\documentclass[12pt]{article}
\usepackage{comment}
%\usepackage{caption}
\usepackage{subcaption}
\usepackage{caption}
\usepackage[utf8]{inputenc}
\usepackage{geometry}
\geometry{verbose,tmargin=2cm}
\usepackage{array}
\usepackage{float}
\usepackage{booktabs}
\usepackage{textcomp}
\usepackage{url}
\usepackage{amsmath}
\usepackage{amsthm}
\usepackage{amssymb}
\usepackage{graphicx}
\usepackage[authoryear]{natbib}
\usepackage[unicode=true,pdfusetitle, bookmarks=true,bookmarksnumbered=false,bookmarksopen=false, breaklinks=false,pdfborder={0 0 0},pdfborderstyle={},backref=false,colorlinks=false] {hyperref}
\usepackage{array,booktabs}
\usepackage{longtable}
\makeatletter
\usepackage{pdflscape}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% LyX specific LaTeX commands.
%% Because html converters don't know tabularnewline
\providecommand{\tabularnewline}{\\}
%\usepackage{adjustbox}


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% User specified LaTeX commands.
\usepackage{amsfonts}
\usepackage[singlespacing]{setspace}
\usepackage[bottom]{footmisc}
\usepackage{indentfirst}
\usepackage{endnotes}
\usepackage{rotating}
\usepackage[auth-lg]{authblk}
\usepackage[authoryear]{natbib}
%\usepackage[numbers]{natbib}
%\setcitestyle{authoryear,open={(},close={)}}
\setcounter{MaxMatrixCols}{30}
\def\@biblabel#1{\hspace*{-\labelsep}}
\usepackage{apalike}
\usepackage[english]{babel}
%\date{}
%\usepackage[nolists,tablesfirst]{endfloat}
%\renewcommand{\efloatseparator}{\mbox{}}
\usepackage{amsmath}
%\usepackage{subfigure}% Support for small, `sub' figures and tables


%\usepackage{array,booktabs}
\usepackage{graphicx}

\makeatother



%Place figures at the end, but  not tables
%\usepackage[figuresonly]{endfloat}
%\renewcommand{\efloatseparator}{\mbox{}}
%\DeclareDelayedFloatFlavor{sidewaysfigure}{figure}

\setlength{\topmargin}{0.1in}
\setlength{\headheight}{0in}
\setlength{\headsep}{0in}
\setlength{\topskip}{0in}
\setlength{\textheight}{8.5in}
\setlength{\oddsidemargin}{0in}
\setlength{\evensidemargin}{0in}
\setlength{\textwidth}{6.5in}
\newcommand\inv[1]{#1\raisebox{1.15ex}{$\scriptscriptstyle-\!1$}}

\makeatletter
\setlength{\@fptop}{0pt}
\newenvironment{nodelay}[1][tp]{\@float{figure}[#1]}{\end@float}
\makeatother

\begin{document}


\thispagestyle{empty}
\fontfamily{phv}\selectfont 

\begin{center}
{\Large \textbf{The Multinational Revenue, Employment, and Investment Database (MREID)}\\}
%{\Large Comparing Armington Elasticities Across Methods and Aggregations: The Case of Fish\\}
\vspace{0.75in}
{\Large Saad Ahmad}\\
\vspace{0.25in}
{\Large Jeffrey Bergstrand}\\
\vspace{0.25in}
{\Large Jordi Paniagua} \\
\vspace{0.25in}
{\Large Heather Wickramarachi} \\
\vspace{0.75in}

{\large ECONOMICS WORKING PAPER SERIES}\\ 
Working Paper 2023--11--B \\ \vspace{0.5in}
U.S. INTERNATIONAL TRADE COMMISSION \\
500 E Street SW \\
Washington, DC 20436 \\
\vspace{0.25in}
Novemeber 2023
\end{center}
\vfill

\noindent The authors thank Bill Powers, David Riker, and Peter Herman for helpful comments and suggestions. Office of Economics working papers are the result of ongoing professional research of USITC Staff and are solely meant to represent the opinions and professional research of individual authors. These papers are not meant to represent in any way the views of the U.S. International Trade Commission or any of its individual Commissioners. Please address correspondence to saad.ahmad@usitc.gov.

\newpage
\thispagestyle{empty} % remove headers, footers, and page numbers from cover page
\begin{flushleft}
The Multinational Revenue, Employment, and Investment Database (MREID)  \\
Saad Ahmad, Jeffrey Bergstrand, Jordi Paniagua and Heather Wickramarachi \\
Office of Economics Working Paper 2023--11--B\\ 
Novemeber 2023\\~\\
\end{flushleft}

\vfill

\begin{abstract}
\noindent  This paper presents a new Multinational Revenue, Employment, and Investment Database (MREID). MREID offers comprehensive and consistent information on international and domestic revenue, employment, and investment variables of multinational enterprises (MNEs) for 185 countries, 25 industries, and (initially) a 12-year annual time series. The database covers a range of industries, including agriculture, mining, energy, manufacturing, and services, enabling a nearly complete description of each economy's foreign direct investment (FDI) activity. MREID currently covers the period 2010 through 2021 and is constructed using reported administrative data from Orbis.
\end{abstract}

\vfill
\begin{flushleft}
Saad Ahmad, Office of Economics\\
\href{mailto:saad.ahmad@usitc.gov}{saad.ahmad@usitc.gov}\\
\vspace{0.25in}
Jeffrey Bergstrand, University of Notre Dame\\
\href{mailto:jbergstr@nd.edu}{jbergstr@nd.edu}\\
\vspace{0.25in}
Jordi Paniagua, University of Valencia\\
\href{mailto:jordi.paniagua@uv.es}{jordi.paniagua@uv.es}\\
\vspace{0.25in}
Heather Wickramarachi, Office of Industry and Competitiveness Analysis\\
\href{mailto:heather.wickrama@usitc.gov}{heather.wickrama@usitc.gov}\\

\vspace{0.75in}

\end{flushleft}


\clearpage
\newpage 
\doublespacing
\setcounter{page}{1}

\section{Introduction}

This paper outlines the development of the Multinational Revenue, Employment, and Investment Database (MREID), a comprehensive source of information on multinational enterprises' (MNE) foreign direct investment (FDI) related activities with cross-border affiliates across 185 countries, 25 industries, and a period of 12 years. MREID includes annual data from 2010 through 2021, with 2021 being the latest year with complete information for all sectors. MREID offers bilateral sector-level data on numerous MNE variables, covering the four major sectors of agriculture, mining and energy, manufacturing, and services, and provides an exhaustive overview of FDI within each economy.

The main purpose of constructing MREID is for the statistical analysis of bilateral FDI-related variables, and we achieve this by aggregating firm-level data from ORBIS. The database contains international and domestic bilateral FDI-related variables. We do not use estimation models (such as the gravity framework) to fill in any missing observations in MREID, thus making it ideal for estimation purposes.\footnote{We include, however, estimated data from Orbis when data are unavailable.} However, MREID is not balanced and includes missing observations for some years and countries.

MREID stands out from other datasets of FDI due to its extensive sectoral coverage at the bilateral level, the number of countries covered, the inclusion of more recent years, and its overall suitability for estimation purposes. Existing FDI or multinational production datasets such as OECD and FDIMarkets only cover certain countries or FDI types. Other datasets, like that of the U.S. Bureau of Economic Analysis (BEA), provide detailed FDI data, but limited to a single economy.

The rest of the paper is organized as follows. Section \ref{sec:background} reviews the relevant literature. Section \ref{sec:overview} describes the Orbis dataset and also discusses the search strategy for constructing the MREID database. 
%describes the approach to constructing the database and provides more details on the dimensions of MREID in terms of countries, industries, and years. Section \ref{sec:search}, describes the search strategy used to construct MREID. 
Section \ref{sec:descriptiondatabase} provides a detailed description of the MREID database. Section \ref{chap:Coverage-and-Validity} compares MREID to existing multinational datasets for coverage and validity. Section \ref{sec:Conclusions} concludes and outlines future work on the database. 

\section{Background and literature review \label{sec:background}}

In a recent survey of the effects of international investment agreements (IIAs), \citet{egger2023international} highlight the limitations of FDI measurement. They note that there is no high-quality bilateral FDI data available. The existing datasets from UNCTAD, IMF, and OECD have known limitations, such as heterogeneous reporting standards and the lack of differentiation between financial (e.g., portfolio) and real FDI transactions. For example, \citet{guvenen2022offshore} show that accounting engineering practices such as profit shifting are common among US affiliates and impact the aggregate measurement of economic variables. Some official sources like the BEA are also compiled through surveys, which might be prone to measurement error.

There is a growing consensus around the advantages of using firm-level data instead of national (income) accounts data for measuring FDI. As discussed in \citet{wildmer2019}, due to profit-shifting motives, there is a divergence between FDI reported in National Income Accounts and their representation of productive activities and investments. For instance, \citet{damgaard2019} find that nearly 40 percent of reported inward FDI results from financial and tax engineering, which does not effectively benefit the ``real'' economy. By contrast, firm-level financial data, supplemented with ownership details, can be a much more reliable measure of cross-border investments and multinational firms' activities. Another advantage is the use of  uni-directional bilateral data (e.g., American investment in Spain and Spanish investment in America) instead of net bilateral data,  an average of the two-way FDI, or country-specific aggregating data from all origins. 



With its considerable coverage of countries and sectors and its detailed ownership information, Orbis has been relied upon increasingly in the recent literature for cross-country firm-level analysis. \citet{gopinath2017capital} used Orbis to examine the productivity of manufacturing firms in Spain from 1999-2012. \citet{cravino2017} used Orbis to investigate how multinational firms contribute to transmitting economic shocks across countries. \citet{alfaro2018} used Orbis to analyze the nature of productivity gains arising from multinational production in the host country. \citet{kalemliconstruct} rely on Orbis to construct a representative firm-level dataset for European countries using financial statements from the Orbis database. The authors show that small-and-medium-sized firms (SMEs) account for a large share of aggregate economic activity.

Orbis has also been used to identify the ownership links between firms. \citet{aminadav2020} use Orbis, among other sources, to investigate ownership concentration and the types of corporate control across countries. \citet{alabrese2020} rely on Orbis to map the complex linkages between parent firms and foreign affiliates and their broader implications for investment and tax policy. Applying a network framework on Orbis ownership data, \citet{rungi2017} assessed direct and indirect control of corporations within and across national borders. \citet{fonseca2023globalization} used 22,000 listed firms in Orbis to study the globalization of corporate control employing a gravity framework. 

The Bureau van Dijk phased out the dataset Zephyr, which tracked  merger and acquisition (M\&A) and ``greenfield'' transactions; several papers used this dataset to analyze FDI, e.g., \citet{liu2021missing}. More recently, Bureau van Dijk launched the Orbis Crossborder Investment Monitor, a similar dataset to FDIMarkets, which tracks greenfield investment ``announcements''; \citet{Linask2023trends} summarized the trends and features of this dataset.

Despite having detailed ownership information that allows researchers to distinguish between domestic and foreign affiliates, few studies have taken advantage of the Orbis data to build a database that captures several dimensions of multinational enterprise (MNE) activities at the bilateral level over numerous sectors and years. One exception is the EU Foreign Ownership (FOWN) dataset, constructed using Orbis firm-level data and described in detail in \citet{wildmer2019}.\footnote{Another exception is the nationally representative firm-level dataset of European countries created by \citet{kalemli2022construct, kalemliconstruct} using Orbis data.} Focusing on foreign-controlled firms that operated in the European Union (EU) for the period 2007 to 2016, the FOWN dataset allows researchers to track how investment in the EU has changed over time and which EU sectors are the ones targeted for foreign investment.\footnote{A  firm is considered as foreign-controlled if its Global Ultimate Owner in the Orbis database is registered in a country outside the EU.} Financial variables for EU countries track the revenues, total assets, and the number of employees of firms and are aggregated to the NAICS (Revision 2) two-digit sector level. Compared with official data sources on foreign investment in the EU, \citet{wildmer2019} find that the FOWN dataset provides similar trends for the number of firms and sales after 2008, but underreports slightly smaller firms before 2008. Beyond the evolution of foreign ownership in the EU, the FOWN database also provides information on  M\&A and greenfield activity in the EU by relying on some other financial data products released by Moody's. However, the FOWN dataset's insufficient coverage of certain countries and years limits its usefulness for a broader analysis of cross-border investment and MNE activities \textit{worldwide}.  

An additional limitation of existing FDI datasets is the lack of accounting for domestic investment, which is important for empirical estimates that generally rely upon structural gravity frameworks. For example, domestic investment is important to identify country-specific variables using the structural gravity equation as shown by \cite{heid2021estimating} for trade and \cite{carril2022does} for greenfield FDI, and \cite{carril2022border} for M\&As.

The MREID dataset, however, is unique by including \textit{comparable} information on revenues, employment, and assets by ownership and by type of investment.
 
\section{MREID: An Overview\label{sec:overview}}
\subsection{Data source: Orbis \label{sec:construction}}

Research on FDI activity is challenged due to different measures
of FDI, types of FDI, and data limitations. Establishing a foreign affiliate can be recorded in many ways (e.g., capital investment, employment) and executed in various ways (e.g., greenfield investment or merger and acquisitions (i.e., M\&As)). We employ a search strategy from Orbis to overcome several of these limitations. 

Orbis is Bureau van Dijk's (a Moody's Analytics company) flagship-company database with data from more than 425 million companies worldwide. It focuses on private company information and 
presents companies' variables in comparable formats.\footnote{The MREID database we construct will consist of publicly owned and privately owned corporate firms with assets or sales larger than USD 1 million; hence, most will be publicly owned. It excludes state-owned enterprises and banks. FDI requires ownership of 50.01 percent or larger. Banks are excluded. International generally accepted accounting standards are used.} The sources of information come from over 170 different providers, which are standardized into comparable cross-country information.  Figure \ref{fig:Orbis-overview-1} summarizes Orbis' linkages and country coverage.


\begin{figure}[h]

\caption{Orbis linkages and geographical coverage\label{fig:Orbis-overview-1}}
% Alt Text: Figure 1 is an illustration summarizing Orbis' linkages and country coverage.
\centering
\includegraphics[scale=0.8]{orbis2}

Source: \url{www.bvdinfo.com}
\end{figure}

Orbis is a popular resource among economists. \citet{kalemli2015construct}
where the first to describe the standard benchmark-search strategy to construct nationally
representative firm-level data from the Orbis global database. Using
this search strategy, \citet{gopinath2017capital} studied capital
stock (fixed assets), output (sales), and employees. These authors
show that Orbis data coverage is comparable to Spanish administrative
data. \citet{osnago2019deep} used Orbis to construct an FDI dataset
for several European countries and were able to distinguish vertical
and horizontal FDI. \citet{garcia2017uncovering} used Orbis data
to unravel offshore financial centers.

\subsection{Search strategy\label{sec:search}}

Our search strategy in Orbis to construct a representative FDI 
dataset from firm-level data follows the best practices in the literature. The
key variable to foreign identity ownership in Orbis is the variable
\textquotedblleft global ultimate owner\textquotedblright (GUO).\footnote{Focusing on the GUO lets us bypass some of the offshore issues that plague official FDI statistics that are based on the direct owner.}
This variable allows us to track firms that invest in foreign countries.
One of the limitations of the Orbis web interface is that the variable
GUO is only available for the \textit{current day}. This constraint has resulted in
incorrect M\&As during the last decade. To overcome this limitation,
\citet{kalemli2015construct} proposed using yearly historical data
(in disk format) to track these complex changes in ownership. More
recently, in an updated version of their original working paper, \citet{kalemli2022construct, kalemliconstruct}
used the M\&A module in Orbis to track these changes. Following this procedure, we can
obtain accurate FDI data without accessing historical data (with the limitation of the ten-year rolling period). 

This procedure also allows us to construct a comparable companion dataset recording M\&A data. Whenever an affiliate enters the dataset within the observation period (2010-2021), we flag it as a greenfield investment. This way, we construct a second comparable companion dataset recording Greenfiled data.

We limited our search to affiliates with more than USD 1 million in turnover (i.e., sales) or in total assets in at least one year in the sample. Consequently, we reduce the number of affiliates with no \textquotedblleft real'' activity. Other FDI datasets have similar thresholds (e.g., the BEA established its threshold at USD 25 million). A key feature of our search strategy is that we also include \textit{domestic establishments} (i.e., domestic affiliates). We established an ownership threshold of 50.01\%.\footnote{The Fifth Edition of the IMF Balance of Payments Manual defines the owner of 10\% or more of a company's capital as a direct investor. However, the majority control threshold (50.01\%) aligns with the IMF and OECD definition of FDI to obtain a \textit{lasting} interest by a resident entity of one economy in another.}

We selected economically active affiliates, as recorded by Orbis. We use Orbis' variable date of incorporation to fix the entry criteria of an affiliate into the MREID dataset. We have implemented criteria to detect exits from the market. Affiliates with more than four consecutive years without reports on any of the key financial variables are marked as having exited. The attrition rate with this strategy is around 8 percent of affiliates per year.

Some data in Orbis contains errors and typos from the original source. For example, some key financial variables contain negative values coded incorrectly or reflect local accounting practices. Following \citet{kalemli2022construct}, we drop all negative values.

In the Appendix (section \ref{sec:search_detail}), we outline a search example that captures the search details in ORBIS. 


\subsection{FDI Variables\label{sec:ORBIS-Variables}}

We selected the following as the key variables to obtain from Orbis for each subsidiary at the closing date of each year per 2-digit NAICS 2017 (core code). Our variables are the key financial variables selected from the global format accounting balance sheet, consolidating US and non-US accounting practices.\footnote{Detailed accounting items and formulas are accessible here: \url{https://help.bvdinfo.com/mergedProjects/65_EN/Data_Osiris/Understanding_Osiris_data_and_formats/DataFormulas/globalformatalltemus_nonus.htm}}



\begin{itemize}

\item Investment: Investment is measured as either total assets or fixed assets.
\begin{itemize}
 \item Total assets: The sum of current assets and fixed assets, including intangibles.

\item Fixed assets: Tangible fixed assets, intangible fixed assets, and other fixed assets (exploration, long-term receivables, investments, long-term associated companies, investment properties, and other long-term assets).
\end{itemize}

\item Revenue (Turnover or Sales): Total operating revenues (= net sales + other operating revenues + stock variations\footnote{The stock variation is the difference between the value of the initial inventory and the end of the fiscal year. According to international accounting practices if the stock valuation at the beginning of the fiscal year is lower than at the end of the fiscal year, this difference must be reflected as income.}) excluding taxes. However, for some companies, no information is provided on value added taxes (VAT); alternatively, the figure is stated as after indirect taxes or excluding sales-related taxes.\footnote{Some reported turnover might contain negative values. As stated earlier, we have dropped them.}

\item Number of employees: Total number of employees included in the company's payroll.

%\item Costs of employees. Personnel Expenses.



%\item Intangible fixed assets. A combined account in the Global format, including goodwill (book value of the company's reputation and name) and other intangible fixed assets (e.g. patents, trademarks, and customer lists)

%\item Tangible fixed assets. Land, Total Land Depreciation, Net Stated land, Buildings, Total Buildings Depreciation, Net Buildings, Plant \& Machinery20130 Plant \& Machinery Depreciation20135 Net Stated Plant \& Machinery, Transportation Equipment, Transportation Equipment Depreciation, Net Transportation Equipment, Leased Assets, Leased Assets Depreciation, Net Leased Assets, Other Property Plant \& Equipment, Other Property Plant \& Equipment Depreciation, Net Other Property Plant \& Equipment, Accumulated Depreciation.

%\item Material costs. Cost Of Materials. \footnote{the item 'Material Costs' does not appear in the Anglo template and is also not available for US companies.} 

\end{itemize}


Orbis uses estimates for turnover, number of employees, and total assets when these data are not available. The estimation procedure uses country and industry averages to impute missing data and does \textit{not} use gravity estimates.\footnote{The estimation procedure is described in detail here: \url{https://help.bvdinfo.com/mergedProjects/65_EN/Data/Financial/Estimates.htm}}

\section{Description of the Database}\label{sec:descriptiondatabase}

\subsection{Country, industry, and year overview\label{sec:ORBIS-Variables}}

The procedures implemented guarantee that each country within MREID has a sufficient number of meaningful observations in each industry for estimation purposes. The dimensions of our database are as follows: MREID (initially) spans 12 years from 2010 through 2021. The dataset contains the financial data of 362,845 parent companies (or Global Ultimate Owners) of 1,132,707 affiliates. Of those, 351,600 are foreign affiliates from 70,661 parent companies, and the rest are domestic. Raw data from the 25 sectors are combined, and after undergoing data cleaning, we have approximately 27,000 raw observations per year at the country-sector (two-digit) level.

MREID provides data on FDI for 186 countries, including 11 countries that only have outward FDI\footnote{These countries are Aruba, Antigua and Barbuda, Brunei, Central African Republic, Dominica, Korea, North, San Marino, Suriname, Turkmenistan, Saint Vincent and the Grenadines, Samoa, and Yemen.} and  14 countries that only have inward FDI.\footnote{These countries are Burundi, Benin, Burkina Faso, Cameroon,  Republic of the Congo, Djibouti, Guinea, Grenada, Kyrgyzstan, Maldives, Mali, Mauritania, Sao Tome and Principe, and Swaziland.} Therefore, the dataset covers data from 175 countries that host affiliates from 172 countries. Table \ref{tab:country} in the Appendix (section \ref{sec:other}) displays the list of countries MREID covers. It also shows each country's average and maximum number of affiliates. 

As noted earlier, domestic affiliates are included in the dataset. These are affiliates of a parent MNE located in the same country as the parent. There are 47 countries for which data is not available for domestic affiliates. Therefore, MREID coverage of multinational domestic and international investment is limited to 139 countries.\footnote{Domestic flows are a relevant element of structural gravity estimation, cf., \cite{Bergstrandetal2015} and \cite{yotov2022role}. Domestic investment is also needed to merge the MREID dataset with other trade datasets that include domestic trade (e.g., ITPD-E).}


\subsection{Countries\label{subsec:countries}}

\subsubsection{Statistics and distributions\label{subsec:stats}}

Table \ref{tab:statcountry} reports summary statistics for foreign affiliates at the country-pair level (averages of years 2010-2021). Panel A reports (time-averaged) total statistics for all country-pairs where there are positive observations. Panel B reports revenues, employees, and total and fixed assets \textit{per affiliate}. 

As noted above, MREID has data on FDI for 186 countries; hence, there are potentially 34,410 (=186x185) FDI measures (for each year). However, FDI investments are characterized by a large number of zeros. As noted in Table 1, there are only 4,817 country-pairs with at least one foreign affiliate investment. The mean number of active foreign affiliates across country-pairs in the sample is 90. 

%Interestingly, from the theoretical number of possible country pairs ($185\times185=34,410$), we only observe foreign investment in a reduced subset of country pairs (4271). This is consistent with the high sparsity of FDI data, characterized by many zeros. 

%Table 1

\begin{table}[H]
\centering
\caption{Summary statistics for foreign affiliates at the country-pair level}\label{tab:statcountry}
\begin{tabular}{l*{2}{cccc}}
\toprule 
 & \multicolumn{3}{c}{Panel A: Totals} &  & \multicolumn{3}{c}{Panel B: Average per affiliate}\tabularnewline
 & mean & max & sd & &  mean & max & sd\tabularnewline
\midrule 
No. of For. Affiliates   &          77&      19,873&         428&    &      &      &         \\
Revenue     &       3,940&     609,312&      20,362&     &     59&       10,782&         293\\
Employees   &       7,029&   1,735,375&      43,965&      &   200&     156,239&       2,666\\
Total assets &      14,480&   6,309,828&     132,300&      &   218&      56,616&       1,472\\
Fixed assets &       5,198&   1,615,221&      48,817&       &   66&      22,530&         610\\
Revenue/employee &      48,251&    65,794,332&   1,282,092&      &     &           &           \\
\midrule
\(N\)       &         &            &            &       & &      4,273      &            \\
\midrule 
\multicolumn{8}{l}{Notes: $N$ denotes number of country-pairs with foreign affiliates.}\tabularnewline
\multicolumn{8}{l}{In both panels, revenue and total and fixed assets are in million USD.}\tabularnewline
\multicolumn{8}{l}{In Panel A, revenue per employee is in thousands of USD.}\tabularnewline
\multicolumn{8}{l}{In both panels, employees denotes the actual number.}\tabularnewline
\bottomrule
\end{tabular} 
\end{table}


Table \ref{tab:statcountry_host} reports summary statistics on (time-averaged) revenues, employees, and total and fixed assets by ownership (i.e., domestic vs. foreign). Domestic affiliate statistics include all affiliates of parent companies from the same country.  As discussed earlier, only 139 countries in the sample report domestic affiliates. Countries have 5,687 active domestic affiliates, on average. Foreign affiliate statistics include all affiliates of parent companies from different countries; hence, statistics in Table \ref{tab:statcountry_host} (Panels A and B) are at the country level. As expected, aggregate values are higher for domestic than foreign affiliates. 

%Table 2

\begin{table}[H]
\centering
\caption{Summary statistics at the host country by ownership (totals)}\label{tab:statcountry_host}
\begin{tabular}{lccccccc}

\toprule 
 & \multicolumn{3}{c}{Panel A: Domestic} &  & \multicolumn{3}{c}{Panel B: Foreign}\tabularnewline
 & mean & max & sd &  & mean & max & sd\tabularnewline
 \midrule 
No. of Affiliates   &       5,141&     128,363&      17,232&   &    1,704&      44,747&       4,729\\
Revenue     &     136,628&   3,570,717&     471,000&    &  86,441&   1,666,594&     238,122\\
Employees   &     246,864&   4,783,207&     764,243&    &  152,329&   3,968,938&     482,269\\
Total assets &     763,302&    28,438,464&   3,351,904&   &  316,189&    12,108,262&   1,174,622\\
Fixed assets &     132,133&   5,199,483&     540,606&    & 113,942&   4,000,906&     473,700\\
Revenue/employee &       1,029&      21,801&       2,667&   &    3,583&     227,384&      21,773\\
\midrule
\(N\)       &         &   123         &            &   &     &   164         &            \\
\midrule 
\multicolumn{8}{l}{Notes: Revenue and assets in million USD . Revenue/employee in thousands USD.}\tabularnewline
\multicolumn{8}{l}{Foreign statistics are at the host country level.}\tabularnewline
\multicolumn{8}{l}{$N$ denotes number of countries in the sample.}\tabularnewline
\bottomrule
\end{tabular} 
\end{table}


Table \ref{tab:statcountry_host_ave} reports summary statistics on (time-averaged) revenue, number of employees, and total and fixed assets \textit{per affiliate} and \textit{by ownership} (i.e., domestic vs. foreign). Note that the average foreign affiliate tends to be larger in (per affiliate) revenues, number of employees, and assets than the domestic one. Moreover, the largest foreign affiliates (max) are larger than the domestic ones in (per affiliate) revenues, number of employees, and fixed assets.


%Table 3

\begin{table}[H]
\centering
\caption{Summary statistics at the host country by ownership (per affiliate)}\label{tab:statcountry_host_ave}
\begin{tabular}{lccccccc}

\toprule 
 & \multicolumn{3}{c}{Panel A: Domestic} &  & \multicolumn{3}{c}{Panel B: Foreign}\tabularnewline
 & mean & max & sd &  & mean & max & sd\tabularnewline
 \midrule 
Revenue &          76&         970&         171&    &      93&       1,224&         188\\
Employees&         250&       3,829&         624&    &     282&       5,095&         697\\
Total assets&         424&      11,394&       1,224&   &      431&       5,505&         749\\
Fixed assets&          51&       1,490&         160&    &      94&       3,915&         428\\
\midrule
\(N\)       &         &   137         &            &   &     &   172         &            \\
\midrule 
\multicolumn{8}{l}{Notes: Revenue and assets in millions of USD. }\tabularnewline
\multicolumn{8}{l}{Foreign statistics are at the host country level.}\tabularnewline
\multicolumn{8}{l}{$N$ denotes number of countries in the sample.}\tabularnewline
\bottomrule
\end{tabular} 
\end{table}


However, means, maximum values, and standard deviations provide only a limited picture. Figure \ref{fig:distributions} shows the distributions of the (time-averaged) variables in Table \ref{tab:statcountry_host} (totals per host country). Figure \ref{fig:den_revenu} shows that the distribution of foreign affiliate sales is similar to that of domestic affiliate sales; this figure confirms visually that a larger share of the distribution of foreign affiliate revenues is at smaller values relative to domestic revenues. However, the left tail of the domestic revenue's distribution is longer than that of the foreign distribution. This means that the mass of very small domestic affiliates is larger than that of foreign affiliates. Not surprisingly, Figure \ref{fig:den_employees} shows similarly that a larger 
 share of the distribution of the number of foreign affiliate employees is at smaller values than domestic employees. Although foreign and domestic total (and fixed) assets show similar distributions, foreign affiliates have a larger share of their assets at lower levels than domestic affiliates. The distributions of revenue per employee are similar for foreign and domestic affiliates.

Figure \ref{fig:distributions_ave} shows the distributions of the variables of Table \ref{tab:statcountry_host_ave} (averages per affiliate and per host country). On a per affiliate basis, the revenue and numbers of employee distributions reveal a different story for foreign and domestic affiliates relative to aggregate values in Figure \ref{fig:distributions}. On a per affiliate basis, the share of revenues per affiliate in panel \ref{fig:den_revenue_ave} is thicker for foreign affiliates relative to domestic affiliates. The left tail of the distribution of dometic affiliates' revenue is much thicker than that of foreign. Conversely, the right tail is longer for foreign than for domestic revenues. Foreign affiliates are more concentrated around the (larger) mean of revenue per affiliate and the largest foreign affiliates exhibit higher revenues than domestic affiliates.
While the share of employees in panel \ref{fig:den_employees_ave} per affiliate is also relatively larger for foreign affiliates (and with similar left and right tails), the evidence is suggestive that profits per foreign affiliate may exceed profits per domestic affiliate, which is consistent with theoretical models' hypotheses that foreign affiliates need to recover larger profits than domestic affiliates to cover the extra fixed costs of establishing a foreign affiliate, cf., \cite{BergstrandEgger2007}, \cite{RamondoRodriguezClare2013}, and \cite{Arkolakisetal2018}.

\begin{figure}
        \caption{Distributions per host country (aggregates)}
        % Alt Text: Figure 2 is comrpised of six distribution graphs that include distributions of revenue, employee, total assets, fixed assets, number of affiliates, and revenue per employee for foreign affiliates and domestic affiliates.
        \label{fig:distributions}
     \centering

     \begin{subfigure}[b]{0.475\textwidth}
         \centering
         \caption{Revenue (million USD)}
         \label{fig:den_revenu}
         \includegraphics[width=\textwidth]{figures/den_revenue.png}
         
     \end{subfigure}
     \hfill
     \begin{subfigure}[b]{0.475\textwidth}
         \centering
         \caption{Employees}
         \label{fig:den_employees}
         \includegraphics[width=\textwidth]{figures/den_employees.png}
         
     \end{subfigure}
      \begin{subfigure}[b]{0.475\textwidth}
         \centering
         \caption{Total Assets (million USD)}
         \label{fig:den_Totalassets}
         \includegraphics[width=\textwidth]{figures/den_Totalassets.png}
         
     \end{subfigure}
     \hfill
     \begin{subfigure}[b]{0.475\textwidth}
         \centering
         \caption{Fixed Assets (million USD)}
         \label{fig:den_Fixedassets}
         \includegraphics[width=\textwidth]{figures/den_Fixedassets.png}
         
     \end{subfigure}
\begin{subfigure}[b]{0.475\textwidth}
         \centering
         \caption{Number of affiliates}
         \label{fig:den_extensive}
         \includegraphics[width=\textwidth]{figures/den_extensive.png}
         
     \end{subfigure}
     \hfill
\begin{subfigure}[b]{0.475\textwidth}
         \centering
          \caption{Revenue per employee }
         \label{fig:den_Revperemploye}
         \includegraphics[width=\textwidth]{figures/den_Revperemployee.png}
        
     \end{subfigure}
     \hfill


\end{figure}

\begin{sidewaysfigure}
        \caption{Distributions per host country (average per affiliate)}
        % Alt Text: Figure 3 is comrpised of four distribution graphs showing the distributions of the variables per affiliate in Table 3: revenues, employees, fixed assets, and total assets.
        \label{fig:distributions_ave}
     \centering

\hfill
     \begin{subfigure}[b]{0.475\textwidth}
         \centering
          \caption{Revenue (million USD)}
         \label{fig:den_revenue_ave}
         \includegraphics[width=\textwidth]{figures/den_revenue_ave.png}
        
     \end{subfigure}
     \hfill
     \begin{subfigure}[b]{0.475\textwidth}
         \centering
          \caption{Employees}
         \label{fig:den_employees_ave}
         \includegraphics[width=\textwidth]{figures/den_employees_ave.png}
        
     \end{subfigure}

\hfill
       
      \begin{subfigure}[b]{0.475\textwidth}
         \centering
          \caption{Total Assets (million USD)}
         \label{fig:den_Totalassets_ave}
         \includegraphics[width=\textwidth]{figures/den_Totalassets_ave.png}
        
     \end{subfigure}
     \hfill
     \begin{subfigure}[b]{0.475\textwidth}
         \centering
         \caption{Fixed Assets (million USD)}
         \label{fig:den_Fixedassets_ave}
         \includegraphics[width=\textwidth]{figures/den_Fixedassets_ave.png}
         
     \end{subfigure}


\end{sidewaysfigure}












\subsubsection{Country coverage and maps\label{subsec:maps}}

Figure \ref{fig:worldmap} provides a heatmap of the spatial distribution of the multinational activity flows in each country. Panel \ref{fig:map_extensive_in_ave} in the top shows the number of inward affiliates by country. This refers to the number of affiliates owned by foreign parents in that country. Panel \ref{fig:map_extensive_out_ave} in the top shows the number of outward affiliates. This refers to the number of affiliates in foreign countries owned by parents of the designated country. The bottom figures show the number of parent firms in a country (panel \ref{fig:map_guo}) and the number of domestic affiliates of parents in a country (panel \ref{fig:map_extensive_dom}). Since the figures are readily interpretable, we need not provide extensive commentary. However, a few results are worth noting. First, while the United States has one of the largest number of outward affiliates (owned by US parents), it is not among the countries with the largest number of inward affiliates (but the United Kingdom is). Second, though China has fewer outward affiliates than the United States, China is close in numbers to the United States in number of inward affiliates. Third, the United States and China are similar in size in terms of domestic affiliates.


\begin{sidewaysfigure}
        \caption{Affiliates world map}
        % Alt Text: Figure 4 is comprised of four heatmaps showing the spatial distribution of multinational activity flows in each country. Panel 4a shows the number of inward affiliates by country. Panel 4b shows the number of outward affiliates by country. Panel 4c shows the number parent firms in a country. Panel 4d shows the number of domestic affiliates of parents in a country.
        \label{fig:worldmap}
     \centering
     \begin{subfigure}[b]{0.475\textwidth}
         \centering
        \caption{Inward affiliates}
         \label{fig:map_extensive_in_ave}
         \includegraphics[width=\textwidth]{figures/map_extensive_in_ave.png}

     \end{subfigure}
     \hfill
     \begin{subfigure}[b]{0.475\textwidth}
         \centering
           \caption{Outward affiliates}
         \label{fig:map_extensive_out_ave}
         \includegraphics[width=\textwidth]{figures/map_extensive_out_ave.png}
       
     \end{subfigure}
\begin{subfigure}[b]{0.475\textwidth}
         \centering
          \caption{Parent firm (Global Ultimate Owner)}
         \label{fig:map_guo}
         \includegraphics[width=\textwidth]{figures/map_guo.png}
             \end{subfigure}
     \hfill
     \begin{subfigure}[b]{0.475\textwidth}
         \centering
            \caption{Domestic Affiliates}
         \label{fig:map_extensive_dom}
         \includegraphics[width=\textwidth]{figures/map_extensive_dom.png}
           \end{subfigure}

\end{sidewaysfigure}


Figure \ref{fig:worldmap2} provides a heatmap of the spatial distribution by (parent firm) country of the revenues earned in foreign countries (panel \ref{fig:map_revenue_in_ave}), employees based at foreign affiliates (panel \ref{fig:map_employees_in_ave}), and revenues per employee in foreign affiliates (panel \ref{fig:map_Revperemployee_in_ave}). We note a couple of insights. First, China, Germany, and the United Kingdom are among the countries with the largest revenue earned from foreign countries, and not the United States. Second, China is also among the countries with the highest number of employees in foreign countries. Third, China and the United States earn comparable levels of revenue per employee in foreign countries; however, Chad and Tunisia are among the highest in revenue per employee in foreign countries, presumably for natural resource reasons.

\begin{sidewaysfigure}
        \caption{Revenue and employees maps}
        % Alt Text: Figure 5 is comprised of three heatmaps showing the spatial distribution by (parent firm) country of the revenues earned by foreign affiliates, employees in foreign affiliates, and revenues per employee in foreign affiliates.
        \label{fig:worldmap2}
     \centering
     \begin{subfigure}[b]{0.475\textwidth}
         \centering
         \caption{Revenue in foreign countries}
         \label{fig:map_revenue_in_ave}
         \includegraphics[width=\textwidth]{figures/map_revenue_in_ave.png}
         
     \end{subfigure}
     \hfill
     \begin{subfigure}[b]{0.475\textwidth}
         \centering
           \caption{Employees in foreign countries}
         \label{fig:map_employees_in_ave}
         \includegraphics[width=\textwidth]{figures/map_employees_in_ave.png}
       
     \end{subfigure}
\hfill
     \begin{subfigure}[b]{0.475\textwidth}
         \centering
           \caption{Revenues/employee in foreign countries}
         \label{fig:map_Revperemployee_in_ave}
         \includegraphics[width=\textwidth]{figures/map_Revperemployee_in_ave.png}
       
     \end{subfigure}
\end{sidewaysfigure}

Figure \ref{fig:worldmap3} provides a heatmap of the spatial distribution of the total and fixed assets owned by foreigners in a country or that country's liabilities to foreigners (top panels \ref{fig:map_Totalassets_in_ave} and \ref{fig:map_Fixedassets_in_ave} respectively) and total and fixed assets owned in foreign countries by the designated country (bottom  panels \ref{fig:map_Totalassets_out_ave} and \ref{fig:map_Fixedassets_out_ave} respectively). 


\begin{sidewaysfigure}
        \caption{Foreign Assets}
        % Alt Text: Figure 6 is comprised of four heatmaps showing the spatial distribution of total and fixed assets owned by foreigners in a country and total and fixed assets owned in foreign countries by the designated country. 
        \label{fig:worldmap3}
     \centering
     \begin{subfigure}[b]{0.475\textwidth}
         \centering
         \caption{Foreign Total Assets (liabilities)}
         \label{fig:map_Totalassets_in_ave}
         \includegraphics[width=\textwidth]{figures/map_Totalassets_in_ave.png}
         
     \end{subfigure}
     \hfill
     \begin{subfigure}[b]{0.475\textwidth}
         \centering
          \caption{Foreign Fixed Assets (liabilities)}
         \label{fig:map_Fixedassets_in_ave}
         \includegraphics[width=\textwidth]{figures/map_Fixedassets_in_ave.png}
             \end{subfigure}
        \hfill     
\begin{subfigure}[b]{0.475\textwidth}
         \centering
         \caption{Total Assets in Foreign Countries}
         \label{fig:map_Totalassets_out_ave}
         \includegraphics[width=\textwidth]{figures/map_Totalassets_out_ave.png}
              \end{subfigure}
     \hfill
     \begin{subfigure}[b]{0.475\textwidth}
         \centering
         \caption{Fixed Assets in Foreign Contries}
         \label{fig:map_Fixedassets_out_ave}
         \includegraphics[width=\textwidth]{figures/map_Fixedassets_out_ave.png}
              \end{subfigure}
\end{sidewaysfigure}




\subsubsection{Bilateral flows \label{subsec:flows}}


Figure \ref{fig:sankey} shows a bilateral flow diagram for our sample's ``top 25'' home and host countries. This particular figure illustrates the (time-averaged) relative numbers of affiliates created by a parent in a country on the left-hand-side (LHS) into a foreign country on the right-hand-side, or RHS. We note several points. First, as expected, the USA is the relatively largest investor (in terms of number of foreign affiliates), followed in size by Japan, Germany, the United Kingdom, and France. Second, the figure indicates that the largest FDI flow is to the United Kingdom, with significant flows to Japan, Germany, and France. Third, one can see the importance of distance from the figure. For example, the sizes of the Japanese FDI flows to Thailand and the United States are relatively similar despite Thailand's relatively smaller economic size, because of Thailand's relative proximity to Japan.

\begin{sidewaysfigure}
\caption{Foreign direct investment flows (number of foreign affiliates) \label{fig:sankey}}
% Alt Text: Figure 7 is a bilateral flow diagram illustrating the (time-averaged) relative number of affiliates created by a parent in a country on the left-hand-side into a foreign country on the right-hand-side. The USA is largest investor, while the United Kingdom is the largest destination for investment.
\includegraphics[width=\textwidth]{figures/sankey_extensive.png}
\end{sidewaysfigure}

Figure \ref{fig:sankey_financials} provides four flow figures similar to that in figure \ref{fig:sankey}. Here, we show various other measures of parent activities in foreign affiliates. Top panels \ref{fig:sankey_revenue} and \ref{fig:sankey_employees} show, respectively, revenues earned by parents on the LHS from affiliates in the RHS countries and numbers of employees at such affiliates. The bottom two panels \ref{fig:sankey_Totalassets} and \ref{fig:sankey_Fixedassets} show, respectively, the total bilateral total and fixed assets of parent countries on the LHS in foreign countries on the RHS.


\begin{sidewaysfigure}
        \caption{Foreign investment flows}
        % Alt Text: Figure 8 is comprised of four bilateral flow diagrams illustrating revenues earned by parents on the left-hand side from affiliates in the right-hand side countries and numbers of employees at such affiliates (in the top two panels). The bottom two panels show, respectively, the total bilateral total and fixed assets of parent countries on the left-hand side in foreign countries on the right-hand side.
        \label{fig:sankey_financials}
     \centering
   

    
\hfill
       \begin{subfigure}[b]{0.475\textwidth}
         \centering
          \caption{Revenue}
         \label{fig:sankey_revenue}
         \includegraphics[width=\textwidth]{figures/sankey_revenue.png}
        
     \end{subfigure}
     \hfill
     \begin{subfigure}[b]{0.475\textwidth}
         \centering
         \caption{Employees}
         \label{fig:sankey_employees}
         \includegraphics[width=\textwidth]{figures/sankey_employees.png}
         
     \end{subfigure}
 \hfill

      \begin{subfigure}[b]{0.475\textwidth}
         \centering
         \caption{Total assets}
         \label{fig:sankey_Totalassets}
         \includegraphics[width=\textwidth]{figures/sankey_Totalassets.png}
         
     \end{subfigure}
     \hfill
     \begin{subfigure}[b]{0.475\textwidth}
         \centering
         \caption{Fixed assets}
         \label{fig:sankey_Fixedassets}
         \includegraphics[width=\textwidth]{figures/sankey_Fixedassets.png}
         
     \end{subfigure}
\end{sidewaysfigure}


\subsection{Sectors}\label{subsec:sectors}

This section describes the sectoral distribution of the data. As discussed earlier, we look at 25 two-digit SITC sectors. In panels \ref{fig:sector_revenue} -- \ref{fig:sector_fixed_assets} of Figure \ref{fig:sectors_total}, we provide four measures of (aggregate) FDI/MNE activity for \textit{each} of the 25 sectors. The data is time-averaged across the 12 years of data, as earlier. We note several results. We note that -- similar to data at the aggregate level reported in Table \ref{tab:statcountry_host} -- domestic affiliates' (total) revenues, number of employees, and total and fixed assets at the sectoral level are typically larger than those for foreign affiliates. Similarly, in panel \ref{fig:sector_affiliates}, the number of affiliates domestically by sector exceeds that abroad. Also similar to the aggregate data, as shown in panel \ref{fig:sector_Revperemployee}, at the sector level, revenues per employee are typically larger for foreign affiliates than their domestic counterparts.



\begin{figure}[H]
        \caption{Aggregate Foreign Revenue, Employees, Assets (total and fixed) by Sector}
        % Alt Text: Figure 9 is comprised of six bar charts showing time-averaged distribution of revenue, employment, total assets, fixed assets, sector affiliates, and revenues per employee in each of the 25 sectors.
        \label{fig:sectors_total}
        \centering
 \centering
   

        
     \begin{subfigure}[b]{0.475\textwidth}
         \centering
         \caption{Revenue (million USD)}
         \label{fig:sector_revenue}
         \includegraphics[width=\textwidth]{figures/sector_revenue.png}
         
     \end{subfigure}
     \hfill
     \begin{subfigure}[b]{0.475\textwidth}
         \centering
         \caption{Employees}
         \label{fig:sector_employees}
         \includegraphics[width=\textwidth]{figures/sector_employees.png}
         
     \end{subfigure}
      \hfill
      \begin{subfigure}[b]{0.475\textwidth}
         \centering
         \caption{Total Assets (million USD)}
         \label{fig:sector_Totalassets}
         \includegraphics[width=\textwidth]{figures/sector_Totalassets.png}
         
     \end{subfigure}
     \hfill
     \begin{subfigure}[b]{0.475\textwidth}
         \centering
         \caption{Fixed Assets (million USD)}
         \label{fig:sector_fixed_assets}
         \includegraphics[width=\textwidth]{figures/sector_Fixedassets.png}
         
     \end{subfigure}
 \hfill
  \begin{subfigure}[b]{0.475\textwidth}
         \centering
       \caption{Sector Affiliates \label{fig:sector_affiliates}}
\includegraphics[width=\textwidth]{figures/sector_extensive.png}
         
     \end{subfigure}
     \hfill
     \begin{subfigure}[b]{0.475\textwidth}
         \centering
         \caption{Revenues per employee}
         \label{fig:sector_Revperemployee}
         \includegraphics[width=\textwidth]{figures/sector_Revperemployee.png}
         
     \end{subfigure}
     

\end{figure}

Figure \ref{fig:sectors_ave} shows the (time-averaged) distribution of the sectors' total revenues (\ref{fig:sector_revenue_ave}), employees (\ref{fig:sector_employees_ave}), and total (\ref{fig:sector_total_assets_ave}) and fixed assets (\ref{fig:sector_fixed_assets_ave}) \textit{per affiliate} for domestic and foreign affiliates. Similar to the comparable data on total revenues, employees, and total and fixed assets per affiliate for aggregate data in Table \ref{tab:statcountry_host_ave}, this figure shows that foreign affiliates' revenues, employees, and total and fixed assets per affiliate typically exceeds those of domestic affiliates.

\begin{sidewaysfigure}
        \caption{Sectors: Foreign Revenue, Employees, Assets per affiliate and by ownership}
        % Alt Text: Figure 10 is comprised of four bar charts showin the time-averaged distribution of the sectors’ total revenues, employees, and total and fixed assets per affiliate for domestic and foreign affiliates. 
        \label{fig:sectors_ave}
     \centering
    
     
\hfill
 \begin{subfigure}[b]{0.475\textwidth}
         \centering
         \caption{Revenue (Million USD)}
         \label{fig:sector_revenue_ave}
         \includegraphics[width=\textwidth]{figures/sector_revenue_ave.png}
         
     \end{subfigure}
     \hfill
     \begin{subfigure}[b]{0.475\textwidth}
         \centering
          \caption{Employees}
         \label{fig:sector_employees_ave}
         \includegraphics[width=\textwidth]{figures/sector_employees_ave.png}
        
     \end{subfigure}
\hfill
 \begin{subfigure}[b]{0.475\textwidth}
         \centering
         \caption{Total Assets (million USD)}
         \label{fig:sector_total_assets_ave}
         \includegraphics[width=\textwidth]{figures/sector_total_assets_ave.png}
         
     \end{subfigure}
     \hfill
     \begin{subfigure}[b]{0.475\textwidth}
         \centering
          \caption{Fixed Assets (million USD)}
         \label{fig:sector_fixed_assets_ave}
         \includegraphics[width=\textwidth]{figures/sector_fixed_assets_ave.png}
        
     \end{subfigure}

\end{sidewaysfigure}

%\subsubsection{Flows per country pair and sector}\label{subsec:sectors}


Figure \ref{fig:sankey_sectors} provides a similar bilateral flow diagram for the top 25 countries in our sample to that earlier for aggregate FDI. However, by considering the sectoral decomposition, this figure adds \textit{another dimension} to the analysis. As before for Figure \ref{fig:sankey}, the LHS lists the top 25 (largest) outward FDI countries. Also as before, the RHS lists the top 25 FDI countries. The difference in this figure is the \textit{sector} identification in the middle column. 

To appreciate insights from this figure, consider US outward FDI; the United States is the largest FDI investor abroad (cf., Figure \ref{fig:worldmap}, panels b and c). However, consider the middle sector identification for the Wholesale industry (in green, at the top). The United States is a large foreign direct investor in the Wholesale industry. Moreover, the number of ultimate foreign affiliate destinations is large, as shown by inward US wholesale FDI into the United Kingdom (GBR, at the top) down to South Korea (KOR, at the bottom).

\begin{sidewaysfigure}
\caption{Foreign investment flows per country and sector (affiliates) \label{fig:sankey_sectors}}
% Alt Text: Figure 11 is a flow diagram of aggregate FDI from investor countries on the left-hand side, through a sector column in the middle, to destination countries on the right-hand side. 
\includegraphics[width=\textwidth]{figures/sankey_sectors.png}
\end{sidewaysfigure}

While the previous figure displays the bilateral flow pattern by sector for\textit{numbers} of foreign affiliates (which we label the extensive margin), Figure \ref{fig:sankey_sectors2} provides a flow diagram for the top 25 countries in our sample for revenues (i.e., turnover for foreign affiliate sales), employees, total assets and fixed assets per country pair and sector in panels \ref{fig:sankey_sectors_revenue}, \ref{fig:sankey_sectors_employees}, \ref{fig:sankey_sectors_Totalassets}, and \ref{fig:sankey_sectors_Fixedassets} respectively. While a detailed examination of each of the flows within these figures is beyond the scope of this paper, we note -- consistent with Figure \ref{fig:sankey_sectors} -- that the United States (in panel \ref{fig:sankey_sectors_revenue}) has relatively large foreign affiliate sales in the Wholesale industry as well. However, the richness of our data set is that it reveals that US FDI activity by employment is highest in Metals, by total assets is highest in Finance, and by total fixed assets is highest in Management. Thus, our data set contributes information across \textit{various dimensions} (or measures) of MNE/FDI activity.


\begin{sidewaysfigure}
        \caption{Foreign investment flows per country and sector}
        % Alt Text: Figure 12 is a four panel flow diagram for the top 25 countries in our sample for revenues, employees, total assets, and fixed assets per country pair and sector.
        \label{fig:sankey_sectors2}
     \centering


 \hfill
          \begin{subfigure}[b]{0.475\textwidth}
         \centering
         \caption{Revenue}
         \label{fig:sankey_sectors_revenue}
         \includegraphics[width=\textwidth]{figures/sankey_sectors_revenue.png}
         
     \end{subfigure}
     \hfill
     \begin{subfigure}[b]{0.475\textwidth}
         \centering
          \caption{Employees}
         \label{fig:sankey_sectors_employees}
         \includegraphics[width=\textwidth]{figures/sankey_sectors_employees.png}
        
     \end{subfigure}

     \hfill
      \begin{subfigure}[b]{0.475\textwidth}
         \centering
           \caption{Total assets}
         \label{fig:sankey_sectors_Totalassets}
         \includegraphics[width=\textwidth]{figures/sankey_sectors_Totalassets.png}
       
     \end{subfigure}
     \hfill
     \begin{subfigure}[b]{0.475\textwidth}
         \centering
         \caption{Fixed assets}
         \label{fig:sankey_sectors_Fixedassets}
         \includegraphics[width=\textwidth]{figures/sankey_sectors_Fixedassets.png}
         
     \end{subfigure}
\end{sidewaysfigure}








\begin{comment}
    \begin{figure}[H]
\caption{Revenue (Turnover) \label{fig:turn}}
\includegraphics[scale=0.75]{OperatingrevenueTurnover.pdf}
\end{figure}


\begin{figure}[H]
\caption{Number of employees \label{fig:affiliates}}
\includegraphics[scale=0.75]{Numberofemployees.pdf}
\end{figure}


\begin{figure}[H]
\caption{Total Assets \label{fig:affiliates}}
\includegraphics[scale=0.75]{TotalassetsthUSD.pdf}
\end{figure}

\begin{figure}[H]
\caption{Fixed Assets \label{fig:affiliates}}
\includegraphics[scale=0.75]{FixedassetsthUSD.pdf}
\end{figure}

\end{comment}


\subsection{Greenfield FDI}\label{sub:greenfield}

A relevant and distinctive feature of the MREID dataset is information regarding the entry mode in foreign markets. Greenfield FDI refers to an investment from a parent company into a \textit{new} affiliate abroad during the sample period. Table \ref{tab:statcountry_green} reports the summary statistics for greenfield FDI for foreign affiliates at the country-pair level (averages of years 2010-2012). Panel A  reports (time-averaged) total statistics for all country-pairs with positive observations. Panel B reports revenues, employees, and total and fixed assets \textit{per affiliate.} As noted in Table \ref{tab:statcountry_green}, there are only 
3,008 country-pairs with at least one foreign affiliate investment. The mean number of active foreign affiliates across country-pairs in the sample is 7.

%Table 4

\begin{table}[H]
\centering
\caption{Summary statistics at the country-pair level, Greenfield FDI}\label{tab:statcountry_green}
\begin{tabular}{l*{2}{cccc}}
\toprule 
 & \multicolumn{3}{c}{Panel A: Totals} & &  \multicolumn{3}{c}{Panel B: Average per affiliate}\tabularnewline
 & mean & max & sd & &  mean & max & sd\tabularnewline
\midrule
No. of Affiliates   &           7&       1,129&          36&         & &       &          \\
Revenue     &          33&       7,917&         269&         & 10&       6,678&         178\\
Employees   &          92&       8,867&         425&         & 32&      28,546&         614\\
Total assets &         376&     114,551&       3,749&        &  47&      24,526&         662\\
Fixed assets &         206&      42,850&       1,950&       &   25&      13,388&         370\\
Revenue/employee &       3,944&   2,293,292&      78,316 & &           &           &           \\
\midrule
\(N\)       &         &            &            &       & &      3,008     &            \\
\midrule 
\multicolumn{8}{l}{Notes: $N$ denotes number of country-pairs with foreign affiliates.}\tabularnewline
\multicolumn{8}{l}{In both panels, revenue and total and fixed assets are in millions of USD.}\tabularnewline
\multicolumn{8}{l}{In Panel A, revenue per employee is in thousands of USD.}\tabularnewline
\multicolumn{8}{l}{In both panels, employees denotes the actual number.}\tabularnewline
\bottomrule
\end{tabular} 
\end{table}


Table \ref{tab:statcountry_host_green} reports summary statistics for (time-averaged) revenues, employees, and total and fixed assets by ownership, i.e., domestic vs. foreign. Domestic affiliate statistics include all affiliates of parent companies from the same country. As discussed earlier, only 139 countries in the sample report domestic affiliates. Foreign affiliate statistics include all affiliates of parent companies from different countries; hence, statistics in Table \ref{tab:statcountry_host_green} (Panels A and B) are at the country level. As expected, aggregate greenfield values are higher for domestic than foreign affiliates.


%Table 5

\begin{table}[H]
\centering
\caption{Summary statistics by ownership (totals),  greenfield FDI}\label{tab:statcountry_host_green}
\begin{tabular}{lccccccc}

\toprule 
 & \multicolumn{3}{c}{Panel A: Domestic} &  & \multicolumn{3}{c}{Panel B: Foreign}\tabularnewline
 & mean & max & sd &  & mean & max & sd\tabularnewline
 \midrule 
Extensive   &         423&      10,200&       1,322& &        175&       2,050&         321\\
Revenue     &       1,002&      21,276&       3,100&  &       775&      13,470&       2,015\\
Employees   &       2,690&      37,893&       6,683&   &    1,943&      19,036&       4,137\\
Total assets &       6,786&     130,465&      21,171&   &    8,706&     188,444&      27,415\\
Fixed assets &       2,870&      69,970&       9,165&    &   4,938&      85,471&      16,318\\
Revenue/emp &         570&       7,637&       1,111&   &    8,528&     405,700&      49,269\\
\midrule
\multicolumn{8}{l}{Notes: Revenue and assets in million USD. Revenue/employee in thousands USD.}\tabularnewline
\multicolumn{8}{l}{Foreign statistics are at the host country level.}\tabularnewline
\bottomrule
\end{tabular} 
\end{table}


Table \ref{tab:statcountry_host_ave_green} reports summary statistics for (time-averaged) revenue, number of employees, and total and fixed assets \textit{per affiliate} and \textit{by ownership} (i.e., domestic vs. foreign). Note that the average foreign affiliate tends to be larger in (per affiliate) revenues, number of employees, and assets than the domestic one. Moreover, the largest foreign affiliates (max) are larger than the domestic ones in (per affiliate) revenues, number of employees, and total and fixed assets.


%Table 6


\begin{table}[H]
\centering
\caption{Summary statistics by ownership (per affiliate),  greenfield FDI}\label{tab:statcountry_host_ave_green}
\begin{tabular}{lccccccc}

\toprule 
 & \multicolumn{3}{c}{Panel A: Domestic} &  & \multicolumn{3}{c}{Panel B: Foreign}\tabularnewline
 & mean & max & sd &  & mean & max & sd\tabularnewline
 \midrule 
Revenue &           8&         211&          27&  &        40&       1,446&         193\\
Employees&          36&       1,257&         157&  &        65&       3,005&         350\\
Total assets&          54&       1,355&         166& &        172&       5,885&         750\\
Fixed assets&          28&       1,118&         121&  &        99&       4,467&         546\\
\midrule 
\multicolumn{8}{l}{Notes: Revenue and assets in millions of USD.}\tabularnewline
\multicolumn{8}{l}{Foreign statistics are at the host country level.}\tabularnewline
\bottomrule
\end{tabular} 
\end{table}


Figure \ref{fig:distributions_ave_green} shows the distributions (average per affiliate) of revenues, employees, and assets (total and fixed) in the host country per ownership (domestic vs. foreign) for greenfield FDI. These distributions have some similarities to the total distributions per affiliate shown in Figure \ref{fig:distributions_ave}, with some relevant differences. As in the general case, the average greenfield foreign affiliate is larger in terms of revenues, employees, and assets than the domestic greenfield affiliate. However, foreign greenfield affiliates exhibit thicker right \textit{and} left tails in the distribution of revenues (panel \ref{fig:den_revenue_ave_green}). In the general case (shown in Panel \ref{fig:den_revenue_ave}), foreign affiliates had shorter left tails. This means that foreign greenfield affiliates are more heterogeneous in terms of revenue dispersion than domestic affiliates or established affiliates. This is compatible with initial greendfield investments at lower levels and scaling up afterward. We observe the same pattern for employees (panel \ref{fig:den_employees_ave_green}) and assets (panels \ref{fig:den_Totalassets_ave_green} and \ref{fig:den_Fixedassets_ave_green}).


\begin{sidewaysfigure}
        \caption{Distributions in the host country (average per affiliate) per ownership, greenfield}
        % Alt Text: Figure 13 is comrpised of four line charts of the distributions (average per affiliate) of revenues, employees, and assets (total and fixed) in the host country per ownership (domestic vs foreign) for greenfield FDI. 
        \label{fig:distributions_ave_green}
     \centering


 \begin{subfigure}[b]{0.475\textwidth}
         \centering
          \caption{Revenue (million USD)}
         \label{fig:den_revenue_ave_green}
         \includegraphics[width=0.9\textwidth]{figures/den_revenue_ave_green.png}
        
     \end{subfigure}
     \hfill
     \begin{subfigure}[b]{0.475\textwidth}
         \centering
          \caption{Employees}
         \label{fig:den_employees_ave_green}
         \includegraphics[width=0.9\textwidth]{figures/den_employees_ave_green.png}
        
     \end{subfigure}
     \hfill
      \begin{subfigure}[b]{0.475\textwidth}
         \centering
          \caption{Total Assets (million USD)}
         \label{fig:den_Totalassets_ave_green}
         \includegraphics[width=0.9\textwidth]{figures/den_Totalassets_ave_green.png}
        
     \end{subfigure}
     \hfill
     \begin{subfigure}[b]{0.475\textwidth}
         \centering
         \caption{Fixed Assets (million USD)}
         \label{fig:den_Fixedassets_ave_green}
         \includegraphics[width=0.9\textwidth]{figures/den_Fixedassets_ave_green.png}
         
     \end{subfigure}

\end{sidewaysfigure}


Figure \ref{fig:sankey_greenfield} shows the bilateral flows between the top 25 home and host countries for greenfield FDI. The country ranking is similar to the general case shown in Figure \ref{fig:sankey}. We do observe a significant difference in the case of the Cayman Islands (CYM), which is ranked as the third country in terms of greenfield outward FDI. This highlights the presence of tax havens and possible profit shifting in greenfield operations.
\pagebreak

\begin{sidewaysfigure}
\caption{Greenfield FDI flows (number of affiliates) \label{fig:sankey_greenfield}}
% Alt Text: Figure 14 is a bilateral flow chart of greenfield FDI between the top 25 home and host countries. The USA is the largest investor and the United Kingdom is the largest destination.
\includegraphics[width=\textwidth]{figures/sankey_greenfield.png}
\end{sidewaysfigure}

\begin{figure}[H]
        \caption{Greenfield FDI, Sectors}
        % Alt Text: Figure 15 is a two panel figure that includes a bar chart illustrating the sectoral breakdown of greenfield investments for foreign and domestic affiliates and a bilateral flow diagram for the top 25 host and home countries by sector for greenfield investment.
        \label{fig:Greenfield}
     \centering
     \begin{subfigure}[b]{0.9\textwidth}
         \centering
         \caption{Greenfield (new) affiliates per sector}
         \label{fig:sector_greenfield}
         \includegraphics[width=\textwidth]{figures/sector_greenfield.png}
         
     \end{subfigure}
     \hfill
     \begin{subfigure}[b]{0.9\textwidth}
         \centering
          \caption{Greenfield flows per country pair and sector}
         \label{fig:sankey_sectors_greenfield}
         \includegraphics[width=\textwidth]{figures/sankey_sectors_greenfield.png}
        
     \end{subfigure}
\hfill

\end{figure}

Figure \ref{fig:Greenfield} shows the sectoral breakdown for greenfield investments (new affiliates) in panel \ref{fig:sector_greenfield}. The sectoral distribution (with top sectors real estate, legal services, and construction) is similar to the general sectoral distribution shown in  panel \ref{fig:sector_affiliates}. The bilateral flows for the top 25 host and home countries by sector are shown in panel \ref{fig:sankey_sectors_greenfield}. The bulk of affiliates from the Cayman Islands are unclassified, followed by real estate and finance sectors.

Figure \ref{fig:worldmap_green} depicts the spatial distribution of greenfield FDI in a world heatmap for inward (panel \ref{fig:map_greenfield_in_ave}) and outward (panel \ref{fig:map_greenfield_out_ave}) foreign affiliates. The spatial distribution of greenfield investment is qualitatively similar to the general spatial distribution shown in Figure \ref{fig:worldmap}.



\begin{figure}[H]
        \caption{Affiliates world map (greenfield, new)}
        % Alt Text: Figure 16 is comrpised of two global heatmaps showing the spatial distribution of greenfield FDI for inward (panel 16a) and outward (panel 16b) foreign affiliates.
        \label{fig:worldmap_green}
     \centering
     \begin{subfigure}[b]{0.8\textwidth}
         \centering
        \caption{Inward affiliates}
         \label{fig:map_greenfield_in_ave}
         \includegraphics[width=\textwidth]{figures/map_greenfield_in_ave.png}

     \end{subfigure}
     \hfill
     \begin{subfigure}[b]{0.8\textwidth}
         \centering
           \caption{Outward affiliates}
         \label{fig:map_greenfield_out_ave}
         \includegraphics[width=\textwidth]{figures/map_greenfield_out_ave.png}
       
     \end{subfigure}


\end{figure}


\subsection{Mergers and Acquisitions FDI}\label{sub:mergers}

The second type of entry mode in foreign markets is mergers and acquisitions (M\&As) FDI. This investment refers to a parent firm's acquisition or merger with a foreign firm.

Table \ref{tab:statcountry_mergers} reports the summary statistics at the country-pair level for foreign M\&As (time-averaged). Panel A reports statistics for all country-pairs where there are positive observations. Panel B reports statistics on (time-averaged) revenues, employees, and total and fixed assets \textit{per affiliate}. As noted in Table \ref{tab:statcountry_mergers}, there are only 1,498 country-pairs with at least one foreign affiliate investment. The mean number of active foreign affiliates across country-pairs in the sample is 5.

%Table 7

\begin{table}[H]
\centering
\caption{Summary statistics at the country-pair level, M\&A FDI}\label{tab:statcountry_mergers}
\begin{tabular}{l*{2}{cccc}}
\toprule 
 & \multicolumn{3}{c}{Panel A: Totals} & &  \multicolumn{3}{c}{Panel B: Average per affiliate}\tabularnewline
 & mean & max & sd & &  mean & max & sd\tabularnewline
\midrule
No. of Affiliates   &           5&         454&          17&  &         &         &          \\
Revenue     &         534&      95,466&       3,165&   &      132&      14,161&         622\\
Employees   &       1,568&     453,219&      14,020&    &     365&      31,683&       1,967\\
Total assets &       1,642&     207,900&      10,695&     &    361&      27,777&       1,741\\
Fixed assets &         395&      28,668&       1,636&      &   109&      14,675&         593\\
Revenue/employee&      15,947&    12,498,113&     385,518&     &      &           &           \\
\midrule
\(N\)       &         &            &            &       & &      1,498     &            \\
\midrule 
\multicolumn{8}{l}{Notes: $N$ denotes number of country-pairs with foreign affiliates.}\tabularnewline
\multicolumn{8}{l}{In both panels, revenue and total and fixed assets are in millions of USD.}\tabularnewline
\multicolumn{8}{l}{In Panel A, revenue per employee is in thousands of USD.}\tabularnewline
\multicolumn{8}{l}{In both panels, employees denotes the actual number.}\tabularnewline
\bottomrule
\end{tabular} 
\end{table}


Table \ref{tab:statcountry_host_mergers} reports summary statistics on (time-averaged) revenues, employees, and total and fixed assets by ownership, i.e., domestic vs. foreign. Domestic affiliate statistics include all affiliates of parent companies from the same country. Foreign affiliate statistics include all affiliates of parent companies from different countries; hence, statistics in Table \ref{tab:statcountry_host_mergers} (Panels A and B) are at the country level.  Aggregate mean values are higher for domestic than foreign firms for revenues, total assets, and fixed assets. However, aggregate numbers of employees on average are higher for foreign affiliates.

%Table 8


\begin{table}[H]
\centering
\caption{Summary statistics by ownership (totals), M\&A FDI}\label{tab:statcountry_host_mergers}
\begin{tabular}{lccccccc}

\toprule 
 & \multicolumn{3}{c}{Panel A: Domestic} &  & \multicolumn{3}{c}{Panel B: Foreign}\tabularnewline
 & mean & max & sd &  & mean & max & sd\tabularnewline
 \midrule 
No. of Affiliates   &          97&       1,268&         230&  &        65&       1,009&         142\\
Revenue     &      12,874&     363,971&      47,339&   &    6,652&     118,368&      14,778\\
Employees   &      18,635&     395,166&      59,496&    &  18,744&     536,559&      57,889\\
Total assets &      63,648&   1,706,169&     234,256&     & 20,162&     246,789&      43,171\\
Fixed assets &       7,601&     166,733&      22,277&     &  4,961&      37,842&       8,664\\
Revenue/employee&       5,980&     379,564&     41,963&      &   950&      25,006&       2,730\\
\midrule
\multicolumn{8}{l}{Notes: Revenue and assets in million USD. Revenue/employee in thousands USD.}\tabularnewline
\multicolumn{8}{l}{Foreign statistics are at the host country level.}\tabularnewline
\bottomrule
\end{tabular} 
\end{table}


Table \ref{tab:statcountry_host_ave_mergers} reports summary statistics on (time-averaged) revenue, employees, and assets \textit{per affiliate} and \textit{by ownership} (i.e., domestic vs. foreign). Note that the average foreign affiliate tends to be larger in revenue and employees, but smaller in total and fixed assets.

%Table 9

\begin{table}[H]
\centering
\caption{Summary statistics by ownership (per affiliate), M\&A  FDI}\label{tab:statcountry_host_ave_mergers}
\begin{tabular}{lccccccc}

\toprule 
 & \multicolumn{3}{c}{Panel A: Domestic} &  & \multicolumn{3}{c}{Panel B: Foreign}\tabularnewline
 & mean & max & sd &  & mean & max & sd\tabularnewline
 \midrule 
Revenue &         116&       1,498&         213&  &       170&       3,767&         426\\
Employees&         299&       3,309&         600&  &       616&      11,292&       1,602\\
Total assets&         873&      21,036&       2,608& &        660&      12,517&       1,532\\
Fixed assets&         137&       4,729&         526&  &       149&       3,611&         420\\
\midrule
\midrule 
\multicolumn{8}{l}{Notes: Revenue and assets in million USD.}\tabularnewline
\multicolumn{8}{l}{Foreign statistics are at the host country level.}\tabularnewline
\bottomrule
\end{tabular} 
\end{table}


Figure \ref{fig:distributions_ave_mas} shows the distributions (average per affiliate) of revenues (\ref{fig:den_revenue_ave_mergers}), employees (\ref{fig:den_employees_ave_mergers}), and assets (total, \ref{fig:den_Totalassets_ave_mergers}  and fixed, \ref{fig:den_Fixedassets_ave_mergers}) in the host country per ownership (domestic, foreign) for M\&As. Some interesting traits surface from observing these distributions. First, the average foreign M\&A is very similar to the average domestic M\&A (both distributions overlap closely). In terms of revenue, the tails follow the same pattern as in greenfield investment: both tails of foreign M\&As are thicker than domestic M\&As. However, only the right tail of employees is thicker for foreign M\&As than domestic M\&As. Foreign M\&As tend to concentrate less on smaller affiliates in terms of employees.


\begin{sidewaysfigure}
        \caption{Distributions in the host country (average per affiliate) per ownership, M\&As}
        % Alt Text: Figure 17 is comprised of four distribution charts of revenues, employees, total assets, and fixed assets in the host country per ownership (domestic, foreign) for M\&A affiliates.
        \label{fig:distributions_ave_mas}
     \centering
\begin{subfigure}[b]{0.475\textwidth}
         \centering
          \caption{Revenue (million USD)}
         \label{fig:den_revenue_ave_mergers}
         \includegraphics[width=\textwidth]{figures/den_revenue_ave_mergers.png}
        
     \end{subfigure}
   %  \hfill
     \begin{subfigure}[b]{0.475\textwidth}
         \centering
          \caption{Employees}
         \label{fig:den_employees_ave_mergers}
         \includegraphics[width=\textwidth]{figures/den_employees_ave_mergers.png}
        
     \end{subfigure}
    % \hfill
      \begin{subfigure}[b]{0.475\textwidth}
         \centering
          \caption{Total Assets (million USD)}
         \label{fig:den_Totalassets_ave_mergers}
         \includegraphics[width=\textwidth]{figures/den_Totalassets_ave_mergers.png}
        
     \end{subfigure}
     %\hfill
     \begin{subfigure}[b]{0.475\textwidth}
         \centering
         \caption{Fixed Assets (million USD)}
         \label{fig:den_Fixedassets_ave_mergers}
         \includegraphics[width=\textwidth]{figures/den_Fixedassets_ave_mergers.png}
         
     \end{subfigure}
  
\end{sidewaysfigure}


Figure \ref{fig:sankey_mas} shows the bilateral flows between the top 25 home and host countries for M\&A FDI. The ranking of the sending are receiving countries is very similar to the general case shown in Figure \ref{fig:sankey}.


\begin{sidewaysfigure}
\caption{Mergers \& Acquisitions Foreign investment flows (affiliates) \label{fig:sankey_mas}}
% Alt Text: Figure 18 is a bilateral flow chart between the top 25 home and host countries for M\&A FDI. The USA is the top investor and the United Kingdom is the top destination for M\&A FDI.
\includegraphics[width=\textwidth]{figures/sankey_mergers.png}
\end{sidewaysfigure}


Figure \ref{fig:sectorsmas} shows the sectoral breakdown for M\&A FDI.  The most popular sectors (wholesale, metal, legal services) for M\&As are different from the general sectors and greenfield FDI, where real estate was more prominent and had fewer affiliates in the metal sector. The bilateral flows for the top 25 host and home countries by sector are shown in panel \ref{fig:sankey_sectors_mergers}.

\begin{figure}
        \caption{M\&As, Sectors}
        % Alt Text: Figure 19 is comprised of two charts: a bar chart showing the sectoral breakdown of M\&A FDI and a bilateral flow chart for the top 25 home and host countries by sector for M\&A FDI.
        \label{fig:sectorsmas}
     \centering
     \begin{subfigure}[b]{0.8\textwidth}
         \centering
         \caption{Mergers \& Acquisitions (new) affiliates per sector}
         \label{fig:sector_mergers}
         \includegraphics[width=\textwidth]{figures/sector_mergers.png}
         
     \end{subfigure}
     \hfill
     \begin{subfigure}[b]{0.8\textwidth}
         \centering
          \caption{Mergers \& Acquisitions flows per country pair and sector}
         \label{fig:sankey_sectors_mergers}
         \includegraphics[width=\textwidth]{figures/sankey_sectors_mergers.png}
        
     \end{subfigure}
\hfill

\end{figure}

Figure \ref{fig:worldmap_mas} depicts the spatial distribution of M\&A FDI in a world heatmap for inward (panel \ref{fig:map_fdimarkets_d_mergers}) and outward (panel \ref{fig:map_mergers_out_ave}) foreign affiliates. These maps are qualitatively similar to those of the general case (Figure \ref{fig:worldmap}) and greenfield investment (Figure \ref{fig:worldmap_green}).



\begin{figure}
        \caption{Affiliates world map (M\&As)}
        % Alt Text: Figure 20 is comprised of two global heatmaps showing the spatial distribution of M\&A FDI for inward and outward foreign affiliates. 
        \label{fig:worldmap_mas}
     \centering
     \begin{subfigure}[b]{0.8\textwidth}
         \centering
        \caption{Inward affiliates}
         \label{fig:map_fdimarkets_d_mergers}
         \includegraphics[width=\textwidth]{figures/map_mergers_in_ave.png}

     \end{subfigure}
     \hfill
     \begin{subfigure}[b]{0.8\textwidth}
         \centering
           \caption{Outward affiliates}
         \label{fig:map_mergers_out_ave}
         \includegraphics[width=\textwidth]{figures/map_mergers_out_ave.png}
       
     \end{subfigure}


\end{figure}




\subsection{Total vs. M\&As vs. greenfield FDI}\label{sub:mavsgreen}

This subsection compares total FDI, greenfield FDI and M\&A FDI. Total affiliates include new affiliates entering a foreign market in the period (greenfield), affiliates that changed ownership during the period (M\&As) and all other affiliates that entered a foreign market before 2010 (and may or may not change in ownership before 2010). Figure \ref{fig:time_green_mas_total} shows that, in line with other standard FDI datasets,  the sum of the volumes of greenfield FDI and M\&A FDI is lower than total FDI. This means the bulk of affiliates entered before 2010 and explains the difference.

 Figure \ref{fig:den_revempl_total_g_m} reveals that, while the revenues per employee are similar for total and M\&A FDI, they are lower for greenfield FDI.  Greenfield and total FDI have thicker tails than M\&A FDI, which are more concentrated around the mean.

 We observe interesting patterns for several \textit{per affiliate} measures of MNE activity in Figure \ref{fig:masvsgreen_all}. The pattern seems to be that the average affiliate in terms of revenue, employment and assets is higher for M\&As FDI, followed by total FDI and greenfield FDI. M\&As and total FDI seem to be similar in the shape of the distribution and greenfield FDI flatter and shifted to the left (i.e., lower mean). In terms of tails, M\&A seems to have shorter tails than the other two types of FDI. The distribution of greenfield FDI seems to be more spread and heterogeneous than M\&A and total FDI.

\begin{figure}[H]
        \caption{Total vs. M\&As vs. greenfield}
        % Alt Text: Figure 21 is comprised of two charts. The first is a line graph showing the total number of affiliates, the number of affiliates for greenfield projects, and the number of affiliates for M\&A deals. The second is a distributional graph showing the distribution of revenue per employee by total affiliates, greenfield affiliates, and M\&A affiliates. 
        \label{fig:masvsgreen}
     \centering
     \begin{subfigure}[b]{0.8\textwidth}
         \centering
          \caption{Affiliates}
         \label{fig:time_green_mas_total}
         \includegraphics[width=\textwidth]{figures/time_green_mas_total.png}
        
     \end{subfigure}
     \hfill
     \begin{subfigure}[b]{0.8\textwidth}
         \centering
          \caption{Revenue per employee}
         \label{fig:den_revempl_total_g_m}
         \includegraphics[width=\textwidth]{figures/den_revempl_total_g_m.png}
        
     \end{subfigure}


\end{figure}





\begin{sidewaysfigure}
        \caption{Distributions in the host country (average per affiliate) Total vs. M\&As vs. greenfield}
        % Alt Text: Figure 22 is comprised of four distributional graphs showing the distribution of revenues, number of employees, total assets, and fixed assets by total affilates, greenfield affiliates, and M\&A afiliates. 
        \label{fig:masvsgreen_all}
     \centering

       \begin{subfigure}[b]{0.475\textwidth}
         \centering
          \caption{Revenue (million USD)}
         \label{fig:den_revenue_ave_total_g_m}
         \includegraphics[width=\textwidth]{figures/den_revenue_ave_total_g_m.png}
        
     \end{subfigure}
     \hfill
     \begin{subfigure}[b]{0.475\textwidth}
         \centering
          \caption{Employees}
         \label{fig:den_employees_ave_total_g_m}
         \includegraphics[width=\textwidth]{figures/den_employees_ave_total_g_m.png}
        
     \end{subfigure}
   
      \begin{subfigure}[b]{0.475\textwidth}
         \centering
          \caption{Total Assets (million USD)}
         \label{fig:den_Totalassets_ave}
         \includegraphics[width=\textwidth]{figures/den_Totalassets_ave_total_g_m.png}
        
     \end{subfigure}
     \hfill
     \begin{subfigure}[b]{0.475\textwidth}
         \centering
         \caption{Fixed Assets (million USD)}
         \label{fig:den_Fixedassets_ave}
         \includegraphics[width=\textwidth]{figures/den_Fixedassets_ave_total_g_m.png}
         
     \end{subfigure}
\hfill



\end{sidewaysfigure}



\subsection{Time trends: Annual data \label{subsec:trends}}

\subsubsection{Aggregate data}
The next set of figures shows the evolution over time (years) of the number of affiliates, revenue, number of employees, total assets, and fixed assets. Figure \ref{fig:revemp} shows the evolution over time of the core MREID variables: aggregate revenues (\ref{fig:OperatingrevenueTurnover}), aggregate employees (\ref{fig:Numberofemployees}), total assets (\ref{fig:TotalassetsthUSD}), fixed assets (\ref{fig:FixedassetsthUSD}), number of affiliates (\ref{fig:extensive}), and revenues per employee (\ref{fig:Revperemployee}). The overall trend is upward, as expected. Total revenues and revenues per employee dipped around 2015 with the world's (especially Europe) economic slowdown. In nominal terms, world GDP grew from 2010-2014. After peaking in 2014, it fell and then did not recover past its 2014 peak until 2017. The world shutdown during Covid's surfacing in 2020 explains the last period's dramatic decline.

\begin{figure}
        \caption{MREID variables over time, aggregates}
        % Alt Text: Figure 23 is comprised of six line charts showing aggregate revenue, employees, total assets, fixed assets, number of affiliates, and revenue over time, from 2010 to 2022. 
        \label{fig:revemp}
        \centering
 \centering
   
      \hfill

        
     \begin{subfigure}[b]{0.475\textwidth}
         \centering
         \caption{Revenue (million USD)}
         \label{fig:OperatingrevenueTurnover}
         \includegraphics[width=\textwidth]{figures/OperatingrevenueTurnover.png}
         
     \end{subfigure}
     \hfill
     \begin{subfigure}[b]{0.475\textwidth}
         \centering
         \caption{Employees}
         \label{fig:Numberofemployees}
         \includegraphics[width=\textwidth]{figures/Numberofemployees.png}
         
     \end{subfigure}
      \hfill
      \begin{subfigure}[b]{0.475\textwidth}
         \centering
         \caption{Total Assets (million USD)}
         \label{fig:TotalassetsthUSD}
         \includegraphics[width=\textwidth]{figures/TotalassetsthUSD.png}
         
     \end{subfigure}
     \hfill
     \begin{subfigure}[b]{0.475\textwidth}
         \centering
         \caption{Fixed Assets (million USD)}
         \label{fig:FixedassetsthUSD}
         \includegraphics[width=\textwidth]{figures/FixedassetsthUSD.png}
         
     \end{subfigure}
\hfill
  \begin{subfigure}[b]{0.475\textwidth}
         \centering
       \caption{Affiliates \label{fig:extensive}}
\includegraphics[width=\textwidth]{figures/extensive.png}
         
     \end{subfigure}
     \hfill
     \begin{subfigure}[b]{0.475\textwidth}
         \centering
         \caption{Revenues per employee}
         \label{fig:Revperemployee}
         \includegraphics[width=\textwidth]{figures/Revperemployee.png}
         
     \end{subfigure}

\end{figure}

Figure \ref{fig:revempperaff} shows the evolution over time of the core MREID variables \textit{per affiliate}. The time evolutions of the variables are very similar to the aggregates, but scaled down in absolute sizes.

\begin{sidewaysfigure}
        \caption{MREID variables time evolution, aggregates per affiliate}
        % Alt Text: Figure 24 is comrpised of four line charts showing aggregate revenue, employees, total assets, and fixed assets per affiliate over time, from 2010 to 2022.
        \label{fig:revempperaff}
     \centering
     \begin{subfigure}[b]{0.475\textwidth}
         \centering
          \caption{Revenue (million USD)}
         \label{fig:rev-in}
         \includegraphics[width=\textwidth]{figures/line_OperatingrevenueTurnover_ave_ave_year.png}
        
     \end{subfigure}
     \hfill
     \begin{subfigure}[b]{0.475\textwidth}
         \centering
         \caption{Employees}
         \label{fig:emp-out}
         \includegraphics[width=\textwidth]{figures/line_Numberofemployees_ave_ave_year.png}
         
     \end{subfigure}
      \begin{subfigure}[b]{0.475\textwidth}
         \centering
         \caption{Total Assets (million USD)}
         \label{fig:totas-in}
         \includegraphics[width=\textwidth]{figures/line_TotalassetsthUSD_ave_ave_year.png}
         
     \end{subfigure}
     \hfill
     \begin{subfigure}[b]{0.475\textwidth}
         \centering
          \caption{Fixed Assets (million USD)}
         \label{fig:fixedas-out}
         \includegraphics[width=\textwidth]{figures/line_FixedassetsthUSD_ave_ave_year.png}
        
     \end{subfigure}

\end{sidewaysfigure}

Figure \ref{fig:revemp_own} shows the evolution over time of the core MREID variables by ownership (domestic versus foreign). Domestic affiliates dominate in terms of aggregate revenue, number of employees, and total assets, shown in panels \ref{fig:OperatingrevenueTurnover_domfore}, \ref{fig:Numberofemployees_domfore}, \ref{fig:TotalassetsthUSD_domfore}, respectively. By contrast, foreign affiliates dominate in terms of revenues per employee (panel \ref{fig:Revperemployee_domfore}).

\begin{figure}
        \caption{MREID variables time evolution, aggregates by ownership}
        % Alt Text: Figure 25 is comprised of six line charts showing aggregate revenue, employees, total assets, fixed assets, number of affiliates, and revenue by ownership (foriegn and domestic) from 2010 to 2022. 
        \label{fig:revemp_own}
        \centering
 \centering
    

        
     \begin{subfigure}[b]{0.475\textwidth}
         \centering
         \caption{Revenue (million USD)}
         \label{fig:OperatingrevenueTurnover_domfore}
         \includegraphics[width=\textwidth]{figures/OperatingrevenueTurnover_domfore.png}
         
     \end{subfigure}
     \hfill
     \begin{subfigure}[b]{0.475\textwidth}
         \centering
         \caption{Employees}
         \label{fig:Numberofemployees_domfore}
         \includegraphics[width=\textwidth]{figures/Numberofemployees_domfore.png}
         
     \end{subfigure}
      \hfill
      \begin{subfigure}[b]{0.475\textwidth}
         \centering
         \caption{Total Assets (million USD)}
         \label{fig:TotalassetsthUSD_domfore}
         \includegraphics[width=\textwidth]{figures/TotalassetsthUSD_domfore.png}
         
     \end{subfigure}
     \hfill
     \begin{subfigure}[b]{0.475\textwidth}
         \centering
         \caption{Fixed Assets (million USD)}
         \label{fig:FixedassetsthUSD_domfore}
         \includegraphics[width=\textwidth]{figures/FixedassetsthUSD_domfore.png}
         
     \end{subfigure}
\hfill
 \begin{subfigure}[b]{0.475\textwidth}
         \centering
       \caption{Affiliates \label{fig:extensive_domfore}}
\includegraphics[width=\textwidth]{figures/extensive_domfore.png}
         
     \end{subfigure}
     \hfill
     \begin{subfigure}[b]{0.475\textwidth}
         \centering
         \caption{Revenues per employee}
         \label{fig:Revperemployee_domfore}
         \includegraphics[width=\textwidth]{figures/Revperemployee_domfore.png}
         
     \end{subfigure}
      \hfill

\end{figure}


However, focusing instead on the variables \textit{per affiliate} as shown in Figure \ref{fig:revempperaff_own}, we observe that foreign affiliates tend to earn higher revenues (panel \ref{fig:line_OperatingrevenueTurnover_ave_ave_year_domfer}) and are larger in terms of employees and assets as seen in panels \ref{fig:line_Numberofemployees_ave_ave_year_domfer} and \ref{fig:line_TotalassetsthUSD_ave_ave_year_domfer}.


\begin{sidewaysfigure}
        \caption{MREID variables time evolution, aggregates per affiliate by ownership}
        % Alt Text: Figure 26 is comrpised of four line charts showing aggregate revenue, employees, total assets, and fixed assets per affiliate by ownership (foreign and domestic) from 2010 to 2022.
        \label{fig:revempperaff_own}
     \centering
     \begin{subfigure}[b]{0.475\textwidth}
         \centering
          \caption{Revenue (million USD)}
         \label{fig:line_OperatingrevenueTurnover_ave_ave_year_domfer}
         \includegraphics[width=\textwidth]{figures/line_OperatingrevenueTurnover_ave_ave_year_domfer.png}
        
     \end{subfigure}
     \hfill
     \begin{subfigure}[b]{0.475\textwidth}
         \centering
         \caption{Employees}
         \label{fig:line_Numberofemployees_ave_ave_year_domfer}
         \includegraphics[width=\textwidth]{figures/line_Numberofemployees_ave_ave_year_domfer.png}
         
     \end{subfigure}
      \begin{subfigure}[b]{0.475\textwidth}
         \centering
         \caption{Total Assets (million USD)}
         \label{fig:line_TotalassetsthUSD_ave_ave_year_domfer}
         \includegraphics[width=\textwidth]{figures/line_TotalassetsthUSD_ave_ave_year_domfer.png}
         
     \end{subfigure}
     \hfill
     \begin{subfigure}[b]{0.475\textwidth}
         \centering
          \caption{Fixed Assets (million USD)}
         \label{fig:line_FixedassetsthUSD_ave_ave_year_domfer}
         \includegraphics[width=\textwidth]{figures/line_FixedassetsthUSD_ave_ave_year_domfer.png}
        
     \end{subfigure}

\end{sidewaysfigure}



Another way to contextualize the time evolution is by showing the growth rate of the MREID variables. In Figure \ref{fig:revempgrowth} all variables have been normalized to a value of 100 in the year 2010. This way, we can appreciate that the growth rate of foreign affiliates has been larger than that of domestic affiliates.


\begin{figure}
        \caption{MREID variables growth rate by ownership}
        % Alt Text: Figure 27 is comprised of six line charts showing the growth rates of revenue, employees, total assets, fixed assets, number of affiliates, and revenue by ownership (foriegn and domestic) from 2010 to 2022. 
        \label{fig:revempgrowth}
     \centering
 

     
     \begin{subfigure}[b]{0.475\textwidth}
         \centering
         \caption{Revenue (Million USD)}
         \label{fig:rev-in}
         \includegraphics[width=\textwidth]{figures/line_OperatingrevenueTurnover_domforeignyear.png}
         
     \end{subfigure}
     \hfill
     \begin{subfigure}[b]{0.475\textwidth}
         \centering
         \caption{Employees}
         \label{fig:emp-out}
         \includegraphics[width=\textwidth]{figures/line_Numberofemployees_domforeignyear.png}
     \end{subfigure}
     
     \hfill
      \begin{subfigure}[b]{0.475\textwidth}
         \centering
          \caption{Total Assets (million USD)}
         \label{fig:totas-in}
         \includegraphics[width=\textwidth]{figures/line_TotalassetsthUSD_domforeignyear.png}
        
     \end{subfigure}
     \hfill
     \begin{subfigure}[b]{0.475\textwidth}
         \centering
          \caption{Fixed Assets (million USD)}
         \label{fig:fixedas-out}
         \includegraphics[width=\textwidth]{figures/line_FixedassetsthUSD_domforeignyear.png}
        
     \end{subfigure}
 \hfill
\begin{subfigure}[b]{0.475\textwidth}
         \centering
     \caption{Affiliates growth rate by ownership \label{fig:affiliatesgrw}}
\includegraphics[width=\textwidth]{figures/line_extensive_domforeignyear.png}
         
     \end{subfigure}
     \hfill
     \begin{subfigure}[b]{0.475\textwidth}
         \centering
         \caption{Revenue per employee}
         \label{fig:emp-out}
         \includegraphics[width=\textwidth]{figures/line_Revperemployee_domforeignyear.png}
     \end{subfigure}

\end{figure}


Figure \ref{fig:revempgrowth_ave} confirms that the growth rate of the MREID variables per affiliate has been larger for foreign firms than that of domestic affiliates.

\begin{sidewaysfigure}
        \caption{MREID variables growth rate, per affiliate by ownership }
        % Alt Text: Figure 28 is comrpised of four line charts showing the growth rates of revenue, employees, total assets, and fixed assets per affiliate by ownership (foreign and domestic) from 2010 to 2022.
        \label{fig:revempgrowth_ave}
     \centering
     \begin{subfigure}[b]{0.475\textwidth}
         \centering
         \caption{Revenue (Million USD)}
         \label{fig:rev-in}
         \includegraphics[width=\textwidth]{figures/line_OperatingrevenueTurnover_ave100_domforeign_ave.png}
         
     \end{subfigure}
     \hfill
     \begin{subfigure}[b]{0.475\textwidth}
         \centering
          \caption{Employees}
         \label{fig:emp-out}
         \includegraphics[width=\textwidth]{figures/line_Numberofemployees_ave100_domforeign_ave.png}
        
     \end{subfigure}
      \begin{subfigure}[b]{0.475\textwidth}
         \centering
         \caption{Total Assets (million USD)}
         \label{fig:totas-in}
         \includegraphics[width=\textwidth]{figures/line_TotalassetsthUSD_ave100_domforeign_ave.png}
         
     \end{subfigure}
     \hfill
     \begin{subfigure}[b]{0.475\textwidth}
         \centering
          \caption{Fixed Assets (million USD)}
         \label{fig:fixedas-out}
         \includegraphics[width=\textwidth]{figures/line_FixedassetsthUSD_ave100_domforeign_ave.png}
        
     \end{subfigure}

\end{sidewaysfigure}


\subsubsection{Separate time trends for greenfield and M\&As FDI}

This subsection shows the time evolutions and growth rates of the MREID variables for the two entry modes: greenfield FDI in Figure \ref{fig:Greenfield_time} and M\&As FDI in Figure \ref{fig:mas_time}. There was a notable fall of greenfield FDI during the COVID years (2019, 2020, 2021). In contrast, M\&As FDI was flat in 2021.


\begin{sidewaysfigure}
        \caption{Greenfield FDI time trends}
        % Alt Text: Figure 29 is comprised of three line charts. The first line chart shows the number of new greenfield affiliates (in USD$ thousands) from 2010-2022. The second chart shows the number of new greenfield affiliates from 2010 to 2022 by ownership (foriegn and domestic). The third chart shows the growth rate of new greenfield affiliates by ownership (foreign and domestic) over time from 2010 to 2022. 
        \label{fig:Greenfield_time}
     \centering
     \begin{subfigure}[b]{0.4\textwidth}
         \centering
         \caption{Greenfield (new) affiliates}
         \label{fig:}
         \includegraphics[width=\textwidth]{figures/greenfield.png}
         
     \end{subfigure}
     %\hfill
     \begin{subfigure}[b]{0.4\textwidth}
         \centering
          \caption{Greenfield, domestic vs. foreign}
         \label{fig:greenfield_domfore}
         \includegraphics[width=\textwidth]{figures/greenfield_domfore.png}   
     \end{subfigure}
  %   \hfill
     \begin{subfigure}[b]{0.4\textwidth}
         \centering
         \caption{Domestic vs. foreign, growth}
         \label{fig:line_greenfield_domforeignyear}
         \includegraphics[width=\textwidth]{figures/line_greenfield_domforeignyear.png}
         
     \end{subfigure}
\end{sidewaysfigure}


\begin{sidewaysfigure}
        \caption{Mergers Acquisitions FDI time trends}
        % Alt Text: Figure 30 is comprised of three line charts. The first line chart shows the number of new M\&A affiliates (in USD$ thousands) from 2010-2022. The second chart shows the number of new M\&A affiliates from 2010 to 2022 by ownership (foriegn and domestic). The third chart shows the growth rate of new M\&A affiliates by ownership (foreign and domestic) over time from 2010 to 2022. 
        \label{fig:mas_time}
     \centering
     \begin{subfigure}[b]{0.4\textwidth}
         \centering
         \caption{Mergers \& Acquisitions (new) affiliates}
         \label{fig:}
         \includegraphics[width=\textwidth]{figures/mergers.png}
         
     \end{subfigure}
   %  \hfill
     \begin{subfigure}[b]{0.4\textwidth}
         \centering
          \caption{Mergers \& Acquisitions, domestic vs. foreign}
         \label{fig:mergersfield_domfore}
         \includegraphics[width=\textwidth]{figures/mergers_domfore.png}
     \end{subfigure}
    % \hfill
     \begin{subfigure}[b]{0.4\textwidth}
         \centering
         \caption{Domestic vs. foreign, growth}
         \label{fig:line_mergersfield_domforeignyear}
         \includegraphics[width=\textwidth]{figures/line_mergers_domforeignyear.png}
     \end{subfigure}
\end{sidewaysfigure}






\section{Validity\label{chap:Coverage-and-Validity}}

We take three approaches to validate the data in the MREID dataset. First, we corroborate that the data follows the aggregate trends of FDI assets and liabilities. Second, we correlate the greenfield investment with an independent dataset. Third, we correlate the MREID data with administrative sources and other independent datasets.

\subsection{Aggregates: External Wealth of Nations Comparison}\label{sub:agregatre}

The External Wealth of Nations (EWN) dataset is a comprehensive database of country-level estimates of external financial assets and liabilities. The Brookings Institution compiles the data based on various sources, including balance of payments data, international investment position data, and other statistical sources. The EWN dataset covers 1970 to 2020 and includes data for over 200 economies.\footnote{The EWN is available here: \url{https://www.brookings.edu/research/the-external-wealth-of-nations-database/}}



 
A comparison of MREID Figure \ref{fig:worldmap3} with Figure \ref{fig:worldmap_EWN} shows the correlation between MREID and EWN. Figure \ref{fig:map_FDIassets_EWN} shows the heatmap for the FDI Assets in foreign countries. Their correlation with MREID assets in foreign countries is 0.70.  Figure \ref{fig:map_FDIliabilities_EWN} shows the heatmap for the FDI liabilities from foreign countries. Their correlation with MREID liabilities in foreign countries is 0.77. Last, Figure \ref{fig:map_GDPUS} shows the heatmap of GDP. Their correlation with the MREID crude measures of output, aggregating domestic and foreign revenue and domestic fixed assets, is 0.74.




\begin{sidewaysfigure}
 \caption{EWN}
 % Alt Text: Figure 31 is comprised of three global heatmaps showing the spacial distribution of FDI assets, FDI liabilities, and GDP using Brookings' External Wealth of Nations dataset.


        \label{fig:worldmap_EWN}
     \centering
     \begin{subfigure}[b]{0.475\textwidth}
         \centering
          \caption{FDI Assets}
         \label{fig:map_FDIassets_EWN}
         \includegraphics[width=\textwidth]{figures/map_FDIassets.png}
        
     \end{subfigure}
     \hfill
     \begin{subfigure}[b]{0.475\textwidth}
         \centering
          \caption{FDI liabilities}
         \label{fig:map_FDIliabilities_EWN}
         \includegraphics[width=\textwidth]{figures/map_FDIliabilities.png}
        
     \end{subfigure}
 \hfill
 \begin{subfigure}[b]{0.475\textwidth}
         \centering
 \caption{GDP \label{fig:map_GDPUS}}
\includegraphics[width=\textwidth]{figures/map_GDPUS.png}
        
     \end{subfigure}
    
\end{sidewaysfigure}





Figure \ref{fig:ORBISvsliab} shows the joint time series of the MREID variables and the EWN FDI assets. The correlation between MREID assets (total and fixed) and EWN FDI assets time series is 0.96 and 0.95, respectively. Figure \ref{fig:GDP_corr} shows the time evolutions of our crude measure of real output. The correlation between this measure of output and the world's GDP is 0.93.




\begin{sidewaysfigure}
        \caption{MREID vs EWN, time correlation}
        % Alt Text: Figure 32 is comprised of three line charts comparing MREID variables with EWN variables over time. The first line chart shows the value of total assets (in USD$ trillions) from 2010-2022. The second chart shows the value of fixed assets (in USD$ trillions) from 2010 to 2022. The third chart shows the value of MRIED output versus GDP over time from 2010 to 2022. 
        \label{fig:ORBISvsliab}
     
     \begin{subfigure}[b]{0.475\textwidth}
         \centering
         \caption{Total Assets}
         \label{fig:total_FDI_liab_corr}
         \includegraphics[width=\textwidth]{figures/total_FDI_liab_corr.png}
         
     \end{subfigure}
     \hfill
      \begin{subfigure}[b]{0.475\textwidth}
         \centering
         \caption{Fixed Assets}
         \label{fig:totas-in}
         \includegraphics[width=\textwidth]{figures/Fixed_FDI_liab_corr.png}
         
     \end{subfigure}
     \hfill
     \begin{subfigure}[b]{0.475\textwidth}
         \centering
\caption{MREID output vs GDP \label{fig:GDP_corr}}
\includegraphics[width=\textwidth]{figures/GDP_corr.png}
         
     \end{subfigure}
     \hfill
\end{sidewaysfigure}

\subsection{FDIMarkets}\label{sub:agregatre}
FDIMarkets is a service from the Financial Times that provides real-time monitoring of cross-border greenfield investment announcements. The database covers all countries and sectors worldwide and includes announcements on investment projects, capital investment, and job creation. FDIMarkets also provides tools for tracking and profiling companies investing overseas and conducting in-depth analyses to uncover trends.





Figure \ref{fig:worldmap_fdimarkets} shows the spatial distribution of greenfield FDI announcements using FDIMarkets projects and its correlation with MREID greenfield affiliates. The correlation between the number of affiliates (MREID) and projects (FDIMarkets) is 0.70 for inward FDI (shown in panel \ref{fig:map_fdimarkets_d_green}) and 0.94 for outward FDI (shown in panel \ref{fig:map_fdimarkets_o_green}).



\begin{figure}[H]
        \caption{Greenfield Investments, FDI Markets (Projects)}
        % ALt Text: Figure 33 is comprised of two heatmaps using fDiMarkets data. The first heat map shows the spacial distribution of inward greenfield affilates. The second heat map shows the spacial distribution of outward greenfield affiliates.
        \label{fig:worldmap_fdimarkets}
     \centering
     \begin{subfigure}[b]{0.8\textwidth}
         \centering
          \caption{Inward greenfield FDI}
         \label{fig:map_fdimarkets_d_green}
         \includegraphics[width=\textwidth]{figures/map_fdimarkets_d_green.png}
        
     \end{subfigure}
     \hfill
     \begin{subfigure}[b]{0.8\textwidth}
         \centering
          \caption{Outward greenfield FDI}
         \label{fig:map_fdimarkets_o_green}
         \includegraphics[width=\textwidth]{figures/map_fdimarkets_o_green.png}
        
     \end{subfigure}
 \hfill

    
\end{figure}




Figure \ref{fig:orbisvsfdimarkets} shows the time evolutions and correlations between MREID's greenfield FDI variables and FDIMarkets. The correlation between the number of affiliates (MREID) and projects (FDIMarkets) is 0.68 (shown in panel \ref{fig:fdimarkets_corr_greenfield}). The correlation between fixed assets (MREID) and capital expenditure (FDIMarkets) shown in panel \ref{fig:fdimarkets_corr_greenfield_capex} is somewhat weaker, 0.45. The correlation between the number of employees (MREID) and jobs (FDIMarkets) is 0.59 (shown in panel \ref{fig:fdimarkets_corr_greenfield_jobs}).



\begin{sidewaysfigure}
        \caption{Greenfield FDI: MREID vs. FDI Markets}
        % Alt Text: Figure 34 is comprised of three line charts comparing MRIED and fDiMarkets variables over time. The first chart shows the total number of affiliates/projects from 2002-2022. The second chart shows the value of fixed assets/captial expenditures (USD$ million) from 2002-2022. The third chart shows the total number of employess/jobs from 2002-2022.
        \label{fig:orbisvsfdimarkets}
     \centering
     \begin{subfigure}[b]{0.475\textwidth}
         \centering
          \caption{Number of affiliates/projects}
         \label{fig:fdimarkets_corr_greenfield}
         \includegraphics[width=\textwidth]{figures/fdimarkets_corr_greenfield.png}
        
     \end{subfigure}
     \hfill
     \begin{subfigure}[b]{0.475\textwidth}
         \centering
          \caption{Fixed Assets/Capital expenditure}
         \label{fig:fdimarkets_corr_greenfield_capex}
         \includegraphics[width=\textwidth]{figures/fdimarkets_corr_greenfield_capex.png}
        
     \end{subfigure}
 \hfill

  \begin{subfigure}[b]{0.475\textwidth}
         \centering
          \caption{Employees/Jobs}
         \label{fig:fdimarkets_corr_greenfield_jobs}
         \includegraphics[width=\textwidth]{figures/fdimarkets_corr_greenfield_jobs.png}
        
     \end{subfigure}
 \hfill

    
\end{sidewaysfigure}



\subsection{Country-level: USA affiliates vs. other datasets and administrative data}

This section starts by documenting MREID coverage and correlations of American affiliates vis-a-vis other administrative and private data sources. Table \ref{tab:US-FDI-in} reports the aggregate values of several FDI measures of the US investment in Spain in 2019 in three datasets: ORBIS, BEA, and FDIMarkets.

We focus on the US investment in Spain to showcase
the search strategy and point out further issues. We choose US-Spanish
investment due to our knowledge of both countries and the fact that
\citet{gopinath2017capital} showed that ORBIS does a good job of tracking
Spanish data.

According to the BEA, there are 708 US affiliates established in Spain. Orbis records 2,490 US affiliates in Spain. However, the BEA records statistics for affiliates with more than 25 million USD in assets or sales. When we limit the ORBIS search to those quantities, the number of affiliates reported by ORBIS is similar to BEA: 711 affiliates. 

According to the BEA, the total assets of American affiliates are USD 185 million, which is slightly higher than the number recorded by ORBIS (USD 158 million). Compared with Spanish administrative data, Orbis also seems to underestimate fixed assets. The BEA does not report fixed assets, only Net Property, Plant \& Equipment (PPE); therefore, we cannot compare the magnitudes of fixed assets.
\\

\begin{table}[H]
\caption{US FDI in Spain in 2019. ORBIS vs BEA \& FDIMarkets \label{tab:US-FDI-in}}

\scalebox{0.45}{%

\begin{tabular}{>{\centering}p{4cm}>{\centering}p{2cm}>{\centering}p{2cm}>{\centering}p{2cm}>{\centering}p{2cm}>{\centering}p{2cm}>{\centering}p{2cm}>{\centering}p{2cm}>{\centering}p{2cm}>{\centering}p{2cm}>{\centering}p{2cm}>{\centering}p{2cm}>{\centering}p{2cm}}
\hline 
\textsf{\addlinespace}{\Large{} }\foreignlanguage{spanish}{\ } & {\Large{}Number of parent companies} & {\Large{}Number of affiliates} & {\Large{}Total assets} & {\Large{}Net property, plant \& equipment} & {\Large{}Fixed assets} & {\Large{}Capital expenditures} & {\Large{}Sales} & {\Large{}Net income} & {\Large{}Value added} & {\Large{}Cost of emloyees} & {\Large{}Number of employees} & {\Large{}M\&A value}\tabularnewline
\hline 
\hline 
\addlinespace{\Large{} }\foreignlanguage{spanish}{\ }{\Large{}BEA
($>$ 25M assets)} & {\LARGE{}N/A} & {\LARGE{}708} & {\LARGE{}185,260} & {\LARGE{}17,830} &  & {\LARGE{}2,367} & {\LARGE{}92,507} & {\LARGE{}7,735} & {\LARGE{}18,022} & {\LARGE{}11,589} & {\LARGE{}180.6} & \tabularnewline
\addlinespace{\Large{} }\foreignlanguage{spanish}{\ }{\Large{}Spanish
Admin. Data ($>$ 0 assets)} &  &  &  &  & {\LARGE{}92,264} &  & {\LARGE{}94,380} & {\LARGE{}5,050} &  &  & {\LARGE{}317.0} & \tabularnewline
\hline 
\addlinespace{\Large{} }\foreignlanguage{spanish}{\ }{\Large{}ORBIS
($>$ 25M assets)} & {\LARGE{}359} & {\LARGE{}711} & {\LARGE{}149,574} &  & {\LARGE{}72,206} &  & {\LARGE{}104,940} & {\LARGE{}2,664} & {\LARGE{}22,704} & {\LARGE{}14,436} & {\LARGE{}248.1} & \tabularnewline
\addlinespace{\Large{} }\foreignlanguage{spanish}{\ }{\Large{}ORBIS-vertical
($>$ 25M assets)} & {\LARGE{}146} & {\LARGE{}323} & {\LARGE{}58,205} &  & {\LARGE{}34,104} &  & {\LARGE{}33,018} & {\LARGE{}898} & {\LARGE{}7,402} & {\LARGE{}4,371} & {\LARGE{}76.4} & \tabularnewline
\addlinespace{\Large{} }\foreignlanguage{spanish}{\ }{\Large{}ORBIS-horizontal
($>$ 25M assets)} & {\LARGE{}213} & {\LARGE{}388} & {\LARGE{}91,369} &  & {\LARGE{}38,103} &  & {\LARGE{}71,922} & {\LARGE{}1,765} & {\LARGE{}15,303} & {\LARGE{}10,065} & {\LARGE{}171.7} & \tabularnewline
 &  &  &  &  &  &  &  &  &  &  &  & \tabularnewline
{\Large{}ORBIS ($>$ 0 assets)} & {\LARGE{}1,902} & {\LARGE{}2,490} & {\LARGE{}158,059} &  & {\LARGE{}74,971} &  & {\LARGE{}115,090} & {\LARGE{}2,484} & {\LARGE{}25,922} & {\LARGE{}17,447} & {\LARGE{}298.7} & \tabularnewline
 &  &  &  &  &  &  &  &  &  &  &  & \tabularnewline
\hline 
\addlinespace{\Large{} }\foreignlanguage{spanish}{\ }{\Large{}Spanish
Admin. Data ($>$ 0) Greenfield} &  &  &  &  &  & \textbf{\LARGE{}3.628{*}{*}} &  &  &  &  &  & \tabularnewline
\addlinespace{\Large{} }\foreignlanguage{spanish}{\ }{\Large{}FDIMarkets
($>$ 25M assets); Greenfield} & {\LARGE{}14} & \textbf{\LARGE{}22} &  &  &  & \textbf{\LARGE{}1,462} & \textbf{\LARGE{}3034{*}} &  &  &  & \textbf{\LARGE{}5.3} & \tabularnewline
\addlinespace{\Large{} }\foreignlanguage{spanish}{\ }{\Large{}ORBIS
($>$ 25M assets); Greenfield} & {\LARGE{}24} & \textbf{\LARGE{}27} & {\LARGE{}10,142} &  &  & \textbf{\LARGE{}2,941{*}{*}} & \textbf{\LARGE{}2,669} & {\LARGE{}330} & {\LARGE{}910} & {\LARGE{}352} & \textbf{\LARGE{}5.4} & \tabularnewline
\hline 
\addlinespace{\Large{} }\foreignlanguage{spanish}{\ }{\Large{}ORBIS
($>$ 25M assets); M\&A} & {\LARGE{}1} & {\LARGE{}1} & {\LARGE{}72} &  & {\LARGE{}69} &  & {\LARGE{}3} & {\LARGE{}39} & {\LARGE{}59} & {\LARGE{}5} & {\LARGE{}0} & {\Large{}78}\tabularnewline
\hline 
\multicolumn{13}{l}{Notes: {*}Constructed sales. {*}{*}New fixed assets. Data in million
USD, employee in thousands}\tabularnewline
\hline 
\end{tabular}}
\end{table}

In Appendix \ref{sec:search_detail}, we describe the details to obtain the data presented in Table \ref{tab:US-FDI-in}. Sales are fairly similar in all three data sources: BEA (USD 92 million), Spanish administrative data (USD 94 million), and Orbis (USD 104 million). Some quantitative differences arise in the values of net income and value added. It is worthwhile noting that there are differences in the definition of these variables. The BEA defines value added as ``The gross output of an industry or a sector less its intermediate inputs''\footnote{\url{https://www.bea.gov/help/glossary/value-added}}. Value added in Orbis is taken from the corporate balance sheet's Profit and Loss (P/L) account, which is calculated by deducting the cost of capital from the operating profit. Similarly, Orbis' definition of net income, which is directly the balance sheet P/L for the period, might be different from BEA's definition of Net income. While the cost of employees is similar in the BEA and ORBIS, the total number of employees is much lower in the BEA (180,000) than in Orbis (248,000) and Spanish administrative data (317,000). 

The BEA reports capital expenditures as the change in property, plant and equipment (PPE). We constructed a similar measure with Spanish administrative data and Orbis. All three measures are qualitatively similar, with some quantitative differences. One limitation of the BEA statistics is that capital expenditure is the only measure of greenfield FDI. We can easily track greenfield investments with Orbis and compare them with other independent datasets, particularly FDIMarkets. The last three columns of Table \ref{tab:US-FDI-in} focus on greenfield investment. Although the capital expenditure is relatively overestimated in Orbis vs. FDIMarkets, sales seem to be relatively close.\footnote{FDIMarkets do not report directly sales, we constructed this measure as a Cobb-Douglas function of capital and labor.} Interestingly, the number of affiliates and the number of new employees is practically identical in both datasets.

Orbis can overcome a standard limitation in all other datasets related to the activity of the subsidiary. Orbis allows us to distinguish between the activity sector of the parent firm and the subsidiary. Horizontal FDI occurs when those activities are similar (i.e., the subsidiary replicates the parent's activity). Vertical FDI occurs when the activity of the parent and subsidiary are different (i.e., the parent splits the production process along the value chain). Table \ref{tab:US-FDI-in} reveals a relatively even split between both affiliates. However, most assets and sales (over 60\%) are concentrated in horizontal FDI. 



To validate our search approach, we inspected the linear trends and
correlations between ORBIS and several independent sources for the key
FDI measures (sales, employees, assets, and fixed assets). Figure \ref{tab:US-FDI-in-val}
shows that the correlations between ORBIS and administrative Spanish data
are positive and strong.

Notably, the main use of the dataset is to perform econometric estimations. Therefore, what is relevant to obtaining accurate estimates (e.g., estimates of policy changes like economic integration agreements) is that the values follow similar trends. That is, the difference in the \textit{levels} we observed in the previous sections should not be an issue to estimate accurately partial effects.

\begin{sidewaysfigure}
\caption{US FDI in Spain in 2003-2019. ORBIS vs. Spanish admin. data \label{tab:US-FDI-in-val}}
% Alt Text: Figure 35 is comprised of four line charts comparing Orbis data to Spanish administrative data over time. The variables compared are sales, employees, fixed assets, and total assets. 
\centering
\includegraphics[width=\textwidth]{\string"Obris BEA FDI in ESP2\string".pdf}
\end{sidewaysfigure}

Figure \ref{tab:US-FDI-in-val-1} shows the correlations between ORBIS
and BEA data for the major American investment locations (Canada,
UK, Germany, China, and Spain). The positive correlations between sales, employees, and assets are high.

\begin{sidewaysfigure}
\caption{US FDI in Canada, UK, Germany, China, and Spain in 2003-2019. ORBIS vs. BEA\label{tab:US-FDI-in-val-1}}
% Alt Text: Figure 36 is comprised of four line graphs comparing Orbis and BEA data for total US FDI in Canada, UK, Germany, China, and Spain. The variables compared are sales, employees, fixed assets, and total assets. 
\centering
\includegraphics[width=0.9\textwidth]{\string"Obris BEA FDI in ALL2\string".pdf}
\end{sidewaysfigure}

When we zoom in  on individual countries (UK and China) in Figure
\ref{tab:US-FDI-in-val-1-1}, we observe some heterogeneity. Some
variables like assets or sales have higher correlations than others
like fixed assets.\footnote{Recall that BEA does not report fixed assets, but rather Property,
Plant and Equipment}


\begin{sidewaysfigure}
       \caption{US FDI China and UK, 2003-2019. ORBIS vs. BEA\label{tab:US-FDI-in-val-1-1}}
       % Alt Text: Figure 37 is comprised of eight line graphs comparing Orbis and BEA data for US FDI in China and the UK. The variables compared are sales, employees, fixed assets, and total assets.
     \centering
     \begin{subfigure}[b]{0.4\textwidth}
         \centering
          \caption{China}
         \label{fig:China}
         \includegraphics[width=\textwidth]{\string"Obris BEA FDI in CHN\string".pdf}
        
     \end{subfigure}
     \hfill
     \begin{subfigure}[b]{0.4\textwidth}
         \centering
          \caption{UK}
         \label{fig:UK}
         \includegraphics[width=\textwidth]{\string"Obris BEA FDI in gbr2\string".pdf}
        
     \end{subfigure}
\end{sidewaysfigure}




\section{Conclusions \label{sec:Conclusions}}

FDI can be characterized by numerous alternative measures such as total assets, fixed
assets, employment, and foreign affiliate sales (FAS) and by various 
types (total, greenfield, M\&As FDI). This paper has 
described and validated a search strategy to adequately construct
a firm-level panel dataset from Orbis that captures many of the complexities and richness
of FDI-related variables.

The breadth of coverage of the Orbis data is broad when compared to administrative
data of individual countries (e.g., USA, Spain) and greenfield FDI announcement data of FDIMarkets (Financial Times). The search strategy allows us to capture FDI data in 185
countries and initially in 12 years, which can be expanded
to subsequent years. Historical data is available (with a subscription cost)
to cover time series periods from the past. The search strategy is validated by strong positive correlations across time and space with individual countries' administrative data sources.



\pagebreak 
\singlespacing
\bibliographystyle{apalike}
\bibliography{memo}
\doublespacing
\phantomsection\addcontentsline{toc}{section}{\refname}

\processdelayedfloats
\csname efloat@restorefloats\endcsname

\newpage 
\appendix
\section*{Appendix}\label{sec:app}

%Appendix A

%\renewcommand{\thetable}{A.\arabic{table}}

%\addtodelayedfloat{table}{\renewcommand{\thetable}{A.\arabic{table}}}

    

%\addtodelayedfloat{figure}{\renewcommand{\thefigure}{A\arabic{figure}}%
%  \setcounter{figure}{0}}%

%\renewcommand{\thepostfigure}{A\arabic{postfigure}}

\section{Search details example}\label{sec:search_detail}
\setcounter{table}{0}
\setcounter{figure}{0}
%\setcounter{postfigure}{0}
\counterwithin{figure}{section}
\counterwithin{table}{section}
\renewcommand\thefigure{\thesection\arabic{figure}}
\renewcommand\thetable{\thesection\arabic{table}}

Figure \ref{fig:Orbis-search-1} describes the boolean search steps
to obtain the 3,787 American subsidiaries in Spain. Each search step
limits the number of firms captured in each part of the search. In the second part of the search process, we limited the total assets
to 25 million USD to ease the comparison with BEA's administrative
data, which only reports subsidiaries with assets above the USD 25 million 
threshold.\footnote{Note that the MREID's threshold is USD 1 million. Sometimes USD 25 million is taken as a threshold for comparison with the BEA data.}

\begin{figure}[H]
\caption{ORBIS search \label{fig:Orbis-search-1}}
% Alt Text: Figure A1 is a screenshot that shows the boolean search steps to obtain the 3,787 Americansubsidiaries in Spain. 
\includegraphics[scale=0.75]{search_strategy}
\end{figure}

Figure \ref{fig:Orbis-search} provides a screenshot of the search results
as seen on Orbis' interface. American companies that operate in Spain, like Ford, ALCOA, HP, and Dow Chemical, appear in this sample.

\begin{figure}[h]
\caption{ORBIS search result\label{fig:Orbis-search}}
% Alt Text: Figure A2 is a screenshot of the search results as seen on Orbis’ interface.

\includegraphics[scale=0.75]{\string"search result\string".png}
\end{figure}

The identification of greenfield investment during the last ten years
is relatively simple. The variable ``date of incorporation'' allows
us to identify new greenfield investments. For example, Netflix entered
the Spanish market in 2018. In Figure \ref{fig:Orbis-search-2-1},
we see that Netflix's assets, employees, and sales were zero
before 2018.

\begin{figure}[h]
\caption{ORBIS greenfield FDI example: Netflix in 2018\label{fig:Orbis-search-2-1}}
% Alt Text: Figure A3 is a screenshot of search results for Netflix in Spain.
\includegraphics[scale=0.75]{netflix}
\end{figure}


Orbis' search module allows us to identify M\&As during the sample
period and resolve changes in ownership. 

Figure \ref{fig:Orbis-MA} shows that during the last ten years, American companies have acquired
26 subsidiaries.

\begin{figure}[h]
\caption{ORBIS M\&A deals\label{fig:Orbis-MA}}
% Alt Text: Figure A4 is a screenshot of search results for US M\&A deals in Spain. 
\includegraphics[scale=0.75]{MAS}
\end{figure}

\noindent For example, Facebook acquired the Spanish company Playgiga in 2019
as shown in Figure \ref{fig:Orbis-face}. This company was owned by
Spanish investors until that date, and the initial search result would
have included it (incorrectly) as an American subsidiary throughout the period. The M\&A search strategy resolves this issue.

\begin{figure}[h]
\caption{ORBIS M\&A deal example: Facebook\label{fig:Orbis-face}}
% Alt Text: Figure A5 is a screenshot of Facebook M\&A activity in Spain.
\includegraphics[scale=0.75]{playgiga}
\end{figure}

This search strategy allows us to identify complex changes in ownership,
like the investment through shell companies. For example, Costco Spain
appears to be a M\&A 2013 as shown in Figure \ref{fig:Orbis-cosco}.

\begin{figure}[h]
\caption{ORBIS M\&A deal example: Cosco\label{fig:Orbis-cosco}}
% Alt Text: Figure A6 is a screenshot of Cosco M&A activity in Spain. 

\includegraphics[scale=0.75]{cosco}
\end{figure}

However, a careful inspection of the changes in ownership in Figure
\ref{fig:Orbis-cosco-1} reveals that Costco Spain was owned by a
shell company \textquotedblleft AUXADI SERVICIOS DE MEDIACION SL\textquotedblright{}
in 2012 and 2013, who registered the name on behalf of Costco Inc.
The initial \citet{kalemli2015construct} procedure would have identified
Costco as M\&A from a Spanish firm, when in fact, it was a greenfield
investment. The M\&A search strategy resolves this issue since Costco
was not identified as an M\&A.

In sum, our search strategy is salient in three ways. First, using the Global Ultimate Owner (GUO) allows us to overcome shell company issues. Second, the one million threshold is eliminated from the sample of non-active affiliates. Third, the use of M\&A data allows us to easily overcome changes in ownership within the sample period.



\begin{figure}[h]
\caption{ORBIS M\&A deal example: Cosco (cont.)\label{fig:Orbis-cosco-1}}
% Alt Text: Figure A7 is a screenshot of Cosco M&A activity in Spain. 
\includegraphics[scale=0.75]{cosco3}
\end{figure}




Figure \ref{fig:Orbis-search-2} shows in detail the information that
ORBIS has for a specific company, in this case, Ford, which operates
in Valencia (Spain) since the mid-1970s. We can follow the yearly
evolution of its sales, employees, total assets, and the rest of the
variables described in section \ref{sec:ORBIS-Variables}.

\begin{figure}[h]
\caption{ORBIS search result example in detail: Ford\label{fig:Orbis-search-2}}
% Alt Text: Figure A8 is a screenshot of search results for a Ford affiliate in Valencia, Spain.
\includegraphics[scale=0.75]{ford}
\end{figure}


\section{Other Tables}\label{sec:other}
\renewcommand{\thetable}{B.\arabic{table}}
\setcounter{table}{0}



MREID covers 25 NAICS 2-digit industries reported in Table \ref{tab:industry}. We report each industry's average and maximum number of affiliates per country pair. Table B2 provides country coverage and summary statistics.
\clearpage
\singlespacing
\begin{longtable}[c]{c>{\centering}p{6cm}cc}
\toprule 
\caption{Industry coverage and summary statistics (number of affiliates per country pair).}
\label{tab:industry}
\tabularnewline
\toprule 
NAICS2  & NAICS\textbackslash\_desc  & mean & max\tabularnewline
\midrule 
11  & Agriculture, Forestry, Fishing and Hunting  & 3  & 128 \tabularnewline
\addlinespace
21  & Oil and Gas Extraction  & 4  & 201 \tabularnewline
\addlinespace
22  & Utilities  & 14  & 1,227 \tabularnewline
\addlinespace
23  & Construction of Buildings  & 8  & 1,863 \tabularnewline
\addlinespace
31  & Food Manufacturing  & 2  & 131 \tabularnewline
\addlinespace
32  & Wood Product Manufacturing  & 2  & 224 \tabularnewline
\addlinespace
33  & Fabricated Metal Product Manufacturing  & 2  & 236 \tabularnewline
\addlinespace
42  & Wholesale Trade  & 3  & 603 \tabularnewline
\addlinespace
44  & Food and Beverage Stores  & 3  & 304 \tabularnewline
\addlinespace
45  & Miscellaneous Store Retailers  & 3  & 233 \tabularnewline
\addlinespace
48  & Air Transportation  & 3  & 268 \tabularnewline
\addlinespace
49  & Postal Service  & 5  & 222 \tabularnewline
\addlinespace
51  & Information  & 4  & 571 \tabularnewline
\addlinespace
52  & Finance and Insurance  & 5  & 1,298 \tabularnewline
\addlinespace
53  & Real Estate  & 13  & 2,746 \tabularnewline
\addlinespace
54  & Legal Services  & 8  & 1,782 \tabularnewline
\addlinespace
55  & Management of Companies and Enterprises  & 55  & 6,669 \tabularnewline
\addlinespace
56  & Administrative and Support and Waste Management and Remediation Services  & 6  & 1,064 \tabularnewline
\addlinespace
61  & Educational Services  & 9  & 1,294 \tabularnewline
\addlinespace
62  & Health Care and Social Assistance  & 7  & 490 \tabularnewline
\addlinespace
71  & Arts, Entertainment, and Recreation  & 5  & 225 \tabularnewline
\addlinespace
72  & Accommodation  & 9  & 685 \tabularnewline
\addlinespace
81  & Repair and Maintenance  & 3  & 294 \tabularnewline
\addlinespace
92  & Executive, Legislative, and Other General Government Support  & 5  & 209 \tabularnewline
\addlinespace
99  & Unclassified Establishments  & 159  & 61,605 \tabularnewline
\midrule 
\multicolumn{4}{l}{Note: Statistics at the country-pair level.}\tabularnewline
\bottomrule
\end{longtable}
\doublespacing

\begin{longtable}{lccccc} 
\caption{Country coverage and summary statistics for Inward, Outward, Domestic affiliates and Global Ultimate Owners).}\label{tab:country}\\
\hline
\hline
iso3 & Country name & Inward & Outward & Domestic & GUO \\ \hline
ABW & Aruba & 1 & 34 & 0 & 1 \\
AGO & Angola & 124 & 25 & 1 & 12 \\
AIA & Anguilla & 2 & 9 & 0 & 5 \\
ALB & Albania & 59 & 3 & 70 & 46 \\
AND & Andorra & 1 & 30 & 0 & 5 \\
ARE & United Arab Emirates & 292 & 976 & 108 & 242 \\
ARG & Argentina & 77 & 80 & 140 & 77 \\
ARM & Armenia & 10 & 6 & 2 & 7 \\
ATG & Antigua \& Barbuda & 0 & 4 & 0 & 1 \\
AUS & Australia & 4,415 & 4,689 & 2,894 & 1,941 \\
AUT & Austria & 4,412 & 3,649 & 2,440 & 932 \\
AZE & Azerbaijan & 3 & 32 & 0 & 7 \\
BDI & Burundi & 2 & 0 & 0 & 0 \\
BEL & Belgium & 6,576 & 6,347 & 8,699 & 2,660 \\
BEN & Benin & 18 & 0 & 0 & 0 \\
BFA & Burkina Faso & 6 & 0 & 0 & 0 \\
BGD & Bangladesh & 9 & 18 & 2 & 17 \\
BGR & Bulgaria & 1,903 & 115 & 1,800 & 922 \\
BHR & Bahrain & 27 & 58 & 22 & 40 \\
BHS & Bahamas & 5 & 410 & 3 & 52 \\
BIH & Bosnia \& Herzegovina & 276 & 71 & 143 & 158 \\
BLR & Belarus & 20 & 66 & 13 & 44 \\
BLZ & Belize & 3 & 55 & 0 & 21 \\
BMU & Bermuda & 123 & 5,298 & 56 & 424 \\
BOL & Bolivia & 9 & 3 & 1 & 4 \\
BRA & Brazil & 12,025 & 804 & 47,198 & 19,720 \\
BRB & Barbados & 15 & 21 & 1 & 10 \\
BRN & Brunei & 0 & 8 & 0 & 3 \\
BWA & Botswana & 12 & 20 & 2 & 4 \\
CAF & Central African Republic & 0 & 2 & 0 & 1 \\
CAN & Canada & 2,278 & 5,846 & 3,331 & 3,309 \\
CHE & Switzerland & 3,195 & 7,738 & 930 & 991 \\
CHL & Chile & 2,214 & 414 & 4,096 & 2,122 \\
CHN & China & 23,982 & 5,104 & 164,203 & 67,937 \\
CIV & Cote d'Ivoire & 29 & 2 & 5 & 3 \\
CMR & Cameroon & 7 & 0 & 0 & 0 \\
COD & Dem. R. Congo & 6 & 1 & 0 & 2 \\
COG & Congo & 2 & 0 & 0 & 0 \\
COL & Colombia & 1,687 & 153 & 1,220 & 801 \\
CPV & Cape Verde & 19 & 1 & 2 & 4 \\
CRI & Costa Rica & 15 & 9 & 7 & 10 \\
CUB & Cuba & 1 & 1 & 0 & 1 \\
CUW & Curaçao & 6 & 570 & 2 & 30 \\
CYM & Cayman Islands & 110 & 10,977 & 50 & 1,258 \\
CYP & Cyprus & 202 & 3,617 & 112 & 789 \\
CZE & Czech Republic & 5,202 & 1,461 & 11,287 & 4,425 \\
DEU & Germany & 21,421 & 22,858 & 24,082 & 6,676 \\
DJI & Djibouti & 2 & 0 & 0 & 0 \\
DMA & Dominica & 0 & 7 & 0 & 2 \\
DNK & Denmark & 5,506 & 4,416 & 5,060 & 1,407 \\
DOM & Dominican Republic & 5 & 5 & 3 & 3 \\
DZA & Algeria & 40 & 48 & 1 & 22 \\
ECU & Ecuador & 196 & 7 & 8 & 14 \\
EGY & Egypt & 148 & 61 & 94 & 82 \\
ESP & Spain & 12,665 & 7,268 & 27,199 & 11,045 \\
EST & Estonia & 1,329 & 286 & 1,961 & 992 \\
ETH & Ethiopia & 9 & 0 & 0 & 1 \\
FIN & Finland & 2,892 & 3,408 & 8,906 & 4,305 \\
FJI & Fiji & 7 & 1 & 2 & 2 \\
FRA & France & 16,212 & 19,913 & 37,123 & 9,354 \\
GAB & Gabon & 6 & 5 & 3 & 2 \\
GBR & United Kingdom & 56,296 & 24,323 & 80,871 & 19,712 \\
GEO & Georgia & 50 & 6 & 72 & 63 \\
GHA & Ghana & 315 & 0 & 9 & 11 \\
GIB & Gibraltar & 12 & 163 & 0 & 37 \\
GIN & Guinea & 4 & 0 & 0 & 0 \\
GMB & Gambia & 1 & 1 & 0 & 1 \\
GNB & Guinea-Bissau & 1 & 1 & 0 & 1 \\
GRC & Greece & 710 & 476 & 776 & 466 \\
GRD & Grenada & 2 & 0 & 0 & 0 \\
GTM & Guatemala & 17 & 0 & 6 & 3 \\
GUY & Guyana & 1 & 4 & 1 & 3 \\
HKG & Hong Kong & 739 & 4,033 & 80 & 736 \\
HND & Honduras & 7 & 5 & 3 & 2 \\
HRV & Croatia & 1,158 & 232 & 779 & 620 \\
HUN & Hungary & 1,846 & 684 & 1,858 & 1,546 \\
IDN & Indonesia & 90 & 252 & 49 & 128 \\
IND & India & 8,768 & 3,871 & 11,707 & 5,389 \\
IRL & Ireland & 7,002 & 5,812 & 2,501 & 1,254 \\
IRN & Iran & 4 & 26 & 83 & 46 \\
IRQ & Iraq & 8 & 4 & 4 & 5 \\
ISL & Iceland & 139 & 251 & 1,332 & 579 \\
ISR & Israel & 143 & 1,274 & 320 & 351 \\
ITA & Italy & 11,159 & 8,429 & 25,609 & 12,700 \\
JAM & Jamaica & 9 & 26 & 8 & 13 \\
JOR & Jordan & 38 & 17 & 25 & 27 \\
JPN & Japan & 1,507 & 23,628 & 27,326 & 8,282 \\
KAZ & Kazakhstan & 61 & 29 & 42 & 47 \\
KEN & Kenya & 24 & 27 & 10 & 17 \\
KGZ & Kyrgyz Republic & 3 & 0 & 0 & 0 \\
KHM & Cambodia & 15 & 2 & 3 & 6 \\
KNA & Saint Kitts and Nevis & 2 & 47 & 0 & 10 \\
KOR & South Korea & 1,650 & 2,321 & 3,957 & 2,968 \\
KWT & Kuwait & 11 & 415 & 55 & 72 \\
LAO & Laos & 4 & 1 & 0 & 1 \\
LBN & Lebanon & 15 & 74 & 21 & 43 \\
LBR & Liberia & 7 & 35 & 0 & 10 \\
LBY & Libya & 1 & 19 & 0 & 2 \\
LCA & Saint Lucia & 1 & 3 & 1 & 3 \\
LIE & Liechtenstein & 14 & 594 & 1 & 112 \\
LKA & Sri Lanka & 42 & 31 & 84 & 58 \\
LSO & Lesotho & 47 & 0 & 1 & 1 \\
LTU & Lithuania & 877 & 434 & 863 & 514 \\
LUX & Luxembourg & 10,390 & 8,096 & 948 & 919 \\
LVA & Latvia & 1,152 & 162 & 802 & 414 \\
MAC & Macau & 6 & 34 & 0 & 10 \\
MAR & Morocco & 735 & 128 & 211 & 89 \\
MCO & Monaco & 21 & 66 & 0 & 23 \\
MDA & Moldova & 76 & 21 & 58 & 61 \\
MDG & Madagascar & 6 & 3 & 0 & 3 \\
MDV & Maldives & 2 & 0 & 0 & 0 \\
MEX & Mexico & 1,025 & 754 & 220 & 189 \\
MHL & Marshall Islands & 2 & 149 & 5 & 31 \\
MKD & Macedonia & 119 & 19 & 239 & 161 \\
MLI & Mali & 8 & 0 & 0 & 0 \\
MLT & Malta & 1,136 & 504 & 1,030 & 413 \\
MMR & Myanmar & 1 & 1 & 0 & 2 \\
MNE & Montenegro & 95 & 12 & 17 & 27 \\
MNG & Mongolia & 3 & 4 & 1 & 3 \\
MOZ & Mozambique & 107 & 0 & 2 & 4 \\
MRT & Mauritania & 1 & 0 & 0 & 0 \\
MUS & Mauritius & 161 & 500 & 298 & 221 \\
MWI & Malawi & 11 & 1 & 8 & 4 \\
MYS & Malaysia & 4,032 & 2,086 & 12,554 & 4,150 \\
NAM & Namibia & 20 & 2 & 2 & 1 \\
NER & Niger & 3 & 1 & 0 & 1 \\
NGA & Nigeria & 19 & 47 & 13 & 28 \\
NIC & Nicaragua & 3 & 4 & 1 & 1 \\
NLD & Netherlands & 17,387 & 10,954 & 9,955 & 2,889 \\
NOR & Norway & 4,067 & 4,446 & 18,924 & 5,220 \\
NPL & Nepal & 2 & 0 & 12 & 15 \\
NZL & New Zealand & 1,040 & 357 & 131 & 178 \\
OMN & Oman & 43 & 23 & 14 & 25 \\
PAK & Pakistan & 49 & 28 & 51 & 51 \\
PAN & Panama & 24 & 506 & 36 & 102 \\
PER & Peru & 102 & 33 & 20 & 32 \\
PHL & Philippines & 862 & 326 & 773 & 208 \\
PNG & Papua New Guinea & 2 & 3 & 1 & 2 \\
POL & Poland & 8,995 & 671 & 6,484 & 2,868 \\
PRK & North Korea & 0 & 0 & 0 & 1 \\
PRT & Portugal & 4,880 & 1,894 & 10,101 & 4,452 \\
PRY & Paraguay & 175 & 4 & 8 & 12 \\
PSE & Palestine & 6 & 1 & 3 & 3 \\
QAT & Qatar & 25 & 126 & 26 & 41 \\
ROU & Romania & 5,117 & 112 & 1,215 & 830 \\
RUS & Russia & 6,648 & 611 & 11,576 & 4,269 \\
RWA & Rwanda & 8 & 0 & 1 & 1 \\
SAU & Saudi Arabia & 166 & 148 & 203 & 152 \\
SDN & Sudan & 2 & 1 & 0 & 1 \\
SEN & Senegal & 13 & 8 & 0 & 3 \\
SGP & Singapore & 15,801 & 3,186 & 6,901 & 2,248 \\
SLE & Sierra Leone & 1 & 2 & 0 & 2 \\
SLV & El Salvador & 20 & 0 & 3 & 2 \\
SMR & San Marino & 0 & 26 & 0 & 10 \\
SRB & Yugoslavia & 1,632 & 74 & 462 & 353 \\
STP & Sao Tome and Principe & 1 & 0 & 0 & 0 \\
SUR & Suriname & 0 & 1 & 0 & 1 \\
SVK & Slovak Republic & 3,750 & 515 & 2,584 & 1,567 \\
SVN & Slovenia & 587 & 282 & 903 & 629 \\
SWE & Sweden & 8,123 & 10,907 & 42,249 & 12,128 \\
SWZ & Swaziland & 10 & 0 & 0 & 0 \\
SYC & Seychelles & 8 & 71 & 0 & 32 \\
SYR & Syria & 5 & 7 & 0 & 5 \\
TCD & Chad & 2 & 1 & 0 & 1 \\
TGO & Togo & 4 & 32 & 2 & 3 \\
THA & Thailand & 3,888 & 973 & 4,364 & 1,359 \\
TKM & Turkmenistan & 0 & 1 & 0 & 1 \\
TTO & Trinidad and Tobago & 8 & 15 & 5 & 7 \\
TUN & Tunisia & 7 & 36 & 3 & 24 \\
TUR & Turkey & 580 & 476 & 461 & 468 \\
TWN & Taiwan & 74 & 2,875 & 335 & 883 \\
TZA & Tanzania & 26 & 66 & 3 & 6 \\
UGA & Uganda & 24 & 0 & 1 & 2 \\
UKR & Ukraine & 1,950 & 131 & 1,800 & 995 \\
URY & Uruguay & 314 & 137 & 50 & 138 \\
USA & United States & 14,730 & 93,450 & 138,312 & 113,556 \\
UZB & Uzbekistan & 4 & 2 & 5 & 6 \\
VCT & St. Vincent and Gr. & 0 & 6 & 0 & 5 \\
VEN & Venezuela & 1 & 34 & 2 & 4 \\
VGB & British Virgin Islands & 7 & 2,739 & 0 & 551 \\
VNM & Vietnam & 2,045 & 67 & 1,494 & 774 \\
WSM & Samoa & 0 & 42 & 0 & 21 \\
YEM & Yemen & 0 & 1 & 0 & 1 \\
ZAF & South Africa & 25 & 1,464 & 49 & 162 \\
ZMB & Zambia & 19 & 0 & 2 & 3 \\
 ZWE & Zimbabwe & 15 & 1 & 6 & 7 \\ \hline
\end{longtable}



%\section{Extra Appendix}



\begin{comment}
   \section{Misc}
\begin{enumerate}
\item Bilateral FDI \textquotedblleft measures\textquotedblright \\
There are several ways to measure FDI, which include:
\begin{itemize}
\item Foreign affiliate sales (FAS)
\begin{itemize}
\item Value Added
\item Net income (Profit/Loss)
\end{itemize}
\end{itemize}
Examples of statistical dataset sources reporting FAS: Bureau of Economic
Analysis (BEA), ORBIS, FDIMarkets
\begin{itemize}
\item Total Assets\\
BEA, ORBIS
\item Fixed Assets\\
BEA\footnote{BEA reports Net property, plant \& equipment}, ORBIS
\begin{itemize}
\item Capital expenditures\footnote{Capital expenditures are changes in fixed assets}\\
BEA / FDIMarkets
\end{itemize}
\item Number of employees\\
BEA / ORBIS / FDIMarkets
\begin{itemize}
\item Cost of employees
\end{itemize}
\end{itemize}
\item FDI types
\begin{enumerate}
\item Mergers \& Acquisitions (M\&A)\\
ORBIS, Thompson Reuters
\item Greenfield \\
BEA / ORBIS / FDIMarkets
\item Vertical / Horizontal FDI \\
ORBIS
\end{enumerate}
\item Dataset limitations
\begin{itemize}
\item ORBIS
\begin{itemize}
\item Issues tracking complex changes in ownership
\item Less coverage on private firms in the US than in the EU
\item 10-year period window
\item some data is estimated
\item \textquotedblleft Global ultimate owner\textquotedblright{} available
only for the present day:
\end{itemize}
\item BEA
\begin{itemize}
\item Only US FDI
\item No firm-level data
\end{itemize}
\item FDIMarkets
\begin{itemize}
\item Only greenfield investment
\item Only capital and employment measures
\item Some data is estimated
\end{itemize}
\end{itemize}
\end{enumerate}
 
\end{comment}




\end{document}