Version: v2.0 (effective from 1 February 2026)
The previous version of the methodology document is available here
The aim of the Minerva Institute's research is to map the general mood, awareness of problems, sense of security, political attitudes, and AI interaction experience of the adult Hungarian population, as well as to conduct a feasibility study and demonstration of artificial intelligence as a technology in the field of market and opinion research.
Data is collected by telephone using an AI-based voice assistant. The calls are initiated using randomly generated phone numbers — no existing database or pre-defined target groups are used. Participation is voluntary and restricted to persons aged 18 and over.
Data collection runs simultaneously across 100 call channels, with a target of 1,000 successful interviews.
Final results are weighted to match the age, gender, and place of residence distribution of the adult Hungarian population. Weighting is carried out using a raking (iterative proportional fitting) procedure, which allows the distributions to simultaneously approximate official statistical data (e.g. from the Hungarian Central Statistical Office, KSH) across multiple dimensions. A detailed description of the weighting procedure can be found in the following chapter.
After each interview, data is stored in .sav format and processed using SPSS Statistics software. Open-text responses are clustered and coded using automated text analysis.
Respondents always receive a brief data protection notice at the start of the conversation. Data management is fully compliant with current legislation, with particular regard to GDPR requirements. Participation is voluntary and responses are recorded anonymously. The respondent cannot be identified from their answers.
Conversations are stored for 30 days for quality assurance purposes and are then automatically deleted. Interviews are not listened to by humans; all processing is done exclusively by machine.
The full sample used in Minerva Institute research is weighted using composite variables based on the KSH 2022 census data: the sample is adjusted to match known population distributions by gender, age, settlement type, and educational attainment, thereby ensuring the representativeness of the research results for the Hungarian population.
Hungarian Central Statistical Office (KSH) Census Database
Educational attainment (completed) of the population aged 15 and over, by county and settlement type
Census Database – Hungarian Central Statistical Office
The referenced data source presents the Hungarian population based on the 2022 census, broken down by highest completed educational attainment (10 categories), gender (2 categories), settlement type (4 categories), and 5-year age groups (15 categories).
The Minerva Institute uses the following category system based on the dimensions used by the KSH:
| KSH category | Minerva Institute category |
| Primary school below 8th grade | Basic |
| Primary school, 8th grade | Basic |
| Secondary school without school-leaving exam, with vocational certificate | Secondary |
| School-leaving exam (general) | Secondary |
| School-leaving exam with vocational certificate | Secondary |
| Post-secondary vocational qualification | Secondary |
| Higher education (tertiary) vocational qualification | Tertiary |
| College / Bachelor's degree (BA/BSc) | Tertiary |
| University / Master's degree (MA/MSc) | Tertiary |
| Doctoral degree (PhD or DLA) | Tertiary |
| KSH category | Minerva Institute category |
| Male | Male |
| Female | Female |
| KSH category | Minerva Institute category |
| Budapest | Budapest |
| County-seat city/cities | County-seat city |
| Other city/cities | Other city |
| Village(s) | Village |
| KSH category | Minerva Institute category |
| 15–19 years | 18–39 (young) |
| 20–24 years | 18–39 (young) |
| 25–29 years | 18–39 (young) |
| 30–34 years | 18–39 (young) |
| 35–39 years | 18–39 (young) |
| 40–44 years | 40–60 (middle-aged) |
| 45–49 years | 40–60 (middle-aged) |
| 50–54 years | 40–60 (middle-aged) |
| 55–59 years | 40–60 (middle-aged) |
| 60–64 years | 60+ (elderly) |
| 65–69 years | 60+ (elderly) |
| 70–74 years | 60+ (elderly) |
| 75–79 years | 60+ (elderly) |
| 80–84 years | 60+ (elderly) |
| 85 years and older | 60+ (elderly) |
The first step in establishing the Minerva Institute's weighting ratios is to determine the population size by reduced education category. This step involves data consolidation, whereby — by gender, settlement type, and 5-year age groups — the following are summed:
Three years have passed since the KSH data collection (2022), making it appropriate to correct the education data, as some people who were classified as having basic education in 2022 have since attained secondary education, and some who were classified as having secondary education have since attained tertiary education.
In the correction process:
A similar approach was used for the 15–19, 20–24, and 25–30 age groups for those with secondary and tertiary education, with the corrected population numbers calculated as shown in the example for both genders and all settlement types.
In order for the sample to more accurately reflect the current composition of the Hungarian population, it is appropriate to correct the KSH 2022 census data using mortality indicators for each age group. (www.ksh.gov.hu/stadat_files/nep/hu/nep0011.html)
The KSH publishes annual mortality statistics per 1,000 persons by gender and age group, from which the cumulative mortality rates since the 2022 census period can be derived.
Accordingly, the cumulative mortality rates per 1,000 persons for males are as follows:
| Age group | Deaths per 1,000 |
| 15–19 | 2.73 |
| 20–24 | 2.73 |
| 24–29 | 2.73 |
| 30–34 | 4.68 |
| 35–39 | 7.41 |
| 40–44 | 11.50 |
| 45–49 | 19.39 |
| 50–54 | 36.64 |
| 55–59 | 65.25 |
| 60–64 | 110.45 |
| 65–69 | 163.19 |
| 70–74 | 212.95 |
| 75–79 | 288.01 |
| 80–84 | 420.17 |
| 85+ | 646.80 |
The cumulative mortality rates per 1,000 persons for females are as follows:
| Age group | Deaths per 1,000 |
| 15–19 | 1.21 |
| 20–24 | 1.21 |
| 24–29 | 1.21 |
| 30–34 | 2.36 |
| 35–39 | 4.06 |
| 40–44 | 5.88 |
| 45–49 | 9.57 |
| 50–54 | 17.11 |
| 55–59 | 30.52 |
| 60–64 | 51.34 |
| 65–69 | 80.24 |
| 70–74 | 115.40 |
| 75–79 | 183.02 |
| 80–84 | 307.11 |
| 85+ | 562.75 |
By multiplying the population figures by gender, settlement type, education level, and age group with the above mortality indicators, we obtain the mortality-corrected population figures.
The final step is the consolidation of data by age group, yielding the 3 age categories we use. The simple addition of values within each age group should also be corrected for the time elapsed since the 2022 census, as a respondent who was 39 years old at the time of the 2022 data collection and therefore still belonged to the young (18–39 years) age group would today belong to the middle-aged (40–60 years) group.
After consolidation, we have a data table calculated along 4 dimensions, corrected for both the time elapsed since the 2022 census — accounting for changes in education data and mortality — and reflecting the current Hungarian population. This corrected data table across 4 dimensions forms the basis for the composite category weighting of the sample used in our research.
| Male | Female | Male | Female | Male | Female | Total | ||
| Basic | Basic | Secondary | Secondary | Tertiary | Tertiary | |||
| Budapest | 18–39 | 19,566 | 14,389 | 110,372 | 87,552 | 100,819 | 127,278 | 459,976 |
| Budapest | 40–60 | 21,008 | 17,964 | 131,451 | 115,259 | 107,705 | 138,554 | 531,941 |
| Budapest | 60+ | 16,637 | 46,387 | 74,087 | 118,681 | 52,975 | 74,416 | 383,184 |
| County-seat city | 18–39 | 26,903 | 20,110 | 152,760 | 121,333 | 71,229 | 98,361 | 490,696 |
| County-seat city | 40–60 | 27,408 | 24,073 | 184,090 | 161,247 | 85,962 | 121,601 | 604,381 |
| County-seat city | 60+ | 25,090 | 78,758 | 112,133 | 158,669 | 48,954 | 66,581 | 490,185 |
| Other city | 18–39 | 63,736 | 52,841 | 251,146 | 197,721 | 71,216 | 112,774 | 749,434 |
| Other city | 40–60 | 66,545 | 62,628 | 318,669 | 269,566 | 92,541 | 140,337 | 950,285 |
| Other city | 60+ | 59,100 | 166,988 | 191,860 | 218,151 | 48,023 | 67,862 | 751,983 |
| Village | 18–39 | 85,030 | 73,331 | 244,723 | 191,105 | 46,517 | 81,288 | 721,995 |
| Village | 40–60 | 90,789 | 85,908 | 296,517 | 240,788 | 56,752 | 89,138 | 859,892 |
| Village | 60+ | 72,377 | 187,689 | 176,595 | 161,904 | 28,291 | 38,082 | 664,937 |
| Total | 574,187 | 831,066 | 2,244,403 | 2,041,976 | 810,985 | 1,156,272 | 7,658,889 |
| Male | Female | Male | Female | Male | Female | Total | ||
| Basic | Basic | Secondary | Secondary | Tertiary | Tertiary | |||
| Budapest | 18–39 | 0.26% | 0.19% | 1.44% | 1.14% | 1.32% | 1.66% | 6.01% |
| Budapest | 40–60 | 0.27% | 0.23% | 1.72% | 1.50% | 1.41% | 1.81% | 6.95% |
| Budapest | 60+ | 0.22% | 0.61% | 0.97% | 1.55% | 0.69% | 0.97% | 5.00% |
| County-seat city | 18–39 | 0.35% | 0.26% | 1.99% | 1.58% | 0.93% | 1.28% | 6.41% |
| County-seat city | 40–60 | 0.36% | 0.31% | 2.40% | 2.11% | 1.12% | 1.59% | 7.89% |
| County-seat city | 60+ | 0.33% | 1.03% | 1.46% | 2.07% | 0.64% | 0.87% | 6.40% |
| Other city | 18–39 | 0.83% | 0.69% | 3.28% | 2.58% | 0.93% | 1.47% | 9.79% |
| Other city | 40–60 | 0.87% | 0.82% | 4.16% | 3.52% | 1.21% | 1.83% | 12.41% |
| Other city | 60+ | 0.77% | 2.18% | 2.51% | 2.85% | 0.63% | 0.89% | 9.82% |
| Village | 18–39 | 1.11% | 0.96% | 3.20% | 2.50% | 0.61% | 1.06% | 9.43% |
| Village | 40–60 | 1.19% | 1.12% | 3.87% | 3.14% | 0.74% | 1.16% | 11.23% |
| Village | 60+ | 0.95% | 2.45% | 2.31% | 2.11% | 0.37% | 0.50% | 8.68% |
| Total | 7.50% | 10.85% | 29.30% | 26.66% | 10.59% | 15.10% | 100.00% |
After weighting, indicators calculated from the sample (e.g. party preferences, opinions) better represent the views of the overall population. However, weighting can also introduce statistical uncertainty, making it particularly important to interpret confidence intervals and margins of error when analysing weighted data.
The research includes open-ended questions in which respondents are asked to express their opinions in their own words. Responses are recorded verbatim and subjected to automated analysis after the survey is completed.
Text mining methods and a large language model (LLM) are used to process open-text responses.
The methodology is as follows: the complete response corpus for a given question is subjected to a clustering procedure, which identifies semantically based clusters of responses, i.e. collective categories. Each individual response is then assigned to one of these categories, and the free-text response is replaced in the database with the code for that category.