No clear sign that Romanian town halls buy more media before elections

The calendar that matters is the budget’s#

Anyone looking at the weeks above looks for a wave before the elections. There isn’t one. Two things recur in most years: a peak in April and one in December, at the end of the budget year. The local elections of 27 September 2020 and 9 June 2024 fall in years whose curve looks like the others.

The data are what town halls published in Romania’s public procurement register (SEAP), released on data.gov.ro by the country’s digitalisation authority: direct purchases, contracts awarded through tenders and, from 2023, notices for direct purchases made outside the platform. Only purchases by the town hall or county council itself are kept, not those of its schools, hospitals or companies, and only the procurement codes (CPV) for advertising, marketing, radio and television, public relations and news agencies.

The test was written before the result#

The question, raised in the Romanian press, was whether town halls buy more media in the months before elections. Before anything was computed, the rule was fixed and recorded with a digital fingerprint: each town hall is compared with itself. How much of its spending in a year falls in the 120 days before an election, against the same days in years without a local election? Years with a European or presidential vote inside the same window were not used as comparisons.

The chart below gives the answer in its simplest form. In 2020, 33% of the year’s money fell in the window before the vote; in other years, leaving out 2017 (the start of the data), between 25% and 34%. In 2024 it was 46%, against 35%–55%. The election years do not stand out: 2020 is among the highest, but 2019 and 2021, years without local elections, are as high; 2024 sits in the middle.

The full model, run once, says the same. For 2020: +27% against the same days of other years, with an interval from −19% to +99%. For 2024: −39% (−68% to +17%). Both together: −25%. The rule fixed in advance required, among other things, at least +25% with the whole 95% interval above zero, in both years, and a result larger than any fake year treated as an election year. It is not met: 2024 is negative, the 2020 interval includes zero, and, taken in turn as if they were election years, four of the seven years without local elections come out higher than 2020 (three of six higher than 2024).

The rule also required each month of the window to be reported separately. This is where the one signal appears. In the 2020 campaign month (28 August – 26 September), spending was roughly twice that of the same days in other years (+101% after the corrections, +60% without the five values above the ceiling), more than in any fake year (the closest, 2019: +91%); without the five values, only 2017, the start of the data, is higher. But in that month town halls made the fewest purchases of any year (410), and five contracts account for 53% of the money. In 2024 the campaign month was 48% below the same days of other years. A spike in one of two campaigns, driven by a few large contracts, does not show a habit.

The number of purchases in the window is lower than usual (−15% in 2020, −17% in 2024), but fake years swing so widely that the dip cannot be read as a habit. The fair conclusion is a modest one: there is no clear sign of a rise before elections, and a doubling would have shown in the two elections taken together. A moderate rise, of up to several tens of per cent, cannot be ruled out with these data.

Three corrections came after the test was run and are shown beside the original result, not instead of it. The file for the second quarter of 2022 contains no purchases from June; the 2,585 missing purchases were taken from the public SEAP list. Five direct purchases are recorded above the legal ceiling of their year (today 270,120 lei before VAT), the largest at 50.2 million lei. And an independent review of the code found errors: contracts marked “no framework agreement” were treated as framework agreements and left out of spending, a few contracts republished in a later quarter were counted twice, and some tax codes read from spreadsheets came out wrong. With June filled and the errors fixed, the effect is +39% in 2020 and −36% in 2024; without the five values, +12% and −9%. So 2020 leans up and 2024 down, but in every version 2019 and 2021, years without local elections, come out at least as high as 2020, and every interval for the whole 120-day window includes zero.

Since 2023, almost twice as much#

What the data do show clearly does not follow the election calendar. In 2018–2022, town halls spent around 36 million lei (about 7 million euros) a year on media and advertising. In 2023 the total reached 71 million, then 64 million in 2024 and 63 million in 2025. By 30 September 2026, a year with no national election, 61 million had already been spent.

Part of the rise comes from how purchases are published: since 2023 town halls also report purchases made outside the register, which add around 8 million lei a year. Inflation in 2022–2023 and the doubling of the ceiling for direct purchases, from 135,060 to 270,120 lei in September 2022, may also have played a part. But direct purchases inside the register alone rose from 35 to 56 million lei between 2022 and 2023, and have not gone back to the old level. Radio Free Europe’s Romanian service noticed the jump in 2023 and linked it to the coming election year; the data after 2024 show the new level stayed.

What is bought is changing too. In 2018–2021 about half the money went on purchases whose title mentions a notice, a publication or an EU-funded project, things town halls are obliged to make public. Since 2022 such titles account for between 21% and 29% of the total.

Who sells#

Over the ten years town halls bought from 3,295 companies, associations and institutions. The market is fragmented: the ten largest suppliers by value take only 10% of the money. And the suppliers that reach the most town halls are not local newsrooms. The first, with 925 town halls, sells entries in a “Guide to Romania’s town halls”. Next come firms that take the notices required by law and publish them in the Monitorul Oficial, the official gazette, and in newspapers; the gazette’s own publishing house, with 477 town halls; and the Yellow Pages (Pagini Aurii), with 361.

The tables name suppliers only next to their purchases. A click opens the list, with the date, town hall, value and subject as published. This is not an accusation: town halls may buy advertising, and many of these purchases are notices the law requires. Purchases from private individuals appear as “private individual”, without a name.

What any town hall bought#

Search for a commune, town, county council or supplier. The record shows the total for each year, every purchase on the same strip as the elections, and the full list with the register number by which each can be checked.

What the data do not show#

Only published purchases are counted. Before 2023, direct purchases made outside the register do not appear at all, and until mid-2016 the register shows few media purchases by town halls. The file for the last quarter of 2017 stops on 5 December, and today’s register list holds nothing before 2018, so the rest of December 2017 is missing; without the comparison years touched by these gaps (2017 and 2018) and without the five values above the ceiling, the result is +12% in 2020 and −2% in 2024, close to the result without the five values.

Only town halls and county councils are counted. Schools, hospitals and municipal companies are left out, though some of them buy advertising too.

Money that parties pay the press, reported to Romania’s Permanent Electoral Authority, is a separate channel and is not part of this count. Nor are “public information” contracts paid from other budgets.

The files published for contracts in the second and third quarters of 2019 are identical, so one quarter of tendered contracts is missing. Values are as entered by the town halls, before VAT and not adjusted for inflation.

Method, data and sources#

120 quarterly files published on data.gov.ro (direct purchases, contracts, call-off contracts, award notices and awards without prior notice) were read record by record. For test months the result matches the public SEAP list: in April 2019 and April 2022 all 1,771 and 1,928 direct purchases under code 79341000 were found and none extra; in October 2025, all 707; in August 2020 and April 2024 the files hold two extra purchases each, at the edge of the month.

The county comes from each authority’s record in the register and is shown for 1,189 of the 3,069 authorities. The test and decision rule are in the project file, registered before the run, with one addition, also before the run, that made the rule stricter. The model is a Poisson regression with fixed effects for each town hall in each year; 95% intervals are clustered by town hall. On data with dates shuffled at random those intervals proved too narrow, so the result is also read against fake years.

Download the data

• Every purchase (CSV, compressed)
• Totals per town hall and year (CSV)

Sources

1. Romanian Digitalisation Authority: “Public procurement” 2016–2026, quarterly files (direct purchases, contracts, award notices). Licence OGL-ROU-1.0 (2016–2020) and CC BY 4.0 (2021–2026); filtered, cleaned and joined here. Accessed 23 Oct 2026. data.gov.ro
2. SEAP public list of direct purchases (test-month checks and the June 2022 fill). e-licitatie.ro
3. Law 98/2016, art. 7(5): the direct-purchase ceiling, 270,120 lei from 10 Sept 2022 (Law 208/2022); 135,060 lei from 4 June 2018 (Emergency Ordinance 45/2018). legislatie.just.ro
4. Law 135/2020: local elections on 27 September 2020. legislatie.just.ro
5. Emergency Ordinance 21/2024: elections on 9 June 2024. legislatie.just.ro
6. Law 115/2015, art. 64: the campaign starts 30 days before the vote. legislatie.just.ro
7. Radio Free Europe Romania (Simona Cârlugea), 5 July 2023: direct purchases of promotion and advertising rising. romania.europalibera.org