The information is prepared on the basis of data from the information-analytical system RSF, informative part is represented in the author's edition. All rights belong to the authors, the use or reprinting of materials is permitted only with the prior consent of the authors.

Project titleRobust methods and models for resilient markets and efficient production lines

AffiliationFederal State Budgetary Educational Institution of Higher Education "Saint-Petersburg State University",

Research area 08 - HUMANITIES AND SOCIAL SCIENCES, 08-154 - Finance, credit, money circulation, market infrastructure

KeywordsFinancial and economic markets; insurance markets; robust statistical and econometric inference; dependent and heterogeneous data; heavy-tailed distributions; copula dependence models; econometric and mathematical modeling in finance, economics and actuarial science; multivariate statistical methods for economic analysis; stochastic frontier models; nonparametric estimators of production function; endogenous factors of production; stochastic frontier models with environmental factors

Annotation
Fundamentally this project targets the continued development and application of mathematical methods suitable for modelling complex financial markets and production processes, affected by a large number of dependent and extreme random shocks. Examples of such markets and processes include the market for financial derivatives, including mortgage-backed, art market and a large class of production processes where technical efficiency is determined by the so called environmental variables, e.g. in agriculture and energy sectors. The data available to modern markets and firms -- financial, economic, actuarial -- exhibit highly complex dependence patterns, including extreme dependence at times of stress. This proposal presents novel, robust and reliable alternatives to the existing models and methods to analyse such data, expanding the knowledge base in the field of econometrics and, more generally, in the field of statistical inference as it applies to modern business and economic data. Importantly, the project will relax the unrealistic assumptions made in the traditional models of production and market behaviour. The proposed research program is situated in the intersection of at least four disciplines: economics, finance and actuarial science, mathematics, and statistics. As an outcome, the project will produce a system of econometric models and estimation methods, accompanied by a freely available toolbox of prototype software. A related task is to expand the activities of the world-class research laboratory in theoretical and applied econometrics -- Centre for Econometrics and Business Analytics at St.Petersburg State University. The most profound impact will likely take place in the field of econometrics, that is in the intersection of economics and mathematics, where the proposed research will produce a new generation of models and methods that reflect the realities of today's manufacturing and the sophistication of today's financial and actuarial products and markets. To this end, the project will solidify the established productive international research network consisting of leading scientists who will consistently produce world-class research in the field of statistical and mathematical analysis and modeling of markets and firms using a wide range of available statistical resources, including massive data sets. In this way, the proposal will meet the objectives of the Russian Science Foundation by enhancing international collaboration and expanding Russia's research capability.

Expected results
A key scientific outcome of this research project will be a series of papers published in leading Scopus-cited journals in the field, targeting those journals that have a 5-year impact factor of 2 and above, such as the Journal of Banking and Finance (IF 3.53) and the Journal of Empirical Finance (IF 3.05), as well as a series of presentations at leading international conferences in the field. One practical benefit of this work for the Russian economy is that the novel methods will be used to achieve a better understanding and a smoother operation of Russian economic and financial markets, a higher productivity of Russian firms and industries. Unanticipated extreme movements in economic and financial markets have profound detrimental effects on real economy. The increased uncertainty usually slows down growth and reduces productivity. Similar effects follow when production decisions do not account for important dependencies of productivity on other factors than inputs. The new robust models and estimation methods that have been proposed and will be further developed in the project provide means by which market participants can better assess such risks and minimize their effects. Correspondingly, the project has been informing Russia's policy by building capacity in the analysis, management and prevention of extreme market behaviour, ensuring smooth economic development, crucial for the long term prosperity of the Russian economy. The methodology that has been developed and will be further developed in the project will have direct implications for statistical inference by providing robust and reliable alternatives to the existing methods. One of the main planned outcomes of the extended project is to provide a toolbox of statistical methodologies used in many areas of knowledge, beyond economics, finance and insurance. An appealing property of some of the results that have been developed and will be further developed in this project, especially in robust copula inference and multivariate modelling of financial extremes and pricing, is that they are easily accessible, among others, to graduate and undergraduate students and practitioners. It is planned that some of the results of the study will continue to be integrated in the courses the research team members are currently teaching at respective universities. In addition, the project will involve collaborations with Ph.D. students, and open problems related to the project are being used by them as possible thesis topics. The research network of mid-career researchers that has emerge as a result of this research project will continue expanding and will provide spill-over effects outside the outlined proposal with long term academic benefits. Overall, the project will continue strengthening research and research training in a world-class environment and continue enhancing international collaboration in research.

Annotation of the results obtained in 2023
The following results were achieved in the reporting period. New methods of estimating flexible families of copulas and new methods of their statistical application for analyzing and evaluating the productive potential of firms and the level of technical inefficiency of production have been developed. New nonparametric methods and adaptive algorithms for estimating and testing copulas resistant to incorrect specifications of the functional form of dependence were developed. Obtained a model of clustering of series of average-frequency economic indicators, including models of series smoothing, selection of a specific DTW metric and clustering. A scientific article with the results of this research was prepared and sent to a special issue of Emerging Markets Review. A study of Artionyms in Action: A Machine Learning Approach to Naming Art was conducted (authors A. Altynova, D. Grigoriev, A. Semenov, V. Kolycheva and I. Vasiliev). The recognition of artonyms by means of the architecture of the neural network NIC (neural image captioning), consisting of a convolutional neural network pre-trained on the ImageNet dataset, recognizing the content of images, and the next recurrent neural network LSTM, used to generate text. A portrait of an average museum visitor is constructed and its changes after the pandemic are analyzed. The portrait is built by analyzing the results of a survey of museum visitors on the basis of a specially designed questionnaire. A portrait of an average museum visitor is constructed and its changes after the pandemic are analyzed. The portrait is built by analyzing the results of a survey of museum visitors on the basis of a specially designed questionnaire. The model of phase spline analysis of average frequency series is developed.The model includes the following components: smoothing of series by means of regression splines, projection of series on a plane with its values and values of the first derivative, identification of cycles of the obtained phase portrait. An interpretable model for predicting the quality of optical element processing based on machine learning methods is built. An output variable design approach is developed for the model. 3 simulation models were built for 3 variants of toll collection point (TCP) placement on a toll road: at the exit of the toll road; on the main course of the road; and at the exit of the road before a regulated intersection. Discrete event simulation modeling was used as a transportation micro-simulation methodology. Based on the obtained modeling results, recommendations were given to improve the efficiency of the toll system and traffic light facility, reducing the risks of congestion at the study sites. The model of phase spline analysis of average frequency series is developed.The model includes the following components: smoothing of series by means of regression splines, projection of series on a plane with its values and values of the first derivative, identification of cycles of the obtained phase portrait. An interpretable model for predicting the quality of optical element processing based on machine learning methods is built. An output variable design approach is developed for the model. 3 simulation models were built for 3 variants of toll collection point (TCP) placement on a toll road: at the exit of the toll road; on the main course of the road; and at the exit of the road before a regulated intersection. Discrete event simulation modeling was used as a transportation micro-simulation methodology. Based on the obtained modeling results, recommendations were given to improve the efficiency of the toll system and traffic light facility, reducing the risks of congestion at the study sites. The results were disseminated in three journal articles published in leading journals in the field and in a monograph.

Publications

1. Amsler C., James R., Prokhorov A., Schmidt P. Improving Predictions of Technical Inefficiency Advances in Econometrics, - (year - 2024)

2. Gadasina L., Vysotskiy R., & Lovlea N. Prediction of Processing Optical Elements Results Using Machine Learning Proceedings - 2023 International Russian Automation Conference, RusAutoCon 2023, 2023, pp. 758-762 (year - 2023) https://doi.org/10.1109/RusAutoCon58002.2023.10272862

3. Artem Prokhorov Efficiency and Productivity Analysis Using Copulas in Stochastic Frontier Models Routledge Singapore; Taylor and Francis, - (year - 2024)

4. Talavirya A., Laskin M., Dubgorn A. Application of Simulation Modeling to Assess the Operation of Urban Toll Plazas Simulation Modeling - Recent Advances, New Perspectives, and Applications, стр. 1-39 (year - 2023) https://doi.org/10.5772/intechopen.1002003