Abstract. This study examines hidden weaknesses in VAT collection by constructing a transparent benchmark for Romania’s VAT performance and inferring a model based shortfall when observed revenues fall below levels consistent with macro and policy fundamentals. Using annual data for 2000 to 2023, VAT revenue and household consumption (as percent of GDP) are taken from Eurostat, while unemployment and internet use are drawn from the World Bank to capture cyclical conditions and broad digital diffusion. The analysis links changes in VAT performance to the standard rate regime, the post EU institutional environment, and the 2009 to 2010 crisis period, while controlling for macro structural variation. Predicted VAT outcomes are translated into an implied shortfall proxy and used to identify years of unusually weak performance relative to the benchmark. The results indicate that Romania’s shortfall signal is episodic rather than persistent, with the largest implied under collection concentrated in a small set of early years. The average inferred shortfall is about 0.160 percentage points of GDP per year, with a maximum of about 0.752 points in 2004, while 2001 and 2004 emerge as unusually weak performance years. A simple predictive check over 2016 to 2023 suggests higher errors in later years, consistent with structural change. The findings complement official VAT gap assessments by offering a single country time series benchmark suitable for diagnostics and policy discussion.

Keywords: value added tax gap, tax compliance, electronic invoicing, shadow economy, fiscal administration, econometric benchmarking.