Day 1 (Room A, July 1) - Congress Replay - The 6th Congress of Polish Statistics

Objective

The aim of this paper is to present the concept of a new methodology for the generalisation of results in the DG-1 survey using the sampling method. The current methodology, based on a generalisation factor relying on the number of employed persons, does not allow for the full application of the sampling method apparatus for assessing the quality of estimates. The new approach includes modifying the sampling scheme and sample allocation, applying the cut-off sampling method, implementing calibration estimation, and developing a method for assessing the quality of estimates.

Methods

The DG-1 survey covers non-financial enterprises (industry, construction, transport, trade, services) employing 10 or more persons. In January 2025, the population consisted of 113,361 units. Currently, all large units (employing at least 50 persons) are included in the sample, together with a minimum 10% sample of medium-sized enterprises (employing 10–49 persons), drawn with stratification by voivodeship, NACE section, division and group, and ownership sector. In order to evaluate the proposed methodological modifications, a simulation study was carried out using the Monte Carlo method (500 replications). The pseudo-population was constructed by linking the DG-1 register from January 2025 with unit-level data from the VAT register using the REGON identifier (103,028 units, 90.9% correctly linked). The study variable was sales. A total of 128 sampling schemes were tested, differing in sample allocation (proportional to stratum sizes and square-root allocation), minimum stratum sample size, and cut-off sampling thresholds, with a fixed sample size of 25,000 units. Calibration estimation was applied with the number of employed persons as the auxiliary variable, by voivodeship and NACE section. The quality of the estimates was assessed empirically using relative bias, relative standard error, and relative root mean square error.

Results

The analysis of the DG-1 population revealed a strong right-skewed distribution: units employing 10–49 persons account for 81.8% of the population but only 26.2% of employed persons, whereas units employing at least 1,000 persons (0.6% of the population) account for as much as 27.4% of employed persons. The problem of non-response mainly concerns the smallest enterprises – approximately 40% in the group employing 10–19 persons compared with approximately 2% in the group of the largest units. The results of the simulation study indicate that square-root allocation leads to a lower relative standard error than allocation proportional to stratum sizes, and that appropriately chosen cut-off sampling thresholds improve the precision of the estimates with an acceptable level of bias.

Conclusions

The generalisation of results in the DG-1 survey based on the sampling method will be possible only in selected domains in which an acceptable quality of estimates can be achieved. The introduction of the cut-off sampling method together with square-root allocation and calibration estimation makes it possible to improve the precision of the estimates while at the same time reducing the response burden, especially for the smallest enterprises, which are most affected by the problem of non-response. The proposed approach may find application in other business statistics surveys carried out within official statistics.

Keywords

DG-1 survey, sampling method, cut-off sampling, calibration estimation, business statistics

Objective

The objective of this study is to develop a quick-to-estimate, operational indicator of local content for Polish enterprises, based on administrative data sources and applicable across sectors. The research aims to quantify the domestic share of supply chains, incorporate multi-tier supplier structures, and provide a scalable tool for policy analysis, while addressing the limitations of existing international approaches.

Methods

The methodology combines administrative microdata from the JPK_VAT system with a directed-tree representation of supply chains. Suppliers network is identified through REGON identifiers. For each enterprise, the Local Content Score LCS(t1,t2,D) is computed for period (t1,t2) and depth D. For D=1, the score equals the share of domestic purchases in total purchases of goods and services. For D*1, the score incorporates weighted contributions of key domestic suppliers’ own LCS(D-1), where key suppliers are selected using statistical rules (e.g., top-N suppliers or suppliers generating x% of total purchases). The supply chain is modelled as a directed acyclic tree with a unique path from the focal firm to each supplier. When short supply chain is considered, the directed acyclic tree is a very precise model of real flows. The longer the supply chain, the more frequent non-unique paths from supplier to focal firm occur. The indicator is mathematically bounded in [0,1] and monotonically decreasing with depth. Two aggregation strategies: top-down and bottom-up, are evaluated for sector-level analysis. The approach emphasizes automation, scalability, and independence from survey-based data collection.

Results

The presentation will showcase illustrative Local Content Score (LCS) results for selected industries, highlighting how the indicator behaves across different supply chain depths (tiers). For each sector, LCS values will be reported for D=1, capturing direct domestic sourcing, and for higher tiers (D*1), which incorporate domestic inputs embedded in the operations of key suppliers. The examples will demonstrate how the score systematically declines with increasing depth, reflecting the diminishing visibility of domestic components further along the supply chain. Differences across industries will also be discussed, showing how sectors with concentrated supplier structures exhibit higher multi-tier domestic content, while more fragmented or import-intensive sectors display steeper declines.

Conclusions

The proposed LCS offers a practical and statistically robust tool for assessing domestic content in supply chains using existing administrative data. It supports evidence-based policymaking in areas such as industrial strategy, procurement, and economic security. While limited by the absence of product details and nationality of equity in JPK_VAT, the method enables scalable, automated monitoring of domestic participation across sectors. Its flexibility and interpretability make it a valuable contribution to the statistical toolkit for analyzing local content.

Keywords

local content: supply chains: administrative data

Objective

The aim of the paper is to identify and assess the dynamic relationships between producer price inflation and short-term receivables and credit adjustments in the manufacturing sector in Poland. The research problem focuses on whether changes in producer price inflation precede changes in receivables and short-term credit and loans, and whether significant dynamic relationships exist between these financial categories.

Methods

The study is based on quarterly data for Poland covering the period 2014–2025. Producer price inflation is measured by the producer price index. Short-term financial adjustments are captured by two variables derived from RF-01 (quarterly report on financial assets and liabilities) reports of Statistics Poland and analysed at the level of the manufacturing sector: receivables, proxied by the value of the item “other amounts receivable”, and credit and loans with an original maturity of up to one year. The analysis applies a vector autoregression (VAR) model, which makes it possible to examine dynamic interdependencies among three variables treated as endogenous. To identify the direction of relationships, Granger causality tests and the Toda–Yamamoto procedure are used: the latter allows inference under less restrictive assumptions regarding the orders of integration of the analysed time series. Given the relatively small sample size, a wild bootstrap procedure is also employed to strengthen the reliability of statistical inference. Prior to estimation, the time-series properties of the variables and the lag structure of the model are verified in order to reduce the risk of misspecification and improve the comparability of the results.

Results

The conducted analysis indicates that producer price inflation may be related to short-term receivables and credit adjustments in the manufacturing sector in Poland, although the strength and direction of these relationships may vary depending on the model specification. The obtained results provide a basis for assessing whether increases in producer prices precede changes in receivables, whether they are associated with changes in the scale of the use of short-term credit and loans, and whether significant dynamic relationships exist between receivables and short-term financing. The findings also make it possible to compare the relative importance of the two adjustment channels in the sector’s response to price pressure.

Conclusions

The contribution of the paper lies in combining a macroeconomic and a financial perspective in the analysis of short-term adjustment mechanisms in the manufacturing sector. The study fits into the stream of research on the transmission of price impulses to short-term financial positions and may serve as a reference point for further studies on the relationships between price changes and the short-term financial condition of the manufacturing sector in Poland. The findings may also be useful for institutions monitoring the condition of the enterprise sector, particularly manufacturing, in Poland.

Keywords

corporate finance, receivables, short-term credit, vector autoregression, Granger causality

Objective

This paper examines discount retail locations as outcomes of firm location choice and as indirect evidence of local consumer-facing market potential. Building on the consumer city perspective (Glaeser, Kolko, * Saiz, 2001), urban economics of retail allocation, retail agglomeration research (Sevtsuk, 2014: Piovani, Zachariadis, * Batty, 2017) and chain store location evidence (Jia, 2008: Holmes, 2011), it asks whether Lidl and Biedronka supermarkets’ geographies reveal transferable relationships between population, accessibility, amenities, barriers and private location suitability.

Methods

The study uses reproducible spatial sources: OpenStreetMap urban feature channels, an NSP 2021 Census Grid population channel from GUS at 250m granularity and observed supermarket store locations within Polish city boundaries. The empirical object is a rasterized urban surface. Layers are projected to EPSG:2180 and harmonized to a 50 m modelling grid: population is resampled and interpreted at the source-grid level, not as directly observed 50 m population. Lidl and Biedronka OSM tags are excluded from explanatory POI channels and used only as target locations. Separate models are trained for each brand. The explanatory tensor contains seven channels: population, roads, rail corridors, water barriers, point-of-interest density, mobility-access proxy and public-amenity density. To reduce spatial leakage, the design uses cross-city validation: Warszawa, Łódź and Gdańsk form the training set: Kraków, Poznań and Białystok the validation set: Wrocław, Szczecin and Bydgoszcz the test set. Following spatial machine-learning arguments about spatial dependence and transferability (Kopczewska, 2022), the model estimates a brand-specific store location suitability surface. The main estimator is a conditional GAN: the generator maps context rasters into predicted store-presence surfaces, while the discriminator evaluates their spatial plausibility given the same context. Evaluation combines pixel loss with count error, count bias, mass bias, cluster morphology, Moran’s I and Geary’s C.

Results

Preliminary diagnostics indicate that observed discount retail geography is not reducible to residential population alone. Economically informative store presence surfaces emerge where population support coincides with road accessibility, POI intensity, public amenities and limited barrier separation. The outputs therefore translate spatial prediction into measures of revealed local-market suitability: where demand-supporting context concentrates, where brand-specific retail mass is expected and where water or rail barriers weaken otherwise dense surroundings. Cross-city errors identify which components of the local market structure are transferable and which remain city specific.

Conclusions

The paper contributes a framework for measuring local urban market potential through revealed discount retail allocation. Predicted store presence surfaces are interpreted as indicators of private firm location suitability: areas that resemble observed store choices under the spatial structure of population, accessibility, amenities and barriers. The claim is empirical rather than normative: the model evaluates how existing retail networks encode local economic activity and how far this logic generalizes across cities. For urban economics and applied statistics, it links consumer city theory, OSM and Census Grid data, spatial ML and autocorrelation diagnostics.

Keywords

local market potential: retail allocation: consumer city: spatial machine learning: generative adversarial networks

Objective

The aim of the study is to assess whether the transition toward a circular economy contributes to reducing regional disparities in separate municipal waste collection in Poland. The analysis focuses on identifying convergence processes across selected waste fractions and examining the role of socio-economic factors in shaping both the pace and direction of these changes across regions over time.

Methods

The study was based on a panel dataset covering Polish counties for the years 2017–2024, obtained from the Local Data Bank of the Central Statistical Office (GUS). The analysis was conducted in three complementary stages. First, dynamic panel models estimated using the system GMM approach were applied to test the hypothesis of ß-convergence and to capture regional heterogeneity as well as dynamic interdependencies between observations over time. Second, the non-parametric Phillips and Sul log-t test was employed to identify convergence clubs and to assess the existence of a common development path across units, allowing for heterogeneous transitional dynamics. Third, discrete choice models (logit and ordered logit) were used to examine the impact of socio-economic factors on the probability of counties belonging to specific convergence clubs. In panel models, the dependent variable was the volume of municipal waste in four fractions: paper and cardboard, glass, bio-waste, and bulky waste. Explanatory variables included income levels, unemployment rate, degree of urbanization, population density, demographic structure, and tourism intensity. All analyses were conducted using STATA and R software, ensuring robustness and comparability of the results.

Results

The results indicate the presence of conditional ß-convergence for paper and cardboard, glass, and bulky waste fractions which means that regions with higher initial waste levels experience lower rates of waste growth. In contrast, a divergence process was observed for bio-waste, indicating increasing regional disparities over time. The log-t test confirmed the absence of global convergence and revealed the existence of several convergence clubs characterized by clear spatial differentiation. The econometric results further show that socio-economic factors such as income levels, degree of urbanization, tourism intensity, and unemployment significantly influence the likelihood of counties belonging to specific convergence groups.

Conclusions

The transition toward a circular economy in Poland contributes to reducing regional inequalities only in selected waste streams. The absence of global convergence and the presence of convergence clubs indicate persistent heterogeneity in waste management systems across regions. These findings highlight the need for differentiated regional policies tailored to local socio-economic conditions rather than uniform national approaches. The study contributes to the literature by applying advanced convergence analysis methods at the subregional level, providing new insights into the spatial dynamics of waste management and circular economy transitions.

Keywords

waste, convergence, counties

Objective

This study aims to identify latent components of reporting behavior among micro and small enterprises and to assess whether data obtained from indirect survey questions, adjusted for behavioral factors, can be used in an econometric model to estimate the unobserved component of firm revenues. The analysis explores the potential of such data to approximate non-observable aspects of economic activity.

Methods

The study is based on primary survey data covering micro and small enterprises for the years 2022–2023, with more than 10,000 firms surveyed annually. The dataset is representative of the SME population. An indirect survey approach was applied, in which respondents acted as expert informants on conditions within their industries. Firms reported subjectively “satisfactory” levels of revenues, costs, and income, which were used to identify latent components of financial reporting behavior. To account for behavioral heterogeneity, a Sentiment Index was constructed, aggregating declared expectations, motivations, and perceptions of business conditions. This index was incorporated as a correction variable in the modeling process. A structural econometric model was then estimated, decomposing expected revenues into observable and latent components. Estimation was conducted using structural modeling techniques, and uncertainty was assessed through standard errors and parameter significance tests. The latent component is interpreted as a systematic deviation in subjective financial reporting, serving as a proxy for non-observable economic activity rather than a direct measure of the shadow economy

Results

The findings reveal significant differences between reported and adjusted financial performance in the analyzed population. Incorporating the Sentiment Index allows for the identification of an underlying structure in firms’ reporting behavior. The latent component is statistically significant and varies across sectors, indicating a systematic pattern in reporting deviations. The results are partially consistent with existing estimates of the shadow economy published by the Central Statistical Office (GUS), supporting the potential of the proposed approach as an approximate measurement tool for informal economic activity in the SME sector

Conclusions

The results show statistically significant differences between reported and behaviorally adjusted financial performance. The Sentiment Index reveals systematic patterns in reporting behavior, indicating that declarations are influenced by expectations and perceived conditions. The latent revenue component is significant and varies across sectors, suggesting structured rather than random deviations. The magnitude of these differences indicates that the model captures part of unobserved economic activity. The findings are partly consistent with estimates of the shadow economy published by the Central Statistical Office (GUS), supporting the validity of the approach.

Keywords

SMEs, shadow economy, survey data, econometric model, Sentiment Index

Objective

The aim of this article is to analyze changes in the higher education premium in Poland between 2002 and 2022. The research hypotheses are as follows: (1) The increase in demand for skills caused by technological progress translated into an increase in the higher education premium, (2) The strong increase in the labor supply of people with higher education, especially women, was a factor reducing the higher education premium.

Methods

The data source is individual data on individual wages and employer and employee characteristics from the Structure of Wages Survey by Occupation. The data are representative of the population working in firms employing 10 or more people. We use aggregated data across age groups, gender, and ownership sectors. We divided the sample into seven age groups, separately for men and women, and for the public and private sectors – a total of 28 age-gender-sector subgroups. We included fixed effects in the model to account for group-specific effects. The theoretical basis for the analyses is the canonical production function model of Card and Lemieux (2001), which we extend in two ways. First, we add a third dimension (besides age and gender) by varying the education premium between the public and private sectors. Second, we allow for the elasticity of substitution between production factors to vary over time. We are interested in estimating the impact of both the relative aggregate supply of workers with tertiary and secondary education on the size of the tertiary education premium, as well as assessing the role that changes in the relative labor supply across age groups, genders, and sectors have played in explaining the relative increase in returns to tertiary education. We test the hypothesis of imperfect substitution between workers with tertiary and secondary education, controlling for age, gender, and sector.

Results

The results indicate a significant role for changes in the relative aggregate labor supply of individuals with higher or secondary education, as well as a significant role for changes in the relative labor supply across age groups, gender, and sector in explaining changes in the higher education premium. The role of demand factors was small. The estimated values ??of the partial substitution elasticity are significantly higher for women than for men. Furthermore, sectoral differences play a significant role in explaining wage differences. The estimated values ??of the partial substitution elasticity are significantly higher for the public sector than for the private sector. Furthermore, the results indicate an increase in the partial substitution elasticity for individuals with higher or

Conclusions

The research shows that, unlike analyses for the US, where demand factors and technological changes played a significant role in shaping the premium for higher education, in Poland, changes in relative labor supply across age, gender, and ownership sector play a key role. The results show that women across age cohorts and sectors are treated by employers as a more substitutable production factor than men. The results also indicate a higher substitutability of workers with the same education in the public sector than in the private sector.

Keywords

Education premium, wages, labor supply, age cohorts.

Objective

The aim of this paper is to define functional urban areas. The research problem addresses the limitations of approaches used by, among others, Eurostat, which focus on one-way commuting flows into cities while ignoring internal linkages and outbound flows from urban centers. It is assumed that the application of graph methods enables a more comprehensive representation of actual functional structures and relationships between territorial units.

Methods

This study used public statistics data from the commuting survey conducted as part of the 2021 National Census. For comparative purposes, the commuting matrix was transformed to treat urban-rural municipalities as single units, without distinguishing between their urban and rural parts. A network approach was applied, in which municipalities were represented as graph nodes and flows of employed workers between them as weighted edges. Functional urban areas were identified using graph methods, particularly community detection algorithms (Leiden and Louvain), which allow the identification of structurally coherent groups of nodes with strong internal connections while taking bilateral relations into account. Centrality measures were also employed to identify centers playing key roles in the network structure and to determine their importance within the functional system. Additionally, an accessibility matrix developed as part of the experimental statistics of the Central Statistical Office (GUS) was used, containing travel times between municipalities. The obtained results were compared with other delimitations of functional urban areas, including those based on the Eurostat methodology. Differences were assessed using selected structural measures and local graphs.

Results

The results indicate that the use of graph methods leads to the identification of functional urban areas with greater internal cohesion than the traditional approach. The network approach enabled the capture of bilateral relationships, resulting in a more realistic representation of functional structures. Comparative analysis revealed differences from existing delimitations of functional urban areas, especially regarding boundary delineation. Furthermore, the inclusion of the accessibility matrix strengthened the identified connections between municipalities, contributing to greater cohesion of the identified structures and a better reflection of actual transport conditions.

Conclusions

The use of graph methods combined with commuting data provides an effective tool for delimiting functional urban areas. This approach allows for a better understanding of actual functional structures and spatial relationships. The results indicate that network analysis can significantly support spatial planning and regional policy by enabling a more precise definition of urban catchment areas. Furthermore, incorporating other types of flows, such as commuting to school, allows for a more comprehensive understanding of the multidimensional nature of connections between territorial units.

Keywords

functional urban areas, graphs, community detection, spatial accessibility, commuting to work, commuting to school