Over 25 years, the market missed next year's inflation by 1.6 points on average

Every Monday, the Central Bank publishes the Focus survey, a report on what banks, brokerages and consultancies expect for inflation, economic growth and interest rates. The median of those answers is what people usually call "the market," and it appears in the news every week. But how often is it right?

To find out, Dado Bruto compared 25 years of Focus reports, from 2000 to 2025, with what actually happened. For each year, expectations were measured on five dates: January and December of the previous year, and January, July and October of the year itself. The outcome is the IPCA (Brazil's official consumer price index) and GDP figures published by IBGE (the national statistics agency), and the Selic on December 31. The error is the distance, in percentage points, between what was expected and what happened.

On average, the December forecast for the following year missed inflation by 1.6 percentage points, growth by 1.6 and interest rates by 2.5. But the average hides a lot. For inflation, the error was under 1 point in 14 of the 25 years; for the Selic, in only 5. And in all three cases, a few years of shock pull the average up.

Inflation: the market tends to see less than what is coming

In 16 of the 25 years, the December forecast for the following year's inflation came in below what happened; in 9, above. The median error, 0.91 point, is well below the mean of 1.62 because a few years drive the calculation.

The biggest misses were in 2001, 2002, 2015 and 2021. For 2002, the market expected 4.8% inflation, and the IPCA closed at 12.5%, in a year when the dollar rose 52%. For 2021, the expectation was 3.3% and the result 10.1%; that year, the dollar rose 7%.

GDP: errors in both directions

For growth, there is no clear direction. The December forecast came in below the result in 12 years and above it in 13, and the average bias is small, 0.15 point. The problem is the size of the errors when they happen. For 2020, the market expected growth of 2.3%, and GDP shrank 3.3%. For 2015, the gap was 4.0 points: 0.5% expected, -3.5% actual.

One caveat. IBGE revises GDP for years after the first release. The comparison uses today's numbers, which may differ from those reported at the time, and this can increase or decrease the error for a specific year.

Selic: the hardest to get right in advance

Interest rates are where the market missed the most: 2.51 points on average for the December forecast, with only 5 of the 25 years within 1 point. The median error, 2.25, is below the mean, again because of a few years. The worst was 2002: in December of the previous year, the market expected the Selic at 17.00% at year-end, and it ended at 25.00%. In 2021, the gap was 6.25 points: 3.00% expected, 9.25% at year-end.

The error shrinks fast as the year goes by

Getting a year that has not even started right is much harder than getting the current year right. Made in January of the previous year, the forecast missed inflation by 2.0 points on average, GDP by 2.2 and the Selic by 3.6. By October of the year itself, the error falls to 0.75, 0.53 and 0.54. The difference is information: between one date and the other, the economy and the world change, and forecasts get corrected.

Does the market beat the simplest calculation?

One way to tell whether a forecast is worth anything is to compare it with the simplest possible one: assume next year will be the same as the year that just ended. The December Focus missed by less than that calculation in 16 of the 25 years for inflation, 17 for GDP and 14 for the Selic. The advantage exists, but it is moderate. It is clearer for GDP (mean error of 1.61 versus 2.62) than for inflation (1.62 versus 1.96) and the Selic (2.51 versus 3.03).

And now, in 2026?

In the October 9 Focus survey, the median expects 5.22% inflation in 2026, growth of 1.85% and a Selic of 13.25% at year-end. For 2027, 4.30%, 1.40% and 12.00%. At the start of the year, the medians for 2026 were 4.06%, 1.80% and 12.25%: inflation and interest rate expectations rose, and the growth expectation barely changed. Today, the Selic stands at 13.75%.

Three months of IPCA remain. From January to September, the index accumulated 3.95%. For the year to close at 5.22%, inflation from October to December would have to total 1.22%. Over the past ten years, that quarter ranged from 0.39% (2018) to 3.13% (2020); in 2025, it was 0.60%. This is not a Dado Bruto projection, but the calculation implied by the market's expectation.

To get a sense of scale: in October, expectations for the current year missed by an average of 0.75 point on inflation, 0.53 on GDP and 0.54 on the Selic. For the following year, measured in December, 1.6, 1.6 and 2.5. The further away the year, the larger the possible error.

Limits

The expectations analyzed are the median of all responses, which hides who got it right and who got it wrong: no institution is evaluated individually. The analysis also does not explain why each year was missed. Context numbers, such as the change in the dollar, show what changed during the year but do not prove the cause. With a 25-year sample, one or two extreme years change the mean a lot, which is why the median appears alongside it.

Expectations are from the business day closest to each date. IPCA and GDP are the figures currently published by IBGE. None of the forecasts cited is from Dado Bruto.