Clare O'Mahony
• Date: 01 October 2019
• DOI: http://doi.org/10.1111/twec.12836

World Econ. 2019;42:2835–2853. wileyonlinelibrary.com/journal/twec

|

2835

© 2019 John Wiley & Sons Ltd

Received: 18 June 2018

|

Accepted: 7 March 2019

DOI: 10.1111/twec.12836

ORIGINAL ARTICLE

Pitfalls in the use of foreign direct investment

statistics

ClareO'Mahony1

|

FrankBarry2

1Technological University Dublin, Dublin, Ireland

2Trinity College Dublin, Dublin, Ireland

KEYWORDS

balance of payments, foreign direct investment, foreign direct investment statistics, special purpose entity

These data are not clean....... One thing that would help is more information about the

data and how they are collected. What procedures do...... countries follow to collect

these data? What are the pitfalls?

Engel (2003, p. 115).

1

|

INTRODUCTION

Because of their ease of accessibility and extent of coverage, foreign direct investment (FDI) data—

which derive largely from the balance of payments—are widely used both for international compara-

tive purposes and to track the performance of individual economies over time. Although these data

measure financial flows and stocks of FDI, they are also used to proxy for multinational enterprise

(MNE) activity data, which are less widely available and lack standardisation.1 FDI data are also in-

cluded in international metrics such as globalisation indexes and macroeconomic performance

measures.2

1 Though Clegg (1992) describes FDI and MNE activity data as “two sides of the same coin,” the relationship between the

two is in fact very tenuous (Lipsey, 2003). Barry and O'Mahony (2005) and Griffith (1999) find MNE investments to be

larger and less volatile than the FDI data for US investment in Irish manufacturing and the UK transport equipment industry,

respectively. Investments funded from host‐country sources are not counted as FDI. Dunning and Lundan (2008) are

incorrect in suggesting, however, that FDI can be regarded as a lower bound on MNE investment, since much current FDI

passes through conduit entities without impacting on local investment. Blanchard and Acalin (2016) look at this latter issue in

relation to emerging markets. Geographical allocations have also become increasingly problematic as conduit entities

proliferate, partly driven by the increasing importance of intangible assets (Lipsey, 2010).

2 For example, in addition to the inclusion of the Net International Investment Position in the European Union's

Macroeconomic Imbalances Procedure (MIP), the European Commission uses various measures of FDI flows and stocks as

auxiliary indicators of economic performance (O'Farrell, 2015).

2836

|

O'MAHONY ANd BARRY

In the empirical literature, the FDI data have been used to study the impact of inward investment on

growth (see Alfaro, Kalemli‐Ozcan, & Sayek, 2004; Borensztein, De Gregorio, & Lee, 1998; Durham,

2004; Li & Liu, 2005; Van Hulten & Webber, 2010) and on other outcomes of interest (see Demir,

2016; Pica & Rodríguez Mora, 2011). The determinants of FDI flows have also been extensively

studied (see Cleeve, Debrah, & Yiheyis, 2015; De Ménil, 1999; Globerman & Shapiro, 2002; Head &

Ries, 2008; Petroulas, 2007; Razin & Sadka, 2007; Schiavo, 2007; Stein & Daude, 2007; Wei, 2000;

Wood, Yin, Mazouz, & Cheah, 2014).

The main international sources of FDI data are the International Monetary Fund (IMF), the

Organisation for Economic Cooperation and Development (OECD), the United Nations Conference

for Trade and Development (UNCTAD) and the European Statistical Agency (Eurostat). OECD data

are popular in studies that focus on industrialised countries or where geographically disaggregated

data are required (see Borensztein etal., 1998; De Ménil, 1999; De Sousa & Lochard, 2011; Durham,

2004; Hatzius, 2000; Head & Ries, 2008; Pica & Rodríguez Mora, 2011; Razin & Sadka, 2007; Stein

& Daude, 2007; Wei, 2000). Eurostat data are often used when the focus is on the EU or on the effects

of monetary union (see Petroulas, 2007), while IMF or UNCTAD data allow for the inclusion of de-

veloping countries. Alfaro etal. (2004), Durham (2004) and Van Hulten and Webber (2010) use IMF

data; Cleeve etal. (2015), Globerman and Shapiro (2002), Gui‐Diby and Renard (2015), Li and Liu

(2005), and Wood etal. (2014) employ UNCTAD data. Some studies, such as Demir (2016), Durham

(2004), and Taylor (2008), combine data from two or more sources.

Large datasets are often preferred covering a large a number of countries for as long a time period

as possible. Alfaro etal. (2004) have 72 countries in their dataset, Borensztein etal. (1998) 69; Demir

(2016) 134; Durham (2004) 80; Gui‐Diby and Renard (2015) 47; and Li and Liu (2005) 84. Of these

studies, each of which spans at least two decades, only Li and Liu (2005, p.393) refer to concerns

about data quality, partly attributing the mixed results in the literature to “data insufficiency.”3

The present paper examines the consistency in the construction and coverage of the FDI data. To

minimise the extent of methodological variation, we focus on six Western EU economies as a reason-

ably homogenous group of countries with well‐developed data collection and dissemination systems.

Asymmetries in datasets containing broader ranges of countries are expected to be far greater than

those unearthed here. In an earlier study on Ireland's inward FDI, Barry and O'Mahony (2005) found

a large number of structural breaks and a lack of comparability between the various international FDI

data sources due to methodological differences, inconsistencies in items included and varying treat-

ments of financial intermediation. Similar concerns have been raised by Bellak (1998), Fujita (2008)

and Stephan and Pfaffmann (2001).

Among our selection of countries, we include the France, Germany and the UK, typically the

three largest EU recipients of FDI. Since FDI can be more significant for smaller economies, we also

include three small economies: Belgium, Luxembourg and the Netherlands.4 Prior to 2002, Belgium

and Luxembourg produced a common balance of payments as the Belgium‐Luxembourg Union

(BLEU).5

3 Gaps in the data are sometimes filled by substituting partner‐reported outward data for missing inward data (see Hatzius,

2000), the consequences of which are explored below.

4 While each has a tax regime that is conducive to FDI, they differ in other significant ways. Conduit investment is particu-

larly important for Luxembourg and the Netherlands. UNCTAD (2006) estimates that pass‐through FDI accounted for 95% of

FDI inflows into Luxembourg in 2002–05. Our calculations based on Dutch Central Bank data suggest that, for the

Netherlands, inflows including Special Financial Institutions (SFIs) were, on average, more than four times higher per annum

over 1999–2012 than those reported by the OECD, which did not include this type of FDI.

5 The BLEU sometimes appears as “Belgium” in datasets and in empirical papers.