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

Objective

The aim of this paper is to identify long-term trajectories of the spatial concentration of innovation activity in 30 European countries between 1980 and 2022, and to assess whether these changes reflect sustained convergence, persistent divergence, or rather nonlinear dynamics. The analysis addresses the question of whether the observed dispersion of innovation across regions was a permanent process or merely a temporary phase preceding a renewed concentration.

Methods

The study was based on two complete patent datasets rather than a statistical sample: therefore, no procedures were used to generalize results from the sample to the population or to assess random errors. The first dataset is the global geolocated patent database by Rassenfosse, Kozak, and Seliger (2019), covering the years 1980–2014: after restricting the data to Europe, 3,247,672 observations were used, each containing information on the year, region, and exact location (x,y) of the inventor. The second dataset is the OECD Regions and Cities Atlas for the years 1995–2022, containing regionally aggregated patent data for TL2 and TL3 levels. The analysis covered 30 European countries, divided into Western Europe and Central and Eastern Europe. Two complementary measures of spatial concentration were used: the Gini coefficient for regionally aggregated data and the ETA (Entropy–Tessellation–Agglomeration) index for point data, based on Voronoi tessellation and Shannon entropy. To limit the impact of short-term fluctuations, calculations were performed using five-year moving average windows. This comparative approach captures both inequalities between administrative regions and concentration independent of territorial unit boundaries, mitigating the MAUP problem. For the average trajectories of changes, a quadratic function fit was also tested, which allowed for the formal verification of the nonlinear, U-shaped pattern of concentration changes.

Results

The results indicate a U-shaped trajectory of spatial concentration of patents in Europe. Between 1980 and 2010, innovation activity became increasingly dispersed, followed by a subsequent re-concentration. In Central and Eastern European countries, the decline in concentration was more pronounced than in Western European countries, yet they remained more spatially concentrated. A possible explanation for this sequence of changes is a shift in the focus of public policy: earlier cohesion policy favored a broader diffusion of resources and a gradual dispersion of innovation activity, while later excellence policy concentrated support more strongly in centers already saturated with innovation.

Conclusions

The spatial equalization of innovation is temporary rather than permanent. Cohesion policy can expand the territorial reach of innovation, especially during the diffusion and catch-up phases, but it does not eliminate the agglomeration forces associated with innovation at the technological frontier, nor does it offset the effects of excellence policies, which once again concentrate funding and R*D activity in areas already heavily saturated with innovation. The study demonstrates the utility of combining point and regional data, as well as classical measures of inequality with entropy measures based on geolocated data.

Keywords

innovation: patents: spatial concentration: regional disparities: innovation policy

Objective

The aim of the presentation is to introduce a method for estimating the number of residents in Poland at the municipality level using integrated administrative registers. The starting point is the new regulation on European statistics on population and housing (ESOP), which defines requirements concerning the resident population and the use of administrative sources. The study aims to develop a methodology for determining the resident population for ESOP purposes, based on length of stay, place of residence, and traces of administrative activity in registers.

Methods

The study is based on administrative data, including registers used to identify a person’s presence in the country and to assign their place of stay over time. The research process includes the selection of registers and variables, data processing and cleaning, assessment of information completeness, and integration of sources. Records containing a PESEL identifier are linked deterministically, while records without this identifier are integrated probabilistically, using a blocking stage based on approximate nearest neighbours (ANN) and pair classification with machine learning algorithms. This is followed by the imputation of missing data using multiply robust imputation methods, identification of a person’s presence in the country based on signs of life, imputation of missing information on the municipality of stay in subsequent months, and estimation of length of stay. Taking the monthly dimension into account makes it possible to reconstruct changes in place of stay and the continuity of a person’s presence in the country. The final outcome is the classification of persons into length-of-stay categories: below 3 months, from 3 to 12 months, and above 12 months, with the last category corresponding to the definition of a resident.

Results

An assessment of the quality of administrative data used to determine the resident population will be presented, covering the completeness of identifiers, the scale of missing data, the availability of information on the municipality of stay, consistency of information between registers, and the usefulness of individual sources as signs of life. The analysis will be carried out in a cross-sectional and monthly perspective, taking into account successive stages of the research process and their importance for the final classification of persons into the resident population. The results will be presented at the municipality level.

Conclusions

The proposed approach may provide a useful extension of methods for estimating the resident population used in official statistics. Its value lies in the integration of administrative data, the imputation of missing values, and the modelling of length of stay. The results may be important not only from a methodological perspective but also in practical terms, as more accurate estimates of the number of residents affect the interpretation of socio-economic indicators, the planning of public services, and the assessment of demographic processes at the local level.

Keywords

residents: administrative registers: length of stay: signs of life

Objective

The aim of this article is to present a novel method for estimating statistical aggregates for sub-threshold entities that are excluded from mandatory reporting due to their small scale of operation. The research problem addresses the high costs and respondent burden associated with traditional data-gap filling methods, such as censuses and surveys. The thesis posits that leveraging the mathematical properties of truncated empirical distributions allows for the precise reconstruction of parameters for the entire population without the need to collect additional data from small and medium-sized

Methods

The methodology is based on the assumption that the distributions of financial microdata (e.g., income, loan values) conform to normal, log-normal, or other known statistical distributions. A key element of the method is the estimation of truncated distribution parameters using the method of moments and the implicit function theorem. This enables the determination of parameters for the unobservable (sub-threshold) portion of the distribution. The quality of the results and the stability of the estimators can be assessed by comparing the results with data from periodic full censuses (if available) or through sensitivity tests regarding the truncation threshold. Statistical inference includes the analysis of estimation errors of the cumulative distribution function and confidence intervals for the estimated aggregates, allowing for the verification of the reliability of this cost-free approach compared to methods based on random sampling. The method is illustrated with one-dimensional and two-dimensional examples. The study also utilizes simulated data from balance of payments statistics regarding assets and liabilities from trade credits. The study focuses on the population above the reporting threshold (3 million PLN), which by definition creates a truncated set. The authors propose a shift from costly census or survey methods toward stochastic modeling. Furthermore, considerations regarding the generalization of the method to cases requiring the analysis of more than two sta

Results

The primary result of the work is a formalized algorithm for reconstructing a full statistical distribution based on its truncated fragment. Simulations conducted on the problem of sub-threshold entities in trade credit within the balance of payments (using simulated data) demonstrated high precision in the estimates. It was observed that the method of moments combined with the implicit function theorem effectively determines the cumulative distribution function values for sub-threshold entities while maintaining a low mean squared error. The interpretation of the results indicates that the information gap created by reporting thresholds can be systematically closed with minimal statistical error, provided the stability of the distribution type in the population is maintained.

Conclusions

The application of the proposed method allows for a significant reduction in the operating costs of central banks and national statistical offices, as well as a decrease in the bureaucratic burden on the micro and small enterprise sector. The contribution to the development of statistical theory lies in the practical application of truncated distribution properties within the production process of official statistics. The method is universal and can be generalized to multidimensional cases and other areas of economic statistics where reporting thresholds are applied. It serves as a viable alternative to traditional survey research, supporting the concept of modern, cost-free statistics based

Keywords

official statistics. sub-threshold population, estimation, truncated distribution

Objective

The PESEL register provides the backbone of probability-based social research in Poland, including key projects of official statistics as well as academic and international surveys. The aim of this analysis is to empirically assess two layers of degradation of its address component – missing addresses in the register and outdated addresses among formally complete records – and to answer the question of the extent to which the resulting inaccessibility of the adult population is structurally non-random and whether it can be effectively corrected through standard post-stratification weighting.

Methods

The analysis draws on data from twelve monthly omnibus surveys conducted by CBOS each year between 2010 and 2025. Samples are drawn through a probability scheme directly from the PESEL register and cover the adult population of Poland. The total gross sample size over the analysed period amounts to approximately 540,000 sampled individuals (around 36,000 per year on average). Results are generalised to the population through post-stratification weighting by age, sex, locality size, and region, based on Statistics Poland`s current population estimates and the 2021 Polish Census. Inaccessibility was operationalised in two layers: (1) missing address in the list released for sampling – treated as a frame coverage error: (2) address-based inaccessibility confirmed during fieldwork – defined as cases in which the interviewer established that the sampled person did not reside at the PESEL address, and the invitation could not be effectively delivered even through alternative channels (CAWI, CATI, PAPI). This category is distinct from temporary noncontact and captures cases of structural unreachability. The structure of inaccessibility was analysed by age, sex, region, and locality size: its scale was compared against the resident population figures published by Statistics Poland. Confidence intervals for proportions were estimated, and the effectiveness of standard calibration in removing bias from estimates was examined.

Results

Total inaccessibility of the adult population through the PESEL frame currently stands at approximately 20%, comprising about 6% of records with no address and approximately 14–15% of records with an address that does not lead to effective contact even in mixed-mode designs. In earlier years, before CAWI and CATI channels were introduced into mixed-mode schemes, the proportion of fully unreachable persons was higher – mixed-mode mitigates the problem only partially, not fully. Inaccessibility is strongly non-random: it more often affects younger persons, residents of large urban centres, and regions of high mobility. Standard post-stratification weighting only partially reduces estimation bias.

Conclusions

Inaccessibility within the PESEL frame is no longer an operational issue of individual research projects – it is a systemic problem of official statistics and the social sciences in Poland. Mixed-mode designs reduce its impact but do not eliminate the hard core of structurally unreachable persons. Standard calibration is reaching its limits, since the inaccessible differ from the accessible also within weighting strata. The necessary direction of development is the integration of PESEL with other administrative registers, in line with established practices of European official statistics.

Keywords

PESEL register: sampling frame coverage error: address-based inaccessibility: post-stratification weighting: