Tourist Arrivals By Country Data: Dataset Structure and Field Coverage

Three official publishers cover the question at three different resolutions, and all three sit in the ten-dataset pool Datadory catalogs for hotels, resorts and cruise lines.

The broadest is World Bank - International Tourism Arrivals (ST.INT.ARVL), a World Development Indicators series that republishes UN Tourism's annual inbound tourist arrivals. Its bulk table measured during research holds 265 rows across 64 year-columns (1960-2025) - countries plus regional and income-group aggregates such as "Europe & Central Asia" riding alongside Spain and France - flattening to roughly 8-9,000 non-null observations once blank cells drop out.

The most granular single-country source is the US NTTO I-94 International Visitor Arrivals Program, which publishes US inbound counts by country of residence, visa type, port of entry and mode of travel from DHS/CBP ADIS records covering over 355 million travelers.

Farther than most analysts use, but not uniformly. The ST.INT.ARVL grid carries columns from 1960 through 2025, yet populated observations concentrate between 1995 and 2020: 206 economies reported arrivals in 1995, 223 in 2019 and 132 in 2020, with no values published for 2021 onward when this page was checked on 2026-08-22. Treat it as an effective ~1995-2020 series despite what the column headers promise.

One attribution trap: the World Bank asks you to credit upstream, "Yearbook of Tourism Statistics, Compendium of Tourism Statistics and data files, UN Tourism" via the World Bank, not just the World Bank itself.

The four tables answer different questions about the same travelers, and mixing them up is the commonest analysis error:

ST.INT.ARVL carries one number per economy per year - international inbound tourist arrivals, defined as overnight visitors traveling outside their usual environment for less than 12 months for non-remunerated purposes. Where tourist counts are unavailable, visitor counts are shown instead, including same-day visitors, cruise passengers and crew members, which matters here because cruise lines fall inside the definition in several markets. Arrivals count trips, not unique people.

tour_occ_arn2 counts arrivals recorded at accommodation establishments rather than at borders, split by NACE Rev. 2 class - I551 hotels and similar, I552 holiday and other short-stay, I553 camping grounds - and by whether guests are residents or non-residents. Sample values from verification: Veneto (ITC4) logged 7,397,598 foreign hotel arrivals in 2024, Stuttgart (DE11) 1,264,976.

NTTO I-94 decomposes one destination: rows by country or world region crossed with visa type across 19 codes (business B-1/WB/I/E-1/E-2, pleasure B-2/WT, student F-1/M-1), mode of transportation (Air, Sea, Land or Not Reported), age bands, US state of first intended address and roughly 40 named ports of entry. Table B ranks the top twenty overseas markets - United Kingdom, India, Japan, Brazil, China PRC, Germany, France, South Korea, Colombia and Italy lead that list.

ST.INT.RCPT.CD, the receipts sibling, converts people into money: annual tourism receipts in current US$ for the roughly 130-175 economies reporting each year, joinable to arrivals on country code and year.

Cross-country arrival comparisons carry three documented traps.

First, collection methods differ: border statistics, border surveys and tourism accommodation establishments feed national series, and some countries count only air arrivals or only hotel stays. France's 2019 jump reflects a methodology change adding same-day visitors, which is why the Metadata_Country special notes should be read before any ranking is published.

Second, coverage thresholds bite at the edges. Eurostat's ESMS metadata notes that establishments under ten bed places may be excluded (twenty for countries below 1% of EU nights), non-rented accommodation falls outside coverage, and a time-series break sits between the 2011 and 2012 reference periods - so a 2011-versus-2012 growth rate is not comparable within Europe even though it is computable.

All eight Datadory personas register these series in their Hotels, Resorts & Cruise Lines packs, and the World Bank indicator alone appears in every one of them - four personas at maximum relevance 3.

Data scientists pull the clean annual cross-country panel via REST for demand models; market researchers benchmark roughly 200 reporting economies against one harmonized definition; journalists and academics cite UN Tourism figures via the World Bank with clear attribution instead of paywalled estimates; investors use country arrivals as a macro demand proxy while respecting the ~1995-2020 populated window. Sales teams rank destination markets when sizing travel-sector territories, and e-commerce operators gauge travel-adjacent demand.

Two workflow pairings recur. Joining arrivals with receipts (ST.INT.RCPT.CD divided by ST.INT.ARVL) yields spend per visitor by destination. And pairing demand with supply - arrivals against the establishments, bedrooms and bed-places counted in Eurostat tour_cap_nat (24,765 observations back to 1990) or OpenStreetMap's 452,341 tourism=hotel objects - flags markets where demand outruns room stock. The investor framing of these joins continues on the hotels resorts cruise lines data guide.