Angel-Marchev/marchev-synth-data

Synthesized Economic Agents Dataset

The authors would like to extend their gratitude to the University of National and World Economy and the project NID NI 23/2023/V for funding this research, and for the prime administrative assistance in general.

How to Cite: Marchev, V., Marchev JR, A., Haralampiev, K., Efremov, A., Markov, B., Lyubchev, D., Piryankova, M., Filipov, B., Masarliev, D., & Mitkov, V. (2024). Methodological Approaches for Multidimensional Personal Data Creation. Vanguard Scientific Instruments in Management, 20, 108-131. Retrieved from https://vsim-journal.info/index.php?journal=vsim&page=article&op=view&path[]=544

BibTex Citation:

@article{
Marchev_Marchev JR_Haralampiev_Efremov_Markov_Lyubchev_Piryankova_Filipov_Masarliev_Mitkov_2024, 
title={Methodological Approaches for Multidimensional Personal Data Creation}, 
volume={20}, 
url={https://vsim-journal.info/index.php?journal=vsim&page=article&op=view&path[]=544}, 
journal={Vanguard Scientific Instruments in Management}, 
author={Marchev, Vasil and Marchev JR, Angel and Haralampiev, Kaloyan and Efremov, Alexander and Markov, Boyan and Lyubchev, Dimitar and Piryankova, Milena and Filipov, Bogomil and Masarliev, Daniel and Mitkov, Valentin}, 
year={2024}, 
month={Dec.}, 
pages={108-131}
}

In the era of big data, there is a notable scarcity of datasets containing inherently sensitive information. Such datasets include those falling under regulations like GDPR, the Banking Secrecy Act, European data protection legislation, and others. The current research explores the possibility of filling this gap by generating a multidimensional dataset integrating personal characteristics, demographic features, personal preferences, and more, applicable for conducting research in banking, financial markets, and other economic and financial domains. The nature of the data, its complexity, and legal frameworks necessitate alternative approaches to acquire, simulate, and synthesize the required quantity and quality of information.

The aim is to accumulate a wide range of diverse personal characteristics to help build a comprehensive profile of a statistically significant number of individuals. Financially active individuals and their behavior in the financial-economic context are identified. The dataset thus compiled could be applied in a broad range of economic and social studies.

1. PROCESS ESSENTIALS

1.1. Coverage

1.1.1. Geographical Coverage

The country data is based on the Bulgarian Census 2021 (for demographic data) and the National Statistical Institute, while the banking information is based on public information from the Bulgarian National Bank and the banking system. We use information obtained from the Ministry of Finance and the Ministry of Agriculture. Several independent studies and questionnaires have been conducted on the investment preferences of individuals, as well as external research on personality and temperament based on a previously conducted survey with more than 15,000 international participants (Tipatov, 2009).

1.1.2. Temporal Coverage

The data simulation process involves a one-time generation of data based on a temporary snapshot of the variables under consideration. If the data needs to be updated, it is important to update the distributions.

1.1.3. Demographic Coverage

The synthesized data represents the defined research subject as a Bulgarian individual, a non-professional investor, with limited investment experience, while at the same time having available funds for investment.

1.2 Data source

1.2.1. Primary Data Sources

The data is simulated based on previously collected information and generated distributions. The distributions are derived based on the following approaches:

In the presence of data for forming distributions, an assumption is made that there will be no change in the conditions when using this approach. Namely, the preservation of the earlier distributions for each factor for which we have sufficient data (NSI, 2021, Census 2021).

In the absence of sufficient data – to prepare a relevant distribution, the information that is available for the specific variable is used, such as the average value of the data, minimum, maximum, weighted average, etc. After establishing the available information, a partial simulation is performed, based on the data we know and on a priori information.

Assumptions – In cases where we do not have available information about the distributions of the variables under consideration plausible assumption is performed. A simulation is performed based on a priori knowledge about the variables under consideration, an analysis of the group to which the specific indicator belongs, etc.

1.2.2. Data Providers

It is necessary to derive distributions based on publicly available information with a high accuracy level. The following sources were used:

National Statistical Institute (NSI) – the primary state agency responsible for collecting and disseminating statistical data regarding Bulgaria's population, economy, and environment.

Census 2021 – provides detailed demographic information about Bulgaria's population. It included data on population size, distribution, age structure, education levels, and socio-economic characteristics.

Bulgarian National Bank (BNB) – central bank of Bulgaria, overseeing monetary policy, financial stability, and the banking system. It provides crucial data on monetary aggregates, interest rates, exchange rates, and banking statistics.

Banking System in Bulgaria – provide financial services to individuals and businesses while generating a wide set of information on lending practices, deposits, and financial transactions.

Financial Supervision Commission (FSC) – responsible for regulating and supervising the non-banking financial sector in Bulgaria. This includes insurance companies, pension funds, and investment firms.

Ministry of Finance – oversees the country's fiscal policy, public finance management, and budgetary processes.

Ministry of Agriculture and Food – focuses on agricultural policies and rural development in Bulgaria. It collects data on agricultural production, land use, crop yields, and livestock statistics.

A priori information and assumptions – refer to knowledge or assumptions made based on knowledge, experience, and preliminary data about the events under consideration and the environment in which they develop.

1.3 Methodology

1.3.1. Data Collection Methods

Synthetic data refers to information that does not correspond to actual records but is generated algorithmically. This type of data is created through statistical models rather than being collected from real-world observations. Numerous methodologies exist for producing multidimensional synthetic datasets, and many scholars in the field of artificial intelligence have addressed this topic.

From a methodological perspective, various scenarios underscore the need for a structured approach to generating multidimensional datasets. This section examines key situations where data generation is an essential component of the information analysis process, organized according to methodological principles.

Need for Requisite Variety

This concept highlights the necessity of having a diverse and comprehensive array of data inputs to enhance the effectiveness and accuracy of data analysis and modeling. A critical sub-step in this context is feature engineering.

Random Missing Data

In instances where data is missing at random, the process includes specific sub-steps such as Missing at Random (MAR) data imputation.

Missing Data for a Class

When output data is absent for a particular class, it becomes necessary to incorporate an additional module into the model. This module serves to balance the dataset.

Data generation

In certain situations, the unique characteristics of the data, along with its complexity and legal considerations, can hinder the ability to secure models with the requisite quantity and quality of information.

1.3.2. Data Synthesis

The process of synthesizing a multidimensional array of synthetic data has several main phases: variable selection, distribution analysis, business logic extraction, and application of the data generated. The last phase is related to the validation of the newly obtained set of synthetic data.

Variable Selection

Identifying the key individual characteristics of financial service users requires a thorough analysis of various subsets of distinct features. Each carefully curated group of variables enriches our understanding, allowing for a more comprehensive and accurate profile of active users.

Demographic Characteristics

Understanding demographic characteristics is vital, as it sheds light on the primary factors that influence financial behavior across society.

Individual Characteristics

Individual characteristics, influenced by both innate and acquired qualities, play a significant role in shaping how individuals interact with the environment.

Socio-Economic Status

Personal characteristics serve as a window into an individual’s socioeconomic status.

Banking and Financial Characteristics

Banking and financial characteristics under consideration reveal individuals' behaviors within financial services markets.

1.4. Data Processing

Each statistical distribution is defined by specific parameters that describe its characteristics, including shape, central tendency, variability, and skewness.

N	Factor	Code	Variable type	Possible values	Derivation
1	Gender	sex	Nominal	M; F	Simulation
2	Age – completed years	age	Continuous	20 - 85	Correlation
3	Level of education	lv_educ	Ordinal	Incomplete; Primary; Basic; Secondary; Higher	Simulation
4	Employment status	empl_stat	Nominal	Employers; Self-employed; Employed in the private sector; Employed in the public sector; Unpaid family workers; Unemployed	Simulation
5	Marital status	marit_stat	Nominal	Single; Married; Divorced; Widowed	Simulation
6	Number of household members	house_memb	Interval	1; 2; 3; 4; 5; 6; 7+	Simulation
7	Number of children under 18 years	chil_u_18_y	Interval	No children under 18; One child under 18; Two children under 18; Three children under 18; Four children under 18; Five children under 18; Six or more children under 18	Simulation
8	Nationality	nation	Nominal	Bulgaria; EU; Other	Simulation
9	Religion	religion	Nominal	Protestant; Catholic; Orthodox; Muslim; Other; No religion; I do not identify myself	Simulation
10	Profession – Industry	prof_ind	Nominal	Agriculture, forestry, and fisheries; Mining and processing industry; Utilities (electricity distribution and water supply); Construction; Trade, automobile, and motorcycle repair; Transportation, warehousing, and mail; Hospitality and restaurant services; Creation and distribution of information and creative products; Telecommunications; Financial and administrative activities; Public administration; Education and research; Human health and social work; Other activities	Simulation
11	Professional status	prof_stat	Nominal	Management contract; Employment contract; Civil contract; Self-employed; Unemployed; Pensioner	Simulation
12	Number of owned apartments/houses	count_house	Interval	0; 1; 2+	Simulation
13	Land ownership	own_field	Binary	YES/NO	Simulation
14	Cars per household	num_car_house	Interval	0; 1; 2; 3+	Simulation
15	Education	edu	Nominal	Educational Sciences; Humanities; Social, Economic, and Legal Sciences; Natural Sciences, Mathematics, and Informatics; Technical Sciences; Agricultural Sciences and Veterinary Medicine; Health and Sports; Arts; Security and Defense	Simulation
16	Temperament	temp	Nominal	Choleric; Phlegmatic; Sanguine; Melancholic	Simulation
17	Individual risk preference	ind_risk	Continuous	0 - 1	Correlation
18	Previous investment experience in years	invest_exp	Ordinal	0; 1-5; 6-10; 11-15; 16-25	Simulation
19	Investment experience with shares	shares	Binary	YES/NO	Simulation
20	Investment experience with bonds	corp_oblig	Binary	YES/NO	Simulation
21	Investment experience with others	oth	Binary	YES/NO	Simulation
22	Investment experience with investment funds	inv_fund	Binary	YES/NO	Simulation
23	Investment experience with currencies	cash	Binary	YES/NO	Simulation
24	Investment experience with cryptocurrencies	crypto	Binary	YES/NO	Simulation
25	Investment experience with government securities	gov_bond	Binary	YES/NO	Simulation
26	Investment experience with bank deposits	deposits	Binary	YES/NO	Simulation
27	Income	income	Ordinal	Up to 6121; Up to 12001; Up to 27601; Up to 43201; Up to 58801; Up to 74401; Over 90001+	Correlation
28	Personal expenses	pers_exp	Ordinal	up to 4500; up to 5000; up to 5500; up to 6000	Correlation
29	Housing costs	house_exp	Ordinal	up to 500; up to 1500; up to 3000; up to 4000	Correlation
30	Taxes and insurance	taxes	Ordinal	up to 500; up to 1000; up to 2000; up to 2500	Correlation
31	Transport and communications	transp_telecom	Ordinal	up to 500; up to 1000; up to 1500; up to 2500	Correlation
32	Leisure and hobby	hobby	Ordinal	0; up to 1500; up to 2000; up to 3000	Correlation
33	Preferred method of banking	banking	Nominal	Online/Offline	Simulation
34	The average number of bank transactions	bk_oprat	Ordinal	Up to 7; From 8 to 10; From 11 to 13; From 14 to 18; From 19 to more	Simulation
35	Debit card	bk_dc	Interval	Under one; One; Two; Three	Simulation
36	Bank account	bk_acc	Binary	YES/NO	Simulation

Table 1: Data dictionary - full list of variables

To estimate these parameters from a given sample, two statistical techniques are commonly employed: The method of moments and the Generalized method of moments (GMM).

1.4.1. Method of Moments

In the method of moments, the sample moments are matched to the theoretical moments of the distribution. This approach involves solving a set of equations to derive the distribution's parameters.

1.4.2. Generalized Method of Moments (GMM)

The GMM extends the method of moments by offering greater flexibility in selecting moment conditions. This technique is particularly useful when there are more moment conditions than parameters or when the moment conditions cannot be solved directly.

1.5. Accessibility

The specificity, complexity, and regulatory framework surrounding the data present significant challenges in obtaining the necessary quantity and quality of information. The data needed is governed by European regulations such as the GDPR and the Bank Secrecy Act, among others.

The alternative strategy for acquiring the required data. The approach involves generating a multivariate dataset that incorporates a variety of demographic, personal, individual, and banking variables.

1.6. Data Format

The possibilities offered by our model are as follows: Excel, CSV, Pandas, Croissanr, Polars and Parquet.

1.7. Quality Assurance

Business Logic in Data Generation

The business logic applied in the data generation process is defined by identifying potential interdependent factors, their constraints, and the possible and impossible combinations of these factors. When combining distributions, there is a risk of producing unattainable or highly improbable values.

Phases: Selection of Potential Interdependent Factors;

Establishing Possible and Impossible Combinations;

Elimination of Impossible Combinations;

id	Independent feature	Independent feature value	Dependent feature	Dependent feature value filter	Note
1	Marital status	Married	Number of household members	>2	The number of household members in family households is more likely to be greater than 2
2	Profession – Industry	Financial and administrative activities	Bank account	>0	They are more likely to own a bank account
3	Age – completed years	<25	Previous investment experience in years	0	Under 24s are less likely to have investment experience. Between 35-44 and 45-54 are more likely to have extensive investment experience
4	Age – completed years	<21	Level of education	<Higher	Under-21s are less likely to have a university degree
5	Age – completed years	<25	Number of children under 18 years	<2	Given the defined demographic coverage, from 20-24, it is less likely to have more than 1 child under 18
6	Previous investment experience in years	>0	Investment experience with bank deposits	Y	They are more likely to own a bank account
7	Investment in stocks	Y	Previous investment experience in years	>0	If investment in stocks = yes, then previous investment experience in years is >0.
8	Investment in bonds	Y	Previous investment experience in years	>0	If investment in bonds = yes, then previous investment experience in years is >0.
9	Other investments	Y	Previous investment experience in years	>0	If investment in other investments = yes, then previous investment experience in years is >0.
10	Investment in a fund	Y	Previous investment experience in years	>0	If investment in funds = yes, then previous investment experience in years is >0.
11	Currency investments	Y	Previous investment experience in years	>0	If investment in currency = yes, then previous investment experience in years is >0.
12	Investing in cryptocurrencies	Y	Previous investment experience in years	>0	If investment in cryptocurrencies = yes, then previous investment experience in years is >0.
13	Investment in government securities	Y	Previous investment experience in years	>0	If investment in government securities = yes, then previous investment experience in years is >0.
14	Age – completed years	<25	Bank account	N	Under 24s are less likely to have a checking account
15	Age – completed years	<18	Bank account	N	Under 18 is not possible to have a current account
16	Level of education	Higher	Income	>27601	A higher level of education implies earnings in the upper range
17	Number of children under 18 years	>1	Number of household members	>3	The number of household members is directly dependent on the number of children under 18 ages
18	Income	>27601	Taxes and insurance	>2500	Earnings in the upper range correspond to higher taxes and insurance

Table 2: Sample of business logic

1.8. Reliability

Validation process. Crucial step in the data generation process.

Data Analysis. Examination of the synthesized data.

Data Validation. Verifying that the generated dataset aligns with the original statistical distributions.

Adjacent Frequencies. A smooth transition between these values is vital for model validation, as it helps avoid abrupt fluctuations that could indicate issues in the simulation.

Quality Assessment. Evaluate the quality of the information obtained.

1.9. Accuracy

The Kolmogorov-Smirnov (K-S) test is employed as the primary method for data validation.

One-Sample K-S Test

The one-sample K-S test compares the ECDF of a sample with the cumulative distribution function (CDF) of a theoretical distribution. The ECDF represents the proportion of observations in a sample that are less than or equal to a certain value, while the CDF indicates the theoretical probability of obtaining a random observation from that distribution that is also less than or equal to that value.

Two-Sample K-S Test

The two-sample K-S test evaluates whether there is a significant correspondence between two univariate probability distributions. The test statistic D for this test is defined as the maximum absolute difference between the two ECDFs.

Hypotheses

In both the one-sample and two-sample K-S tests, the null hypothesis (H0) posits that the sample(s) conform to the specified distribution (for one sample) or that both samples originate from the same distribution (for two samples).

1.10. Update Frequency

The data should be obtained once. In case of a change in the general conditions for the main groups of variables considered, a re-generation of the data set can be envisaged.

1.11. Contact Information

Corresponding author – Vasil Marchev, vmarchev@unwe.bg

2. METADATA FOR STATISTICAL FEAURE

Concerning the approach considered for simulating a multidimensional array of synthetic data, it is necessary to prepare a detailed description of each of the considered characteristics. The metadata provides essential context and documentation for statistical data. It encompasses structured information that describes the data, its processes, and methodologies, which aids in understanding, interpreting, and utilizing statistical information effectively. A complete list of the detailed variables contained in the generated dataset is available in Table 3.

Feature Name	Description	Calc. method Formula	Calculation Method Data Sources	Unit of Measure	Relevance	Sampling Error	Non-sampling Error	Geo.Disaggregation	Temporal Disaggregation	Comparability – Time	Comparability Regions
Sex/Gender	Shows gender identity	Synthesized variable*	The distribution is derived from the Census 2021 in Bulgaria	Nominal: M/F	The aim is to set up possible correlations between sex/gender & other individual characteristics.	Official data – Census 2021	Potential errors include mis recording	Data for Bulgaria	This data is static and reflects the Sex/Gender at the time of the 2021 Census.	Not applicable	Not applicable
Age	Stands for the number of years.	Synthesized variable	The distribution is derived from the last national Census in Bulgaria conducted in 2021	Continuous: 20 - 85	Age is a fundamental demographic factor essential for analyzing various social dynamics.	Official data – Census 2021	Potential errors include mis recording or incorrect date of birth in administrative records.	Data for Bulgaria	The data is static and reflects the population's age as of the census date (2021)	Not applicable	Not applicable
Level of Education	Completed level of education.	Synthesized variable	The distribution is derived from the Census 2021 in Bulgaria	Ordinal: -Incomplete primary -Primary school -Secondary school -College degree -University degree	Education level is a key demographic characteristic used to analyze individual and community outcomes.	Official data – Census 2021.	Potential errors could stem from incorrect self-reporting or classification during data collection.	Data for Bulgaria	This data is static and reflects the education levels as reported during the 2021 Census.	Not applicable	Not applicable
Employment Status	Indicates the current labor force participation of an individual.	Synthesized variable	The distribution is derived from the Census 2021 in Bulgaria & labor force survey	Nominal: -Employers -Self-employed -Employees in private enterprises -Employees in public enterprises -Unpaid family workers -Unemployed	Employment status is a critical demographic characteristic used to evaluate labor market dynamics.	Official data – Census 2021.	Errors may arise from misclassification or non-response.	Data for Bulgaria	This data is static and reflects the employment status during the reference period of the 2021 Census.	Not applicable	Not applicable
Marital Status	Stands for an individual's legal relationship status.	Synthesized variable	The distribution is derived from Census 2021 Bulgaria	Nominal: -Single -Married -Divorced -Widower	Demographic characteristics for understanding household composition, and social dynamics.	Official data – Census 2021.	Potential errors may arise from misreporting.	Data for Bulgaria	This data is static and reflects marital status at the time of the 2021 Census.	Not applicable	Not applicable
Number of Household Members	Stands for the total number of individuals residing in a household.	Synthesized variable	The distribution is derived from the Census 2021 in Bulgaria.	Interval: 1; 2; 3; 4; 5+	Used to analyze living arrangements, household size trends, & socioeconomic forecasting	Official data – Census 2021.	Potential errors include misreporting household composition.	Data for Bulgaria	This data is static and reflects the number of household members at the time of the 2021 Census.	Not applicable	Not applicable
Number of Children Under 18	Stands for the total number of individuals below 18 years.	Synthesized variable	The distribution is derived from the Census 2021 in Bulgaria, with data collected from household questionnaires.	Interval: 1; 2; 3; 4+	The number of children under 18 assesses dependency ratios, education structure, and understanding of family structures.	Official data – Census 2021.	Potential errors include misclassification of age or omission of household members.	Data for Bulgaria	This data is static and reflects the number of children under 18 at the time of the 2021 Census.	Not applicable	Not applicable
Nationality	Indicates the legal or self-identified national affiliation of an individual.	Synthesized variable	The distribution is derived from the Census 2021 in Bulgaria, with data collected from household questionnaires.	Nominal** From Census 2021	A key demographic characteristic for analyzing population diversity, cultural composition, and community integration.	Official data – Census 2021.	Potential errors include reluctance to show, or data entry mistakes.	Data for Bulgaria	This data is static and reflects the self-identified nationality at the time of the 2021 Census.	Not applicable	Not applicable
Religion	Stands for an individual’s religious affiliation, belief system, or self-identified lack thereof.	Synthesized variable	The distribution is derived from the Census 2021 in Bulgaria	Nominal** From Census 2021	Demographic factor for understanding cultural diversity, social dynamics & its influence on traditions, and community engagement.	Official data – Census 2021.	Potential errors include reluctance to disclose, or data entry mistakes.	Data for Bulgaria	This data is static and reflects religious affiliation as self-identified during the 2021 Census.	Not applicable	Not applicable
Profession/Industry	Stands for the type of occupation or industry in which an individual is employed.	Synthesized variable	The distribution is based on information from the National Statistical Institute.	Nominal**: From NSI	The profession/industry - important demographic factor for understanding employment trends, economic structure, and the distribution of labor across various sectors.	Official data – NSI.	Errors may occur if individuals provide inaccurate responses.	Data for Bulgaria	This data is static and reflects the profession/industry status during the reference period of the 2021 Census.	Not applicable	Not applicable
Professional status	Stands for the type of employment of an individual, categorized based on their role in the labor market.	Synthesized variable	The distribution is based on information from the National Statistical Institute.	Nominal**: From Infostat	Professional status provides information about socio-economic position, its role in the labor market, employment trends & economic inequalities.	Official data – NSI	Errors may occur due to incorrect completion of surveys or errors in data entry or classification.	Data for Bulgaria	These data are static and reflect socioeconomic status during the reference period of the 2021 Census.	Not applicable	Not applicable
Apartment/house numbers	Stands for the number of residential units (apartments or houses) owned by a household.	Synthesized variable	The distribution is based on the 2021 Census in Bulgaria, administrative and statistical reports.	Interval: 0; 1; 2+	Provides information on access to housing and conditions. Helps analyze the distribution of housing resources and living standards in different social groups.	Official data – Census 2021	Errors may occur if individuals provide inaccurate responses.	Data for Bulgaria	These data are static and reflect the number of apartments/houses during the 2021 census reference period.	Not applicable	Not applicable
Plots of Land	Stands for agricultural land owned by an individual highlighting ownership percentage.	Synthesized variable	The distribution is derived from the land registry & data from the Ministry of Agriculture & Assumptions.	Binary: Yes/No	Provides information on land access, ownership, and the distribution of agricultural resources across the regions and social groups.	Assumption discrepancies are possible	Errors can occur due to registration errors, inaccuracies in data entry, or missing information.	Data for Bulgaria	These data are static and reflect the number of land plots during the 2021 census reference period.	Not applicable	Not applicable
Household car	Stands for the number of cars owned by a household.	Synthesized variable	The distribution is derived from the 2021 Census in Bulgaria.	Interval: 0; 1; 2; 3+	The number of cars helps analyze mobility and living conditions. Also revealing socio-economic differences between households.	Official data – Census 2021	Error includes inaccuracies in self-reporting, misunderstanding of questions, or missing data.	Data for Bulgaria	These data are static and reflect the household car during the 2021 census reference period	Not applicable	Not applicable
Education	Education shows the distribution of individuals across different fields of study.	Synthesized variable	The distribution is based on information from the National Statistical Institute.	Nominal**: From Infostat	Reveals trends in the educational structure of the population, highlighting differences in access to educational resources and opportunities for professional development.	Official data – NSI	Misreporting educational levels, non-response bias, data processing mistakes, and inaccuracies in classifying education levels	Data for Bulgaria	These data static and reflect the Education during the 2021 census reference period.	Not applicable	Not applicable
Temperament	Temperament reflects the distribution of individuals across different personality traits.	Synthesized variable	The distribution is derived from an international survey with more than 15k respondents.	Nominal: Choleric Phlegmatic Sanguine Melancholic	Provides information about personality traits and behavioral patterns, highlighting their impact on social interactions.	Minimal possibility in the data from the study	Errors in temperament may include biases in self-assessment, or subjectivity.	Data for Bulgaria	Periodic data updates are needed over a relatively long period (>5 years)	Not applicable	Not applicable
Individual risk	Reflects the distribution of individuals based on their willingness to take investment risks.	Synthesized variable	The distribution is derived from an internal survey with more than 900 respondents.	Continuous: 0 - 1	Individual risk preferences reflect decision-making under uncertainty, highlighting individuals' behavior in financial markets.	Safe environment. Difficulties in assessing behavior in a real situation.	May include inaccurate self-reporting. Safe environment. Difficulties in assessing behavior in a real situation.	Data for Bulgaria	Ongoing research. Stable results. No sharp fluctuations are expected.	Not applicable	Not applicable
Investment exp	Previous investment experience, measured in years, reflects an individual’s history in investment, and financial decision-making.	Synthesized variable	A plausible assumption and a priori simulation	Ordinal*: 0; 1-5; 6-10; 11-15; 16-25 *Interval variable converted into Ordinal	Reflects decision-making in investments, highlighting financial behavior and its impact on economic choices.	Minimal in the data from the study	Errors could include misinterpretation of financial terms, subjective reporting, or inaccurate self-assessment.	Data for Bulgaria	Periodic data updates are needed over a relatively long period (>5 years)	Not applicable	Not applicable
Shares	Shows the distribution of individuals who invest in shares.	Synthesized variable	The distribution is derived from publicly available data from the BNB.	Binary: Yes/No	The shares segment reflects the presence or absence of investments in shares, highlighting the financial behavior of investors.	Official data from BNB	Errors for the shares segment may occur due to inaccurate self-reporting.	Data for Bulgaria	Periodic data updates are needed. No more often than once a year.	Not applicable	Not applicable
Obligations	Shows the distribution of individuals who invest in Obligations.	Synthesized variable	The distribution is derived from publicly available data from the BNB.	Binary: Yes/No	Reflects the investments in obligations, highlighting the financial behavior of investors	Official data from BNB	Potential inaccuracies may stem from misreporting or incomplete representation.	Data for Bulgaria	Periodic data updates are needed. No more often than once a year.	Not applicable	Not applicable
Others	Shows the distribution of individuals who invest in other investment instruments	Synthesized variable	The distribution is derived from publicly available data from the BNB.	Binary: Yes/No	The other investments reflect the presence or absence of investments in other investment instruments.	Official data from BNB	Potential inaccuracies may stem from misreporting or incomplete representation.	Data for Bulgaria	Periodic data updates are needed. No more often than once a year.	Not applicable	Not applicable
Investment funds	It shows the distribution of investors who have investment experience with currencies	Synthesized variable	The distribution is derived from publicly available data from the BNB.	Binary: Yes/No	Provides information for assessing individual investment strategies, wealth accumulation, and financial risk exposure	Official data from BNB	Inaccuracies can arise from incorrect classification.	Data for Bulgaria	Periodic data updates are needed. No more often than once a year.	Not applicable	Not applicable
Cash	It shows the distribution of investors who have investment experience with currencies.	Synthesized variable	The distribution is derived from publicly available data from the BNB.	Binary: Yes/No	Investing in currency provides information about portfolio diversification & knowledge of the forex markets.	Official data from BNB	Inaccuracies may occur from the misclassification of currency investors	Data for Bulgaria	Periodic data updates are needed. No more often than once a year.	Not applicable	Not applicable
Cryptocurrency	Provides information for investors who have experience with cryptocurrencies.	Synthesized variable	The distribution is derived from publicly available data from the BNB.	Binary: Yes/No	Indicator of an individual’s involvement in the digital asset market. It also reflects broader trends in the adoption of decentralized finance.	Official data from BNB	Inaccuracies may arise from misreporting, lack of visibility into private cryptocurrency wallets.	Data for Bulgaria	Periodic data updates are needed. No more often than once a year.	Not applicable	Not applicable
Gov bonds	Provides information on whether the investors under consideration have experience with investments in government bonds.	Synthesized variable	The distribution is derived from publicly available data from the BNB.	Binary: Yes/No	Indicator of individual investment behavior in low-risk, stable financial instruments. They provide insight into financial strategies and trust in government bonds.	Official data from BNB	Inaccuracies may occur due to misreporting.	Data for Bulgaria	Periodic data updates are needed. No more often than once a year.	Not applicable	Not applicable
Deposits	Provides information on whether the investors under consideration have experience with investments in bank deposits.	Synthesized variable	The distribution is derived from publicly available data from the BNB.	Binary: Yes/No	Provides insight into the financial habits of individuals, the penetration of banking products, and consumer trust in the bank system.	Official data from BNB	Inaccuracies could result from misreporting.	Data for Bulgaria	Periodic data updates are needed. No more often than once a year.	Not applicable	Not applicable
Income	Stands for the distribution of income across different income brackets within a population. The data provides information about the corresponding percentage of individuals falling within each one.	Synthesized variable	The distribution of personal income is based on data collected through a survey of 491 people.	Ordinal*: up to 19 200 19 201 - 27 600 27 601 - 54 000 54 001 - 82 800 from 82 801 *Interval variable converted into Ordinal	Income distribution is critical for understanding economic inequality, social stratification, and wealth concentration within a population.	Possible discrepancies in data if the survey sample does not fully stand for the population.	Distortion may occur if data is filled in incorrectly	Data for Bulgaria	Dynamic variable. Data updates are needed once per year.	Not applicable	Not applicable
Personal exp	Stands for the total personal expenses of an individual. This data helps to understand the spending behavior of individuals across different income groups.	Synthesized variable	The distribution of expenses is derived from a priori knowledge and assumptions about the types of household expenses.	Ordinal*: Group 1 Group 2 Group 3 Group 4 *Interval variable converted into Ordinal	Indicator for assessing financial health, economic behavior, and consumption patterns across various demographics. It helps to find potential areas for improvement in savings behavior.	Possible discrepancies in assumptions if the environment changes. Example – inflation.	Potential issues include inconsistent classifications of expenditures or biases in categorization.	Data for Bulgaria	Dynamic variable. Data updates are needed once per year.	Not applicable	Not applicable
House exp	It refers to the total expenditure incurred by an individual on housing-related expenses, including rent, mortgage payments, utilities, and maintenance. These expenses are fundamental to understanding financial stability & behavioral patterns.	Synthesized variable	The distribution of expenses is derived from a priori knowledge and assumptions about the types of household expenses.	Ordinal*: Group 1 Group 2 Group 3 Group 4 *Interval variable converted into Ordinal	Measure of financial well-being, assessing the affordability of housing in different economic contexts. It helps find how much of a household's income is dedicated to housing.	Possible discrepancies in assumptions if the environment changes. Example - inflation.	Inconsistencies in the definition or categorization may occur.	Data for Bulgaria	Dynamic variable. Data updates are needed once per year.	Not applicable	Not applicable
Taxes	Stands for the total taxes and social security contributions paid by an individual, including income taxes, social security, pension contributions, health insurance, and other mandatory payments.	Synthesized variable	The distribution of expenses is derived from a priori knowledge and assumptions about the types of household expenses.	Ordinal*: Group 1 Group 2 Group 3 Group 4 *Interval variable converted into Ordinal	Tax and social security contributions are important for assessing individual financial obligations and understanding the impact of taxation on disposable income.	Possible discrepancies in assumptions if the environment changes. Example - inflation.	Inconsistencies in the definition or categorization may occur.	Data for Bulgaria	Dynamic variable. Data updates are needed once per year.	Not applicable	Not applicable
Transportation & telecom	Stands for the total expenditure an individual spends on transportation (e.g., public transport, car expenses, taxis) and communication services (e.g., mobile phone bills, internet, postal services).	Synthesized variable	The distribution of expenses is derived from a priori knowledge and assumptions about the types of household expenses.	Ordinal*: Group 1 Group 2 Group 3 Group 4 *Interval variable converted into Ordinal	Essential to understanding individuals' mobility patterns and their access to communication. These costs highlight differences in financial capabilities between the groups.	Possible discrepancies in assumptions if the environment changes. Example - inflation.	Inconsistencies in the definition or categorization may occur.	Data for Bulgaria	Dynamic variable. Data updates are needed once per year.	Not applicable	Not applicable
Hobby exp	Stands for the total expenditure an individual allocates towards leisure activities, hobbies, and entertainment. This feature provides insight into an individual's lifestyle, and priorities.	Synthesized variable	The distribution of expenses is derived from a priori knowledge and assumptions about the types of household expenses.	Ordinal*: Group 1 Group 2 Group 3 Group 4 *Interval variable converted into Ordinal	Hobby expenses are important for understanding an individual's discretionary income and lifestyle preferences. An indicator of economic well-being.	Possible discrepancies in assumptions if the environment changes. Example - inflation.	Inconsistencies in the definition or categorization may occur.	Data for Bulgaria	Dynamic variable. Data updates are needed once per year.	Not applicable	Not applicable
Preferred method of banking	Stands for the preferred mode of banking for individuals, whether they prefer online banking or onside banking. This feature provides insights into digital adoption trends and regional or demographic differences in banking behavior.	Synthesized variable	The distribution is derived from NSI data.	Nominal: Online/Offline	This feature is crucial for understanding consumer behavior and guiding decisions on resource allocation, as well as describing the level of trust in digital payment systems.	Official data from NSI	Discrepancies may occur if there are individuals with regular banking both online and offline.	Data for Bulgaria	Data updates are needed once per year.	Not applicable	Not applicable
Avg num banking	Provides information about the active banking user by analyzing the average number of banking transactions performed by an individual in a month. This includes deposits, withdrawals, transfers, bill payments, etc.	Synthesized variable	A priori simulated distribution	Ordinal: up to 10 11 - 14 15 - 20 21 - 26 from 27	The frequency of banking transactions is critical for financial institutions to evaluate the activity of the customers. This data can also assist in determining the most commonly used bank services, and in the customer segmentation process.	Possible discrepancies in assumptions. Possibility of bias in the sample (focusing only on a specific group of customers).	Potential errors include incorrect categorization of transactions or missed transactions that occurred on platforms outside the bank's recorded systems.	Data for Bulgaria	Data updates are needed once per year.	Not applicable	Not applicable
Debit card	It stands for the percentage of people who have one or more debit cards. The function provides information about the penetration of banking services among the population.	Synthesized variable	A priori simulated distribution	Interval: 0; 1; 2; 3	The number of debit cards owned is important for understanding customer behavior. Customers who own multiple debit cards may have different banking needs, such as separate cards for personal and business use, or for different spending categories.	Possible discrepancies in assumptions.	Potential errors include misreporting or lack of clarity.	Data for Bulgaria	Data updates are needed every three/five year.	Not applicable	Not applicable
Bank acc	Stands for the percentage of individuals who own a bank account in Bulgarian Lev (BGN). This feature helps to understand the penetration of basic banking services across different customer segments, particularly about the presence of an account.	Synthesized variable	The distribution is derived from publicly available information about the banking system.	Binary: Yes/No	Ownership of a bank account is the most important indicator of the economically active client. A bank account is a fundamental tool for managing finances and engaging with the broader economy, making this feature critical for understanding financial habits.	Possible discrepancies in data if the survey sample does not fully stand for the population.	Misunderstanding the types of accounts or inaccurately reporting ownership.	Data for Bulgaria	Data updates are needed every three/five year.	Not applicable	Not applicable

Table 3: Full list with described variables

* Synthesized through a simulation method based on distributions and business logic

** Nominal variables with detailed possible values are presented in Table 4

Feature Name	Type	Data source	Possible values
Gender	Nominal	From Census 2021	- M - F
Employment status	Nominal	From Census 2021 & labor force survey	-Employers -Self-employed -Employees in private enterprises -Employees in public enterprises -Unpaid family workers -Unemployed
Marital status	Nominal	From Census 2021	- Single - Married - Divorced - Widower
Nationality	Nominal	From Census 2021	- Bulgarian - European Union - Other
Religion	Nominal	From Census 2021	- Orthodox - Protestant - Catholic - Muslim - Other - No religion - I don't want to answer
Profession/Industry	Nominal	From NSI	- Agriculture, forestry and fishing - Mining, quarrying & Manufacturing - Electricity, gas, steam and air conditioning supply. Water supply, sewerage, waste management and remediation activities - Construction - Wholesale and retail trade; repair of motor vehicles and motorcycles - Transportation and storage - Accommodation and food service activities - Information and communication - Financial and insurance activities. Real estate activities - Education, professional, scientific and technical activities. - Administrative and support service activities. Public administration and defense; compulsory social security - Human health and social work activities - Arts, entertainment and recreation. Other service activities
Professional status	Nominal	From NSI	- Management contract - Employment contract - Civil contract - Self-employed person - Unemployed - Pensioner
Owner of a house	Nominal	From Census 2021 & NSI	- Yes - No
Education	Nominal	From NSI	- Educational Sciences - Humanities - Social, Economic, and Legal Sciences - Natural Sciences, Mathematics, and Informatics - Technical Sciences - Agricultural Sciences and Veterinary Medicine - Health and Sports - Arts - Security and Defense
Temperament	Nominal	International survey (Tipatov, 2009)	- Choleric - Phlegmatic - Sanguine - Melancholic
Preferred method of banking	Nominal	From NSI	- Online - Offline

Table: 4 Nominal variables with detailed possible values