#90 - How You Can Prove Anything With the “Right” Statistics
- Adam Pawel Pietruszewski
- Aug 13, 2025
- 5 min read
Recently I wrote about growing inequalities in Poland. The post was based on data from the World Inequality Database (WID), a respected source of statistics on the measurement of inequality. Their analysis suggests that Poland is among the most unequal countries in Europe, with the bottom 50% falling behind and the gap to the top 10% widening.
A recently published SDR report, however, presents the opposite perspective: Poland is actually doing very well. Inequalities have been addressed in the last two decades, and it has received favourable ratings on both the Gini coefficient and the Palma ratio (the share of all income received by the top 10% divided by the share received by the bottom 40%).
How is this possible? Who is right? Why are the messages so confusing?
I decided to investigate. It took more than a week of intensive work, and the results essentially confirm that both are correct — and that the differences are explainable. The divergence comes from differences in:
How income is defined (pre-tax vs post-tax; national income vs disposable income)
How households and individuals are counted (equivalised households vs per-adult income)
How top earners are covered (survey under-reporting vs top-income adjustments)
What worries me is that the complexity of statistical analysis makes it very easy to misinterpret — or to interpret in whichever way pleases you more. Depending on your agenda, you can use WID or SDR data to prove almost any point you want. One can argue that the policy agenda has been successful based on SDR, or unsuccessful based on WID.
Let me walk you through the key reasons and implications.
METHODOLOGY
SDR uses official survey-based measures of disposable income inequality. These surveys tend to under-report top incomes, which is an important element of the discrepancies discussed later. SDR provides data for post-tax disposable income in the equivalised household (meaning that there is a calculation translating real income into household equivalents, further complicating interpretation [1]).
WID attempts to account for missing data in the official surveys used by SDR by making adjustments to ensure underrepresented top incomes are included. It provides pre-tax and post-tax national income as the key dimensions (national income is different from disposable income, but I will leave that aside here to avoid overcomplicating the discussion). WID also calculates income per adult, without making household adjustments, which makes it closer to actual market incomes. Pre-tax data naturally omit the redistributive effect of progressive taxation, which reduces inequality. In the dataset for Poland, WID also has disposable income data, but this is not easy to select — I had to download the full dataset to choose measures based on a similar income definition as SDR.
I also checked data from the Polish Statistical Office (GUS) to triangulate the results. It provides wage data split by deciles every other year, allowing salaries to be compared across different percentiles of the population. The limitation is that it covers only wages and only employed people, so it does not capture the full picture.
Both SDR and GUS data are available only from the early 2000s, so they do not allow us to understand longer-term patterns. For that, we must rely on WID, which shows a rapid growth of inequality at the beginning of the transformation in the 1990s, flattening in the 21st century.
INEQUALITY MEASURES
I prepared a chart showing key measures from these three sources for the period when all are available (2004 to the latest available 2022 data). Data are normalised and presented as deviation from the same starting point — 100 in 2004.
Gini Coefficient – After adjusting for income definition (I used disposable income data to improve comparability), both SDR and WID show a similar trend: inequality has decreased substantially, although the decline is larger in SDR’s official data. The most plausible explanation is the lack of top earners in official data. WID reports the share of the top 1% of earners, which has increased by over 2 percentage points — more than the increase for the top 10% overall. This suggests that underrepresentation of top earners can skew official data quite substantially. The Gini still shows the same overall trend because it focuses on middle incomes and is rather insensitive to the tails (highest and lowest data points). It is perhaps surprising that this measure is so popular: the number itself is hard to interpret, it is insensitive to critical data points for measuring inequality, and, apart from simplicity, it offers limited insight. Yet our desire to simplify the world has made it a key statistic in inequality discussions.
WID Top 10 / Bottom 50 Ratio and SDR Palma (Top 10 / Bottom 40) – These should show reasonably similar patterns, but in fact move in opposite directions. The divergence is driven by the top earners not included in SDR.
GUS P90/P10 – I complement the analysis with two statistics calculated from GUS data. The 90th percentile is the lowest earner in the top 10%, while the 10th percentile is the highest earner in the bottom 10%. This suffers from the same bias as SDR, as 10% on both tails are excluded. However, it shows a pattern fairly similar to the more sophisticated measures from both WID and SDR.
GUS Mean/P50 – The second GUS measure is the mean salary vs the median salary, available monthly. It focuses exclusively on middle earners and can serve as an early indication of larger trends. The benefit is frequent availability; the drawback is that it tends to be very stable and therefore offers few signals.
Discussion
I walked you through my entire analysis to illustrate how careful we need to be when reading statistical reports. Some conclusions seem to be obvious but in reality the story is much more nuanced.
Poland clearly faces an inequality challenge, but it seems to relate mainly to a small minority of top earners. Among the broader population, inequality appears to have declined, with policy actions such as tax changes at the lower end of the earnings scale and increases in the minimum wage contributing to a more even income distribution.
Should we be happy or worried? From a power-balance perspective, the accumulation of income in the hands of a small minority is not desirable. However, when it comes to inclusive growth and poverty reduction for the majority, SDR may be a better measure — as in many areas of statistics, outliers are often excluded to tell the story for the majority.
Conclusion
One measure is certainly not sufficient. I recommend:
Comparing Top 10 / Bottom 50 Ratio or Palma with wealth concentration (Top 1% earnings)
Being cautious with Gini: while widely used, it can mask important changes at the extremes of the income spectrum
Focusing on income after tax (whether national income or disposable income — both offer valuable insights but I leave this story for another time).
... and most of all, not jumping to conclusions, the reality can be more complex and nuanced.
If you like this post please join the growing community of forward-thinking readers and sign-up to my newsletter. My weekly posts explore how individuals and organizations adapt and evolve. Gain evidence-based insights to boost resilience across domains.
References
Bukowski, P. et al. (2023). Income Inequality in the 21st Century in Poland. World Inequality Lab Working Paper 2023/31.
Litwiński, M., Iwański, R. & Tomczak, Ł. Acceptance for Income Inequality in Poland. Soc Indic Res 166, 381–412 (2023). https://doi.org/10.1007/s11205-023-03072-2
Comments