The role of digital intermediation platforms in opening up international markets

Annex A. Data and empirical specification

Data overview

The analysis draws on a dataset of around 860 digital intermediary platforms (DIPs) operating across three consumer-facing sectors: travel and accommodation booking, ride-hailing and carpooling, and X2C online marketplaces. Platforms are classified as “pure” DIPs, which act solely as intermediaries facilitating transactions for a fee, or “hybrid” DIPs, which combine intermediation with direct retail activity (IMF et al., 2023[1]). Platform identification and screening were conducted in accordance with the definitional framework set out in the Handbook on Measuring Digital Trade, which characterises DIPs by their role in enabling transactions between multiple distinct parties via an online interface. Platforms were systematically identified using public databases (Crunchbase, Pitchbook) and cross-checked with company websites, focusing on platforms with measurable web presence between 2014 and 2024. The sample retains platforms throughout their full period of activity, including those that entered or exited the market, ceased operations, or were subject to mergers and acquisitions during the studied period, to capture platform entry and exit dynamics over time. Platform origin is assigned based on the ownership structure under which the platform operated for the greatest amount of time during the period covered. Table 1.A.1 presents a detailed breakdown of the number and type of platforms by sector, distinguishing between those with a presence in multiple countries and those operating in one or a few countries.

Table A A.1. Number and type of DIPs by sector and area of activity, 2014‑2024

Area of activity	Main sectors affected (ISIC Rev. 4)	DIP type	Number of platforms	Examples
				Present in many countries	Present in one or a few countries
Travel & Accommodation Booking	Travel agency, tour operator, reservation service and related activities (N79) Accommodation (I55) Air transport (H51)	Pure	337	airbnb.com, booking.com, expedia.com, agoda.com	stayz.com.au, szallas.hu, travel.rakuten.co.jp
Ride-Hailing & Carpooling	Land transport and transport via pipelines (H49)	Pure	191	uber.com, bolt.eu, free-now.com, blablacar.com	didiglobal.com, gett.com, liftago.cz, gomore.dk
X2C Online Marketplace	Retail trade (G47)	Pure	256	ebay.com, aliexpress.com, etsy.com, vinted.com, temu.com	onbuy.com, mydeal.com.au, joom.com
		Hybrid	75	amazon.com	kogan.com, linio.falabella.com, bol.com, coupang.com
Total			859

Notes: Pure DIPs act solely as intermediaries, facilitating transactions between buyers and sellers without taking economic ownership of goods or services. Hybrid DIPs combine this role with direct retail operations. Data reflect platforms’ operational presence across OECD Member countries between 2014 and 2024.

Sources: Author’s elaboration based on data from Crunchbase.

Empirical specification

To examine how regulatory environments shape the uptake and diffusion of DIPs, the analysis builds on the framework developed by Costa et al. (2021[5]) and is adapted to a sector-specific context. Regressions are estimated at the country‑sector‑year level using a log-linear panel model, in which the dependent variable – expressed in natural logarithms – captures platform traffic as a proxy for diffusion. Three traffic metrics are used: total traffic, traffic from foreign firms, and the share of traffic from foreign firms, providing a comprehensive view of both the scale and international dimension of platform use.

${ln\left(Traffic\right)}_{c,s,t}={\beta }_0+{{\beta }_{1}Characteristics}_{c,t}+{{\beta }_{2}Policy}_{c,t}+{\theta }_t+{ϵ}_{c,s,t}$ (1)

Where ${ln\left(Traffic\right)}_{c,s,t}$ denotes the natural logarithm of a given traffic metric in country c, sector s, and year t. The vector ${Characteristics}_{c,t}$​ includes structural country-level factors that may influence DIP uptake – namely, GDP per capita (log-transformed), total population (log-transformed), self-employment rates, fixed broadband subscriptions per 100 inhabitants, and the Network Readiness Index (NRI). These controls are included in all regressions to isolate the specific effect of policy environments on platform diffusion.

The variable ${Policy}_{c,t}$ captures the regulatory environment affecting digital services. The focus is on the DSTRI, introduced one component at a time to mitigate multicollinearity among highly correlated measures. Both the overall DSTRI score and its sub-components – covering infrastructure restrictions, electronic transactions, payment systems, intellectual property rights, and other relevant policy areas – are examined. For broader context, two additional structural indicators are also included in separate regressions: the overall OECD Product Market Regulation (PMR) indicator (OECD, 2025[26]) and World Bank’s indicator on the cost of enforcing contracts (World Bank, 2025[27]).

The estimation proceeds in two steps for each combination of dependent variable and DIP type. First, a baseline regression is run with only structural characteristics and year fixed effects. Second, policy indicators are added individually to this baseline to assess their association with DIP diffusion, while controlling for structural characteristics to online platform diffusion (Costa et al., 2021[5]). Country fixed effects are not included in the main specification, as many of the explanatory variables – particularly the regulatory indicators – exhibit limited within-country variation over time. Year fixed effects are included to control for common shocks and global trends (e.g., COVID-19). Robust standard errors are clustered at the country level to account for potential serial correlation within countries over time. The regressions are estimated using data from 2014 to 2024, corresponding to the availability of DSTRI data to date.