Regional economic indicators and econometric forecasts for Surkhandarya province, Uzbekistan, 2010-2032: a reproducible dataset

Published: 12 October 2026| Version 1 | DOI: 10.17632/6knrzdfgrw.1
Contributors:
,
,

Description

This dataset provides the annual economic indicators of Surkhandarya (Surxondaryo) province, Uzbekistan, for 2010-2025, together with the econometric outputs derived from them and the code that reproduces them. It covers gross regional product (GRP, y) and seven explanatory indicators: industrial output (x1), per-capita industrial output (x2), the physical-volume index of per-capita industrial output (x3; previous year = 100), agricultural output (x4), volume of services rendered (x5), number of registered enterprises and organisations (x6), and foreign trade turnover (x7). The monetary indicators (y, x1, x2, x4, x5, x7) are in current prices (nominal); x3 is a real physical-volume index. Beyond the raw series, the deposit provides the Pearson correlation matrix; a three-stage variance-inflation-factor (VIF) selection that, combined with economic reasoning (agricultural output and services are structural components of GRP; per-capita industrial output is almost perfectly collinear with industrial output), retains three non-collinear factors - industrial output (x1), the physical-volume index (x3) and foreign trade turnover (x7), with final VIF 1.044-1.812; descriptive statistics; an ARIMA model-selection table for x1, x3 and x7; a comparison of linear, exponential and log-log regression specifications (log-log selected; in-sample adjusted R2 = 0.973, MAPE = 10.67%); and GRP forecasts to 2032 with a 95% prediction interval and low/high scenarios. Data are provided in tidy (long) and wide CSV formats and as a multi-sheet Excel workbook, with a data dictionary and reproducible R and Python scripts. The dataset offers a consistent, openly documented basis for regional economic analysis of a sub-national unit that is under-represented in open data, and supports reuse in growth modelling, planning and comparative Central-Asian studies. The 2026-2032 values are descriptive model projections, not causal estimates; users requiring causal or policy inference should deflate the series or move to growth rates and test cointegration (see the companion article’s Limitations).

Files

Steps to reproduce

Inputs. Annual values of GRP (y) and seven indicators (x1-x7) for Surkhandarya province, 2010-2025 (16 observations), were compiled from official regional statistics (State Committee of the Republic of Uzbekistan on Statistics / regional statistical office; [specific yearbook/table and access date]). Transcription was verified by recomputing the reported Pearson correlations and VIF values (code/build_dataset.py). Correlation and multicollinearity. Pearson correlations were computed among y and x1-x7 (correlation_matrix.csv). VIF was applied in three stages (vif_analysis.csv); combined with economic reasoning (x4, x5 are structural components of GRP; x2 is almost collinear with x1, r = 0.9996), the retained factors are x1, x3, x7 (final VIF 1.044-1.812). Stationarity and ARIMA. The retained factors were tested with ADF and KPSS; alternative ARIMA(p,d,q) models were compared by AIC/AICc/BIC, in-sample RMSE/MAE/MAPE and the Theil U statistic, with residual diagnostics (Ljung-Box, Shapiro-Wilk/Jarque-Bera, Breusch-Pagan). Selected models: ARIMA(0,2,0) for x1, ARIMA(0,2,2) for x3, ARIMA(0,2,1) for x7; factor forecasts to 2032 with 95% intervals (arima_model_selection.csv; run code/02_arima_models.R). Regression and GRP forecast. Linear, exponential (log-y) and log-log (power) specifications linking GRP to x1, x3 and x7 were compared; the log-log model was selected (in-sample adjusted R2 = 0.973, MAPE = 10.67%). Feeding the ARIMA factor forecasts through it yields GRP to 2032: central 218,880.9 bln soum; low/high scenarios 179,214.5-229,490.0; 95% prediction interval 141,222.1-339,244.7 bln soum (grp_forecast_2032_summary.csv; run code/03_regression_models.R). Software. R (readxl, openxlsx, ggplot2, lmtest, forecast, tseries) and Python (numpy, openpyxl). In each R script set the input path to data/surxondaryo_indicators_2010_2025_wide.csv (columns y, x1..x7).

Institutions

Categories

Economics, Econometrics, Regional Economics, Economic Forecasting Method

Licence