Credibility of Discriminatory Models on the Example of Enterprises from the Lubelskie and Podkarpackie Voivodeships

Theoretical background: The results of the conducted research allowed the classification of early-warning models according to the accuracy of the forecasts received for the last year of the study. Purpose of the article: The aim of the article was verification and prognostic assessment of discriminative models popular among researchers, answer to the question whether the model properly reflects the financial situation of the company. Research methods: The basis of all the methods used in this article was the analysis of existing data and methods of discriminant analysis. Main findings: The selected models properly reflected the financial situation of the 84 enterprises surveyed.


Introduction
In domestic and foreign literature on the subject there are many methods (divided into types) that are used to assess the financial condition of enterprises (financial situation of enterprises) -the terms interchangeably used by the authors.Of the many financial methods, discriminatory models are the most popular tools in the field of early-warning methods.Bankruptcy prediction models (also called "models", "early bankruptcy warning systems") are tools used to assess the economic and financial situation of enterprises, enabling not only forecasting the threat of bankruptcy, but also assessing changes in the condition of the analyzed units and the degree of stability or variability of this condition (Dec, 2009, p. 79).
The purpose of this article concerns verification and prognostic assessment of 10 discriminative models selected for the study.The research sample comprised enterprises from the commercial, production and service industries, originating in the Lubelskie and Podkarpackie voivodeships.Eighty-four enterprises were divided into two groups: 42 bankrupt enterprises and 42 healthy enterprises.For the calculations, the analysis of financial data from the period 2010-2018 was used.Finally, the results obtained and the reliability of the methods used for the study are presented.

Literature review
In the extensive literature on the subject, many researchers attempt to verify early-warning models.Among the available research results, discriminative models are the most popular.The first Polish discriminatory model whose task was the bankruptcy forecast was Mączyńska's model, in which the author used a multiplication model of simplified discriminant analysis to predict the bankruptcy of Polish companies (Mączyńska, 1994).Table 1 presents a summary list of studies conducted in which the authors use the largest number of models and the number of enterprises.Source: (Kitowski, 2017, p. 181).
The analysis of early warning models was carried out based on the collected financial data of enterprises that declared bankruptcy in the years 2010-2018.The enterprises were located in two provinces -Podkarpackie and Lubelskie.
Pobrane z czasopisma Annales H -Oeconomia http://oeconomia.annales.umcs.plData: 14/09/2023 13:39:52 The research sample consisted of enterprises from the commercial, production and service industries.The enterprises were divided into two groups: bankrupt (in poor condition) and healthy (in good condition).Healthy enterprises were selected in a purposeful way, they had a similar business profile in relation to bankrupt enterprises and a similar property and capital structure.Finally, data on 42 entities with poor financial condition -bankrupt from both voivodeships -and the same number of their healthy counterparts was collected.

Results
The prognostic effectiveness of 10 discriminative models was assessed based on the collected financial data over a five-year period.The last year of the survey was the year of bankruptcy by the bankrupt group.The calculations were made adequately for five periods of enterprises included in the healthy group.Finally, attention was focused on the last year of the study.Table 4 contains detailed results obtained for the analysed sample for the last year of the survey.Of the respondents, three models achieved the highest prognostic values, above 80%.Mączyńska and Zawadzki's "G" model turned out to be the best diagnosing model.The Korol model was second in this respect, and the Poznań model came in third.All 10 models had a prognostic value above 50%.Hadasik and Wierzba methods were characterized by the lowest prognostic values.Both models achieved predictive efficacy slightly above 50% -57% and 58%, respectively.As for the effectiveness of forecasts by voivodeships, there were no significant differences in the assessment of individual enterprises from the Podkarpackie and Lubelskie voivodeships.The percentage of accuracy of diagnoses in the assessment of enterprises by voivodeship did not mean significant differences.

Conclusions
The role of discriminant analysis and early warning systems based on it is to make a comprehensive assessment of the company's financial condition and to reveal elements indicating the increasing risk of bankruptcy (Wysocki & Kozera, 2012, p. 169).The results of the conducted research, whose purpose was verification and prognostic assessment of discriminative models popular among researchers for predicting bankruptcy of enterprises from the Lubelskie and Podkarpackie voivodeships confirm the validity of the research.Each of the 10 models used for research obtained prognostic reliability of 57% and more.
None of the discriminant analysis models in the same period had credibility above 90% efficiency.In the authors' opinion, the selected models correctly reflected the financial situation of the 84 enterprises surveyed (the highest prognostic value concerned the "G" model of Mączyńska and Zawadzki, the Korol model and the Poznań model).
Pobrane z czasopisma Annales H -Oeconomia http://oeconomia.annales.umcs.plData: 14/09/2023 13:39:52 In the article, the second degree error was more frequent than the first degree error.However, in a few cases the number of incorrect diagnoses of the first and second degree of the tested models was the same (first degree error: Appenzeller and Szarzec models and Prusak model; second degree error: INE PAN model by Mączyńska and Zawadzki, Poznań model by Hamrol, Hołda model, Maślanka model and Wierzba model).As research shows, the time of creation of a given model does not determine its effectiveness.Therefore, it is difficult to determine the useful life of a particular model.It is similar with the number of indicators used in the studied models, it does not determine the effectiveness of the results.
Based on the review of the literature and the results of the authors' research, it can be concluded that the time in which the model was created does not affect (or clearly does not determine) its efficiency of calculations and thus the reliability of the results obtained.Hence, it is really difficult to determine the usefulness time, use of a specific model for research on bankruptcy of enterprises; similarly, the number of indicators used in the studied models does not prejudge the effectiveness of the results.

Table 1 .
Characteristics of selected studies according to the largest number of discriminatory models used and the number of enterprises surveyed

Table 3 .
Classification of enterprises used for the survey Source: Authors' own study based on collected financial data.

Table 4 .
Classification of early warning models according to the accuracy of forecasts for the last year of the study