FIRM HETEROGENEITY IN SERVICES TRADE: MICRO-LEVEL EVIDENCE FROM EIGHT OECD COUNTRIES

FIRM HETEROGENEITY IN SERVICES TRADE: MICRO-LEVEL EVIDENCE FROM EIGHT OECD COUNTRIES Sebastian Benz, Dorothée Rouzet, Francesca Spinelli Workshop on Trade Policy, Inclusiveness and the Rise of the Service Economy 25-26 April, 2018 - Geneva

In a nutshell.. Confidential firm-level trade and investment data from Belgium, Finland, Germany, Italy, Japan, Sweden, the United Kingdom and the United States. Ø Cross-border exports and imports of services Ø Ownership and sales of affiliates Ø Key financial and employment information Main question: How are the decisions of services exporters and investors influenced by policy conditions in the host markets? Looking at: Ø Trade margin: The decision to enter a market (extensive margin) and the value of exports (intensive margin) Ø Granularity: Heterogeneous impact of restrictions depending on firm size, productivity, experience, main activity Ø Modes of supply: Cross-border trade and foreign affiliate sales Ø Common features: Main patterns that can be identified in most countries..

Related literature Gravity models are able to explain global patterns of services trade: Ø Kox and Lejour (2005), Kimura and Lee (2006), Anderson et al. (2014, 2015), Nordås and Rouzet (2016), Benz (2016) Services trade barriers are high and are significantly associated with firm-level services trade: Ø Kelle et al. (2013), Crozet et al. (2013), Crozet et al. (2016), Christen et al. (2013), Christen and Francois (2015), Ariu (2016b), Ariu et al. (2017) Services exports are concentrated among few large firms, exporting large volumes to a large number of countries: Ø Breinlich and Criscuolo (2011), Kelle and Kleinert (2010), Gaulier et al. (2010), Federico and Tosti (2012), Ariu (2016a, 2017), Walter and Dell mour (2010), Haller et al. (2014)

Methodology Gravity-type regressions on firm-level exports and foreign affiliate sales. Eight datasets, each including exporters from a single country. Exploiting heterogeneous trade patterns of firms across destinations: Interaction of firm characteristics and STRI. Effect of services trade restrictions on exports differs by firm characteristics. Reported results based on median coefficients. Conservative clustering of s.e. (importer level) Firm-level control variables: turnover, productivity, main activity, foreign ownership, services imports

Datasets Micro-data from 8 OECD countries Ø Belgium: Survey on Trade in Services (National Bank of Belgium) Ø Finland: International Trade in Services and Foreign Affiliate Statistics (Statistics Finland) Ø Germany: Statistics on International Trade in Services and Micro-database Direct Investment (Deutsche Bundesbank) Ø Italy: Survey on International Trade in Services (Bank of Italy) Ø Japan: Basic Survey on Overseas Business Activities (Ministry of Economy, Trade and Industry) Ø Sweden: Survey of Foreign Trade in Services (Statistics Sweden) Ø United Kingdom: International Trade in Services Inquiry (Office for National Statistics) Ø United States: Survey of Transactions in Selected Services and Intellectual Property with Foreign Persons (Bureau of Economic Analysis)

Data characteristics Firm-time-importer-service dimension Time: usually seven years of data (2008-2014) with some exceptions. Services: Audio-visual, Construction, Postal and Courier, Banking, Insurance, Accounting, Architecture/Engineering, Legal, Telecommunications, Maritime transport Importers: 42 countries (OECD + BRA, COL, CHN, IDN, IDN, RUS, ZAF)

The OECD STRI Extensive coverage of global services Ø 44 countries Ø 22 sectors Ø 4 years of data available (2014-2017) Restrictions on multilateral basis: Ø Preferential agreements are not taken into account. Standardised set of measures: Ø Minimum: 61 measures in computer services Ø Maximum: Air transport (154 measures), Telecommunications (148), Insurance (143). Indices from 0 (liberal) to 1 (restrictive)

Methodology: PPML Gravity structure using time-variant firm-level i and time-variant country-level c control variables and year dummies. %! "#$ = exp(+ + -. /012 # % + - 3 /012 # % 4 "$ + 56 "$ + 7/ #$ + 8 $ + 9 "#$ 4 "$ indicates firm characteristics of interest Additional trade cost equivalents (e.g. faced by small firms) t " = [exp - 3 >?@A # % (max ln?f@ghij@ ln?f@ghij@ " )/r 1] 100 where max ln?f@ghij@ indicates the log turnover of large firms used as benchmark level of firm size. % )

Methodology: Probit! " & #$% > 0 = F(, +. / 0123 & $ +. 4 0123 & $ 5 #% + 67 #% + 80 $% + 9 % + : & #$% ) How to identify expected change in number of exporters? Find probability cut-off to maximises regression fit using Matthews correlation coefficient: <== = 100?@?ABC@ CA Ø MCC = 100 if all predictions correct?@dc@?@dca?adc@ (?ADCA) Ø MCC = -100 if all predictions incorrect Simulate change in predicted number of exporters

Results: export experience (PPML) 60% Tariff mark-up for firms without export experience in the destination country Additional ad valorem tariff equivalent 45% 30% 15% rho -1.5 rho -3 rho -5 0% 0 0.1 0.2 0.3 0.4 STRI Note: The numbers indicate the additional ad valorem tariff equivalent for firms without export experience, where export experience is defined as having exported the same service to the same country in the previous year. Results are based on the median coefficient from sector-level PPML regressions. Import demand elasticities used for the calculation of the ad valorem equivalent are indicated as rho. Source: Own elaborations on firm-level data from Belgium, Finland, Italy, Germany, Sweden, the United Kingdom, and the United States.

Results: export experience (Probit) Effect of STRI on number of exporters, by firm characteristics Estimated impact of a global 0.1 reduction in STRI % growth of exporter-destination combinations (left axis) % growth of exporter-destination combinations 1.00% 0.75% 0.50% 0.25% 0.00% Group share of all exporters in baseline (right axis) Yes No 80% 60% 40% 20% 0% Group share of all exporters in baseline Export experience to same country in previous year Note: The numbers indicate the median percentage change in the number of exporter-destination combinations when reducing the STRI score by 0.1. Small firms have an annual turnover of less than EUR 20 million, large firms have an annual turnover of more than EUR 150 million. Source: Own elaborations based on firm-level data from Belgium, Finland, Italy, Germany, the United Kingdom and the United States.

Results: Firm size (PPML) Additional ad valorem tariff equivalent of STRI = 0.2 30% 20% 10% 0% Tariff mark-up for small firms on cross-border exports Estimated additional tariff equivalent of an STRI score of 0.2, by turnover in EUR Sectors: audio-visual, banking, insurance, maritime transport, telecommunications 500,000 1 million 5 million 10 million 50 million 200 million Turnover rho -1.5 rho -3 rho -5 Note: The numbers indicate the additional ad valorem tariff equivalent of an STRI score of 0.2 for small and medium sized enterprises. Estimates are based on the median coefficient from sector-level PPML regressions, except professional services. Import demand elasticities used for the calculation of the ad valorem equivalent are indicated as rho. Source: Own elaborations based on firm-level data from Belgium, Finland, Italy, Germany, Sweden, the United Kingdom and the United States.

Results: Firm size (Probit) Effect of STRI on number of exporters, by firm size Estimated impact of a global 0.1 reduction in STRI Sectors: audio-visual, banking, insurance, maritime transport, telecommunications % growth of exporter-destination combinations (left axis) Group share of all exporters in baseline (right axis) 1.00% 40% % growth of exporter-destination combinations 0.75% 0.50% 0.25% 0.00% Small and medium Large Very large 30% 20% 10% 0% Group share of all exporters in baseline Firm size by turnover Note: The numbers indicate the median percentage change in the number of exporter-destination combinations when reducing the STRI score by 0.1. Small and medium firms have an annual turnover of less than EUR 20 million, large firms have an annual turnover of more than EUR 150 million. Thresholds have been chosen to obtain a sufficient number of exporters in each group. Source: Own elaborations based on firm-level data from Belgium, Finland, Italy, Germany, the United Kingdom and the United States.

Results: Firm productivity (PPML) Tariff mark-up for less productive firms on cross-border exports Estimated additional tariff equivalent of an STRI score of 0.2, by labour productivity in EUR Sectors: audio-visual, banking, insurance, maritime transport, postal and courier, telecommunications 6% Additional ad valorem tariff equivalent of STRI = 0.2 4% 2% rho -1.5 rho -3 rho -5 0% 20,000 55,000 110,000 400,000 Labour productivity Note: The numbers indicate the additional ad valorem tariff equivalent of an STRI score of 0.2 for less productive firms. Results are based on the median coefficient from sector-level PPML regressions. Import demand elasticities used for the calculation of the ad valorem equivalent are indicated as rho. Source: Own elaboration based on firm-level data from Belgium, Finland, Italy, Germany, Sweden, the United Kingdom and the United States.

Results: Firm productivity (Probit) Effect of STRI on number of exporters, by firm productivity Estimated impact of a global 0.1 reduction in STRI Sectors: audio-visual, banking, insurance, maritime transport, postal and courier, telecommunications % growth of exporter-destination combinations 1.25% 1.00% 0.75% 0.50% 0.25% 0.00% % growth of exporter-destination combinations (left axis) Group share of all exporters in baseline (right axis) Low and medium High Very high Firm efficiency by labour productivity Note: The numbers indicate the median percentage change in the number of exporter-destination combinations when reducing the STRI score by 0.1. Low and medium productivity firms have a labour productivity of less than EUR 180 000 per worker, very high productivity firms have a labour productivity of more than EUR 360 000 per worker. Source: Own elaborations based on firm-level data from Belgium, Finland, Italy, Germany, the United Kingdom and the United States. 50% 40% 30% 20% 10% 0% Group share of all exporters in baseline

Results: main activity (Probit) Effect of STRI on number of exporters, by firm characteristics Estimated impact of a global 0.1 reduction in STRI 1.00% % growth of exporter-destination combinations (left axis) Group share of all exporters in baseline (right axis) 100% % growth of exporter-destination combinations 0.75% 0.50% 0.25% 0.00% Goods Industry of main activity Services 75% 50% 25% 0% Group share of all exporters in baseline Note: The numbers indicate the median percentage change in the number of exporter-destination combinations when reducing the STRI score by 0.1. Small firms have an annual turnover of less than EUR 20 million; large firms have an annual turnover of more than EUR 150 million. Source: Own elaborations based on firm-level data from Belgium, Finland, Italy, Germany, the United Kingdom and the United States.

Results: foreign affiliate sales Tariff mark-up for small firms on foreign affiliate sales, by parent size Estimated additional tariff equivalent of an STRI score of 0.2, by turnover in EUR Additional ad valorem tariff equivalent of STRI = 0.2 60% 45% 30% 15% rho -1.5 rho -3 rho -5 0% 500,000 1 million 5 million 10 million 50 million 200 million Turnover Note: The numbers indicate the additional ad valorem tariff equivalent of an STRI score of 0.2 for small and medium-sized enterprises. It is based on the median coefficient from sector-level PPML regressions, except professional services. Import demand elasticities used for the calculation of the ad valorem equivalent are indicated as rho. Source: Own elaborations based on firm-level data from Finland, Germany, Japan and the United States.

Results: overview STRI restrictions are particularly detrimental to firms without previous export experience. In a large group of sectors, cross-border exports of small firms and less productive firms are more affected by services trade barriers. Ø For foreign affiliate sales, this holds in all sectors. Ø These findings suggest that a portion of the costs induced by STRI restrictions are fixed costs or sunk costs. Services exporters with a main activity in a manufacturing sector are less affected by services trade barriers. Bundles of goods and services: Services exports follow patterns of goods trade.

Contact us We look forward to hearing from you! Access all of the information from the Trade & Agriculture Directorate at: www.oecd.org/tad You can reach us via e-mail by sending your message to the following address: tad.contact@oecd.org We invite you to connect with us on Twitter by following: @OECDtrade