Forecasting the Profitability of the Textile Sector in Emerging European Countries Using Artificial Neural Networks

The textile and apparel industry is a genuine driving force of the global economy, representing “a highly globalized industry” [1], with a value amounting to USD 1,695.13 billion in 2022, and it is anticipated to grow in the coming years [2].

Despite the fierce competition that low-cost products manufactured in developing countries exert on the global market, the European textile and apparel industry is competitive, thanks to the high purchasing power of Europeans in Western countries, and the openness of companies in the sector to innovation. Emerging European countries play a very important role in the textile and apparel industry because their geographical proximity to Europe's major “fashion consumers” makes them very attractive to investors.

The textile industry is considered one of the oldest and largest industries in the world [3]. According to a report of the Business Research Company, [4] the global textile market was $573.22 billion in 2022, and is expected to grow to $755.38 billion in 2027 at a compound annual growth rate of 5.5%. The same study states that Asia-Pacific was the largest region in the textile market in 2022, followed by Western Europe. From the Asia-Pacific region, China has mass productivity that provides its textile industry with economies of scale unmatched by its rivals [3]. A recent report [5] shows that China, the European Union, India and Turkey comprise 75% of the world's exports. China is the world's biggest textiles exporter with a value of 148 billion U.S. dollars. Some countries such as Italy, France, Germany and the USA produce high-end or technologically advanced fabric.

The European Apparel and Textile Confederation, EURATEX, reports that in 2022 European textile activity fully recovered from the strong contraction caused by the Covid-19 pandemic, while clothing companies almost returned to their pre-pandemic level.

The research was based on the assumption that there is a direct link between the financial performance of companies and the change or variation in economic growth (as measured by the GDP/capita growth rate). The research objectives are as follows: –

generating estimates concerning the evolution of some representative indicators over the period 2023–2025 for the financial performance of public textile companies in the emerging European countries under analysis;

To achieve these objectives, the research tool available in GMDH Sheel Forecasting Software 3.5.8, [6] based on the artificial neural networks, was used. This application allows time series forecasting and regression analysis. We opted to use this application because the accuracy of neural networks and other artificial intelligence methods is superior that of traditional statistical methods. GMDH Shell software was used for forecasting analysis of public companies' pre-tax ROA (Return on Assets) in the textile industry in emerging European countries for the years 2023, 2024 and 2025.

The GMDH Sheel Forecasting Software 3.5.8 [7] research toolkit is based on learning algorithms of the Group Method of Data Handling developed in the 1960's. Over time, the method has been improved and perfected, in consequence of which GMDH Sheel Forecasting Software is a powerful tool for predictive analytics and forecasting of multi-parametric datasets. Detailed information about this algorithm is available at https://gmdhsoftware.com/docs/learning_algorithms [7]. In brief, the algorithm applies a “generator of gradually complicating models and selects a set of models that show the highest forecasting accuracy for previously unseen data. GMDH-type neural networks employ a combinatorial algorithm for optimization of neuron connection. The algorithm iteratively creates layers of neurons with two or more inputs. Every new layer is created using two or more neurons taken from any of the previous layers. Every neuron in the network applies a transfer function that allows the exhaustive combinatorial search to choose a transfer function that predicts testing data most accurately. Since every new layer can connect to previous layers, the layer width grows constantly. The generation of new layers is usually stopped when a new layer cannot show better testing accuracy than the previous one.

The forecast of financial performance indicators for textile companies in emerging European countries was based on average values of the EBITDA (Earnings Before Interest, Taxes, Depreciation and Amortization) margin, operating margin, pretax ROA and pretax ROE (Return on Equity), recorded over the period 2011–2022 for all countries included in the analysis. In accordance with the GMDH Shell Forecasting Software guidelines, the following steps were followed [6]: –

establishing the forecast range: the period 2023–2025;

The accuracy of the prediction models was evaluated using the mean absolute error and root mean square error. For the financial performance indicator (taking into account the average value across all countries included in the analysis) with the best performing forecasting model, the research was deepened for each state. Thus, one-factor linear regression was used, the dependent variable being the financial performance indicator, and the independent variable, GDP per capita growth (annual %).

The Eikon Refinitiv platform [8] (named LSEG Data & Analytics from August 2023) was accessed to collect the data required for the study in September 2023, as a search was done for textile companies in emerging European countries over the period 2003–2022. The results of the search indicated the existence of 127 companies in the textile industry in 12 emerging European countries (including Turkey and Russia). According to the statements of the provider for the financial market (LSEG Data & Analytics), the company database covers 99% of the world's market cap. Data concerning the performance indicators covered by the study are not available for 58 companies. After testing the data series for homogeneity and normality, by using the ANOVA test, 22 companies were eliminated, for which the statistical data volume was small, having been listed on the stock exchange for short periods of time. Thus, we considered relevant companies listed at least in the period 2011–2022, which were as many as 47 (Appendix 1), as the research aimed at the performance indicators of those companies, from the following countries: Turkey, Serbia, Russia, Romania, Poland, Lithuania, Estonia, Croatia and Bulgaria. It should be mentioned that the Eikon Refinitiv platform offers the possibility to access information about public companies (companies whose shares are traded on the stock market).

Our research involves a financial analysis of profitability (cost-benefit analysis). Financial analysis is the selection, evaluation, and interpretation of financial data, along with other pertinent information, to assist in assessing the risk and return associated with an investment [9].

The financial indicators considered in our analysis are profitability ratios:

EBITDA Margin represents annual earnings before interest, taxes and depreciation, expressed as a percentage of the annual total revenue. EBITDA Margin=EBITDATotal revenue×100 EBITDA\,Margin = {{EBITDA} \over {Total\,revenue}} \times 100

The analysis of the evolution of representative indicators for the financial performance of public companies in the textile industry in emerging European countries, in the period 2012–2022 (Table 1), highlights the considerable decrease in the activity of companies in the sector and, therefore, in profits and profitability rates, in the period 2019–2020, under the impact of the COVID 19 pandemic.

The year 2021 marked the beginning of the recovery of economic indicators, but not to the values before the outbreak of the pandemic. In the year 2022, the indicators analyzed are found to have registered significant increases, falling in line with the global trend. A McKinsey and Company report [35] on the global fashion industry showed that 2022 saw the highest profit margins compared to those in all years between 2011 and 2020, except for one.

The above set of indicators provides an overview of the evolution of public textile companies in the emerging European countries surveyed. The higher values of the average EBITDA margin compared to the average operating margin suggest that the share of operating costs in total costs is very high. The upward trend of the average pretax ROE was higher than that of the average pretax ROA, but it showed stronger fluctuations during the period under analysis, which suggests that equity performance is more volatile than asset performance in the textile industry. Median values are useful in financial analysis because they tend to reflect a more accurate view than averages; medians are typically less affected than averages by large deviations or outliers in the data. The analysis of the textile industry in emerging European countries in 2022, based on the median values of 4 economic performance indicators (Table 2), shows the following situation:

Clothing manufacturers and the production of coated fabrics, ecological leather, and other textiles are the specialties that have obtained the best economic results, in terms of the indicators analyzed. Increased consumer purchasing power in developing regions leads to profitability rates in the two specialties mentioned being higher than the general market profitability. For example, compared to the S&P Global ROE in 2022 (around 17%), the median pretax ROE in clothing manufacturers was double, and compared to the S&P Global ROA in 2022 (around 8.5%), the median pretax ROA in clothing manufacturers was triple.

The forecasts for the evolution of representative economic indicators for the textile industry in emerging European countries, for the period 2023–2025, indicate a slight decrease in the operating profits obtained by the companies that were the subject of the observations, with direct implications for their financial performance (Table 3).

The likelihood of this development is high if we take into account the estimates of rising production costs, so that even if the textile industry market grows in the future, the rates of return may be on a falling course.

An assessment of the performance of the forecasting models was carried out using errors, and is presented in Table 4. The low error values indicate a good performance of the forecasting models, the best performance being observed for the average pretax ROA. All the forecasting models show small errors.

The general template used to forecast the time series pretax ROA is suitable for multivariate and univariate time series analysis and forecasting. We considered it appropriate to complement the performance profile of the analyzed public companies in emerging European countries by analyzing the sensitivity of ROA to changes in economic growth (GDP/capita growth rate). Pretax ROA developments of public companies in the textile industry and GDP per capita growth (annual %) for the emerging European countries researched are highlighted below (Figure 1).

Fig. 1.

Trends in pretax ROA and GDP per capita growth (annual %) in emerging European countries

Source: own compilation based on data from Refinitiv and World Bank [36]

The unifactorial regression problem was solved using the GMDH Shell application, which generated valid econometric models for 8 of the 9 countries studied. In the case of the available data for Croatia, no unifactorial regression model could be identified.

The regression models constructed using the GMDH Shell application, which can be used for forecasting analyses of public companies' pretax ROA in the textile industry in emerging European countries, are presented in graphical form in the following figures (Figure 2).

Fig. 2.

Public companies' pretax ROA forecasting in the textile industry in emerging European countries

The analysis of the correlation between public companies' pretax ROA in the textile industry and GDP per capita growth is shown in Table 5.

For most of the countries analyzed, there is a direct correlation between public companies' pretax ROA in the textile industry and GDP per capita growth, which is in line with the results of other previous research. In most cases the strength of the link between the two indicators is strong.

The regression functions show that the change in GDP per capita growth generates a change in public companies' pretax ROA in the textile industry.

The analysis of performance indicators EBITDA margin, operating margin, pretax ROA, and pretax ROE, is imperative in order to ensure an optimal business framework needed to achieve the desired results by companies in the textile industry, within the context of volatile international markets. At the same time, the investigation of key performance indicators allows to highlight the situation of the textile industry in a specific country or region, namely Turkey, Serbia, Russia, Romania, Poland, Lithuania, Estonia, Croatia and Bulgaria.

The dynamic analysis of the average values of the above-mentioned key performance indicators revealed the continuous improvement of the financial performance of public textile companies in emerging European countries in the period 2012–2018, followed by a steep decline in the years 2019–2020, caused by the decrease in economic activity during the pandemic period, and some recovery after 2021.

By 2022, the textile industry in emerging European countries is found to have outperformed as compared to the results recorded before the onset of the Covid 19 pandemic. In the context of the difficult situation in which the European textile industry has been placed, due to globalization and competition from developing countries, the good dynamics of the key performance indicators researched demonstrate that the textile industry in emerging European countries has a high competitive potential on the international market. Therefore, despite the difficult international political context (the Russian-Ukrainian conflict), which has led to rising energy and transport prices and therefore higher production costs, the textile industry is able to successfully face the economic challenges.

In 2022, the best performing specialties were clothing manufacturers, production of coated fabrics, ecological leather, and other textiles, as well as production of polyester staple fiber, filament yarn and polymer.

The use of GDMH Sheel techniques allowed us to obtain valid econometric models to forecast public companies' pretax ROA in the textile industry in emerging European countries and to highlight the sensitivity of ROA to changes in economic growth (GDP/capita growth rate). For most of the cases investigated, it was observed that economic growth (measured by GDP/capita growth rate) has a positive effect on financial performance (measured by ROA). The percentage of variability in the dependent variable explained by an independent variable is between 45% – 68%. In terms of estimates of ROA evolution over a three-year forecast period, the results of the GMDH Shell Forecast application suggest a slight downward trend in the financial performance of public companies in the countries, surveyed, with the exception of Turkey, compared to 2022. One of the reasons for the recovery of the Turkish textile industry immediately after the Covid-19 pandemic (the country showing pretax ROA on a major increasing trend in 2021 and 2022) is the technical textile sector (including masks and personal protective clothing). In addition to this, the geographical location of Turkey between Europe and Asia is a favorable issue in the recovery and financial performance of its textile sector. After the Covid-19 pandemic, many European companies adopted a nearshoring strategy to optimize their operations, thus, Turkish textile suppliers were preferred over Chinese ones.

The most important limitation of the research stems from the limited content of the database, which is limited to public companies, since the Refinitiv platform collects information on companies listed on stock exchanges. As the impact of the activity of unlisted companies on the textile industry in emerging European countries has not been taken into account, it is possible that the overall picture of the field studied is to some extent biased. Another limitation is related to the short observation period, but we consider that this shortcoming has been largely overcome by using neural networks for prediction.