Appendix D

Technical data issues and discrepancies

This appendix documents the most significant technical issues encountered when using production and trade statistics as inputs to the UNITAR plastic model. While the datasets used are official and high-coverage, several types of uncertainty limit their direct use in plastic-flow modelling without careful pre-processing and validation (Chen et al., 2022; Eurostat, 2024; UNECE, 2022). These issues highlight the importance of further methodological development and dialogue with national statistical agencies before routine use of the model.

Overview of common data quality concerns

After evaluating the underlying data, in combination with previous research on the topic, it’s clear that using trade and production data as a foundation for calculating apparent consumption of plastic comes with its innate challenges. The datasets contain inconsistencies in terms of extreme or missing values, units or extreme outliers – however it’s a difficult validation process. The most common data issues are:

Data is misreported by producers: Misreporting at the enterprise level is common. Firms frequently enter incorrect quantities or misclassify goods in customs declarations, leading to large errors in aggregate totals (Eurostat, 2020, 2024).
Unit mismatches regularly occur: Typical examples include declaring litres, pieces or square metres but recording them as kilograms, or omitting weight fields altogether (Eurostat, 2020, 2024; UNSD, 2011). These errors can inflate or deflate plastic-flow estimates by orders of magnitude.
Missing values and confidentiality gaps: Quantity or value fields are frequently blank
In conversation with Eurostat, we confirmed the difference between data being flagged as zero (“0”), “Data not available”. That data is not available does not mean that the value is zero, nor does it mean that the data is missing due to confidentiality. It simply means that “the data were not provided by the relevant member state to Eurostat” (E. Firth, personal communication, May 20, 2025). This highlights the importance of directly including stakeholders from the relevant statistical agencies in development of these datasets – to be able to provide, validate and distinguish missing data from other causes of data issues.
or suppressed, increasing uncertainty (Chen et al., 2022; Eurostat, 2024; UNSD, 2011).

Previous analyses of global plastic trade confirm that these three issues—misreporting, unit errors and confidentiality gaps—propagate uncertainty through plastic-flow models (Barrowclough et al., 2020). This highlights the need for thorough validation of unit codes and plausibility thresholds. It also showcases the importance of allocating both time and resources to making sure the underlying input data is sound – even before evaluating the assumptions of plastic content.

Large unexplained year-to-year shifts

Several countries show substantial changes in reported trade and production volumes between adjacent years, which may reflect real shifts in trade but could also be caused by changes in reporting practice, classification, or data quality.

As shown in figure 15 and figure 16, imports and exports for countries like Sweden and Denmark vary considerably year to year for the included trade codes (before using the model to calculate plastic content). Some of these fluctuations are very sharp as seen for Sweden before 2011, and Norway around 2010.

Figure 15 Reported import volumes (tonnes) for included HS codes before using the model to calculate plastic content, 1988–2024. Large fluctuations are visible in Sweden (yellow line) and Norway (dark blue line), particularly around 2007–2010.

Figure 16 Reported export volumes (tonnes) for included HS codes before using the model to calculate plastic content, 1988–2024. Denmark’s reported export volumes show a sharp rise after 2015, possibly linked to changes in classification or re-export practices.

Figure 17 displays calculated plastic volumes in domestic production revealing several extreme outliers. In some instances, volumes increase by several thousand percent from one year to the next—levels that far exceed what can reasonably be attributed to changes in actual plastic volumes.

This is illustrated in the Norwegian data for specific product groups. In the Building and Construction category (figure 18), plastic content spikes to over 2 million tonnes around 2010, compared to typical levels below 200,000 tonnes. A similar outlier is visible in the Packaging category (figure 19), where a single-year jump exceeds 10 million tonnes—orders of magnitude above baseline trends for the country totals. These anomalies are not isolated to Norway and appear to stem from either unit conversion errors, inconsistent reporting practices, or the use of inappropriate plastic share proxies in the source data.

Such outliers strongly distort plastic flow modelling and highlight the need for systematic outlier detection and validation protocols before production data can reliably support harmonised statistics.

Figure 17 UNITAR model plastic content from production (no removed outliers).

Figure 18 Plastic content in Building and Construction - Norway

Figure 19 Plastic content in Packaging - Norway

Due to project constraints, this pilot only removed the most extreme outliers and did not undertake a detailed validation of the underlying trade and production data. As a result, it remains unclear whether the observed spikes reflect actual material flows—potentially requiring custom plastic share assumptions—or stem from statistical issues such as unit mismatches, misreporting, or confidentiality-related gaps. Nonetheless, the magnitude and frequency of such anomalies—and their disproportionate impact on modelled plastic flows—underscore the need to integrate robust outlier detection and data validation in future modelling protocols.

Discrepancies between Eurostat and national production statistics

A clear example of inconsistencies in the data came from comparing the PRODCOM production data acquired from Eurostat as well as directly from SSB. As we can see in figure 20 significant differences for some years were observed between production data obtained from Eurostat and from Statistics Norway (SSB) under the same PRODCOM codes. As shown in figure 20 differences exceeded 10 million tonnes in some years.

Figure 20 Differences between Norwegian PRODCOM production data (Eurostat vs. SSB). Notably even if most years, apart from two, are quite similar, only 2023 show the exact same numbers.

To combat this in the analysis manual removal of extreme outliers was performed where deemed necessary and most often replaced with linear trends between the surrounding years. This finding reinforces the importance of involving national agencies directly in the data collection, validation and interpretation process.