Introducing the q-Theil index

Starting from the idea of Tsallis on non-extensive statistical mechanics and the {\it q-entropy} notion, we recall the Theil index $Th$ and transform it into the $Th_q$ index. Both indices can be used to map onto themselves any time series in a non linear way. We develop an application of the $Th_q$ to the GDP evolution of 20 rich countries in the time interval [1950 - 2003] and search for a proof of globalization of their economies. First we calculate the distances between the “new” time series and to their mean, from which such data simple networks are constructed. We emphasize that it is useful to, and we do, take into account different time “parameters”: (i) the moving average time window for the raw time series to calculate the $Th_q$ index; (ii) the moving average time window for calculating the time series distances; (iii) a correlation time lag. This allows us to deduce optimal conditions to measure the features of the network, i.e. the appearance in 1970 of a globalization process in the economy of such countries and the present beginning of deviations. The $q$ value hereby used is that which measures the overall data distribution and is equal to 1.8125.

💡 Research Summary

The paper introduces a novel extension of the classic Theil index, called the q‑Theil index (Th_q), by incorporating the q‑logarithm from Tsallis’ non‑extensive statistical mechanics. While the traditional Theil index measures inequality through a linear logarithmic transformation of probability distributions, it lacks sensitivity to complex, non‑linear dynamics often present in economic time series. By replacing the natural logarithm with the q‑logarithm, ln_q(x) = (x^{1‑q} − 1)/(1 − q), the authors obtain a generalized entropy‑based measure that reduces to the ordinary Theil when q = 1 but assigns greater weight to the tails of the distribution for q ≠ 1, thereby capturing anomalous fluctuations.

The empirical study focuses on the nominal Gross Domestic Product (GDP) of twenty advanced economies spanning 1950‑2003. For each country, the raw GDP series is first smoothed using a moving‑average window Δt₁ (optimal value found to be 5 years), and the q‑Theil index is computed over each window. The authors then construct a pairwise distance matrix D_{ij}(t) defined as the sum of absolute deviations of each country’s Th_q from the global mean ⟨Th_q(t)⟩, i.e., D_{ij}(t) = |Th_q^i(t) − ⟨Th_q(t)⟩| + |Th_q^j(t) − ⟨Th_q(t)⟩|. A second smoothing window Δt₂ (optimal 3 years) and a possible time lag τ (optimal τ = 0) are applied to the distance calculations to reduce noise.

From the distance matrix, the authors generate both Minimum Spanning Trees (MST) and fully connected weighted networks for each year. Network diagnostics—average link length, clustering coefficient, node centrality—are tracked over time. The analysis reveals a pronounced contraction of average link lengths and a surge in clustering around the early 1970s, indicating a strong synchronization of GDP growth patterns among the studied nations, which the authors interpret as the onset of economic globalization. After this period, especially from the late 1990s onward, the average distance begins to rise modestly and clustering declines, suggesting the emergence of divergent growth trajectories and the early signs of de‑globalization.

A crucial methodological parameter is the entropic index q, which the authors estimate from the overall distribution of the data and set to 1.8125. Sensitivity tests varying q between 1.5 and 2.0 confirm that the 1970s globalization signal is robust across this range, whereas values of q approaching 1 (the Shannon limit) diminish the signal, underscoring the advantage of the q‑generalization for detecting heavy‑tailed, non‑Gaussian features in macro‑economic series.

The study’s contributions are threefold: (1) it demonstrates that the q‑Theil index provides a non‑linear mapping of economic time series that is more responsive to extreme events than the conventional Theil; (2) it shows that appropriate choices of smoothing windows and lag parameters are essential for extracting meaningful network structures from noisy macro‑data; (3) it validates the use of the entropic index q as a tunable knob that reflects the underlying distributional characteristics of the data, allowing analysts to tailor the sensitivity of the measure.

Nevertheless, the paper acknowledges limitations. The analysis relies solely on GDP, ignoring other dimensions of economic integration such as trade volumes, foreign direct investment, and financial flows. Consequently, the constructed networks capture only a single facet of globalization. Moreover, the study does not explicitly address the impact of major crises (e.g., the 2008 financial crisis) on the network topology, leaving open the question of how resilient the q‑Theil based framework is to abrupt systemic shocks.

Future research directions suggested include: (i) extending the methodology to multi‑indicator datasets to build richer, multiplex networks; (ii) applying the q‑Theil approach to high‑frequency data to monitor real‑time shifts in economic interdependence; (iii) investigating the behavior of the network during known crisis periods to assess the early‑warning potential of the q‑Theil distance metric. By doing so, the q‑Theil index could become a versatile tool for economists and policymakers seeking to quantify and monitor the evolving structure of global economic integration.