Estimation of the Fiscal Elasticities for Cyprus

This paper presents updated estimates of short-run elasticities for the fiscal variables that depend upon the economic cycle for Cyprus. First, we construct a ‘policy-neutral’ time-series dataset, adjusting fiscal data for the estimated or assumed impact of discretionary policy measures. Then, for each fiscal variable, we estimate various specifcations using different macro bases and control variables (year dummies, structural breaks, or tax reforms), applying two econometric models (first differences and error correction model) to annual data for the period 1995-2023. In addition, we employ rolling-window regressions to detect possible trends in the fiscal-to-base elasticities. For each fiscal variable, we select the specifcations that satisfy the pre-specifed model-selection criteria and identify the most suitable specification. Based on the estimation results, we propose the use of updated fiscal elasticities for our forecasts. The proposed changes to the fiscal elasticities involve adjustments to the macro bases for three main revenue categories – personal income tax, social security contributions, and value added tax – but most importantly, they generally suggest the use of higher elasticities compared to those previously applied, particularly for revenue from direct taxes (personal and corporate income taxes). The impact of these changes on the June 2024 fiscal projections of the CBC is signifcant. All else equal (i.e. without any changes in judgment), public revenue would rise by 0.1 percent of GDP in 2024, 0.3 percent of GDP in 2025 and 0.7 percent of GDP in 2026. The largest impact would stem from the changes in the elasticities of personal and corporate income taxes, with average annual impacts of 0.12 percent of GDP and 0.10 percent of GDP, respectively. The updated estimates of fiscal elasticities could improve the accuracy of our fiscal projections and provide a clearer understanding of the factors driving government revenue and unemployment-related expenses.