Expanded Summary

Georgia’s electricity system depends heavily on hydropower and is exposed to seasonal and interannual variation in river discharge. Hydropower supplied about 80.5% of domestic electricity generation in 2021. Generation usually declines in winter, when demand, thermal generation, and electricity imports increase. During high-discharge periods, additional hydropower may reduce thermal generation, lower imports, or support exports. This study estimates how Georgia’s electricity system responds to an unexpected hydropower-supply shock.

Previous studies have examined electricity balances, seasonal deficits, cross-border trade, renewable-energy development, and institutional reform. These descriptive relationships do not isolate unexpected hydropower changes because generation may respond simultaneously to demand, weather, reservoir management, and regional market conditions. This study addresses that identification problem using a proxy structural vector autoregression with an external hydrological instrument.

Monthly electricity data from the Electricity System Commercial Operator of Georgia cover January 2012 through December 2025 and include hydropower generation, thermal generation, imports, exports, and domestic demand. Net imports equal gross imports minus gross exports, while transit flows are excluded. The descriptive sample contains 168 observations. The proxy-SVAR is estimated from March 2012 through May 2025 because of weather lags and the availability of consolidated GloFAS v4 discharge data. After three endogenous lags, the reduced-form VAR uses 156 observations.

Hydropower-supply shocks are identified with a standardized river-discharge anomaly from a GloFAS v4 grid cell upstream of the Enguri reservoir. Daily discharge data were obtained through the Open-Meteo Flood API using the consolidated_v4 product. The series is generated by the LISFLOOD model with ERA5 forcing and represents modelled discharge, not observed reservoir inflow. Daily values were averaged by month and standardized separately for each calendar month using 2009–2024 moments.

Weather controls include precipitation and 2-m air temperature. The precipitation index combines standardized anomalies for the Enguri, Rioni, Mtkvari, and Aragvi basins. Three upstream locations were averaged within each basin, after which the basin indices were equally weighted. The temperature index is a population-weighted anomaly for Tbilisi, Kutaisi, Batumi, and Rustavi. The model includes contemporaneous weather controls, two lags, calendar-month indicators, and squared temperature.

The endogenous system contains logged hydropower generation, inverse-hyperbolic-sine-transformed thermal generation, and net imports in GWh. VAR(3) minimized the Akaike information criterion among stable specifications without significant residual autocorrelation.

The Enguri anomaly instruments the reduced-form hydropower innovation. The first-stage Newey-West HAC Wald F-statistic was 16.0, with an  of 0.074. The shock is normalized to increase hydropower generation by 100 GWh on impact. Impulse responses are estimated over 24 months. Uncertainty is assessed using 2,000 joint moving-block-bootstrap replications with 12-month blocks, supplemented by weak-identification-robust Fieller confidence sets.

Hydropower represented 78.4% of cumulative domestic generation, with median monthly generation of 720.0 GWh. Median thermal generation was 207.4 GWh, and median net imports were 58.6 GWh. In the lowest hydropower quartile, median thermal generation was 361.0 GWh and median net imports were 171.3 GWh. In the highest quartile, these values fell to 4.2 GWh and −165.5 GWh. Hydropower was negatively correlated with both thermal generation and net imports, with Spearman coefficients of −0.79.

The identified shock produced a temporary increase in hydropower generation. Following the normalized 100 GWh impact response, the hydropower response was 62.0 GWh after one month (95% CI 31.5 to 111.8), 44.7 GWh after two months (95% CI 20.3 to 83.9), and 35.7 GWh after three months (95% CI −8.8 to 77.4). The month-three 95% confidence interval included zero; the response was therefore not statistically significant at the 5% level.

Thermal generation declined by 80.1 GWh on impact, although the 95% confidence interval included zero. The point estimate is consistent with substitution between hydropower and thermal generation, but the effect is imprecisely estimated.

Net electricity imports declined by 56.6 GWh on impact, with a 95% confidence interval from −137.6 to −9.8 GWh. This is the clearest statistically supported response. It may reflect lower imports, higher exports, or both. The cumulative response over months 0–12 remained negative, but its confidence interval included zero.

Additional sensitivity analyses considered alternative hydrological instruments, GloFAS products, estimation samples, lag orders, bootstrap block lengths, and weather-control specifications.

The baseline instrument is based on modelled discharge from one Enguri grid cell rather than observed reservoir inflow or a nationwide measure. The analysis also cannot fully account for regional electricity prices, transmission constraints, reservoir management, plant outages, or other operational conditions. The exclusion restriction cannot be tested directly, and the linear monthly VAR does not capture seasonal, nonlinear, or regime-dependent responses.

The results identify cross-border electricity trade as Georgia’s clearest short-run adjustment margin following a positive hydropower-supply shock. Thermal generation may also decline, but the estimate is imprecise. Storage, diversified generation, improved hydrological forecasting, coordinated reservoir management, and adequate cross-border transmission may strengthen system flexibility. 

Keywords: hydropower-supply shocks; river discharge; electricity trade; net electricity imports; proxy-SVAR.

JEL Classification: C32; F14; Q25; Q41.