How high is the rate of unemployment in Hungary now?

In the three months from November 2022 to January 2023, the unemployment rate in Hungary rose to 4% on a rolling average basis. This is officially the headline unemployment figure. Meanwhile, the model-estimated monthly unemployment rates were 3.8% in Nov, 3.8% in Dec and 3.9% in Jan, respectively, i.e. they have never reached 4.0%.

The 3-month average for December-February already shows an unemployment rate of 4.1%, while the new KSH model estimate for February was 4%. The raw unemployment figure, calculated from the 3-month average data, however, rose to U-rate of 4.4% in the same month. That is, the gap between the estimated and raw data widened to 0.4 percentage points this month.

Last year, the unemployment rate was close to 3%, but from there it rose to 4% and beyond, meaning that the jobless rate went higher in the midst of the economic slowdown and recession. But as you can see, the rates are different. We asked the experts at KSH for details and the reasons for the differences.

Katalin Janák, Head of the Quality of Life Statistics Department of the KSH, told Portfolio that there is no significant difference in the trends between the two, and the three-month averages of the Labour Force Survey and the monthly model estimates show the same picture. Both data have relevance in different respects: the monthly data are more timely, but the three-month averages provide a deeper breakdown, she added.

About the model changes

The source of data is the Labour Force Survey, which collects demographic, education related and labour market data of the population living in private households. The survey involves nearly 38,000 private households per quarter.

Monthly data are estimated using model estimation to mitigate variation from individual monthly samples and may therefore differ from the three-month average data estimated directly from the survey.

From the reference month of January 2023 onwards, estimates are produced using administrative data from the National Tax and Customs Administration (NAV) and the National Employment Service (NFSZ), with a higher degree of precision than before.

The main indicators of labour market participation (number of employed, employment rate, number of unemployed, unemployment rate, number of economically inactive, potential labour reserve) remained quarterly and three-monthly moving average data from the Labour Force Survey and were not affected by the change, the stats office said in a methodological note.

In essence, the model estimation was amended to get a clearer picture of labour market developments. The KSH switched to publishing monthly figures in addition to the three-monthly figures during the coronavirus pandemic, as things were changing rapidly. At that time, only one raw figure was published, then a model estimate was introduced in mid-2021, followed by another from the beginning of 2023.

"Users want data as quickly as possible, as soon as possible, while at the same time it is important that the accuracy of the data is as high as possible. It is easy to see that these two aspects work against each other," Roland Kadlecsik, Head of the Earnings Statistics Department of the Quality of Life Statistics Department at KSH, told Portfolio.

"One of the tasks of official statistics is to find an optimum that takes these two aspects into account and harmonises them as much as possible.

"The solution to this dilemma is usually revision, which, in addition to the introduction of the new model, has also been applied to these statistics when we have retrospectively replaced the results of the monthly estimates with the results of the new model, which provides more accurate data," he added.

As the model the stats office has switched to this year gives a more accurate estimate, we have revised our previously published monthly data back to 2011. This is what we call in statistics major revisions, which are not systematic, not routine, occur infrequently and usually over a longer time series, said Roland Kadlecsik.

He stressed that the main reason for introducing the new model was that it is better able to keep up with changes. In other words, it can describe the trends behind the three-month average in a more meaningful way than the old model. However, it is important to stress that the difference between the old and the new model is not significant.

The old model overestimated employment by 15,000 in August and underestimated it by 31,000 in September. Subsequently, the new model shows a decline in employment due to seasonal effects.

Looking at December (as at the beginning of the year), the two model estimates were essentially identical. So it is not that the two models describe very different employment dynamics, but that the new model describes monthly changes more accurately and sensitively than the old model, explained Kadlecsik.

Overall, however, the difference between the old and the new model is not significant, with the average difference between the two figures being less than 0.5% over the whole revised period. It should also be borne in mind that the three-monthly data are also estimates, with a sampling error of around 30,000 (0.6%) at the 95% confidence level.

The difference in unemployment is not significant either, he said. "Between November and January, the unemployment rate was 4% based on 3-month average data, while the average of the monthly model-estimated data is 3.8%. We also indicated in the estimation methodology that the result of the estimation may differ from the 3-month average data (see box above). It is very important that the average absolute deviation between the two models is less than 0.15%, reaching 0.2% in this 3-month period. But even this is not significant, such deviations are possible in the future," concluded Kadlecsik.

Cover photo: Getty Images