One price panel for Uganda's construction economy
Phase 0 turns nine years of Uganda Bureau of Statistics construction-price releases and the Ministry of Finance macro portal into a single, audited, monthly panel — the foundation under every CiO product, projection and claim.
How the data moves
Every stage is a script in the repository; counts are read from the current buildWhat the panel is for
Three target products, each with a live prototype in the estimation layerAccurate bidding and pricing
Material-level price indices, spliced across every UBOS rebasing, so a bill of quantities can be priced against the real trajectory of cement, steel, diesel and labour — not a rule of thumb.
Realistic works programmes
Volatility clusters and structural-break dates show which inputs move together and when regimes changed, so programmes carry the right float where price shocks actually land.
Claim avoidance
The Dynamic Adjustment Mechanism replaces disputed price-fluctuation claims with a contract clause that revises the contract sum when a calibrated volatility index breaches its band.
Two public sources, one harmonised panel
Only obtainable, citable sources are used. Every raw release is kept frozen in the repository; the panel is rebuilt from them by one command.
Coverage of the 29 panel series
Solid bars are observed months; hatched segments mark gaps inside a series' span. The short bars (aluminium, buildings and civil-works sub-indices) are structurally short — UBOS published them for six months of 2022 only — not parser failures.
Material prices, July 2017 to April 2026
Pick any materials and macro series. Everything is re-indexed to 100 at the first month shown so different units share one axis; switch that off to see the raw index values.
Every number can be traced back to a release
The panel ships with its own evidence: how vintages were chained, how it reconciles with official aggregates, and the time-series properties a reviewer will ask about first.
Reconciliation with official aggregates
Overlap window, monthlyCorrelation of the CIPI headline with the consumer price index and the manufacturing producer price index over the same months. High agreement confirms the spliced series tracks the official picture; the residual index-point gap is the construction-specific signal.
Stationarity verdicts
ADF and KPSS, level and first differenceMost usable series are integrated of order one — difference-stationary — which is what the SVAR and forecasting layers assume. Level-stationary and inconclusive cases are the short or sparse series.
Structural breaks by series
Bai–Perron-style scan, BIC-selectedEach dot is a detected break in a series' level. The 2020 and 2022 columns are the pandemic and the commodity shock; the explorer above overlays the same dates.
Vintage splicing
Chain-linking across UBOS rebasingsA factor of 1.000 means consecutive releases agreed on the overlap; factors below one record a rebasing. Links with no overlapping months are flagged for manual review rather than fabricated.
Full stationarity table (29 series)
Five estimation models already run end to end
Each paper in the portfolio has a starter model wired to the panel. These are its current results, read straight from the repository outputs.
P4 · Which inputs move the market — the SIMI ranking
Diebold–Yilmaz spillovers from a VAR on monthly log-returnsNet spillover: the share of other materials' forecast-error variance a material explains, minus the share explained by others. Cement dominates the system; sand, timber and paint are mostly receivers.
Spillover table: row = receiving material, column = source. The diagonal is own-variance; brighter off-diagonal cells are stronger transmission.
P3 · Volatility clusters
Dynamic-time-warping k-means on log-returnsMaterials whose price paths share a shape sit in the same cluster — the basis for procurement risk grouping and programme float.
P6 · Macro shocks into construction costs
Recursive SVAR, 24-month horizonForecast-error variance of the CIPI headline attributed to each shock as the horizon lengthens. The exchange rate is the dominant external driver.
P6 · Impulse responses of the CIPI headline
Response to a one-unit shock, months after impactP8 · The Dynamic Adjustment Mechanism, back-tested
Flagship product · stylised road, building and water projectsUnder a fixed-price contract the contractor's margin erodes one-for-one as input costs drift. The DAM is a price-adjustment clause: when a Weighted Macro-Volatility Index — the basket's rolling material volatility plus the exchange-rate volatility scaled by the P6 sensitivity — breaches its trigger band, the contract sum is revised upward, capped per period and cumulative for the remainder of the contract.
The back-test picks the most volatile 18-month window in the panel (the 2022 commodity shock) and compares the final margin and its variance under a fixed price against the DAM, for three reference baskets built from CIPI weights. These are simulations on transparently stylised projects, not real contract data.
P9 · Does the network help forecast the CIPI?
Strict walk-forward evaluation, ridge regression on lagsThe network-augmented model adds cross-material dispersion and pairwise-correlation features to the own-lag baseline. A Diebold–Mariano statistic above about 1.96 means the accuracy gain is unlikely to be noise.
The public purse behind public works
A companion monthly panel from the MoFPED fiscal operations tables — the demand side of the construction market, since government is the sector's largest client.
Revenue and expenditure
UGX billion, monthlyTotal public debt
UGX billion, monthly, as publishedPhase 0 is complete. What it unlocks
The foundation is reproducible, tested and documented. The next phases deepen the data, finish the paper portfolio and turn the prototypes into products.
Phase 0 · Foundation
- UBOS CIPI harvested and parsed across every layout change
- MoFPED macro block joined; CPI/PPI reconciliation passed
- Panel v1.0 with splice log, coverage, break and stationarity diagnostics
- Five estimation starters running; regression suite green
Phase 1 · Depth
- Extend the panel backwards with historical UBOS and Bank of Uganda series
- Add the wider drivers: forex, weather, tax-law changes, Hansard
- Automate monthly refresh from the source portals
- Deposit versioned panels with a citable DOI
Phase 2 · Papers
- Complete P1 methodology and the P5 network-centrality paper
- Replace ridge with the GRU forecaster in P9
- Full DCC-GARCH estimation for P4
- Peer review and publication of the portfolio
Phase 3 · Products
- Bid-pricing service on live indices
- Programme risk tool from clusters and breaks
- DAM clause templates and calibration service
- Partner access for UBOS, BoU and the host university