Economic Policy Uncertainty Index for Poland: Data, Construction, and Validation
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: Economics. The day in economics: theory, mechanisms and the evidence beneath them.
Ruth: We're looking at the paper: "Economic Policy Uncertainty Index for Poland".
Warren: Here's the gist of the paper. The paper constructs and validates the Economic Policy Uncertainty Index for Poland using newspaper-based data and corrects for panel composition changes to provide a continuous monthly measure spanning January 2000 to June 2026, which is important because it establishes a systematically constructed and validated measure of economic policy uncertainty in Poland The central substantive finding is that EPU Index PL captures a predominantly domestic dimension of economic policy uncertainty rather than simply mirroring international uncertainty Index Construction and Methodology The index is constructed following the Baker–Bloom–Davis methodology, built from six deep-history Polish newspaper titles and scaled by article volume The construction involves five steps, including scaling raw counts by article volume, standardizing each newspaper series to unit standard deviation over a strict 25-year fixed baseline of 2000–2024, and then averaging across the available newspapers A crucial methodological contribution is the construction of a consistent historical series from an unbalanced newspaper panel, demonstrating that changes in newspaper coverage can generate artificial level shifts and that multiplicative chain-splicing preserves non-negativity while better retaining historical variation Validation of Measurement Quality The measurement quality is systematically validated through four complementary exercises, including a human audit and precision assessment The human audit involves trained annotators coding 600 articles from the E&U universe, yielding a precision on the flagged cell of 70.2% and an 87.5% pooled inter-annotator agreement Furthermore, tests confirm systematic validation of measurement quality through sensitivity to editorial orientation and responses to major historical events Robustness Checks on Index Construction The construction procedure includes an adjustment for newspaper composition changes using the BBD Appendix-A splicing procedure, which is implemented via a backward chain-splice from the richest six-paper anchor panel Robustness checks confirm that the multiplicative convention resolves both the negative-value problem and the GFC attenuation problem identified in Section 2.3 The index is also tested for media slant, showing that removing either partisan pole leaves the index materially unchanged, with correlations remaining above 0.
Title and authors: Ruth: So we're looking at this paper now, the "Economic Policy Uncertainty Index for Poland: Data, Construction, and Validation." It’s by Alina Landowskaa and Robert Kłopotekb. The main thing here is they’ve put together a continuous monthly measure of economic policy uncertainty for Poland from January two thousand all the way to June two thousand twenty-six <ref:2611.05640#pg1>.
Warren: A continuous index over that long period, spanning twenty-six years, that's a big data set to work with. What I find interesting is the methodology they used, because it’s based on newspaper data rather than some kind of economic survey or financial market indicator.
Iris: Exactly. The paper uses six specific deep-history Polish newspaper titles—like Rzeczpospolita and Gazeta Wyborcza—and they scale those by article volume to build the index. It’s a news-based index, which immediately makes me wonder what the assumptions are about how much news actually reflects policy uncertainty.
Ruth: Right, so it’s not pulling from GDP or inflation reports; it’s using the actual textual coverage in these specific papers. The core finding they highlight is that this EPU Index PL captures a predominantly domestic dimension of economic policy uncertainty, rather than just mirroring some kind of international uncertainty.
Warren: That sounds like a significant distinction to make. So, instead of looking at global trends, they are saying the biggest source of uncertainty for Poland is internal policy decisions and their immediate news coverage. What does that imply for how we should be thinking about Polish economic volatility?
Iris: It implies that if you want to track uncertainty accurately in this context, you can't just look at external shocks; you have to focus on the domestic media narrative. But they also spend a lot of time on the technical side of making sure this index is actually reliable.
Ruth: That’s right, because it builds this measurement using the Baker–Bloom–Davis methodology, which is a specific way of scaling and standardizing that newspaper data. They do this through five steps including scaling raw counts by article volume, standardizing each newspaper series to unit standard deviation over a strict twenty-five year fixed baseline from two thousand to twenty twenty four, and then averaging across all the available papers for each month.
Warren: That standardization step is crucial. Setting that fixed baseline period helps remove some of the noise from how much news a paper might cover in a given year versus another, right? But I wonder how robust that twenty-five year window holds up when we look at recent years like two thousand twenty-six and beyond.
Iris: That’s where the construction method gets really interesting because they address panel composition changes. Since the newspapers enter the index at different dates, a simple average would jump unnaturally if a new paper starts contributing, which is called an artificial level shift in their text.
Title and authors: Ruth: Exactly. So they introduce this correction using what they call the BBD Appendix A splicing procedure. It basically takes a backward chain-splice from the richest six paper anchor panel and matches the dispersion through a multiplicative scale factor and matches the level through a multiplicative ratio over each overlap window to keep it level continuous across all newspapers.
Warren: So they’re dealing with this issue where simply adding a new newspaper at different times could create fake jumps in their uncertainty measure, which is a real headache for any time series analysis. That splicing method sounds like a necessary engineering fix for the data structure itself.
Iris: And they also show that the multiplicative convention they use resolves two specific problems they identified earlier in the paper: the negative-value problem and what they called the GFC attenuation problem. These are technical issues related to how you handle those scaling factors when you're working with these news counts.
Ruth: The validation part of this work is pretty thorough too. They don’t just stop at building the index; they systematically validate its measurement quality through four complementary exercises including a human audit and a precision assessment.
Warren: A human audit sounds like the real test for any index based on text mining, because you can’t always trust an algorithm to perfectly tag every piece of relevant information in news articles. How precise were they finding that uncertainty?
Iris: The human audit involved trained annotators coding six hundred articles from the E andU universe, and they found a precision of seventy point two percent on the flagged cell and an eighty-seven point five percent pooled inter-annotator agreement. That’s a pretty high score for cross-checking that kind of data.
Ruth: And they also test this measurement quality by checking its sensitivity to editorial orientation and how it responds to major historical events, which suggests the measure isn't just capturing some random noise from the news cycle but is actually tied to real shifts in policy coverage.
Warren: So we have a continuous index, corrected for panel changes, validated by human auditors on a large sample, and tested against editorial bias. It sounds like they’ve built something quite solid here for tracking Polish policy uncertainty over that long timeframe. What do you think this means when we look at the larger picture?
Iris: It means we have a way to systematically constructed and validated measure of economic policy uncertainty in Poland because it’s not just some ad hoc calculation based on a snapshot; it’s built on a rigorous methodology. The central substantive finding is that EPU Index PL captures a predominantly domestic dimension of economic policy uncertainty rather than simply mirroring international uncertainty.
Ruth: That's the big takeaway for me—the focus remains internal. It tells us that when we analyze Polish economic shifts, we should be looking more at the local political and regulatory environment reflected in these specific newspapers than just what’s happening over there globally.
Title and authors: Warren: If you were running a strategy based on this index, would you rely on this domestic measure for your risk assessment? Or does it still feel like you need to overlay some external factors, like global interest rates or commodity prices?
Iris: The paper itself doesn't suggest we replace other measures entirely; instead, it provides a specific tool that isolates the policy uncertainty component from the broader noise. It’s an input for a larger model.
Ruth: So the index is an important piece of evidence showing how policy news shapes uncertainty locally. If you wanted to test this, you’d look at how these newspaper counts move alongside actual changes in government policy announcements or legislation in Poland over that twenty-six year span.
Warren: It’s a good tool for isolating that specific type of risk. But the underlying assumption is still about the quality of the news coverage itself, which is always a variable we can’t perfectly control. That's where Warren comes in, because if everyone starts tracking this index, how does that affect those newspaper counts?
Iris: That leads us to their suggestions for improvement. The paper points out things they could do next. One key area is the dynamic adjustment of their policy classification dictionary based on audit results, which they show helps minimize false positives in identifying relevant policy terms, as seen when version five of that dictionary attained the lowest gross error and was adopted as the base dictionary.
Ruth: So refining the language used to define 'policy uncertainty' itself is an active part of improving this system. It’s not just about gathering data; it’s about making sure you are measuring what you think you are measuring.
Warren: That makes sense for a quantitative approach, because if the definitions drift, the signals drift with them. And I like the idea of incorporating tests against external shocks and non-policy controls to make sure that when uncertainty spikes, it’s actually because of policy news and not some unrelated media event.
Iris: Yes, that’s about discriminant validity against external factors. They need to test if the index is truly sensitive only to policy signals and not just general media volatility. They want tests that show how the index responds when you introduce known economic events that aren't strictly policy-driven.
Ruth: And they also discussed analyzing media slant, which means they looked at correlation matrices between ideologically opposed titles to see if a title’s relative coverage moves with the party in power or something similar. It tests for bias, which is always a concern with news-based indices.
Warren: That would be useful for understanding if the index is capturing actual policy uncertainty or just partisan coverage shifts. But what about connecting this index to broader financial indicators? Could it correlate with sovereign risk premiums or exchange rates, and how do they distinguish that from measures like the VIX?
Title and authors: Iris: They also have a suggestion for a multi-dimensional validation framework against established economic indicators. That would involve correlating the EPU Index PL with things like sovereign risk premiums and exchange rates while trying to separate those out from financial volatility measures like the VIX, which is basically market sentiment.
Ruth: So, to summarize these improvements, they’re focusing on making the dictionary smarter, adding external shock tests for validity, analyzing media slant through correlation matrices between opposed titles, and building a framework that connects it to standard economic indicators. It’s moving from just a descriptive index to a more analytically robust tool.
Warren: If you take all of that—the construction details and these proposed improvements—what is the practical implication for someone who only listens to the show, someone trying to understand Poland's economy? Does this index give you a clearer picture than what’s available elsewhere?
Iris: It gives you a specific lens. For someone outside the field, it shows that uncertainty in Poland has a measurable component tied directly to how policy is discussed in domestic news. It shifts the focus from abstract macroeconomic variables to concrete media coverage patterns.
Ruth: Ultimately, it establishes a systematic way of measuring this specific type of uncertainty for Poland over that long period. It’s not just an interesting data point; it’s a structured measurement tool built on careful methodology and validated against human judgment and technical checks.
Warren: It sounds like the paper, "Economic Policy Uncertainty Index for Poland: Data, Construction, and Validation," provides a concrete foundation for tracking this specific risk factor using newspaper data. It's about getting continuity right through that splicing procedure to ensure the time series is clean.
Iris: And they make it clear that while this index is useful for isolating policy uncertainty, it’s not meant to replace all other economic indicators; it’s a specialized measure for one specific kind of information flow. That's the key caveat they put out about its scope.
Ruth: So, we have this detailed look at how to build a continuous index using BBD methodology and then how the authors are suggesting they can make that index more robust by improving their dictionary, testing against external noise, and checking for media slant.
Warren: It’s a lot of technical work to get that continuity right over twenty-six years with newspaper data, but if the results hold up under those validation tests, it gives us a reliable way to monitor policy risk in Poland.
Iris: And the final word is that they’ve done the heavy lifting on construction and validated it thoroughly against several checks before claiming its predictive power. That level of detail is what makes this paper worth reading for anyone serious about measuring uncertainty.
Ruth: We’ll wrap up here with this discussion on the "Economic Policy Uncertainty Index for Poland: Data, Construction, and Validation." It’s a specific contribution to how we measure policy risk using textual data.
The paper's summary: Ruth: So we're looking at how they put together this Economic Policy Uncertainty Index for Poland from two thousand to twenty twenty six, and what that actually tells us about policy risk there.
Warren: Basically, they’ve built a continuous monthly measure of uncertainty using news data from Polish newspapers. It’s not some abstract economic number; it’s based on how much policy stuff is being talked about in the press.
Iris: That index captures what they call a domestic dimension of policy uncertainty. The authors argue it doesn't just reflect international stuff happening elsewhere, but rather what’s going on inside Poland.
Ruth: Exactly. So, the big finding is that this measure focuses heavily on internal policy decisions and how those are reflected in the news here in Poland specifically.
Warren: That means if you want to understand economic shifts in Poland, you should be looking at the local political and regulatory environment reflected in domestic news coverage rather than just global financial trends.
Iris: They also spend a lot of time on the technical side to make sure this measurement is sound. They use a method called the Baker–Bloom–Davis approach, which involves scaling those newspaper counts by article volume and standardizing them over a fixed twenty-five year period from two thousand to twenty twenty four.
Ruth: That standardization step is key for cleaning up the data because it helps remove some of the noise related to how much news a paper covers in one year versus another.
Warren: But I’m interested in how they handle panel changes, since newspapers don't always stay in the index for the whole twenty-six years. They address this by using a splicing procedure to make sure there are no artificial jumps when a new newspaper starts contributing data.
Iris: That splicing method is important because it ensures continuity even when papers enter the index at different times, which prevents those weird level shifts you see in unbalanced panels.
Ruth: They also rigorously validated this measurement quality through four different tests, including a human audit where trained people checked a hundred percent of the articles they coded.
Warren: So they found that the human auditors were pretty accurate, getting seventy point two percent precision on the specific cells they flagged. That’s how you know it’s not just some flawed calculation from an algorithm.
Iris: And they also tested sensitivity by checking how it reacts to changes in editorial orientation and major historical events, which shows that the measure is tied to real shifts in policy coverage, not just random news noise.
Ruth: So the main thing is that they’ve built a systematic way of measuring this specific type of uncertainty for Poland over that long period, using newspaper data as the input.
Warren: It’s a structured measurement tool built on careful methodology and tested against human judgment and technical checks to isolate policy risk.
Iris: And they make it clear that while this index is useful for isolating policy uncertainty, it’s not meant to replace all other economic indicators; it’s a specialized measure for one specific kind of information flow.
Ruth: So the focus remains on making sure we have a reliable, continuous way to track how policy talk in Polish newspapers shapes the uncertainty picture locally.
Warren: If you want to test this, you’d look at how these newspaper counts move alongside actual changes in government policy announcements or legislation over that twenty-six year span.
Iris: That’s where the real test is, connecting this index to other things, like sovereign risk premiums or exchange rates, while trying to separate that from general financial volatility measures like the VIX.
The paper's improvements: Tom: So we're looking at what the authors suggest to make this Economic Policy Uncertainty Index better, beyond just building it as they did for the first time.
Ruth: They point out that they need to make their policy classification dictionary smarter by dynamically adjusting it based on audit results. It seems like refining the language used to define policy uncertainty is an active part of improving this system.
Warren: That makes sense from a quantitative standpoint because if the definitions drift, the signals in your data drift with them too. I also like the idea of adding tests against external shocks and non-policy controls to make sure that when uncertainty spikes, it’s actually because of policy news and not some unrelated media event.
Iris: That’s about discriminant validity against external factors. They want tests that show how the index responds when you introduce known economic events that aren't strictly policy-driven to prove the measurement is isolating what it claims to be.
Ruth: And they also look at media slant by analyzing correlation matrices between ideologically opposed titles. This tests for bias, which is always a concern with any news-based index, checking if a title’s coverage moves with the party in power.
Warren: That would be useful for understanding if the index is capturing actual policy uncertainty or just partisan coverage shifts that happen along political lines. But they also suggest integrating it into a bigger picture by correlating it with established economic indicators like sovereign risk premiums and exchange rates.
Iris: They want to see how the EPU Index PL correlates with those things, but they need a framework to distinguish those from general financial volatility measures like the VIX, which is basically market sentiment.
Ruth: So the plan is to move from just building this descriptive index to creating a more analytically robust tool that connects it to broader economic signals.
Warren: It sounds like they’re trying to bridge the gap between a specific news-based measure and standard macroeconomic risk assessment tools.
Iris: The goal is to make sure the index has strong links both internally, with its own data structure, and externally, with established financial metrics.
Conclusion: Ruth: So we’re wrapping up our look at the Economic Policy Uncertainty Index for Poland: Data, Construction, and Validation paper. It’s been a deep dive into how they built this continuous monthly measure of policy risk using newspaper data spanning twenty-six years.
Warren: Yeah, it’s a solid piece of work because they didn't just throw numbers together; they actually built a system that handles the complexity of unbalanced data and panel changes.
Iris: The core result is that this EPU Index PL captures a domestic dimension of uncertainty, which is really important because it shows us where the immediate policy noise is coming from inside Poland.
Ruth: Exactly. It’s not just some abstract global measure; it’s tied to the specific news flow in Polish newspapers, and they validated this measurement quality through human audits and sensitivity checks to editorial orientation.
Warren: From a strategy side, that means if you are tracking risk in Poland, you should focus on those domestic policy signals reflected in the press rather than just looking at broad international financial indicators.
Iris: And their method for handling unbalanced panels using that splicing procedure is technically sound, addressing those artificial level shifts they worried about when newspapers start contributing data at different times.
Ruth: It shows how much care goes into the methodology—from scaling raw counts by article volume to making sure the multiplicative convention resolves problems like negative values.
Warren: It’s a lot of technical work to get that continuity right across twenty-six years with text data, but if their validation tests hold up, it gives us a reliable way to monitor this specific type of policy risk.
Iris: I just want to stress that while this is a great tool for isolating policy uncertainty, it’s important to remember its limitations; it doesn't replace all other economic indicators.
Ruth: Right. It’s a specialized measure for one specific kind of information flow, and they also pointed out the need to keep refining their dictionary and testing against external noise.
Warren: That leads us nicely into what we could do next—how this type of index actually connects to broader financial risk, like sovereign risk premiums, while still keeping that policy signal clean.
Alina M. Landowskaa, Robert Kłpotekb, He Zhangc
Kozminski University · Cardinal Stefan Wyszyśki University in Warsaw · The Uncertainty Institute
econ.GN, q-fin.EC
Submitted: 2026-11-03
Updated: 2026-11-03
Comments: 45 pages, 4 figures. JEL: D80, E32, E66, G18, L50. EPU Index Poland Working Paper No. 01/2026. Data and code: https://doi.org/10.17605/OSF.IO/EP759
Code: https://github.com/rakubaku/polish-epu-index
Project page: https://www.epu-i
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Importance score: 91/100
The gist: The paper constructs and validates the Economic Policy Uncertainty Index for Poland using newspaper-based data and corrects for panel composition changes to provide a continuous monthly measure…
Key concepts
- Economic Policy Uncertainty Index (EPU) PL
- This is a newly constructed measure for Poland based on newspaper coverage. It aims to quantify how uncertain people are about the government's future economic policies. The index is built systematically using data from six Polish newspapers.
- Baker–Bloom–Davis Methodology
- This is the specific mathematical framework used to build the index. It involves scaling raw article counts by volume and standardizing newspaper series over a fixed historical period (2000-2024) to ensure a consistent, comparable measure across different sources.
- Panel Composition Adjustment
- Since newspapers enter and leave the study panel at different times, this adjustment method corrects for these changes. It uses a specific splicing procedure to prevent artificial shifts in the index levels caused by changes in which newspapers are included over time.
Terminology
Summary
The paper constructs and validates the Economic Policy Uncertainty Index for Poland using newspaper-based data and corrects for panel composition changes to provide a continuous monthly measure spanning January 2000 to June 2026, which is important because it establishes a systematically constructed and validated measure of economic policy uncertainty in Poland The central substantive finding is that EPU Index PL captures a predominantly domestic dimension of economic policy uncertainty rather than simply mirroring international uncertainty
Index Construction and Methodology
The index is constructed following the Baker–Bloom–Davis methodology, built from six deep-history Polish newspaper titles and scaled by article volume The construction involves five steps, including scaling raw counts by article volume, standardizing each newspaper series to unit standard deviation over a strict 25-year fixed baseline of 2000–2024, and then averaging across the available newspapers A crucial methodological contribution is the construction of a consistent historical series from an unbalanced newspaper panel, demonstrating that changes in newspaper coverage can generate artificial level shifts and that multiplicative chain-splicing preserves non-negativity while better retaining historical variation
Validation of Measurement Quality
The measurement quality is systematically validated through four complementary exercises, including a human audit and precision assessment The human audit involves trained annotators coding 600 articles from the E&U universe, yielding a precision on the flagged cell of 70.2% and an 87.5% pooled inter-annotator agreement Furthermore, tests confirm systematic validation of measurement quality through sensitivity to editorial orientation and responses to major historical events
Robustness Checks on Index Construction
The construction procedure includes an adjustment for newspaper composition changes using the BBD Appendix-A splicing procedure, which is implemented via a backward chain-splice from the richest six-paper anchor panel Robustness checks confirm that the multiplicative convention resolves both the negative-value problem and the GFC attenuation problem identified in Section 2.3 The index is also tested for media slant, showing that removing either partisan pole leaves the index materially unchanged, with correlations remaining above 0. p. 4
Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.
Improvements for AI systems
1
The AI system can be improved by incorporating a mechanism to dynamically adjust its policy classification dictionary based on audit results. This allows for continuous refinement of terms to minimize false positives, as shown by v5 attaining the lowest gross error and was adopted as the base dictionary.
2
The system can perform robust validation checks against external shocks and non-policy controls. This enables the AI to distinguish between genuine economic policy uncertainty and fluctuations caused by unrelated media events, using Discriminant validity: non-policy controls
tests.
3
The system can incorporate a sophisticated data adjustment layer for unbalanced panel entry. This addresses the issue of artificial level shifts
by implementing methods like BBD Appendix-A splicing procedure,
ensuring continuity even when newspapers enter the panel at different dates.
4
The system can analyze media slant and partisan robustness using correlation matrices between ideologically opposed titles. This allows the AI to test for bias, as demonstrated by testing whether a title's relative coverage moves with the party in power.
5
The system can integrate a multi-dimensional validation framework against established economic indicators. This enables the AI to correlate EPU Index PL with sovereign risk premiums and exchange rates while distinguishing these from financial volatility measures like the VIX.
Related papers
- Huge If True: Learning from the Size of the Surprise
- The Rebundling of Freelance Work in the AI Age
- Dynamic investment and coordination in teams: Theory and experimental evidence
- Does Frontier AI Strengthen Cyber Defense? Evidence from Project Glasswing
- Generative AI at Work
- What is Work For? Labor, Politics, and Meaning in the Age of AI