[Business] Enterprise demography in Europe — 10 questions about where startups thrive and where they die (2008-2020)

What this dataset contains

eurostat_bd_hgnace2_r3_nuts3 (DataCivicLab, source Eurostat) tracks enterprise births, deaths, survival rates and high-growth firms by NACE sector and NUTS3 region, for all EU countries, from 2008 to 2020.

Coverage: 4.4M rows, 15 columns, 2008-2020
Granularity: provincial (NUTS3), annual by NACE sector
Metrics: number of enterprises, births, deaths, 3-year survival rate, high-growth enterprises

---

Questions the data can answer

1. Which EU regions have the highest business birth rates?

The dataset includes birth rate (V97020) as a percentage of active enterprises. Are capital regions systematically higher? Do Nordic countries really lead?

2. Do Southern European businesses die faster?

Death rate (V97030) varies enormously across Europe. Compare Mediterranean regions vs Northern and Central Europe. Is the gap structural?

3. Do survival rates vary by region and sector?

3-year survival rate (V97043) — compare manufacturing vs services, urban vs rural regions. Which sectors have the highest failure rate?

4. Where are high-growth enterprises concentrated?

High-growth enterprises (V11960) — firms growing 10%+ in employment. Which NUTS3 regions have the most? Tech clusters? Industrial districts?

5. Which NACE sectors have the highest birth/death ratio?

Compare births (V11920) vs deaths (V11930) across sectors. Are ICT and professional services more volatile than manufacturing?

6. Is net business population growth positive everywhere?

Net growth (V97010) shows which regional economies are expanding their enterprise base and which are contracting.

7. How many new jobs do enterprise births create?

Employment in births (V16920) measures the job impact of new enterprises. Do births in manufacturing create more jobs than in services?

8. Are Eastern EU regions catching up in business dynamism?

Compare birth rates and high-growth enterprise shares between Eastern and Western EU. Are Poland and Czechia converging?

9. What was the impact of 2020 on enterprise births and deaths?

The dataset includes 2020 — the first COVID year. Compare 2019 vs 2020 birth/death rates across sectors and regions.

10. How do Italian regions rank on business demography?

Compare birth rate, death rate, survival rate, and high-growth enterprise share across Italian NUTS2/3 regions.

---

How to explore

Via MCP: eurostat_query(slug='eurostat_bd_hgnace2_r3_nuts3', sql='...')

Births in Italy (2020):
SELECT geo_label_en, ROUND(SUM(value)) AS births
FROM data
WHERE indic_sb='V11920' AND year=2020 AND nuts_level='NUTS3' AND geo LIKE 'IT%'
GROUP BY geo_label_en ORDER BY births DESC;

High-growth enterprises in Europe (2020):
SELECT geo_label_en, ROUND(SUM(value)) AS high_growth
FROM data
WHERE indic_sb='V11960' AND year=2020 AND nuts_level='NUTS3'
GROUP BY geo_label_en ORDER BY high_growth DESC;

[Gabrymi93]

Answer to Q1: French regions lead Europe in business birth rates, Bulgaria at the bottom — a 31× gap

I queried birth rate (V97020: births as % of active enterprises, 2020) across all NUTS3 regions:

Highest birth rates:

Region	Birth rate (%)
Territoire de Belfort (FR)	43.2%
Haute-Vienne (FR)	32.7%
Hérault (FR)	32.0%
Haute-Garonne (FR)	31.8%
Panevėžio apskritis (LT)	31.6%
Rhône (FR)	31.4%

Lowest birth rates (excluding zeros):

Region	Birth rate (%)
Gabrovo (BG)	1.4%
Lungau (AT)	1.5%
Sisacko-moslavacka (HR)	1.5%
Cuneo (IT)	1.6%
Macerata (IT)	1.7%
Lecco (IT)	1.7%

Key observations:

1. France dominates — 7 out of top 10 regions are French. The French business registration system is efficient, and micro-enterprise creation is actively encouraged (auto-entrepreneur regime).

2. Nordic countries do not lead — contrary to the common narrative, Swedish, Danish and Finnish regions are not in the top tier for business birth rates. Their high-growth enterprise share might be a different story.

3. Italian regions at the bottom — Cuneo (1.6%), Macerata (1.7%), Lecco (1.7%) have among the lowest birth rates in Europe. These are Northern Italian provinces with mature industrial structures — low churn, stable but not dynamic.

4. The gap is 31× between Territoire de Belfort (43.2%) and Gabrovo (1.4%). But high birth rates don't always mean high growth — survival rates matter too.

Want to explore Q3 (survival rates by sector) next?

[hamidedefr]

@Gabrymi93 as promised, business demography first — and your Q1 answer was the perfect place to start, because it is about France and I could check it against a register I live in.

First, a correction, offered in the spirit of #297: the query behind your top-10 is missing a NACE filter. V97020 varies by nace_r2, so an unfiltered ORDER BY value DESC returns single-sector rows. Every value in your top table is sector H (transportation and storage) — Territoire de Belfort's 43.2% is its transport sector, not its economy. Same for the bottom table: Cuneo's 1.6% is also a NACE H row. With the whole-economy aggregate:

SELECT geo_label_en, value FROM data
WHERE indic_sb='V97020' AND year=2020 AND nuts_level='NUTS3'
  AND nace_r2='B-S_X_K642' AND value IS NOT NULL
ORDER BY value DESC;

the 2020 top ten is all ten Lithuanian counties (17.4–19.6%), France runs 7.0–14.3% with a median of 10.0% (the max is Val-de-Marne, in the Paris inner suburbs), Belfort lands at 12.3%, and Cuneo climbs back to 5.3%. So "France dominates Europe, 31× gap" becomes "Lithuania leads, France is high-normal". I reproduced both versions with DuckDB directly on your public clean parquet — the infrastructure works beautifully, for what it is worth.

Second, the more interesting part: the sectoral signal you stumbled on is real, and it has a name. French NACE H birth rates are the meal-delivery boom. INSEE counted 87,900 new "other postal and courier activities" businesses in 2020 — 97% of them micro-entrepreneurs — inside a record 848,200 total creations (Insee Focus 247, Insee Première 1837). Registering as a micro-entrepreneur is free and takes minutes, and a registration is not an economically active enterprise: SIRENE, the French register these figures ultimately come from, is full of shells that never invoice anyone and of businesses that died years before their deregistration. So even correctly filtered, French "birth rates" in this dataset measure registry churn as much as entrepreneurship. Your point about the auto-entrepreneur regime was right — it just cuts the other way: it inflates the numerator rather than proving dynamism.

Two coverage notes from the same session, for anyone comparing countries: the French NUTS3 series is effectively 2014–2020 (2008–2013 is almost empty), and in the 2020 slice I find no NUTS3 values for Germany, Belgium, Ireland or Greece — worth stating which countries are actually in frame before ranking "Europe".

If it is useful, I would take your Q3 (survival rates) next with the same register-aware lens: French survival figures split very differently for micro-entrepreneurs versus classic companies, and that distinction is invisible in the Eurostat aggregate. Happy to bring the SIRENE side if someone brings the Italian one.

Alla prossima.

[Gabrymi93]

Answer to Q3 — and the Q1 correction, acknowledged

First, thank you @hamidedefr for the catch on Q1. Reproduced on the public parquet with the NACE filter: the 2020 top ten is all Lithuanian counties (17.4–19.6%), France runs high-normal (median ~10%, Val-de-Marne the max), Cuneo climbs back to 5.3%. The "France dominates, 31× gap" claim is withdrawn — the unfiltered ranking was reading single NACE H rows, not whole economies. What survives is your second point: the sectoral signal was real.

Q3: do survival rates vary by sector and region? Yes — and Italy's hierarchy doesn't match Europe's.

Method notes first, because they change the reading: V97043 is a 3-year survival rate on the birth cohort T−3 (the 2020 value describes firms born in 2017, not the COVID year); the NUTS3 series is usable only from 2014 (2011-13: 35 values, 2008-10: none); the 2020 slice covers 18 countries at NUTS3 — no DE, BE, IE, EL, confirming your coverage note; nace_r2 has no pure manufacturing, so B-E (industry ex construction) is the proxy; averages are unweighted across NUTS3.

Sectors, 2020 (NUTS3 avg):

nace_r2	EU	IT	Δ
H — transport & storage	62.1	70.6	+8.5
I — accommodation & food	58.7	63.1	+4.4
B-E — industry ex constr.	64.0	63.9	−0.1
R_S — arts & other services	59.8	62.0	+2.2
K_L — finance & real estate	64.1	60.0	−4.1
J — information & comm.	62.5	59.9	−2.6
F — construction	60.1	56.2	−3.9
G — wholesale & retail	57.5	56.6	−0.9
M_N — prof., sci., tech.	59.5	52.5	−7.0
P_Q — education & health	61.5	51.9	−9.6
Whole economy	59.5	57.0	−2.5

Three findings:

1. In Europe, industry and finance survive best (64.1 / 64.0); trade and hospitality worst. In Italy the ranking is inverted: transport (70.6%) and hospitality (63.1%) lead, while knowledge-intensive sectors — professional services (52.5), education & health (51.9), ICT (59.9) — sit below their EU levels. Same NACE codes, opposite hierarchy.
2. Italy overall: 57.0% vs EU 59.5% — 10th of 18 countries. Not the bottom, but below the EU median; only 16 points separate our best province from our worst.
3. Geography is not North-South. Top: Bolzano 65.1, Prato 63.8, Enna 61.9, Barletta-Andria-Trani 61.7, Sondrio 61.7. Bottom: Massa-Carrara 48.5, Reggio Calabria 48.6, Catanzaro 50.5, Vercelli 51.4, L'Aquila 52.1. Tuscany spans both extremes (Prato first, Massa-Carrara last); Rome (53.1) sits near the bottom.

Trend 2014-2020 (IT, provincial avg): 54.9 → 53.4 → 55.4 → 59.4 → 61.2 → 57.9 → 57.0. The 2017-18 peak and the 2019-20 decline are cohort effects (T−3), not COVID years — worth remembering before someone reads the dip as a pandemic signal.

The Q1↔Q3 link: Lithuania tops EU birth rates yet has the lowest survival in Europe (34.4%). High registration churn, low persistence. Birth rate alone says little about dynamism — the survival lens is the one that separates registry noise from actual persistence.

Limits: register-based data (your SIRENE point applies to the whole dataset, Italy included); unweighted averages; B-E as manufacturing proxy; urban/rural split needs a density join — doable with eurostat_pop_density_nuts3, happy to take it next.

And on the collaboration: yes, please bring SIRENE. The H story is a good place to start — transport is exactly where Italy scores highest (70.6%), and if the French H signal is delivery-boom, the Italian one deserves the same register-aware look.

Query (DuckDB, public clean parquet):

SELECT nace_r2, ROUND(AVG(value),1) AS avg_surv_it
FROM data
WHERE indic_sb='V97043' AND year=2020 AND nuts_level='NUTS3'
  AND geo LIKE 'IT%' AND value IS NOT NULL
GROUP BY nace_r2 ORDER BY avg_surv_it DESC;