# Reconstructed Table 4 contrast

Source: Vriz et al. (2026), DOI 10.1038/s41558-026-02607-y, Extended Data Table 4, original PDF page 14.

Let g(H/L,EU/nonEU,post/pre) denote a reported group coefficient. Baseline L/nonEU/pre is zero. The reconstructed contrast is [(H−L)post−(H−L)pre]EU − [(H−L)post−(H−L)pre]nonEU. The intercept cancels.

Quantity: EU [−.078−(−.052)]−[−.055−(−.047)] = −.018; nonEU [−.022−(−.002)]−[.063−0] = −.083; difference +.065. Revenue: EU [−.193−(−.096)]−[−.155−(−.091)] = −.033; nonEU [−.063−.030]−[.080−0] = −.173; difference +.140.

Inputs are rounded published coefficients for log(1+monthly outcome). The output is a linear contrast on that transformed scale, not a percentage-point or exact percentage effect on original quantities or revenue. Group-level significance stars do not test this contrast. Without coefficient covariances, its standard error and statistical significance are unknown. No proprietary microdata/code were replicated. This limits the table’s EU-specific corroboration claim; it neither proves a reverse causal effect nor invalidates the separate baseline association.