Replication package: Does Development Finance Drive Non-Oil Growth, or Follow the Oil Cycle? Evidence from Azerbaijan, 2000–2024

# Replication package **Paper:** Does Development Finance Drive Non-Oil Growth, or Follow the Oil Cycle? Evidence from Azerbaijan, 2000–2024 This package contains the data, code and outputs needed to reproduce every table and figure in the paper. ## Contents ### `data/` Original source files, downloaded unmodified, plus the compiled dataset. | File | Source | Used for ||---|---|---|| `4-4_1.xls` | State Statistical Committee of the Republic of Azerbaijan (SSC), Table 4.4.1, "Investment in fixed capital by source of financing" | DF_BUDGET, INV_TOTAL || `009__2_.xls` | SSC, Table 9, "Composition of gross national income" | NGDP_AZN || `007_1.xls` | SSC, Table 7.1, oil-gas and non-oil GDP at current prices and volume indices | NONOIL_VA, GDPG_NONOIL, GDPG_OIL, GDPG_DSK || `API_NY_GDP_MKTP_KD_ZG_*.xls` | World Bank WDI, NY.GDP.MKTP.KD.ZG | GDPG || `API_NY_GDP_MKTP_KD_*.xls` | World Bank WDI, NY.GDP.MKTP.KD | RGDP || `API_BX_KLT_DINV_WD_GD_ZS_*.xls` | World Bank WDI, BX.KLT.DINV.WD.GD.ZS | FDI || `API_BX_KLT_DINV_CD_WD_*.xls` | World Bank WDI, BX.KLT.DINV.CD.WD | FDI_USD || `API_NE_TRD_GNFS_ZS_*.xls` | World Bank WDI, NE.TRD.GNFS.ZS | OPEN || `API_FP_CPI_TOTL_ZG_*.xls` | World Bank WDI, FP.CPI.TOTL.ZG | INF || `API_DT_NFL_MLAT_CD_*.xls` | World Bank International Debt Statistics, DT.NFL.MLAT.CD | DF_EXT_USD || `CMO-Historical-Data-Annual.xlsx` | World Bank Commodity Price Data (Pink Sheet), annual | OIL, OIL_REAL || `AZ_FDI_DF_dataset.xlsx` | Compiled by the authors | All estimation | WDI and IDS series are from the 13 July 2026 update; the Pink Sheet file is dated 2 September 2026. All files were downloaded on 11 September 2026. **`AZ_FDI_DF_dataset.xlsx`** has five kinds of sheet: - `data` — the estimation dataset, 2000–2025, one row per year. Green cells are formulas linking to the raw sheets; black cells are computed.- `meta` — one row per variable: full name, unit, source, series code, source file, download date and methodological notes.- `raw_*` — values copied unmodified from each source file, in blue, with grey shading for years outside the estimation sample.- `checks` — internal consistency checks computed as live formulas (see below). ### `code/` Python scripts, run in this order. They require `pandas`, `numpy`, `scipy`, `statsmodels` 0.15, `arch`, `openpyxl`, `xlrd` and `matplotlib`. | Script | Produces ||---|---|| `analysis1.py` | Descriptive statistics, correlations, ADF, PP and Zivot–Andrews tests (Tables 4–7) || `tables1.py` | Formats those results into `AZ_Analysis1_Tables.xlsx` || `ardl_tables.py` | ARDL/UECM estimates, delta-method long-run multipliers, diagnostics (Tables 9–11) || `boot_rec.py` | Recursive wild-bootstrap bounds tests under the no-levels-relationship null (Table 8) || `causality.py` | Bivariate Toda–Yamamoto tests with wild bootstrap (Table 13) || `ty3.py` | Trivariate Toda–Yamamoto tests (Table 14) || `robust.py` | Robustness variants R1–R6 (Table 15) || `lagrob.py` | Lag-sensitivity variants R7–R8 (Table 15) || `cusum.py` | OLS-CUSUM and recursive CUSUM/CUSUMSQ tests (Table 12, Figures 1 and A1) || `revfix.py` | Long-run confidence intervals, lag-order sensitivity, VAR stability check || `write_ardl.py`, `add_t12.py` | Format the estimation results into `AZ_Analysis2_ARDL_Tables.xlsx` | Random seeds are fixed in every bootstrap routine, so results are exactly reproducible. ### `output/` - `AZ_Analysis1_Tables.xlsx` — Tables 4–7- `AZ_Analysis2_ARDL_Tables.xlsx` — Tables 8–15- `AZ_Figure1_OLS_CUSUM.png` — Figure 1- `AZ_FigureA1_Recursive_CUSUM.png` — Figure A1 ## Consistency checks Every series was cross-checked before estimation. The checks are live formulas in the `checks` sheet of the dataset: 1. In SSC Table 4.4.1, the financing sources sum exactly to the reported total in each year, 1999–2024.2. In SSC Table 7.1, oil-gas value added, non-oil value added and net taxes sum exactly to total GDP, and that total matches SSC Table 9.3. In SSC Table 9, GDP equals the sum of its income-side components in every year.4. `FDI_USD / FDI × 100` gives GDP in current US dollars; dividing SSC nominal GDP by it returns an exchange rate of exactly 1.7000 manat per dollar for 2018–2025 (the official peg) and 0.78 for 2013–2014.5. Constant-price GDP (2015 base) equals current-price GDP in 2015: 53.0762 billion US dollars.6. The annual change in constant-price GDP reproduces the published growth rate in every year.7. Brent prices were compared against US Energy Information Administration figures for 2024 and 2025; differences are below 0.2 dollars per barrel, within the range expected from different price assessments. ## Licence and citation Source data are public and remain subject to the terms of their publishers (SSC, World Bank). The compiled dataset, code and documentation in this package are released under CC BY 4.0. Please cite the paper when using them.