So the question behind Friday's headline is still open.

Transcript

1. Brazilian headlines said it on Friday.
2. To reach 1,000 people with ads on Facebook and Instagram,
3. President Lula's campaign paid up to 43% more than his rival, Flávio Bolsonaro's.
4. The study behind those headlines says the public data can't establish that.
5. Lula's campaign had accused Meta of cutting the reach of his posts and blocking his ads.
6. Meta admitted disabling his ad account by mistake for just over an hour.
7. It cited technical faults and legal rules, and denied political bias.
8. A judge ordered equal rules for both campaigns on reach and search.
9. Those rules said nothing about prices.
10. On Sunday, Lula won the runoff.
11. It was an election in which we had to defeat
12. interference coming from outside, through digital platforms.
13. The study's authors in Rio de Janeiro
14. took over 7,600 first-round ads from Meta's ad library.
15. Meta sells ads in an auction, weighing each bid against its own scoring,
16. so every ad gets its own price,
17. which the researchers estimated per 1,000 people reached.
18. Four versions of the comparison allowed for more each time:
19. the ads' settings, then timing and spend, then who they reached.
20. The four versions estimated Lula's ads cost 10, 33, 39 and 43% more.
21. To test whether Meta treated the campaigns differently,
22. the study says the first version is the one to use.
23. That 10 was within the margin of error.
24. 43% became the headline.
25. An outside reviewer, Fabio Votta, found a problem.
26. A campaign can run many ads from one Facebook page,
27. sharing a budget and targeting.
28. Allowing for that, none of the four estimates showed a clear difference.
29. No clear difference isn't the same as equal prices.
30. And with Bolsonaro's campaign on a single page,
31. even that check rests on very little.
32. Meta disputes the study's method.
33. It says it doesn't set prices by who the advertiser is,
34. and that comparing the spending ranges it publishes can mislead.
35. The public record can't settle it.
36. Meta reports spending only as a range, and for over half of these ads it's R$0 to R$99,
37. so any price for them rests on an assumed spend.
38. Meta doesn't publish what each campaign wanted its ads to do, or how it bid.
39. Those choices can change the price without Meta treating anyone differently.
40. The study points to a test of the auction itself:
41. identical ads, identical settings, rival parties.
42. So the question behind Friday's headline is still open.
43. The missing numbers sit with Meta and the campaigns.
44. And to test the auction, the study says both would need to take part.